Mutational load distribution analysis yields metrics reflecting genetic instability during pancreatic carcinogenesis

Abstract

Considering carcinogenesis as a microevolutionary process, best described in the context of metapopulation dynamics, provides the basis for theoretical and empirical studies that indicate it is possible to estimate the relative contribution of genetic instability and selection to the process of tumor formation. We show that mutational load distribution analysis (MLDA) of DNA found in pancreatic fluids yields biometrics that reflect the interplay of instability, selection, accident, and gene function that determines the eventual emergence of a tumor. An in silico simulation of carcinogenesis indicates that MLDA may be a suitable tool for early detection of pancreatic cancer. We also present evidence indicating that, when performed serially in individuals harboring a p16 germ-line mutation bestowing a high risk for pancreatic cancer, MLDA may be an effective tool for the longitudinal assessment of risk and early detection of pancreatic cancer.

Fig. 1.

MLDA profiles in three distinct populations. Each row represents one subject; Top is composed of subjects with no known pancreatic pathology (n = 9), Middle groups patients with chronic pancreatitis at increased risk for pancreatic carcinoma (n = 12), and Bottom depicts the results obtained in patients with pancreatic carcinoma (n = 21). The number above the triplet sequences identifies the codon. Each column represents one allele, and the color in each box denotes the proportion of each allele constituting the population of molecules encoding Ki-ras p21 or the p53 protein. Although many alleles in cancer patients were >5%, the actual representation is cut off to depict the dynamic range of values between 0.000% and 5.000%. The enhanced SI Fig. 5 shows the results from the addition of the fractional values for an individual gene (Ki-Ras or p53) and the total mutational load resulting from the addition of all fractional values (Ki-Ras and p53). The actual fractional values for the each allele are provided http://genecube.med.yale.edu:8080/montebello.

Fig. 2.

Total mutational load. Individual cases (arrayed along the x axis) are presented by increasing total mutational load values (y axis). The total mutational load parameter derived from the MLDA profiles separates the three groups of subjects with a narrow band of overlap between pancreatitis and cancer. The increase in total mutational load can be interpreted as a reflection of progressive genetic instability.

Fig. 3.

Longitudinal MLDA profiling of a population at risk. MLDA profiles from subjects belonging to different families with increased risk for pancreatic cancer due to inherited p16 mutation. Notation is: Family.Subject.Serial sample (SI Fig. 7 provides the enhanced version of the figure indicating the age at which the sample was obtained and the genotype. Two patterns can be recognized in the samples: normal- (e.g., family 5, sample N5) and pancreatitis-like patterns (e.g., family 4, sample N4). For subjects with sequential samples, the profiles change with time from normal- to pancreatitis-like, indicating an increase in risk. Note that in instances when the risk increases, the alleles with high values do not necessarily persist. The total load for Ki-ras and p53, the age at time of sampling, and the p16 genotype are provided for each subject on the right.

Fig. 4.

Simulated mutational load over time. Graphic representation of the MLDA values obtained at each time step in runs for the three classes of outcome. Time series of mutational load at each 25th step over the entire 5,000 iterations (200 measurements per run). Rows represent the mutational load at a single time point proceeding from bottom (t = 0) to top (t = 5,000). Columns represent the mutational load for 10 alleles of three genes as in Fig. 1. (Top) Low risk (undisturbed); (Middle) high risk (no tumor formation for duration of simulation); and (Bottom) tumor. The transition from “low risk” to groups Middle or Bottom is determined by setting the disturbance parameter. Specific synthetic histories are classified a posteriori into “high risk” or “tumor” by outcome. The differences in MLDA profile are clearly apparent. For the run ending in tumor, MLDA provides a time zone of early detection. Note the similarity of the risk and tumor profiles during the early time period preceding the “early detection band,” The simulation indicates that the progressive increase in risk identifies the individual runs marked by the emergence of a “tumor” and enables the prediction of the tumor class.