Gene expression signatures of radiation response are specific, durable and accurate in mice and humans

Abstract

Background: Previous work has demonstrated the potential for peripheral blood (PB) gene expression profiling for the detection of disease or environmental exposures.

Methods and findings: We have sought to determine the impact of several variables on the PB gene expression profile of an environmental exposure, ionizing radiation, and to determine the specificity of the PB signature of radiation versus other genotoxic stresses. Neither genotype differences nor the time of PB sampling caused any lessening of the accuracy of PB signatures to predict radiation exposure, but sex difference did influence the accuracy of the prediction of radiation exposure at the lowest level (50 cGy). A PB signature of sepsis was also generated and both the PB signature of radiation and the PB signature of sepsis were found to be 100% specific at distinguishing irradiated from septic animals. We also identified human PB signatures of radiation exposure and chemotherapy treatment which distinguished irradiated patients and chemotherapy-treated individuals within a heterogeneous population with accuracies of 90% and 81%, respectively.

Conclusions: We conclude that PB gene expression profiles can be identified in mice and humans that are accurate in predicting medical conditions, are specific to each condition and remain highly accurate over time.

Figure 1. Peripheral blood gene expression profiles distinguish irradiated mice within a heterogeneous population

(A) The left panel represents a heat map of a 25 gene profile that can predict radiation status. The figure is sorted by dosage (0 cGy, 50cGy, 200cGy, and 1000cGy). High expression is depicted as red, and low expression is depicted as blue. (B) The right panel is a graph of the predicted capabilities of the irradiation signature across all mice (including C57Bl6 and BALB/c strains, males and females and 3 sampling time points) versus a control, non irradiated sample. All predicted probabilities for the controls are listed.

Figure 2. Impact of sex on murine irradiation profiles

(A) Heat map images illustrating expression pattern of genes selected for classifying control, non-irradiated mice versus 50 cGy, 200 cGy, or 1000 cGy irradiated mice within female (top) and male C57Bl6 mice (bottom). (B) Heat map images illustrating expression pattern of genes found in the female C57Bl6 strain or male C57Bl6 strain predicting the irradiation status of the opposite sex at dosage 50 cGy, 200 cGy, 1000 cGy. High expression is depicted as red, and low expression is depicted as blue. (C) A leave-one-out cross-validation analysis of the classification for control (blue) versus 50 cGy (black), 200 cGy (green), and 1000 cGy (red) for the female C57Bl6 (squares) and male C57Bl6 (circles) samples is shown. The control probabilities for each prediction are shown.

Figure 3. Impact of genotype on murine irradiation profiles.

(A) Heat map images illustrating expression pattern of genes selected for classifying control, non-irradiated samples versus 50 cGy, 200 cGy, 1000 cGy irradiated samples between female C57Bl6 strain (top) and female BALB/c strain (bottom). (B) Heat map images illustrating expression pattern of genes developed in one strain as predicting the other strain (C57Bl6 or BALB/c). High expression is depicted as red and low expression is depicted as blue. (C) A leave-one-out cross-validation analysis of the classification for control versus 50 cGy (black), 200 cGy (green), and 1000 cGy (red) for the female BALB/c and female C57Bl6 samples is shown. The control probabilities for each prediction are shown. BK represents the application of female C57Bl6 metagenes to predict the status of female BALB/c mice, and BC represents using female BALB/c mice metagenes to predict the status of female C57Bl6 mice.

Figure 4. Impact of time on murine irradiation profiles.

(A) Heat map images illustrating expression pattern of genes selected for classifying control, non-irradiated samples versus 50 cGy, 200 cGy, 1000 cGy irradiated samples at time points 6hr, 24hr, and 7days. (B) Heat map images illustrating expression pattern of genes found in the 6hr time point as applied to the dosages 50 cGy, 200 cGy, 1000 cGy at 24 hr and 7 day time points. High expression is depicted as red, and low expression is depicted as blue. (C) A leave-one-out cross-validation analysis of the classification for control (blue) versus 50 cGy (black), 200 cGy (green), and 1000 cGy (red) for the time points 6 hr (circles), 24 hr (squares), and 7 days (triangles) is shown. The control probabilities for each prediction are shown.

Figure 5. Peripheral blood profiles of irradiation and LPS-treatment are highly specific.

(A) Heat maps representing unique metagene profiles are shown which were utilized to distinguish 3 different levels of irradiation (left) and to distinguish LPS-treatment (right) in C57Bl6 mice. (B) The graph at left represents the predictive capabilities of the PB irradiation signatures in the female C57Bl6 mice in predicting dosage profiles at 50 cGy (black), 200 cGy (green), and 1000 cGy (red); the middle graph represents the predictive capabilities of the irradiation signatures when validated against the LPS-treated samples (squares); at right, the LPS signature was validated against the C57Bl6 irradiated mice and the predicted probabilities for 50 cGy (black), 200 cGy (green), and 1000 cGy (red) are shown.

Figure 6. PB metagene profiles of human radiation exposure and chemotherapy treatment are accurate and specific relative to each other.

(A) The heat map on the left depicts the expression profiles of genes (rows) selected to discriminate the human samples (columns); high expression is depicted as red, and low expression is depicted as blue. A leave-one-out cross-validation assay (at right) demonstrated that the PB metagene of radiation was capable of distinguishing healthy donors (black), non-irradiated patients (gray), irradiated patients (red), pre-chemotherapy treatment patients (green), and post-chemotherapy patients (blue). A ROC curve analysis was used to define a cut-off for sensitivity and specificity of the predictive model of radiation. The dotted line represents this threshold of sensitivity and specificity. (B) The heatmap on the left depicts an expression profile of chemotherapy treatment that distinguishes chemotherapy-treated versus untreated patients. A leave-one-out cross-validation assay demonstrated that this PB metagene of chemotherapy treatment could accurately distinguish pre-chemotherapy patients (green), chemotherapy-treated patients (blue), healthy individuals (black), pre-irradiated patients (gray) and irradiated patients (red).