XIII International Charles Heidelberger Symposium and 50 Years of Fluoropyrimidines in Cancer Therapy Held on september 6 to 8, 2007 at New York University Cancer Institute, Smilow Conference Center.

Targeted therapy has become the ideal of therapy in general and of cancer therapy in particular. From this perspective, 5FU and methotrexate (MTX) can be considered as the first rationally designed targeted anticancer drugs. Furthermore, the development of 5FU emerges as a prototype of translational research, not only in the last two decades when such research has been emphasized but, thanks to Heidelberger, even in the first few years after its discovery and introduction into the clinic. 5FU was developed on the basis of findings in the 1950s that cancer cells incorporated a larger amount of the uracil base into the DNA than normal cells ( 1 , 2 ). From the halogen-substituted uracils, 5FU appeared to be the most active and promising drug ( 3 ), and clinical studies were immediately initiated ( 4 ), in which not only plasma concentrations were measured but also tissues’ disposition using radioactive 5FU ( 5 ). It was rapidly recognized that 5FU was metabolized by enzymes from normal pyrimidine metabolism ( 6 ) and that it was incorporated both into RNA and DNA ( 5 ), although it has long been debated what accounted for its antitumor effects. Also degradation pathways of 5FU were similar to the natural bases ( 7 ), demonstrating an important role for dihydropyrimidine dehydrogenase (DPD). 5FU can be activated by the subsequent action of uridine phosphorylase and uridine kinase ( 8 ), while Heidelberger already postulated that a activation via the thymidine phosphorylase and thymidine kinase pathway is unlikely due to lack of the co-substrate deoxyribose-1-phosphate. Later some significant differences were found: in contrast to uracil, 5FU is an excellent substrate for mammalian orotate phosphoribosyltransferase (OPRT) (ref. 9 ), for which an association with 5FU sensitivity was observed both in experimental models and in clinical samples. It was also recognized by Heidelberger’s group that the 5FU metabolite 5-fluoro-2′-deoxyuridine monophosphate (fdUMP) is a very potent suicide inhibitor of the pyrimidine de novo enzyme thymidylate synthase (TS) (refs. 6 , 10 , 11 ) and resistance to 5FU was already associated with disturbances in TS ( 12 ). In the early 1980s a breakthrough in this research was achieved by the Heidelberger laboratory in applying sensitive biochemical assays to determine the role of TS activity and the extent of TS inhibition in relation to the response to 5FU in animal models ( 13 ), followed by similar studies in patients ( 14 ). This led to more extensive studies that clearly demonstrated that response to 5FU was related to TS activity ( 15 , 16 ). Two novel technical developments enabled the laboratory to perform retrospective studies on the role of TS in large patient populations: the development of specific antibodies against TS ( 17 ) and sensitive reverse transcriptase (RT)-PCR assays ( 18 ). In a retrospective study, Peter Danenberg’s group demonstrated that low tumor TS and DPD levels were associated with a very high response rate and survival compared to the high TS and/or DPD group ( 19 ). Application of this knowledge in a prospective study by measurement of pretreatment TS and DPD levels demonstrated that it is possible to double the response and survival to single agent 5FU (with leucovorin) by selecting patients with both a low TS and DPD level ( 20 ). Recent developments in therapy of colorectal cancer have demonstrated the additional benefit of concomitant chemotherapy with e.g., oxaliplatin ( 21 ) and irinotecan ( 22 ), as well as the benefit of anti-angiogenic treatment with the vascular endothelial growth factor (VEGF) antagonist Bevacizumab ( 23 ). Therefore, predicting response to 5FU combination treatment on the basis of TS and DPD analysis in selecting patients may need to be revisited. What can we learn from these studies for current targeted drug development? Sledge ( 24 ) defined targeted therapy in its simplest form as “drug plus a molecular target”; hence targeted therapy implies therapy toward a molecularly defined target. A targeted therapy should attack a biologically important process (usually but not necessarily a single molecule) preferably one central to a hallmark of cancer. The target should be validated in the laboratory but subsequently needs validation in the clinic as well; the target should be measured reliably in order to prove its validity. When 5FU was developed a substantial amount of information on its potential metabolism was known compared to other drugs developed in that period. Its target(s) were rapidly characterized. However, available techniques were insufficient to properly validate the target in patients, although some preliminary measurements were performed ( 5 ). The latter, target validation in the tumor and normal tissues/body fluids, can be considered as essential for developing novel targeted drugs. For this purpose we can now use gene expression, gene polymorphisms, and protein expression [activity, immunohistochemistry, enzyme-linked immunosorbent assays (ELISAs)]. In several studies 5FU’s anticancer activity has been shown to be related to TS expression ( 25 ). However, current research has clearly shown that this relation is not straightforward. In untreated patients high TS is a poor prognostic factor, but in adjuvant therapy this may be favorable because TS is also linked to proliferation, which makes a tumor sensitive to antiproliferative drugs. However, in advanced disease a low TS is favorable because it is easier to inhibit a low TS activity completely compared to a high TS. This inhibition has been shown in various clinical studies ( 26 ), and similar data were also shown in this symposium. Comparison of studies performed in the same patients with different methods (activity, protein expression, gene expression, gene polymorphisms) has also shown that one should be careful in extrapolating data from laboratory models to a patient. For example, in cell-free models a clear linear correlation may exist between activity, gene expression, and immunohistochemistry as was shown for the triple repeat of the TS enhancing region (TSER) in the promoter and gene expression ( 27 ). However, regulation of TS in a tumor is far more complicated, since TS translation is regulated by TS itself ( 28 ), while the transcription may also be regulated by p53 ( 29 ), which is obviously deregulated in tumors with a mutant p53. Hence, TS polymorphisms in lymphocytes are unlikely to predict the response of a tumor ( 30 ). They may, however, predict toxicity. Similarly, a low systemic DPD activity led to severe, sometimes life-threatening 5FU toxicity ( 31 ). Although several polymorphisms were linked with this low activity, the relation was not sufficiently strong to use sequence analysis for screening of patients, leaving enzyme activity analysis as the most reliable biomarker for DPD ( 32 , 33 ). Although single nucleotide polymorphism (SNP) analysis is becoming a major tool in defining differences between individuals, one should be careful in extrapolating data from normal cells to the tumors. Nevertheless, quick and accurate genotypic and phenotypic assays should be made available to allow clinicians to select patients for adverse events. Research with 5FU has clearly shown that analysis of both the tumor and normal tissues is essential to select patients for optimal personalized treatment. One important lesson from the past 50 years is that the target and its regulation should be clearly defined at all levels, such as gene copy number, gene expression, epigenetic regulation, posttranslational and posttranscriptional regulation, differential expression in tumor/normal tissue, and for enzymes kinetic parameters and intracellular regulation by signal transduction, normal metabolites, and other drugs. Next to that, essential pharmacokinetic parameters, such as bioavailability for oral drugs, metabolism, and tissue penetration and retention should be evaluated as well. Since most drugs are given in combinations, interactions with other drugs should be characterized, both at the tumoral level and in normal tissues (e.g., with endothelial cells for anti-angiogenic compounds), in addition to potential chemical interaction. Obviously, defining a novel target is an important first step, leading to successful treatment when carefully developed.

Watanabe described his laboratory’s attempt to study prostate cancer in vitro and replicate the tissue environment in an effort to have more reliable surrogates of sensitivity to therapeutic interventions. In spheroids he has shown that cells make VEGF more than in two-dimensional cultures; also oxaliplatin is more effective in spheroids, although efficacy varies with zone: central, which is not viable; the intermediate zone, which is mostly quiescent and shows increased p27 and PARP; and the proliferating zone; the intermediate zone is under hypoxic stress and shows many methylated genes and increased resistance. In an effort to replicate the in vivo environment, he is attempting to establish spheroid cultures with cocultures of stroma, extracellular matrix, and adipocytes. Peter Houghton focused on insulin-like growth factor-1 (IGFR1) receptor as a therapeutic target in childhood sarcoma. Interest in this area has derived from the known t( 2 ; 13 ) translocation that turns on IGF2. Other sarcomas also may be influenced by this pathway: IGF1 is a mitogen for osteosarcoma and IGF-BP3 is increased in Ewing’s where T( 11 ; 22 ) is present. Accordingly, Houghton has been interested in IGF1R inhibitors that are being considered for clinical trials, and in their possible synergy with mTOR inhibitors: IGFs appear to protect against rapamycin-induced apoptosis; therefore, a dual blockade may enhance the efficacy of rapamycin ( 52 ). Robert Ladner shifted the discussion to one gene: dUTPase. This is an S-phase dependent gene that exerts a marked influence on dUTP pools, that are usually tightly controlled. When dUTPase is overexpressed, small interfering RNA (siRNA) leads to up to 60 × sensitivity to FUDR in MCF7 and SW620 cells. Oxaliplatin also reduces dUTPase expression, possibly an explanation to the known oxaliplatin/5FU synergy. Other inhibitors of dUTPase are being sought ( 53 ). TS polymorphisms have been among several determinants of drug action studied by Heinz-Joseph Lenz. High TS activity is associated with shorter overall survival in various colorectal cancer trials. Its down-regulation by a drug such as vorinostat is being tested clinically in refractory disease. The significance of germ-line polymorphisms in the promoter is also being investigated. Focusing on identifying factors leading to decreased response to fluoropyrimidines may lead one to treat with other agents, in fact, in colon cancer, high TS expression seems to be associated with better response to irinotecan. Three talks focused on folates. David Goldman described the unique features of pemetrexed, and perhaps most relevant that it is a 300 × better substrate of folyl polyglutamyl synthetase (FPGS) than methotrexate. An additional feature of pemetrexed may be that it is a substrate of a newly described protein-coupled folate transporter (PCFT) that is commonly expressed in spleen, liver, and the upper gastrointestinal tract, and functions best at an acidic pH. This channel for pemetrexed was discovered by comparing its activity in wild-type or reduced folate carrier (RFC)-null cells: collateral sensitivity to pemetrexed was found in the latter ( 54 ). Since PCFT is widely expressed in tumors, the selectivity of pemetrexed over other antifolates may be enhanced in these tumors; moreover, its efficacy may be affected by the level of folate. Ann Jackman provided an overview of the potential for TS folate inhibitors tracing its historical origins on CB3717, raltitrexed, and pemetrexed, and then focusing on two novel agents. In addition to their activity against the enzyme and other enzymes where folate are cofactors or being regenerated to reduced folates by DHFR, knowledge of transport and their affinity for FPGS are key features in predicting their spectrum of activity. An antifolate being developed is BGC 9331, which is transported via the RFC, but in contrast with the other TS inhibitors is not a substrate for FPGS. Phase I and II clinical studies are complete with activity observed in several tumor types, e.g., gastric cancer. Another compound, BGC 945, with high affinity to the α folate receptor and very low affinity to the RFC is being developed for treatment of tumors expressing high folate receptor levels; very little systemic toxicity is expected ( 55 ). Clinical studies are planned and a range of biomarkers are being developed (e.g., folate receptor levels). Furthermore, TS inhibition may be studied by using 18Fluorothymidine as a reagent for positron emission tomography (PET) imaging. Richard Moran studied the epigenetic control of tissue-specific expression for FPGS. He examined the contrast between the promoters in liver and kidney, and the promoters in proliferating tissues and tumor that utilized IRES and are unmethylated ( 56 ). Gemcitabine was the subject of Bill Plunkett’s and Fritz Peters’ talks. This fluoropyrimidine drug was selected for development at Lilly by the late Gerald Grindey and was the first drug to show consistent clinical benefit in clinical trials of patients with pancreatic cancer. Plunkett provided the rationale for the development of fixed-dose rate schedules that have been tested with varying results in the clinic. He also pointed out that gemcitabine diphosphates inhibit ribonucleotide reductase, although the major antitumor effects probably derive from activation to triphosphates and DNA incorporation. Peters and his group have done comprehensive studies on gemcitabine nucleotides and their perturbation in various clinically relevant combinations (cisplatin, taxanes, radiation, and more recently bortezomib), although mostly confined to determination in peripheral blood mononuclear cells. He is also exploring ribonucleotide inhibitors to potentiate the effects of gemcitabine, and the role of cytidine deaminase in predicting the likelihood of response; lower levels signify a better response, but it also is a marker of toxicity, and patients with overexpression may be undertreated. Kathy Danenberg’s presentation focused on how to better predict response of various tumors to both chemotherapy and “targeted” therapies. Predicting response from determinations on formalin-embedded tissues has been central to the effort of Response Genetics, a company that uses DNA and RNA arrays to determine emerging predictive markers. For example ERCC1 has recently emerged as a marker of platinum sensitivity in lung cancer. Investigators are utilizing determination of this protein’s expression by RT-PCR to guide clinical trials. Finally, gene expression profiles to direct therapeutic decisions was also the topic of Joffre Baker. Oncotype Dx has been widely incorporated into breast therapeutics and in clinical trials, in the subset of patients that are estrogen receptor positive and lymph node negative. The future for validated gene expression profiles appears bright in the formulation of new treatment strategies. This symposium was representative of the legacy of Charles Heidelberger ranging from cancer causation to mechanisms of drug action and to laboratory-guided clinical applications. This legacy derived not only from his accomplishments as a scientist, but also from making his laboratory a fertile ground for training. A forthcoming issue of Molecular Cancer Therapeutics will expand on items covered in this symposium proving the continued viability of the Heidelberger view of anticancer drug development: recruiting students and collaborators to pursue biological findings in the laboratory and in the clinic.