Biological model of ED01 hepatocarcinogenesis

Identification of biomarkers of chemically induced hepatocarcinogenesis in rasH2 mice by toxicogenomic analysis

Toxicogenomic approaches have been applied to chemical-induced heptocarcinogenesis rodent models for the identification of biomarkers of early-stage hepatocarcinogenesis and to help clarify the underlying carcinogenic mechanisms in the liver. In this study, we used toxiciogenomic methods to identify candidate biomarker genes associated with hepatocarcinogenesis in rasH2 mice. Blood chemical, histopathologic, and gene expression analyses of the livers of rasH2 mice were performed 7 and 91 days after the administration of the genotoxic hepatocarcinogens 2-acetylaminofluorene (AAF) and diethylnitrosoamine (DEN), the genotoxic carcinogen melphalan (Mel), and the nongenotoxic noncarcinogen 1-naphthylisothiocynate (ANIT). Histopathologic lesions and a rise in accompanying serum marker levels were found in the DEN-treated rasH2 mice, whereas no neoplastic lesions were observed in the rasH2 mice. However, biological functional analysis using Ingenuity Pathways Analysis (IPA) software revealed that genes with comparable molecular and cellular functions were similarly deregulated in the AAF- and DEN-treated rasH2 mice. We selected 68 significantly deregulated genes that represented a hepatocarcinogen-specific signature; these genes were commonly deregulated in both the AAF- and DEN-treated rasH2 mice on days 7 and 91. Hierarchical clustering analysis indicated that the expression patterns of the selected genes in the hepatocarcinogen (AAF and DEN) groups were distinctive from the patterns in the control, Mel, and ANIT groups. Biomarker filter analysis using IPA software suggested that 28 of the 68 signature genes represent promising candidate biomarkers of cancer. Quantitative real-time PCR analysis confirmed that the deregulated genes, which exhibited sustained up- and down-regulation up to day 91, are likely involved in early-stage hepatocarcinogenesis. In summary, the common and significant gene expression changes induced by AAF and DEN may reflect early molecular events associated with hepatocarcinogenesis, and these "signature" genes may be useful as biomarkers of hepatocarcinogenesis in mice.

Corrections to: ‘‘Age distribution of cancer in mice’’

We found a crucial error in an earlier paper on cancer in elderly mice, Age distribution of cancer in mice: the incidence turnover at old age (Pompei et al., 2001). That paper’s principal data set, the ED01 records, was scrambled when read and analyzed with a statistical software package. Having done our best to correct the error, and having subjected the data to a more exact extension of originally published methods, we arrive at conclusions significantly different from those proposed in the original article. What appeared to be a dramatic fall off of the cancer mortality rate in mice over 2 years of age is now found to be a continuation or flattening of approximately exponential growth. This new finding is entirely at odds with the old, and does not support our later work on humans. Two of this paper’s authors, F Pompei and R Wilson, contributed to the original article. We are informing authors who have cited our paper in the past and apologize deeply for any wasted time or lost work. We should have subjected the ED01 records to more error checks. We thank Jennifer Blank for helping us discover and correct this error. The ED01 records and our earlier research are available http://physics.harvard.edu/∼wilson/cancer&chemicals/ED01.

Risk assessment of chemical carcinogens and thresholds

The controversial arguments about the existence of "thresholds" for carcinogens are discussed and some conclusions are drawn: (1) The meaning of "threshold" has changed considerably during the last decades. Initially, the discussion focused on the genotoxic properties of chemicals. In dose-response studies the endpoint was tumor incidence. Later, DNA adducts represented the biologically active target dose and whether saturation of metabolic activation could lead to non-linear relationships was tested as a hypothesis. (2) In a next step, the implications of the initiation-promotion model were studied. Carcinogens with tumor-initiating properties showed linear dose-response relationships at low doses without a definable threshold, whereas those with tumor-promoting properties showed non-linear characteristics compatible with the existence of a threshold. However, the results are difficult to transfer to the human situation, and many critical endpoints are subject to other risk factors so that a meaningful value cannot be given. (3) Eventually, it turned out that most carcinogens exhibit genotoxic as well as non-genotoxic properties, and toxicity may be equally important as genotoxicity. (4) In view of the discussion for more than 60 years about the existence of thresholds for carcinogens, it is suggested that the threshold approach not be used to establish acceptable risk limits. (5) Instead of calculating an acceptable risk from cancer risk data, the recommended method is to assess the incremental contribution of exposure to the background of avoidable and unavoidable exposures by using biomonitoring data from human individuals. Such data could help in risk management, in order to reach acceptable limits of exposures on the basis of the "as low as reasonably achievable" or "ALARA" principle.

Rainbow trout (Oncorhynchus mykiss) and ultra-low dose cancer studies

Cancer risk assessment utilizing rodents requires extrapolation across five orders of magnitude to estimate the Virtually Safe Dose (VSD). Regulatory agencies rely upon the Linear Extrapolated Dose (LED) except when sufficient information on mechanism of action justifies alternative models. Rainbow trout (Oncorhynchus mykiss) has been utilized at Oregon State University as a model for human cancer for forty years. Low cost and high capacity, made possible by our unique facility, along with low spontaneous background and high sensitivity, allow design and conduct of statistically challenging studies not possible in rodents. Utilization of custom microarrays demonstrates similarities in gene expression in trout and human hepatocellular carcinoma (HCC). We have completed one study employing over 42,000 trout with dibenzo[a,l]pyrene (DBP) and determined the dose resulting in 1 additional cancer in 5000 animals, a 50-fold enhancement over the mouse ED(01) study. Liver tumor incidence at low dose deviated significantly from linearity (concave down), whereas, DBP-DNA adductions deviated slightly (convex up). A second study is underway with aflatoxin B(1) (AFB(1)). Results to date indicate AFB(1) at low dose, in contrast to DBP, elicits a linear dose-response function on the log-log scale which falls below the LED with a slope slightly greater than 1.0. Such studies demonstrate the statistical power of the trout cancer model and strengthen the case for incorporation of these data-sets into risk assessment for these environmental human carcinogens.

Use of biomarkers in risk assessment

The systematic development and application of biomarkers in environmental health risk assessment is a relatively new field. At first, the major interest was in biomarkers of exposure, borrowing concepts from pharmacology, then it moved from the external estimates of exposure to internal measures of dose, and ultimately, to markers of target dose. While these markers provide evidence of exposures, they do not provide evidence of that toxicological damage has occurred. For this reason, measurements of DNA adducts and protein adducts are of interest, since they may provide bridges between exposures and disease end-points. In parallel, more quantitative and more sensitive end-points for diseases have been sought. Again, with advancing techniques in cytogenetics, extensive studies were conducted on such markers as chromosomal aberrations, micronuclei and other changes deemed to represent genomic damage. However, these types of end-points are quite unspecific for application to new hazards of uncertain human toxic (carcinogenic) potential. Recent work focusing on more specific early-effect markers such as certain oncogenes and tumour-suppressor genes have substantial promise as shown by work with aflatoxins and vinyl chloride. Such studies have also enhanced mechanistic insight. The advances in molecular genetics have led to an upsurge in interest in most susceptibility factors, and identification of polymorphisms of various enzymes has become possible. Ongoing search for "ultra-high risk" individuals may be fruitful, but probably only relevant to a small segment of potentially exposed populations. Factors associated with a small differential risk, however theoretically or mechanistically important, offer only little practical use.

Determinability of model parameters in a two-stage deterministic cancer model

We investigate the determinability of model parameters for a two-stage cancer model, using as an example chemically induced skin cancer in mice. When the time course of the papilloma number and cancer rate are known, the rate of creation of initiated cells and their net-growth rate can be uniquely determined. These two high-level parameters are sufficient to uniquely simulate the experimental papilloma data. However, the mitotic and death rates of initiated and transformed cells cannot be determined from the available experimental data. The rate of creation of transformed cells and their net-growth rate cannot be determined independently. Thus, although the deterministic two-stage cancer model can simulate the kinetics of papilloma formation and skin cancer data, many of the basic underlying biological parameters (i.e., mitotic and death rates) cannot be uniquely determined from the usually available biological data.

Role of the Ha-ras gene in the malignant transformation of rat liver oval cells

We have shown that the oval cell line OC/CDE 22 can be transformed by the highly carcinogenic fjord-region diol epoxides of benzo[c]phenanthrene. Mutational activation of the ras proto-oncogene family has been proposed to be a critical event in the formation of tumors induced by polycyclic aromatic hydrocarbons. Therefore, we investigated whether in the earlier transformed OC/CDE 22 cells any point mutations were detected in the ras proto-oncogene. The results indicate that the malignant transformation of OC/CDE 22 cells by the 4 stereoisomeric benzo[c]phenanthrene diol epoxides in vitro is independent of activation of the Ha-ras proto-oncogene. In addition, Northern and Western blot analyses revealed no overexpression of the Ha-ras protooncogene in the transformed OC/CDE 22 cell lines. However, transfection of the OC/CDE 22 cells with an activated Ha-ras oncogene malignantly transformed the OC/CDE 22 cells, and the transfected cells served as precursor cells of tumors with a cholangiocellular morphology and phenotype. Our latter finding reinforces the view that OC/CDE 22 cells are committed to the bile duct epithelial cell lineage.