What lies ahead in cancer genetics, epigenetics

January 2007
Feature Story

William Check, PhD

Yin and yang, body and soul, nature and nurture—pairs of apparent opposites that are actually different manifestations of the same underlying reality. To this list we can add genetics and epigenetics, two distinct but intimately related molecular modes by which gene expression can go awry and lead to cancer and other diseases.

“We don’t have any example of any cancer where it is pure genetics—cancers are a mixture of genetics and epigenetics, mutations and abnormal DNA methylation,” says Manel Esteller, MD, PhD, director of the Cancer Epigenetics Laboratory of the Spanish National Cancer Centre, Madrid.

At the 2006 annual meeting of the Association for Molecular Pathology, or AMP, two forms of epigenetics—DNA methylation and microRNA genes—were explored in depth in plenary lectures, but genetics was always the inseparable partner.

Recognizing another fundamental connection, each plenary was sponsored jointly by the solid tumors and hematopathology subdivisions of the AMP. Says Janina Longtine, MD, chief of molecular diagnostics at Brigham and Women’s Hospital and associate professor of pathology at Harvard Medical School, “People in each group are interested in cancer pathogenesis and mechanisms. So we thought this would be a good opportunity to join forces to find speakers who could address these interests.” The two speakers proved satisfactory. “They provided relatively new and very interesting perspectives on the initiating events of cancer,” says Dr. Longtine, outgoing chair of the AMP hematopathology subdivision. Both topics addressed the aims of molecular diagnostics—to find signatures of cancer that are important for diagnosis or prognosis. “I’m not sure either of these was something we will be bringing into the clinic at the moment,” Dr. Longtine says. “But they are both topics with which all pathologists interested in molecular testing want to be conversant.

“We talk about cancer initiation and progression occurring because of alterations in the coding regions of tumor suppressor genes or oncogenes,” Dr. Longtine continues. What she found exciting was that both plenary talks addressed other events that influence expression of tumor suppressor genes or oncogenes. One was about microRNAs having silencing effects, the other about hypo- and hypermethylation influencing expression of oncogenes and tumor suppressor genes. “It was like opening another whole field,” Dr. Longtine says.

“Hypermethylation of gene promoters can act as an alternative to mutation in tumorigenesis,” said Stephen B. Baylin, MD, professor of oncology and medicine in the Sidney Kimmel Comprehensive Cancer Center of Johns Hopkins University, opening his plenary talk on this topic. “We expect to see translational applications of this field emerging,” Dr. Baylin said, adding, “Right now there are still lots of questions about how it will play out in the clinic.” Dr. Baylin defined epigenetic changes as those occurring in heritable gene expression patterns based on changes other than those in the primary base sequence of DNA. Epigenetics concerns “how our DNA is packaged,” Dr. Baylin said; epigenetic activity provides the normal underpinning of development, imprinting, differentiation, and adult cell renewal patterns.

As a striking demonstration of epigenetics’ power, Dr. Baylin cited a study of epigenetic changes in monozygotic twins conducted by Dr. Esteller and his colleagues. Twins share a common genotype but are typically phenotypically dissimilar in many ways. To investigate the basis for these nongenetic differences, the investigators looked at two epigenetic changes: DNA methylation and histone acetylation. “Although twins are epigenetically indistinguishable during the early years of life,” they reported, “older monozygous twins exhibited remarkable differences in their overall content and genomic distribution of 5-methylcytosine DNA and histone acetylation, affecting their gene-expression portrait” (Fraga MF, et al. Proc Natl Acad Sci U S A. 2005;102:10604–10609).

“Monozygotic twins can have different diseases,” Dr. Esteller told CAP TODAY. “One can have breast cancer early and the other one never develops it.” Such differences can’t be attributed to genetics. Environmental factors, such as cigarette smoking, can cause mutations, but that is difficult. “It is easier for environmental influences to cause changes in DNA methylation patterns, which are found in lung cancer cells,” Dr. Esteller says.

In his talk, Dr. Baylin expanded on the mechanisms of hypermethylation and related pathways. Nucleosomes, which are the basic chromatin units, “are at the heart of DNA packaging,” Dr. Baylin said. In addition to DNA, nucleosomes include enzymes and histones. Nucleosomes are arranged in a higher order and can be densely packed to prevent illegitimate transcription. Most DNA is in such areas of closed chromatin. Transcription is only allowed for DNA in regions of linear nucleosomes in the euchromatic state with open promoter chromatin. Methylation of DNA differs between open and closed chromatin regions. Methylation typically occurs on cytosine bases in runs of cytosine-guanosine dinucleotides, called CpG islands, which are often found in promoters or regulatory areas. “Methylation of DNA doesn’t by itself change gene expression,” Dr. Baylin said. “It needs to be coupled with modifications of amino acids in the histone tail. These may determine where methyl groups get placed in the genome.”

In the cancer genome one can observe altered DNA methylation status and altered chromatin status. “Abnormal chromatin structure leads to abnormal gene expression and a local change—CpG islands in areas not usually methylated are methylated,” Dr. Baylin said. “There is a spectrum of changes that convey an abnormal expression state.” In cancer, the pattern is global hypomethylation and specific promoter hypermethylation. Because many pathways are involved in methylation, expression of many genes can be altered—Dr. Baylin mentioned E-cadherin, VHL, p16, p53, genes of the Wnt pathway, and mismatch repair genes, among others. Histone modifications may also be markers. “A network of epigenetic events act together to decrease genetic stability and alter gene silencing patterns,” Dr. Baylin summarized.

An even more complex epigenetic pattern appears in cancer cells, one that was first observed in embryonic stem cells: They have repressive and active markers at the same time. A key system in this bivalent state is the Polycomb Group (PcG) of proteins, which are fundamental to embryonic development (Lee TI, et al. Cell. 2006;125:301–313; Bernstein BE, et al. Cell. 2006;125:315–326). PcG proteins are epigenetic gene silencers. Certain PcG proteins occupy a subset of developmental genes in ES cells that must be repressed to maintain pluripotency; the PcG proteins silence the developmental genes while keeping them poised for activation. PcG proteins are also implicated in neoplastic development (Sparmann A, van Lohuizen M. Nat Rev Cancer. 2006; 6: 846– 856). Many genes in cancer cells are found to be marked by PcG proteins. One speculation is that there is such a thing as a bivalent tumor stem cell that can proceed to a full-fledged cancer cell through abnormal clonal expansion.

Dr. Baylin and a colleague reviewed the findings related to hypermethylation of CpG islands in cancer cells and the reactivation of gene expression by removing methylation (Herman JG, Baylin SB. N Engl J Med. 2003;349:2042–2054). In Dr. Baylin’s view, these observations have implications for translational research. “CpG island promoter hypermethylation may be the ideal tumor marker,” he asserted. “A small panel of markers can cover the entire genome of all human cancers.”

Several people working in this area endorse Dr. Baylin’s view. “Over the last few years it has become clear that hypermethylation is as important as mutation to knock out specific genes,” says David Sidransky, MD, professor of otolaryngology and oncology at Johns Hopkins University. “It is just as important as any of the major genetic changes that occur during progression of cancer. And it is unique—while genes have mutations in many areas, generally hypermethylation occurs in promoter areas, which you can define with PCR. So maybe it will be easier than mutational analysis for detection and quantitation.”

Dr. Sidransky has been studying methylation of a gene called GSTP1, which he has found to be the most common epigenetic change in prostate cancer. Methylation of the GSTP1 promoter region is found in 90 percent-plus of prostate cancers but not in benign prostatic hypertrophy, or BPH. “Most men over age 50 get PSA screening,” Dr. Sidransky notes. “If the PSA level is high, a number of biopsies are done to confirm or not that there is disease. But most of these high screening PSA values come when the prostate is enlarged with BPH, not when there is cancer.” If a biopsy is negative in the presence of a high PSA level, a methylation assay can help determine whether that is a true or false negative, Dr. Sidransky says. He has reported that a panel of methylation markers that includes GSTP1 has about 90 percent sensitivity for cancer with “near perfect” specificity (Tokumaru Y, et al. Clin Cancer Res. 2004;10:5518–5522). The patient is taken out of what Dr. Sidransky calls “PSA hell” —the PSA remains high and the patient keeps returning for biopsies with indeterminate results. Dr. Sidransky’s approach is now being tested in a large corporate-sponsored multicenter clinical trial. “I believe they are planning to get FDA clearance and introduce the assay into the clinic by the end of 2007,” he says.

In the area of breast cancer, too, methylation assays appear promising. “Among all molecular markers available so far, methylation carries the strongest possibility of being applied clinically in early detection and in predicting response to therapy and in prognosis,” says Saraswati Sukumar, PhD, Barbara B. Rubenstein professor of oncology and co-director of the Breast Cancer Program at the Johns Hopkins Sidney Kimmel Comprehensive Cancer Center. Dr. Sukumar says the value of methylation is particularly marked when applied to the very small samples so common in breast cancer today, such as fine needle aspirates of breast or ductal fluid or serum—the three most common sample types. “While RNA expression methods or immunohistochemistry tests are all possible on resected tumor tissue, or perhaps even in core biopsies, the possibility of detecting an errant cancer cell among millions of normal cells is very difficult, almost impossible,” Dr. Sukumar says. “One needs a molecular test to be able to detect these abnormalities in serum or other body fluids.”

Methylation markers are especially valuable because many types of cancers have these markers and hundreds of genes are methylated, though not all in the same tumor type or in the same tumor sample. “So we will have a panel of many genes to test at the same time,” Dr. Sukumar says. “That will make a test that is very robust and covers a large percentage of the population, particularly in a tumor like breast cancer with heterogeneous genetics.” Dr. Sukumar believes that a panel of 12 or even five genes will make it possible to detect almost all breast cancers.

To deal with the problem of quantitation—a cytology sample may have one to 10 tumor cells among thousands of normal cells, while a serum sample may contain one nanogram of tumor DNA in many milligrams of total DNA—Dr. Sukumar developed a two-step assay that she calls quantitative multiplex-methylation-specific PCR, or QM-MSP (Fackler MJ, et al. Cancer Res. 2004;64:4442–4452). This assay allows accurate assessment of promoter hypermethylation for many genes simultaneously in small samples. In the first step, multiplex amplification is done with DNA primers flanking CpG islands of interest for several genes. These primers recognize methylated and unmethylated DNA equally well. In the second step, an aliquot of the first amplification reaction is subjected to a quantitative PCR assay using the bisulfite method and primers that discriminate methylated DNA from unmethylated DNA. Comparing the amount of methylated DNA for a particular gene relative to the unmethylated form of that same gene—using a gene as its own internal control—provides a great advantage over using housekeeping genes as a reference, Dr. Sukumar says, since different genes amplify differently.

With QM-MSP, the methylation status of several genes can be determined simultaneously. A quantitative methylation score is obtained by adding up methylation ratios for all measured genes and expressing that number as a cumulative methylation index. Using the methylation status of four genes, Dr. Sukumar and colleagues detected carcinoma with 84 percent sensitivity and 89 percent specificity in 28 samples (19 carcinoma, nine normal) of 50 to 1,000 epithelial cells collected from breast ducts during endoscopy or by lavage. In a comparative evaluation using cells from 37 ductal lavage samples from women undergoing mastectomy (27 with cancer and three without), QM-MSP doubled the sensitivity of detection of cancer compared with cytology, 62 percent versus 33 percent (Fackler, MJ, et al. Clin Cancer Res. 2006;12:3306–3310). Specificities were 82 percent and 99 percent, respectively.

Dr. Sukumar sees this assay as a potential adjunct to cytology. “We hope it will also become an adjunct to mammography,” she adds. Methylation would not be a diagnostic test but a risk-assessment assay: When mammography does not detect a lesion, can the methylation assay indicate the likelihood of a very small tumor? In general, Dr. Sukumar emphasizes, methylation would be used for risk assessment or determining intermediate response to therapy. “Can we detect abnormalities in high-risk individuals and flag them for prevention trials or close surveillance?” she asks. Also, at the start of treatment for metastatic breast cancer there are high levels of gene methylation in serum. “Will a drop in methylation markers predict that that patient is responding to therapy?” she wonders. Large-scale testing and validation will begin shortly using serum, and FNA samples with cytology as the gold standard.

(Dr. Sukumar notes that the use of methylation markers occupies a different niche from Genomic Health’s Oncotype DX assay, which is claimed to predict the risk of recurrence and response to therapy in diagnosed hormone receptor-positive breast cancer.)

“DNA methylation is a critical event in the etiology of cancer,” Dr. Esteller agrees. Mutations are important in familial cancers, but sporadic cancer makes up 90 percent of cases and methylation is very important in many of these sporadic cancers. For example, Dr. Esteller says, one-fourth of primary breast tumors have abnormal methylation of the familial breast cancer gene BRCA1. And, while the mismatch repair gene hMLH1 is mutated in cases of familial colon cancer, in sporadic colon cancer methylation of hMLH1 is the operative oncogenic mechanism.

Dr. Esteller is working to characterize cancers with DNA methylation signatures. “It is like a new fashion to typify tumors according to expression profiles with microarrays,” he says. “We can also do this with DNA methylation markers.” For instance, he found that DNA methylation signature profiles correspond to different progression rates of colorectal tumors (Frigola J, et al. Hum Mol Genet. 2005;14:319–326). His group has also characterized DNA methylation profiles of neuroblastoma in 150 children. “Several DNA methylation markers were able to elaborate different prognostic profiles,” he says (Alaminos M, et al. J Natl Cancer Inst. 2004;96:1208–1219).

In addition to its potential uses as a marker of early disease and predictor of response to therapy, Dr. Esteller emphasizes that DNA methylation is also a target for drug development (Gore SD. J Natl Compr Canc Netw. 2006; 4: 83– 90). Two drugs for the treatment of leukemia and myelodysplastic syndromes—5-azacytidine and 5’-aza-2-deoxycytidine—are demethylating agents that restore normal levels of DNA methylation and reactivate tumor suppressor genes. In addition, histone deacetylase inhibitors, or HDACs, are now approved for treating advanced cutaneous T-cell lymphoma and Sezary syndrome and are being evaluated for wider indications. “These drugs restore normal patterns of histones,” Dr. Esteller says. The first approved drug in this class, vorinostat, was discovered in part because it was shown to inhibit proliferation in cell culture and had a similar structure to molecules that inhibited HDACs. Clinical trials combining a demethylating agent and an HDAC inhibitor are underway (Gore SD. Nat Clin Pract Oncol. 2005; 2 Suppl 1: S30– S35). Proteins of the PcG system are also altered in hematological and epithelial cancers, and agents that act on these proteins are being developed as specific chemotherapeutic agents (Raaphorst FM. Hum Mol Genet. 2005; 14: R93–R100).

Steven Belinsky, PhD, director of the Lung Cancer Program at the Lovelace Respiratory Research Institute, Albuquerque, NM, is working with Dr. Baylin and others to study DNA methylation as a method for early detection of lung cancer in high-risk persons—heavy smokers who have chronic obstructive pulmonary disease. In one study, they analyzed sputum samples that had been collected before diagnosis from 98 people who developed lung cancer and from 92 who did not. Promoter methylation of six of 14 genes tested was associated with a greater than 50 percent increased lung cancer risk. If any three of these six genes were simultaneously methylated, the patient had a 6.5-fold increased risk of lung cancer; having three genes methylated predicted lung cancer with a sensitivity and specificity of 64 percent (Belinsky SA, et al. Cancer Res. 2006;66:3338–3344). “We are continuing to evaluate more cases and more controls and to test new genes in our panel,” Dr. Belinsky says. “Our goal is to increase sensitivity and specificity to the 80 percent range, which we believe will put us into the clinical validation setting. We are getting very close to being able to do that with our revised panel.”

A major question is whether early diagnosis and treatment will lead to increased survival. Dr. Belinsky says lung tumors detected at stage IA (<3 cm, no invasion) have a five-year survival rate of 70 percent to 80 percent, compared with 50 percent for those diagnosed at stage IB. (Currently, most lung tumors are detected at stage III.) “Intuitively you would conclude that if we detect lung cancer earlier, people will survive longer,” he says. With a methylation marker panel, this idea can be tested. “Our goal is to use methylation as a front-loading test to identify people at highest risk and refer them to clinical intervention—spiral CT and possibly bronchoscopy,” Dr. Belinsky says.

To complicate the methylation story further, in a subset of colorectal cancers it appears that aberrant methylation of some CpG islands occurs at an exceptionally high frequency. This has been proposed as a distinct epigenetic trait called CIMP—CpG island methylator phenotype—and CIMP-associated methylation of MLH1 has been suggested to be the basis of sporadic microsatellite instability and mismatch repair deficiency (Weisenberger DJ, et al. Nat Genet. 2006;38:787–793). CIMP is anything but “CIMPle.” “Nobody knows the exact underlying molecular mechanism for the CIMP phenotype, unlike mismatch defects for microsatellite instability phenotype,” says Shuji Ogino, MD, PhD, assistant professor of pathology at Harvard Medical School and staff pathologist at Brigham and Women’s Hospital and Dana-Farber Cancer Institute. Even in normal tissue, there may be increased methylation in some promoter CpG islands as individuals age. Dr. Ogino says some scientists speculate that the protective effect of folate may be mediated through prevention of the CIMP pathway: In cohort studies, people who eat more vegetables have a lower risk of colorectal cancer.

Dr. Ogino has taken part in large-scale prospective cohort studies using quantitative real-time PCR measurement of eight methylation markers and has shown that four to eight markers fairly accurately identified the CIMP cases (Ogino S, et al. Gut. 2006;55:1000–1006; Ogino S, et al. J Mol Diagn. In press.). “This is analogous to MSI [microsatellite instability] testing, where there are many markers but the best markers are BAT25 and BAT26,” Dr. Ogino says. “Not all CpG islands perform in a similar way, and some methylation markers are better than others.” They also found a clear bimodal distribution of 130 MSI-high tumors according to the number of methylated promoters, which led them to conclude that “CIMP is a distinct epigenotype of colorectal cancer.”

How to make clinical benefit of these findings is still under investigation. Dr. Ogino suggests that the CIMP trait may help predict therapeutic response/resistance and prognosis. “One group showed an association between CIMP-positive colorectal cancer and survival benefit from chemotherapy [van Rijnsoever, et al. Clinical Cancer Res. 2003;9:2898–2903], while another group showed that CIMP predicted worse prognosis in microsatellite-stable colorectal cancer [Ward, et al. J Clin Oncol.2003;21:3729–3736]. There is still much to be learned about effects of CIMP on tumor behavior and patient survival,” he says. Dr. Ogino and colleagues are attempting to clarify the biological effects of CIMP, using robust standardized methods and a large group of patients at different stages of disease. Another potential use of DNA methylation testing is to favor one way or the other for the diagnosis of hereditary nonpolyposis colorectal cancer, or HNPCC. CIMP positivity is rare in HNPCC cases. However, CIMP negativity can be seen in some sporadic MSI colorectal cancers. Thus, “nothing is perfect at this point, and further investigation is clearly necessary in cancer epigenetics,” Dr. Ogino says.

In the other plenary presentation, which focused on the role of changes in the expression of microRNA genes in cancer, Carlo M. Croce, MD, told the assembled molecular pathologists, “You will all be looking at microRNA profiles within five years because they are tissue-specific.” Dr. Croce has shown that expression profiles of microRNA (miRNA) genes on a microarray chip can distinguish various tissues (Liu CG, et al. Proc Natl Acad Sci U S A. 2004;101:9740–9744). He noted that about 15 percent of human cancers appear as metastases from a primary of unknown type. “The tissue of origin can be identified with this chip,” said Dr. Croce, who is John W. Wolfe chair in human cancer genetics, director of the Institute of Genetics, and director of the Human Cancer Genetics Program at the Ohio State University Comprehensive Cancer Center.

Dr. Croce was the first to identify miRNA genes as important factors in human oncogenesis. While attempting to find genetic alterations in B cell chronic lymphocytic leukemia (B-CLL), he sequenced the chromosome 13q14 region, which is deleted in more than half of B-CLL cases. However, his group could not identify a responsible mutation, not even when Dr. Croce himself went into the laboratory to do hands-on analysis. “Even when we cloned two mb of DNA and sequenced all the genes in the responsible region, we failed to find any gene specifically altered in CLL,” Dr. Croce said. “People reached the same conclusion in other labs around the world.”

The answer came from a B-CLL case that Dr. Croce obtained from a physician in a CLL consortium, a case in which the translocation breakpoint was contained in a 31.4 kb segment that was deleted in another CLL case. Though no mutated gene was found, Dr. Croce did notice that two miRNA genes—miR15 and miR16—were located right at the breakpoint, 13q14.3. MicroRNAs had been identified in 1993 as small (21-22 nucleotide) double-stranded RNAs that control the timing of developmental events in the nematode Caenorhabditis elegans by inhibiting translation of target mRNAs through an antisense RNA-RNA interaction. In 2001, miRNA genes were found to be abundant in many species, including humans (Lee RC, Ambros V. Science. 2001; 294: 862– 864). Dr. Croce and his colleagues went on to show that miR15/ 16 genes are deleted or down-regulated in the majority (approximately 68 percent) of CLL cases (Calin GA, et al. Proc Natl Acad Sci U S A. 2002; 99:15524–15529). He called loss of miR15/16 “a very early event in the indolent form of B-CLL.” Since miR15/16 transcripts inhibit translation of protein coding genes, in particular the apoptotic regulator gene Bcl2, and their loss leads to cancer, they can be considered tumor suppressor genes.

Dr. Croce’s laboratory went on to show that different patterns of expression of miRNA genes distinguish normal B cells from malignant B cells and that miRNA gene expression profiles correlate significantly with progressiion interval. In B-CLL, the miRNA signature is associated with prognostic factors such as mutations in the immunoglobulin heavy-chain variable-region gene or high expression of ZAP-70 protein (Calin GA, et al. N Engl J Med. 2005;353:1793–1801). Moreover, they reported, “Germ-line or somatic mutations were found in five of 42 sequenced microRNAs in 11 of 75 patients with CLL, but no such mutations were found in 160 subjects without cancer.”

Many miRNA genes map to fragile regions of the genome involved in cancer, according to work from Dr. Croce’s laboratory (Calin GA, et al. Proc Natl Acad Sci U S A. 2004;101:2999–3004). Alterations in miRNA gene expression profiles are found in solid tumors as well as hematopoietic cancers (Volinia S, et al. Proc Natl Acad Sci U S A. 2006;103:2257–2261). In some cases it is overexpression of a miRNA gene that bodes poorly. In lung cancer, for example, high expression of the miR155 gene is associated with poor prognosis (Yanaihara N, et al. Cancer Cell. 2006;9:189–198). “We have defined miRNA gene signatures for six human solid cancers—stomach, prostate, lung, colon, breast, and pancreas,” Dr. Croce said. “If this work is verified, these deranged miRNA genes would make wonderful therapeutic targets.” He is now working to develop miRNAs or corresponding antisense molecules as therapeutic entities.

While Dr. Longtine believes more work is needed on both of these mechanisms—altered methylation and miRNAs—before they emerge into clinical pathology, she notes that methylation now has “a broader base.” She considers the evidence for the importance of methylation changes in cancer “quite compelling.” However, it does face a technical challenge. Dr. Longtine calls a methylation-specific PCR assay marketed by Oncomethylome Sciences (Dr. Baylin developed this assay and is a consultant to this company) “quite laborious. It is not so easy to do in the clinical lab,” she says. Like many other fields, methylation assays are moving toward real-time PCR.

Turning to miRNAs, Dr. Longtine says that Dr. Croce’s argument about the role of miR15/16 in CLL is also compelling. She thinks miRNA profiles might eventually be useful for prognostication. However, just as with RNA expression microarrays, it is not at all clear now which signatures will be most helpful for patient care. “We need a clinical trials approach to this question,” she says. She is also not convinced by Dr. Croce’s idea to use miRNA profiles to identify the tissue of origin of metastases. “Do we really need to know the primary tissue for mets of unknown origin?” she wonders.

Another issue arises from the redundancy built into the miRNA system. miRNAs bind to their targets with imperfect complementarity. This is great for their biological activity because it allows one miRNA to regulate several genes. However, it may not be so great for diagnostic accuracy. And it could pose problems for using miRNAs as therapeutic agents.

Dr. Sidransky’s perspective on the clinical application of these two epigenetic mechanisms is to look at them as boiling pots of water. “Right now they are percolating,” he says. “There are lots of what I call studies of convenience.” In these studies different markers or panels of markers are being explored in a clinical setting. “At some point the work will take steam—clinical utility or benefit will be demonstrated—and the method will get out of the pot” and into clinical practice.