When I arrived in Baltimore, I learned that the GI path group was working closely with Bert. I began working on the clinical correlation of specific chromosomal defects with the appearance and behavior of colorectal adenomas and cancers. Within a little over a year, I joined Bert's lab group for a term of three years. I worked mostly on the biochemistry of the p53 protein. When I formed my own research lab in 1992, I had never sequenced a gene to look for mutations, and I had to do some catching up while studying the mutation patterns found in my own research interest, colitis-associated dysplasia.

Interestingly, a chance encounter of my chairman (now Dr. Yardley) with the chair of surgery (Dr. John Cameron) on a cross-country flight proved the most influential. Dr. Cameron alerted Dr. Yardley that the number of pancreatic cancer resections was about to skyrocket at our institution due to surgical advances. He forcefully asked why there were no molecular biologists waiting to take advantage of it. Dr. Yardley put a bug in my ear and also into that of Dr. Ralph Hruban, a new faculty member in pathology. Ralph and I spoke briefly in the hall and talked in general terms about possible future studies. Dr. Cameron passed the same enthusiasm on to a young surgeon, Dr. Charles Yeo.

Charlie established a multidisciplinary group that has met bimonthly since 1991. At first, we had not much to talk about. Charlie soon submitted a grant occasioned by a special surgical RFA (request for applications) that provided half a technician's salary for Ralph and me, which we used to bootstrap our studies of pancreatic cancer. The McDonnell Foundation, upon the recommendation of our Cancer Center, awarded the money needed to properly set up my lab and do the preliminary studies. We obtained a SPORE (Specialized Program of Research Excellence) grant in 1993 on the subject. Over the past 13 years, Ralph and I have used the SPORE grant, the strengths of our surgeons, and many, many talented collaborators to foster new careers and productive research in pancreatic cancer.

For the first few months upon entering Bert's lab, I joined the ongoing mapping of a locus on chromosomal arm 18q. Eric Fearon, a brilliant graduate student of Bert, had found a homozygous deletion there. We all were familiar with the story of the retinoblastoma gene having been found as the target of homozygous deletions. Conceptually, the genes found within the boundaries of a homozygous deletion were candidate tumor-suppressor genes, and homozygous deletions were thought otherwise to be rare. The DCC gene was the only gene identified within the deletion, and the published report became a citation classic.

Subsequent sequencing of the gene as published by the original lab and others, however, yielded few mutations. People make their own conclusions from data, but many investigators found that DCC raised as many questions as it answered. Nonetheless, herd mentality sometimes takes hold of the literature (see our discussion in Cancer Biol. Ther. 3:903-910, 2004), and most of the citing authors referred solely to the initial finding of DCC rather than to the subsequent and brutally honest report of the low mutation rate by the same group. The citations also omitted the fact, although described in the initial paper, that one end of the homozygous deletion had not been mapped (due to having run out of tumor tissues). Thus, the full gene content of the 18q deletion was still unknown.

It took a series of failed strategies before my lab could readily study the deletion patterns in pancreatic cancer, but we finally found that growing the patient's tumor in mice, as a xenograft, was fruitful. Even after multiple passages, the tumors remained incredibly faithful to the patterns of mutations found in the patient's initial tumor. Stephan Hahn, a postdoctoral fellow in my lab provided by Wolf Schmiegel and funded by the Deutsche Forschungsgemeinschaft, decided to study chromosome 18q with every marker he could obtain—even 100 markers if needed (which was a lot back then). Although this was a fairly unbiased approach, after only about 23 markers he had found a hotspot of homozygous deletions in a very familiar place. They clustered on the very side of DCC on which the earlier mapping had not been completed. Nearly a third of all the pancreatic cancers had these deletions. Toward the end of this project, the NIH wrote us to inform us that the study section rated our approach as fundamentally flawed and within the bottom 50% of applications. They explained that the proposal was not a significant improvement over the prior grant we had submitted concerning the localization of BRCA2, at a locus we termed DPC1/DPC2, for Deleted in Pancreatic Carcinoma at loci 1 and 2.

Only months later, we published the 18q gene in Science. We termed it DPC4, for it was found at homozygously deleted locus 4. DPC4, now known officially as SMAD4, had both homozygous deletions and intragenic mutations expected to inactivate the gene. Based on sequence-similarity, we proposed that it mediated signals similar to the Mad genes of Drosophila and the TGFbeta-like signals of vertebrates. The report became the most-cited cancer research paper of 1996. About the same time, the BRCA2 gene was cloned by our collaborators in the UK and in Utah who were using our inside information from the deletion boundaries we were defining. The journal Science reported this news accompanied by a map—the same map of the pancreas cancer homozygous deletion on chromosome 13q that had twice failed to impress the expert panel of NIH reviewers.

It is important to recognize that the surgeons, pathologists, our Cancer Center, Dr. Schmiegel, the McDonnell Foundation, and the NIH SPORE system recognized the importance of this translation of research that bridged the clinic and the lab bench. I have since repeatedly observed that translational science can be mishandled by the standard NIH review system operated by CSR (Center for Scientific Review) and by some levels of the NCI. Scientists do not always agree on what is valuable.

DPC4/SMAD4 was subsequently applied to clinical, basic science, and translational uses. With collaborators, we found a family of other Smad genes in humans, and we found a DNA sequence bound by DPC4/SMAD4, establishing it as a transcription factor and permitting functional assays and the identification of genes downstream of and directly regulated by DPC4/SMAD4. DPC4/SMAD4 was identified to be the cause of the inherited syndrome, juvenile polyposis. Further study of DPC4/SMAD4 emphasized its role in transmitting signals from TGFbeta, activin, and bone morphogenic protein receptors.

Owing to this realization, we found tumor mutations in the TGFbeta receptor type I and activin receptor types I and II played a role in pancreatic tumorigenesis. Mutations in a BMP receptor were found to be another cause of juvenile polyposis. Engineered knockouts of DPC4/SMAD4 showed it to play a role in vertebrate development. An immunohistochemical assay for DPC4/SMAD4 loss in tumorigenesis confirmed the causative relation of precursor lesions to the invasive cancers in the pancreas. The same assay now provided a clinically useful assay to identify cancer in otherwise difficult and small clinical biopsies of pancreatic masses. The lessons from the mapping of the DPC4 homozygous deletions were applied successfully in the cloning of other tumor-suppressor genes, including PTEN, by colleagues.

Pancreas cancer remains difficult to diagnose in an asymptomatic stage. Once symptoms occur, the tumor is almost invariably at a fatal stage, having created distant micro-metastases. Our studies of mutations in the precursor lesions in the pancreas have fleshed out a model, which in turn conveys some hope. In this model, the most important precursor lesions arise in the ducts in adulthood and a minority progress over a period of decades to become the more aggressive forms of the precursor and eventually invade to constitute the lethal stage, ductal cancer. It is likely, however, that high-grade precursor lesions probably exist for some years before the stage of invasion and metastasis. This intermediate stage provides an opportunity for surgical removal at a potentially curable stage.

Patients in high-risk families can now be enrolled in an early-detection screening program that, while experimental, has identified early cancerous lesions in asymptomatic patients. In the highest-risk families, nearly 18% were found to have inherited mutations in BRCA2, allowing us to inform the families of the precise reason for their affliction. Other inherited gene mutations are also now known to cause pancreatic cancer as well, although BRCA2 remains the most common known cause.

As introduced above, difficult biopsies are now studied immunohistochemically for markers of pancreatic cancer, permitting a firm diagnosis where, only a few years ago, additional diagnostic techniques might have been required and delayed the institution of appropriate therapy. Some of the proteins found to be overexpressed in pancreatic cancer have reached a practical stage of clinical evaluation, including ongoing clinical trials of antibody therapy against specific surface proteins, development and human testing of a post-surgically administered cancer cell vaccine that is now shown to generate immune responses in the patients against tumor-specific proteins, and active development of new immunologic agents designed to elicit more direct immune responses against the same tumor markers.

In experimental trials, the treatment of ductal cancer is now taking advantage of genetic signatures of the cancer cells for which one would expect a high responsiveness of the tumor cells to specific and available drugs. For example, mutations in the BRCA2 gene and the BRCA2 cellular pathway are being used investigationally to assign individual patients to specific families of drugs, or to explain the different responses of individual patients to such drugs.

Our genetic studies also identified subsets of tumors which appeared not to follow the usual patterns of mutations. One major subset was subsequently recognized as a new diagnostic classification of pancreatic cancer, the medullary cancers. This subtype is genetically, histologically, and clinically different from the pancreatic cancer with which it was previously classified.

Pancreas cancer is unique in a numbers of features. Most strikingly for the researcher is the characteristic desmoplastic, or scarring, reaction of the normal tissues to the cancer cells within the tumor. Thus, the majority of the tumor volume comes from the non-cancerous reaction. This impairs the study of pancreas cancer biology, for the cancer cells within a pancreas cancer must often be separated from the normal tissues prior to study; this was the advance provided by xenografting the tumors to mice in order to facilitate genetic analysis.

Pancreas cancer also engenders systemic effects which seem exaggerated in relation to the volume of the tumor itself; these include cachexia, loss of appetite, malaise, and weight loss, and suggest the presence of a circulating factor released by the cancer cells or by the body's reaction to the cancer cells.

Pancreas cancers are highly resistant to virtually all forms of chemotherapy. This is especially unfortunate, since surgery alone will not provide a true cure for the disease. In contrast, half of colorectal cancers are cured by surgical steel alone, and more than half of breast cancers are also.

Radiographic imaging cannot distinguish the cancers from the density of the surrounding normal tissue, and cannot distinguish the cancer reaction from a benign pancreatitis. Thus, the cancers are not readily screened by external imaging. A moderately invasive procedure, endoscopy, is required to see the disease in its early forms, before it has become large and its spread has become clinically apparent.

The mortality of pancreas cancer is about 33,000, just a hair behind breast cancer, which kills about 41,000 in the U.S. annually. Yet research funding for pancreatic cancer is relatively minor when compared to other major cancer killers. There are still extremely few research scientists studying pancreas cancer full-time.

Unlike breast and other cancers, pancreatic cancer is rather homogeneous genetically. Nearly all have a mutation in the p16 gene, over 90% mutate the KRAS gene, 75% mutate the p53 gene, and just over 50% inactivate DPC4/SMAD4. These rates are all among the highest seen in any malignancy, and many pancreas cancers have mutated all four genes. This severity of the genotype may, if we be allowed to speculate, explain the severity of the clinical course. Perhaps no other cancer is genetically as "screwed up."

Although about 95% of the cells of a normal pancreas are "acinar," meaning that they produce digestive enzymes, well over 90% of the cancers are of ductal cell type. Thus, research comparisons of pancreatic cancer cells to "normal pancreas" are seldom of any value! This fact continues to amaze and even accost those of us in this field. This fact has greatly impaired the speed of research and impairs the value even of some of the published studies.

Scott E. Kern, M.D.
Sidney Kimmel Comprehensive Cancer Center
Johns Hopkins University School of Medicine
East Baltimore, MD, USA