Final Report: Molecular Characterization of a Biological Threshold in Developmental Toxicity

EPA Grant Number: R827445
Title: Molecular Characterization of a Biological Threshold in Developmental Toxicity
Investigators: Knudsen, Thomas B.
Institution: Jefferson Medical College
EPA Project Officer: Hahn, Intaek
Project Period: October 1, 1999 through September 30, 2002 (Extended to September 30, 2003)
Project Amount: $207,170
RFA: Children's Vulnerability to Toxic Substances in the Environment (1999) RFA Text |  Recipients Lists
Research Category: Children's Health , Health Effects , Human Health , Health


The objective of this research project was to characterize, in molecular terms, the biological basis of a threshold in developmental toxicity and place this into the context of a quantitative dose-response model for risk assessment. We have undertaken microarray analysis of gene expression in the early mouse embryo with the goal of uncovering functional relationships between gene expression and developmental toxicity at the low end of the dose-response curve. The research prototype is 2-chloro-2'-deoxyadenosine (2CdA), an ocular teratogen in mouse embryos exposed on day 8 of gestation. Susceptibility to 2CdA-induced microphthalmia is dependent on the administered dosage of 2CdA to the pregnant dam (Wubah, et al., 2001), the p53 tumor suppressor genotype of the embryo (Wubah, et al., 1996), and the pharmacological status of the mitochondrial peripheral-type benzodiazepine receptor (Charlap, et al., 2003). Microarray analysis was used to address changes in gene expression under three experimental contexts: (1) dosages across the threshold for microphthalmia; (2) trajectories across the critical period of p53 protein induction in early embryos; and (3) effects sensitive to mitochondrial benzodiazepine receptor (Bzrp) ligands that suppress the p53 response.

Summary/Accomplishments (Outputs/Outcomes):

This project had 32 microarray hybridizations of RNA from headfold-stage mouse embryos collected on day 8 of gestation under 16 different experimental conditions. For dose response analysis, we ran eight microarray hybridizations (four dosages independently replicated) of whole embryos from pregnant CD-1 mice. Benchmark doses (BMDs) were calculated from microphthalmia rates fit to a probit model using BMD software (BMDS) ( Exit ). Test doses were selected as follows: (1) 5.0 mg/kg, induces microphthalmia in 24.4 percent of CD-1 mouse fetuses; (2) 2.5 mg/kg, the dose modeled for an extra 5 percent risk (BMD5) of microphthalmia; (3) 1.25 mg/kg, just below the no-observed adverse-effect-level (NOAEL) for developmental toxicity; and (4) 0.625 mg/kg, well below the developmental NOAEL. Thus, test doses 1.25- and 2.5 mg/kg flank the threshold for developmental toxicity, which is modeled as the BMD level (BMDL) = 2.0 mg/kg. For the time-course study, we ran eight microarray hybridizations from the headfold of day 8 mouse embryos (replicates collected at three time intervals after low-dose 2CdA exposure and time 0). Treatment of pregnant CD-1 mice with 2.5 mg/kg 2CdA on day 8 of gestation provided exposed embryos. This dose, unlike 5.0 mg/kg, causes little or no apoptosis in the early embryo, and was chosen to reflect important changes occurring near the threshold for developmental toxicity (BMDL). Posttreatment time intervals of 3.0, 4.5, and 6.0 hours were chosen for this study based on the rationale that p53 protein induction is first demonstrable between 3.0 and 4.5 hours in the embryonic headfold. Thus, the 3.0 hour time point precedes this critical event, whereas 4.5 hours coincides with, and 6.0 hours follows, this critical event in the teratogenic mechanism of 2CdA-induced microphthalmia. Control RNA samples were collected from untreated embryos at corresponding times and at time zero. For the Bzrp intervention study, microarray data were collected for two sets of experiments. The first compared 2CdA + PK11195 cotreated samples (test) with 2CdA alone (reference). PK11195 is a partial antagonist of the Bzrp and can rescue mouse embryos from microphthalmia when coadministered with 2CdA. When PK11195 treatment is delayed beyond the time of 2CdA exposure, 4.5 hours emerges as the longest delay that could rescue fetuses from the increased risk for microphthalmia. We ran six microarray hybridizations for the headfold (four-six somite pair stage) at 3.0 hours, 4.5 hours, and 6.0 hours postexposure to 2CdA (2.5 mg/kg) + PK11195 (4.0 mg/kg). The second experiment used 10 microarray hybridizations to compare intervention using PK11195 versus Ro5-4864 at 4.5 hours posttreatment. In contrast to PK11195, Ro5-4864 has agonist properties at the Bzrp and is not antiteratogenic.


Twenty embryos or 10-20 microdissected headfolds were pooled in each microarray sample. Quality assurance of RNA was determined by absorbance A260/280 ratio of qreater than or equal to 1.7, 1 percent agarose gel electrophoresis, and assessment by RNA 6000 Nano microchip electrophoresis on the Agilent 2100 Bioanalyzer to rule out degradation of the isolated RNA and/or DNA contamination, and reverse transcriptase-polymerase chain reaction (RT-PCR) amplification of positive control genes (beta-actin, 16S ribosomal RNA). We isolated, for each measurement, 15-30 µg of total RNA. RNA (1 µg) was converted to cDNA with incorporation of biotin-dUTP, dinitrophenol-dUTP, or fluorescein-11-dCTP. Approximately 15-20 ng labeled cDNA from each sample was hybridized competitively (56-59°C) to the MPS621 microarray probe printed with 2,400 or 4,800 sequence-verified cDNA elements (PerkinElmer Life Sciences). The genes represent a cross section of human diseases, metabolic and regulatory pathways, and control spots representing more than 10 different tissue sources (80 percent brain derived, more than 40 percent full-length cDNA). Histochemical detection used peroxidase-conjugated anti-dinitrophenol (DN) or anti-fluorescein (FL), and cyanine-3-tyramide (Cy3) reaction followed by peroxidase-conjugated streptavidin (BN) and cyanine-5-tyramide (Cy5) reaction. Each array was repeated with reversal of the labeling scheme (e.g., FL test/biotin reference versus biotin test/FL reference) and independent sample replication. Fluorescent images of Cy3 and Cy5 channels were generated with a laser confocal scanner (ScanArray 5000) at 10-micron pixel resolution and analyzed with QuantArray software (GSI Luminomics). Units of intensity for background-corrected signals encompassed three log scales of dynamic response. The log-intensity ratio was normalized with locally weighted regression, standardized using absolute deviation of data points from their mean, and equalized between paired swaps for each gene in the matrix. Data reduction and visualization used GeneSpring software (version 6.0, Silicon Genetics, Redwood City, CA). Workflow was designed to select genes that passed several quality control criteria, including an absolute signal threshold for minimal detection and concordance for the test/reference ratio between replica samples. The results included in this progress report were derived from analysis of 30 microarray chips. Data sets were deposited into the Gene Expression Omnibus repository as part of a larger 100-array data set scheduled for release March 1, 2004, ( under platform GPL560.

Figure 1. Competitive Fluorescent Hybridization of Embryonic RNA with Micromax-I Microarray Probe (59°C). Test RNA was procured from the headfold of day 8 embryos 3 hours after intrauterine exposure to 5.0 mg/kg 2CdA. The reference was control mouse embryos at the same developmental age. (Note: Replacing the dinitrophenol labeling reaction with fluorescein isthiocyanate improved the specificity of the data considerably; however, this report will use data from the original formulation).

Biochip Performance

Of the 2,400 genes (4,800 elements) spotted on the MPS621 biochip, 2,338 genes yielded signal above the detection threshold. When RT-PCR was applied to a panel of 50 target genes that were positive by microarray, this amplified the correct size PCR band in 42 cases (84 percent success rate). The success rate in endpoint PCR varied with intensity of the microarray signal, reflecting relative transcript abundance, sequence homology of cDNA elements, cellular heterogeneity, and labeling efficiencies. This endpoint PCR panel reproduced the qualitative effect of the microarray in 32 of 50 target genes (64 percent success rate). A representative sample of 232 elements from the DNA biochip produced a 79.8 percent average sequence homology between human and mouse (confidence interval = 77.0 – 82.7%). Therefore, genes representing the lower quartile (190-1,123 median absolute signal) were deemed unreliable and filtered from the microarray data set.


Dose Response. Understanding the toxicogenomic responses associated with physiological mechanisms linking environmental exposures to disease requires data reduction standards and practices that have not yet been unified in the field. Because changes in gene expression reflect the primary vulnerability of embryonic systems to toxic insult as well as cellular adjustments that may or may not contribute to the disease process, an immediate goal was to score gene expression profiles as a function of 2CdA dose. The differentiality profile at 3.0 hours posttreatment showed dose dependence in both complexity and magnitude (see Figure 2). We saw minimal effects with 0.625 mg/kg 2CdA and a broad response at the teratogenic dose level (5.0 mg/kg) for 3.0 hours posttreatment. Doses flanking the developmental NOAEL were intermediate in their effect on the embryonic transcriptome.

BMDS predicts the threshold for microphthalmia occurrence at 2.0 mg/kg 2CdA. This dose represents the lower 95 percent confidence limit for modeled doses associated with an increased 5 percent risk for the defect (BMD5), and is intermediate between the NOAEL (1.5 mg/kg) and the BMD5 (2.5 mg/kg) modeled from the dose response curve. We identified 182 genes changing by at least 1.5-fold in the 2CdA dose response (-0.625 to 5.0 mg/kg), and we partitioned them by K-means clustering (see Figure 2). This showed a segmental response of the transcriptome that seemed to flicker at intermediate levels and burn above the threshold dose for microphthalmia; however, somewhat unexpectedly, the transcripts could be grouped as to those spiking at the threshold dose for developmental toxicity (2.5 mg/kg) versus those altered only at the overtly teratogenic dose (5.0 mg/kg). Genes varying across the dose vector showed two principal components of response in diametric opposition. The two behaviors diverged as the 2CdA dosage exceeded the BMDL for microphthalmia (e.g., 2.0 mg/kg). This implies a novel cellular regulation invoked as the toxicant exposure crosses the exposure threshold for developmental toxicity.

Figure 2. K-Means Clustering of 182 Genes That Changed 1.5-Fold in the Mouse Embryo at 3.0 Hours After Intrauterine Exposure to 2CdA on Day 8 of Gestation. Independent replicate analysis was performed on total RNA labeled with BN-dCTP or DN-dCTP and Cy5-tyramide or Cy3-tyramide signal amplification, respectively. Each segmental dose represents a true replicate sample with reversal of the label-dye assignments. Doses were 0.625, 1.25, 2.5, and 5.0 mg/kg. The threshold (BMDL) is modeled at 2.0 mg/kg (arrows).

What metabolic and regulatory pathways correspond to a “molecular benchmark signature” for 2CdA developmental toxicity? With the hope of using disease signatures to improve our understanding of how risks of exposure to environmental contaminants are assessed, important questions pertain to our own abilities to interpret the commonality and differentiality of complex multigenic responses within the context of chemical mode of action. Preliminary analysis identified a number of responsive genes associated with ocular defects in human infants or experimental animals. These included the group IV Pax gene (Pax6) for microphthalmia, which is downregulated with 2CdA exposure. The drop in Pax6 levels correlated with independent analysis of PAX6 protein in the murine optic vesicle of 2CdA treated embryos (Figure 3). If we define "molecular benchmark signature" as the collective set of changes that characterize developmental toxicity, then suites of genes clustering with the candidate disease genes should provide a more complete picture of 2CdA's mode of action.

Figure 3. Confirmation That 2CdA Exposure Leads to a Drop in Abundance of Transcripts for Pax6, the Master Gene of Eye Development. Left panel: RT-PCR analysis performed on optic pit or headfold of day 8 mouse embryos 3.0 hours after exposure to 5.0 mg/kg 2CdA. Right panel: PAX6 protein immunoperoxidase staining in the optic vesicle of untreated (left) and treated (right) embryos harvested 24 hours posttreatment.

Time Course. To address the trajectories of gene expression at critical time points after exposure, dams were treated with 2.5 mg/kg 2CdA. Embryos cotreated with 4.0 mg/kg PK11195 provided parallel samples of embryos in which p53 protein induction is delayed or blocked by the Bzrp ligand. The test period was 3.0 to 6.0 hours postexposure to encompass events on either side of p53 protein induction at 4.5 hours. We identified 180 genes sensitive to 2CdA and/or PK11195 cotreatment that fell into three principal behaviors by K-means clustering (Figure 4). The 2CdA time course showed minor phasing of gene trajectories that was clearly promoted by PK11195 cotreatment. Genes varying across the time vector showed three principal components of response, although the effect of PK11195 on trajectories of gene expression was most pronounced at 3.0 hours postexposure. Temporal changes following 2CdA exposure should be interpreted within the context of p53 protein induction because this is a critical event in the pathogenesis of microphthalmia. Because p53 begins to accumulate in embryos between 3.0 hours and 4.5 hours after 2CdA exposure, all three temporal behaviors were evident before p53 protein induction. Thus, PK11195 seemed to alter the regulation of the embryonic transcriptome during a period of time that ordinarily would signal p53 protein induction (upstream intervention).

Figure 4. K-Means Clustering of 180 Genes That Changed in the Mouse Embryonic Headfold After Intrauterine Exposure to 2.5 mg/kg 2CdA Alone or in Combination With 4.0 mg/kg PK11195 Cotreatment on Day 8 of Gestation. Each time interval represents a true replicate sample with reversal of the label-dye assignments. Times were 3.0, 4.5, and 6.0 hours after treatment. The critical time based on p53 protein induction is 4.5 hours postexposure (arrows).

Specificity of Intervention. Microarray analysis compared PK11195 with another well-studied Bzrp ligand, Ro5-4864. These ligands differ from one another with respect to physiological activities at the Bzrp, the former being a partial antagonist and the latter an agonist. Teratological evaluation showed Ro5-4864 and PK11195 also differ substantially from one another with respect to several activities on the embryo: PK11195 blocked 2CdA-induced p53 protein accumulation whereas Ro5-4864 did not; PK11195 suppressed 2CdA-induced microphthalmia whereas Ro5-4864 did not; and Ro5-4864 had weak teratogenic activity at 4.0 mg/kg whereas PK11195 does not. RNA was collected from the headfold of early mouse embryos 4.5 hours after exposure to PK11195 (4.0 mg/kg) or Ro5-4864 (4.0 mg/kg) in combination with 2CdA (2.5 mg/kg). Microarray analysis revealed that Ro5-4864, unlike PK11195, did not substantially influence the patterns associated with the 180 genes of the 2CdA time course. We conclude that differential activities of these Bzrp ligands on the “molecular benchmark signature” for 2CdA were consistent with the physiological responses in the teratogenic mechanism.

Integration of Data. Because changes in gene expression may reflect the primary vulnerability of embryonic systems to toxic insult or cellular adjustments that may or may not contribute to the disease process, we sought to refine the 2CdA core response to genes that comprehensively describe the covariance across the different experimental parameters. One solution to this problem derived from a Pearson covariance matrix generated from a core response signature of genes that revealed orthogonal symmetry. The orthogonal blocks of genes correlate (white) or anticorrelate (blue) with one another as principal components of response. Assuming these genes arrive at their position because of parameter-dependent variance, the regularity between experimental parameters can be regarded as a metric for coherent expression patterns as well as variation across conditions. We then can look for "resonance" between the different experiments, as demonstrated in Figure 5.

Figure 5. "Resonance": A Different View of the Expression Phenotype. The expression data from 2,400-gene (4,800 spotted element) arrays were decomposed into a reduced set of 162 genes having a regular Pearson correlation clustering profile. In the gene x gene correlation matrix, white tiles represent clusters of genes whose expression levels changed qualitatively in perfect correlation, and blue tiles represent clusters of perfect anticorrelation. The matrix revealed data structure across the various conditions of the experiment evident for the dose-response at 3.0 hours posttreatment, the temporal profile at a BMD5 dose, and intervention with 2CdA cotreatment with Bzrp ligands that are active (PK11195) and inactive (Ro5-4864) with respect to rescue of microphthalmia. Trajectories resonated between different experiments, with the temporal profile at the threshold describing the greatest amount of this data structure.

Several general characteristics are worth noting in the covariance display (see Figure 5). One pertains to the regularity contributed by gene expression data derived from each experimental parameter. The covariance matrix is based on expression profiles of the core response for all three experimental parameters studied here (dose, time, intervention). Patterns generated for covariance as a function of time were complete for the time parameter, but were incomplete for the other two experimental parameters. Thus, data derived for the 2CdA dose response or PK11195 cotreatment time plot each matched distinct domains of the overall structure of the data set, but not the data structure in its entirety. From this, we propose that a temporal response at the threshold dose for an adverse response holds more information regarding the responsiveness of the embryonic transcriptome to an acute exposure. If so, then the temporal response signature of the precursor target cell populations to a low dose exposure may be a better predictor of developmental toxicity than a wider but static dose response signature at only one particular time point. This has clear implications for risk assessment of developing systems, which are dynamic in nature, and hence for children’s vulnerability to toxicant exposures.

Normal embryonic development is dependent on multiple genetic signals and responses that control cellular decisions to proliferate, differentiate, migrate, or die. Animal development uses approximately 17 distinct signaling pathways that integrate morphogenetic processes with the more fundamental regulatory processes that govern cellular growth and metabolism. Critical questions thus pertain to the relativity between gene expression clusters and coherent pathways, as well as to how altered expression of coherent pathways can be used to predict chemical mode of action. Annotating the genes associated with 2CdA's core response signature by linkages to cell signaling pathways revealed striking representation of the tyrosine kinase (RTK) signaling pathway. Early embryos rely heavily on RTK signaling to regulate cell growth, shape, and motility. Many birth defects, including microphthalmia, are linked to altered function of the RTK pathway in humans and experimental animals. One main flow of signals is fibroblast growth factor (FGF). RTK receptors FGFR1 and FGFR2 stood out because they are differentially sensitive to the exposure conditions studied here. RT-PCR analysis confirmed expression of these receptors, and further suggested presence of at least three splice variants of FGFR1 in the mouse embryo (all affected by 2CdA exposure). FGFR3 was detected, but unaffected, and FGFR4 was not detected. Genes having syn-expression with FGFR1 included amyloid protein A4, DiGeorge's critical region 2 (GR2DC), and Treacher-Collins syndrome (TCOF-1). These genes transiently upregulated after 2CdA and further with PK11195 cotreatment before irrevocable pathogenesis of the eye. On the other hand, several genes had con-expression, including Pax6 and retinitis pigmentosa-3, which all are associated with ocular malformations in humans and mice. The natural trajectory of these genes with teratogen exposure was reinforced with PK11195, suggesting adaptive changes prior to signals leading to p53 protein induction (see Figure 6).

Journal Articles on this Report : 1 Displayed | Download in RIS Format

Other project views: All 14 publications 2 publications in selected types All 1 journal articles
Type Citation Project Document Sources
Journal Article Charlap JH, Donahue RJ, Knudsen TB. Exposure-disease continuum for 2-chloro-2'-deoxyadenosine, a prototype ocular teratogen. 3. Intervention with PK11195. Birth Defects Research (Part A): Clinical and Molecular Teratology 2003;67(2):108-115. R827445 (2002)
R827445 (Final)
  • Abstract from PubMed
  • Full-text: InterScience Full Text
  • Other: InterScience PDF
  • Supplemental Keywords:

    risk assessment, health effects, teratogen, genetics, microarray, mouse, mitochondrion, mechanisms, low-dose extrapolation, functional assays, bioenergetics, children, children's vulnerability, children's health, epidemiology, health risk assessment, molecular biology, genetics, susceptibility, sensitive population, genetic susceptibility, biological threshold, biomedical research, birth defects, human malformation, developmental disorders, growth and development, developmental toxicity, dose response model, dysmorphogenesis, environmental hazard exposures, exposure, human exposure, molecular characterization., RFA, Scientific Discipline, Health, Toxicology, Health Risk Assessment, Susceptibility/Sensitive Population/Genetic Susceptibility, Children's Health, Molecular Biology/Genetics, genetic susceptability, biological threshold, birth defects, health effects, sensitive populations, developmental toxicity, dose response model, molecular characterization, exposure, human malformation, children, assessment of exposure, children's vulnerablity, dysmorphogenesis, functional assys, growth & development, biomedical research, developmental disorders, environmental hazard exposures

    Relevant Websites: Exit

    Progress and Final Reports:

    Original Abstract
  • 2000 Progress Report
  • 2001 Progress Report
  • 2002 Progress Report