Exposure-disease continuum for 2-chloro-2?-deoxyadenosine, a prototype ocular teratogen. 1. Dose-response analysis

TERATOLOGY 64:154 –169 (2001)

Exposure–Disease Continuum for 2-Chloro-2ⴕDeoxyadenosine, a Prototype Ocular Teratogen. 1. Dose-Response Analysis JUDITH A. WUBAH,1 R. WOODROW SETZER,2 CHRISTOPHER LAU,2 JEFFREY H. CHARLAP,1 AND THOMAS B. KNUDSEN1* 1 Department of Pathology, Anatomy and Cell Biology, Jefferson Medical College, Philadelphia, Pennsylvania 19107 2 National Health and Environmental Effects Research Laboratory, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina 27711

ABSTRACT Background: Treatment of pregnant mice with 2-chloro2⬘-deoxyadenosine (2CdA) on day 8 of gestation induces microphthalmia through a mechanism coupled to the p53 tumor suppressor gene. The present study defines 2CdA dosimetry with respect to exposure (pharmacokinetics), p53 protein induction, and disease (microphthalmia). Methods: Pregnant CD-1 mice dosed with 0.5–10.0 mg/kg 2CdA on day 8 provided fetuses for teratological evaluation; 2CdA was measured by HPLC in the antimesometrium through 180 min postexposure, and p53 was assessed with immunostaining of the embryo through 270 min. 5⬘-/3⬘-RACE was used to sequence the candidate gene for 2CdA bioactivation from target cells. Results: Microphthalmia appeared first in the dose-response curve. The highest 2CdA dose having no observable adverse effect (NOAEL) was 1.5 mg/kg; the benchmark dose that produced an extra 5% risk of microphthalmia (BMD5) was 2.5 mg/kg, and the lower confidence limit (BMDL) was 2.0 mg/kg. Pharmacokinetic parameters for doses encompassing the threshold (1.5– 2.5 mg/kg) were modeled at 1.0 –1.8 ␮M (Cmax) and 30 – 80 ␮M-min (AUC). The p53 response was not detected below the BMDL; however, a low-grade response appeared 4.5 hr after a teratogenic dose (5.0 mg/kg), and high-grade induction followed an embryolethal dose (10.0 mg/kg). RACE identified a novel splice variant of mitochondrial deoxyguanosine kinase, dGK-3, as the likely candidate for 2CdA bioactivation in the embryo. Conclusions: Microphthalmia represented the critical effect malformation of 2CdA. The findings suggest a mitochondrial mechanism for 2CdA bioactivation, leading to an embryonic p53 response only after 2CdA elimination and implying pharmacodynamic coupling to the exposure– disease continuum.

Teratology 64:154 –169, 2001. Published 2001 Wiley-Liss, Inc.†

INTRODUCTION Biologically based dose-response (BBDR) models represent a new stage in the evolution of risk assessment † This article is a US government work and, as such, is in the public domain in the United States of America.

Published 2001 WILEY-LISS, INC.

for developmental toxicity (Kavlock and Setzer, ’96). These models integrate the pharmacokinetic and pharmacodynamic properties of a chemical with the underlying biomolecular changes leading to disease states in the embryo (Gaylor and Razzaghi, ’92; O’Flaherty and Clarke, ’94; Shuey et al., ’94). BBDR models are iterative, incorporating new information that becomes available in cell and developmental biology. Consequently, BBDR models may improve the scientific basis of risk assessment in several possible ways, including (1) formulation of testable hypotheses pertaining to critical modes of action of developmental toxicants; (2) building the framework for serial translation of empirical dose-response relationships into a mechanistic model for toxicant-induced perturbations leading to dysmorphogenesis; (3) predicting interspecies homology of responses to environmental chemicals; (4) reducing uncertainties in the shape of the dose-response curve for exposure levels below experimental observation; (5) providing risk prediction in different model systems; and (6) expressing, in biomathematical terms, the quantitative estimation of human risk (Lau et al., ’00). To model complex processes in developmental toxicity, it makes sense to work with prototype agents that invoke fundamental disease pathways in the developing embryo. One candidate is 2-chloro-2⬘-deoxyadenosine (2CdA). This purine nucleoside analogue is cytotoxic by virtue of its similarity to 2⬘-deoxyadenosine, a metabolic toxin that invokes deoxyadenylate (dATP) stress imbalance within specific cell types (Cohen et al.,

Grant sponsor: National Institute of Environmental Health Sciences; Grant number: T32 ES07282; Grant sponsor: NRSA; Grant sponsor: National Institute of Child Health and Human Development; Grant number: F31 HD08167; Grant sponsor: Environmental Protection Agency; Grant number: EPA-CR 824 445-01. *Correspondence to: Thomas B. Knudsen, Department of Pathology, Anatomy and Cell Biology, Jefferson Medical College, 1020 Locust Street, Philadelphia, PA 19107. E-mail: [email protected] Received 28 August 2000; Accepted 7 May 2001

DOSIMETRY OF 2CdA-INDUCED MICROPHTHALMIA ’78; Hershfield et al., ’82). Early mouse embryos are vulnerable to dATP stress imbalance during late gastrulation and early neurulation (Knudsen et al., ’89; Airhart et al, ’93; Blackburn et al., ’97). Pathological accumulation of 2⬘-deoxyadenosine in the antimesometrium (Blackburn et al., ’92; Knudsen et al., ’92) invoked dATP accumulation and excessive programmed cell death (apoptosis) in the mouse embryo on day 8 of gestation (Gao et al., ’94). Intrauterine exposure to 5.0 mg/kg 2CdA similarly invoked excessive apoptosis, an effect that could be reproduced in culture by directly exposing embryos to 5 ␮M 2CdA (Wubah et al., ’96). Teratological effects of 2CdA were first reported in TSG-p53 mice on a C57/6J genetic background (Wubah et al., ’96). Malformations such as retinal coloboma, narrowing of the pupillary ring, microphthalmia, and anophthalmia appeared in a significant percentage of exposed fetuses. The determination of 2CdA-induced eye reduction defects within a mixed population of normal and p53-deficient fetuses in the TSG-p53 strain yielded malformation incidences of 73.3%, 52.5%, and 2.2% among p53⫹/⫹, p53⫹/⫺, and p53⫺/⫺ genotypes, respectively (Wubah et al., ’96). Because the effect on the developing eye was intermediate among mouse embryos heterozygous for p53 deficiency and essentially absent among nullizygotes, a critical exposure– disease relationship may exist between 2CdA exposure and its mode of action through p53. Eye reduction defects are common outcomes in experimental teratogenesis (Spaeth et al., ’82), and the induction of micro-/anophthalmia with excess p53 mRNA injected into blastomeres of Xenopus embryos (Hoever et al., ’94) corroborates a fundamental disease pathway connected with the embryonic p53 response. The Trp53 gene contributed to the intrinsic genetic susceptibility of mouse embryos to radiation-induced digital reduction defects (Wang et al., ’00). By contrast, p53 deficiency promoted the teratogenicity of benzo[a]pyrene (Nicol et al., ’95), 4-hydroperoxycyclophosphamide (Moallem and Hales, ’98), and ␥-radiation (Kato et al., ’01). In addition, p53-deficient mouse embryos are susceptible to spontaneous malformations such as anterior neural tube defects (Sah et al., ’95) and persistent hyperplastic primary vitreous (Reichel et al., ’98; Ikeda et al., ’99). Therefore, the embryonic p53 pathway has intrinsic importance to mechanisms of differential teratogenic susceptibility but can function as either a suppressor (Nicol et al., ’95; Moallem and Hales, ’98) or a mediator (Wubah et al., ’96; Wang et al., ’00) of teratogenesis depending on the nature and dose rate of exposure (Kato et al., ’01). Systematic analyses of 2CdA and other developmental toxicants acting through the cellular p53 pathway provide a framework upon which to build an embryobased BBDR modeling core. As a purine nucleoside analogue, 2CdA complements an existing BBDR modeling core being developed in the rat with the pyrimidine analogue 5-fluorouracil (5-FU) (Shuey et al., ’94; Lau et al., ’01; Setzer et al., ’01). By modeling fundamental disease pathways in mechanistic terms, the

155

hope is to capture the significance of toxicant-induced perturbations at the low end of the dose-response curve. To establish continuity between exposure and disease, we have undertaken a comprehensive analysis of 2CdA developmental toxicity and will present these findings in three parts, focused on (1) dose-response analysis, (2) cross-species differences, and (3) intervention. In the present study we define the dosimetry for 2CdA in outbred CD-1 mice on day 8 of gestation, specifically with respect to exposure (pharmacokinetics), p53 protein induction, and disease (microphthalmia). MATERIALS AND METHODS Materials 2CdA was a gift from Dr. James Oldham of R.W. Johnson Pharmaceutical Research Institute (Spring House, PA). Standard biochemicals purchased from Fisher Scientific (Fairlawn, NJ) or Sigma Chemical Company (St. Louis, MO) were of the highest grade available. Sheep polyclonal antiserum Ab-7 to recombinant human p53 was purchased from Oncogene Research Products (Cambridge, MA). SK-4100 DAB peroxidase substrate was purchased from Vector Laboratories (Burlingame, CA). Micro-spin columns were from Qiagen (Valencia, CA). Polymerase chain reaction (PCR) reagents, RNase-free DNase I, and other reverse-transcriptase polymerase chain reaction (RT-PCR) components were from Gibco-BRL (Gaithersburg, MD); pGEM-T Easy vector and restriction enzymes were from Promega (Madison, WI). Protein assay dye kit was from Bio-Rad Laboratories (Melville, NY). Marathon complementary DNA (cDNA) amplification kit for 5⬘ and 3⬘ rapid amplification of cDNA ends (RACE) was from Clontech (Palo Alto, CA). Water for solutions and buffers was collected at 18.2 M⍀-cm from the Milli-Q Plus Ultrapure water system from the Millipore Corporation (Milford, MA). Animals The Institutional Animal Use and Care Committee at Thomas Jefferson University approved all protocols describing the animal research reported here. Outbred CD-1 mice (20 –30 g) purchased from Charles River Breeding Laboratories (Wilmington, MA) were housed on a 12-hr photoperiod (07.00 –19.00 hr light) and fed Purina mouse chow and water ad libitum. Timed pregnancies were generated by caging a male with nulliparous females starting at 07.30 – 08.30 hr. Detection of a vaginal plug at 12.30 –13.30 hr signified coitus (gestational day [GD] 0). Pregnant mice were euthanized with carbon dioxide on GD8 for pharmacokinetic analysis or else on GD17 for teratological evaluation. In the former experiment, embryodecidual units were removed from the uterus. and the antimesometrial half was harvested by freeze fixing on a stainless steel plate precooled with dry ice to preserve extracellular nucleoside pools (Knudsen et al., ’92).

156

WUBAH ET AL. 2CdA treatment

Gravimetric 2CdA stock solutions were calibrated by absorbance at 264 nm using 15.0 as the millimolar (mM) extinction coefficient (Liliemark, ’97). Stock solutions diluted with sterile water were prepared to deliver the prescribed dose when injected in 0.2 ml per 0.03 kg of maternal body weight. All 2CdA stock and dosing solutions were stored frozen (⫺20°C) in singleuse aliquots. Pregnant mice received 2CdA by a single intraperitoneal (i.p.) injection 09.30 hr on GD8. Test doses of 2CdA were 0.5, 1.5, 5.0, 7.5, and 10.0 mg/kg. Pharmacokinetic analysis used seven incremental postinjection times between 5 and 180 min. Teratological evaluation After recording implantations and resorptions, fetuses were examined for gross malformations, weighed, euthanized, and fixed in formalin (to cloud the lens). Means were calculated from total fetuses and weighted for litter size. Eyes were inspected for gross structural change to the optic globe under a widefield stereoscope. An eye was classified as microphthalmic if the lens was visibly smaller than normal and anophthalmic when extremely reduced or absent. To grade eye reduction defects (ERD), a value was assigned to each eye: 0 if normophthalmic, 2 if microphthalmia, and 3 if anophthalmic; the cumulative value for the entire population was then divided by the number of eyes. Subtle defects such as coloboma and narrowing of the pupillary ring, which were previously detected in pigmented mice and graded as 1 (Wubah et al., ’96), were not easy to pick up in the nonpigmented CD-1 mice and thus were not included in this evaluation. The 95% confidence interval (CI) was calculated by Rao-Scott transformation that takes litter effects into consideration (Rao and Scott, ’92; Krewski and Zhu, ’95). Statistical tests used Fisher’s protected least significant differences test, in which one-way analysis of variance (ANOVA) was used to demonstrate group differences taking the unweighted litter mean as the sampling unit (n ⫽ 10), and post hoc differences were localized by the unpaired t-test (P ⱕ 0.05). p53 immunostaining Embryodecidual units were fixed in neutral-buffered formalin overnight at 4°C and embedded in paraffin (Wubah et al., ’96). Semiserial sections were cut at 5 ␮m and collected on positively charged Superfrost/plus microscope slides. Deparaffinated sections were rehydrated and subjected to microwave retrieval at 95–100°C for 10 min in 0.01 M sodium citrate, pH 6.0. Sections were rinsed in phosphate-buffered saline (PBS) and subjected to peroxidase block. Sheep polyclonal antiserum Ab-7 was applied at a dilution of 1:500 in 0.5% bovine serum albumin (BSA)–PBS. Incubation was overnight at 4°C. Normal sheep serum provided the negative control, and human breast carcinoma sections provided a positive control. Primary antiserum was localized by sequential incubation with biotinylated rabbit anti-sheep serum (1: 10,000 PBS) and streptavidin–OR03L– horseradish per-

oxidase (HRP) (1:200 PBS). Color development used DAB peroxidase substrate. Cell nuclei were scored as p53positive or p53-negative, based on whether the specific staining was more (positive) or less (negative) intense than the corresponding cytoplasm. The findings reported in the present article were derived from independent analysis of several tissue sections each from a minimum of two replicate embryos per dose group. High-performance liquid chromatography Five antimesometrial halves were pooled per sample. Tissue samples were extracted with ice-cold 0.6 N perchloric acid, neutralized, and centrifuged. Protein content was determined in the pellet (Bio-Rad dye-binding assay) and yielded an average of 23.1 ⫾ 4.3 mg protein per sample. The supernatant was analyzed directly by reversed-phase high-performance liquid chromatography (HPLC) with photodiode array detection (Waters Associates, Millipore). Separation was performed on a Partisphere C18 reversed-phase cartridge column (4.6 inner diameter [ID] ⫻ 12.5 cm) protected with a 1.7-cm reversed-phase guard cartridge (Whatman, Maidstone, UK). The mobile phase was 50 mM ammonium phosphate, pH 5.1, containing 10% methanol and 5% acetonitrile run at a flow rate of 1.5 ml/min. The 2CdA peak was identified in biological samples by co-retention and spectral congruency with authentic 2CdA, and integrated at 264-nm absorbance. The 2CdA standard (0.1 mM) was prepared and stored as described above for animal dosing solutions. 2CdA was stable under these conditions, as evidenced by the absence of significant breakdown to 2-chloroadenine. HPLC calibration became a problem as 2CdA levels approached the detection limits. When a calibration curve was developed for 2CdA between 5 and 100 pmol per HPLC injection, the relationship between milliabsorbance units (mAU) at 264 nm, and the amount of 2CdA (pmol injected) was linear for mAU between 0.297 and 20 mAU, but nonlinear below that range. There were 54 of 94 noncontrol values of ⬍0.297 mAU, i.e., 57.4% of the data values are in the nonlinear part of the curve. Therefore, a hybrid cubic–linear standardization curve was constructed from the standard data set, fitting a cubic function to account for the slight curvature at the low end of the calibration curve, and a linear function smoothly joined to the cubic to account for the rest of the calibration curve. The formula for this calibration is If mAU ⬎ 0.297, or

then y ⫽ 0.00843x

y ⫽ 关x2 ⫺ 共x3 /105.793兲 ⫺ 35.264x ⫹ 414.525兴 共⫺0.000127兲 ⫹ 0.00843x

where y is mAU and x is pmol injected. This equation places the detection limit at about 4.24 pmol per injection (if y ⫽ 0, then x ⫽ 4.24) and sets the bound of linearity at 35.2 pmol per injection (if y ⫽ 0.297, then x ⫽ 35.20). Samples falling below the lower limits of

DOSIMETRY OF 2CdA-INDUCED MICROPHTHALMIA detection were assigned one-half the value (2.12 pmol per injection). In terms of average mg protein per sample the lower limits of detection corresponded to a 2CdA concentration of 0.184 pmol per mg protein and the bound of linearity at 1.524 pmol per mg protein. The regression function was inverted to predict the pmol of 2CdA from mAU measurements in the mouse pharmacokinetic data set. Two replicate measures were taken per litter, and at least two replicate litters were taken for each dose ⫻ time combination. The variability of the measurement of 2CdA levels within dose in this nested design is thus analyzable into a component attributable to variance among litters, and a component due to variance within litters. A further feature of these data (and most other data on biochemical endpoints) is that the variance among measurements is proportional to the square of the mean (that is, the coefficient of variation is constant). The function lme in S-Plus (version 3.4, Mathsoft, Seattle, WA) implements the method of linear mixed effects models (Laird and Ware, ’82; Lindstrom and Bates, ’88; Davidian and Giltinan, ’95) and allows mean response levels to be estimated, and their statistical uncertainty to be quantified properly in such nested designs. It was used to compute dose- and time-specific means and 95% confidence intervals. Because of this design, and because the variance of the measurements increased with the mean, mean values of the ratio of 2CdA (pmol) to the amount of protein (mg) in the antimesometrium were estimated by applying a mixed effects model to log-transformed 2CdA/mg protein values. This results in a mean at each dose ⫻ time point, as well as an overall estimate of the standard deviation among litters and among samples within litters. The estimates of standard deviation result from pooling information across all dose ⫻ time groups. Uncertainty in the estimates of the means was expressed as 95% confidence intervals. Means and confidence intervals were exponentiated to give results on the original pmol/mg protein scale. Pharmacokinetic parameters and their standard errors were determined for each dosage based on the estimates of means and their standard errors. Area under the curve (AUC) was estimated using the trapezoid rule; Cmax is the greatest observed level, and Tmax is the time for which the concentration equaled Cmax. Terminal half-life was determined by fitting an exponential to the terminal portion of the fetal data. 5ⴕ/3ⴕ-RACE amplification and sequencing Day 8 mouse embryos (approximately 4 – 6 somite pairs) provided a source of RNA for complementary DNA (cDNA) synthesis. Embryos were harvested in ice-cold Hank’s balanced saline solution (HBSS) and microdissected into the prosencephalon and primitive heart. These tissues represent rudiments known to be sensitive and resistant, respectively to the cytotoxic effects of 2⬘-deoxyadenosine (Gao et al., ’94) and 2CdA (Wubah et al., ’96). RNA was purified with microspin columns and removal of DNA contamination with RNase-free DNase I. An absorbance A260/280 ratio of

157

ⱖ1.7 was acceptance criterion for RNA. Quality assurance was further determined by formaldehyde containing 1% agarose gel electrophoresis to rule out degradation of the isolated RNA and/or DNA contamination. Visualization of sharp 28S and 18S rRNA bands that migrate at approximately 5 kb and 1.9 kb, respectively, and no evidence of smearing or high-molecular-weight bands were acceptance criteria for the RNA preparation. For reverse transcription, 1 ␮g RNA was annealed to random primers (HAS701 from Gibco-BRL) and reverse-transcribed with SuperScript II Reverse Transcriptase (Gibco-BRL) at 42°C. Control reactions included replacing SuperScript II by an equal volume of DEPC-treated water (negative control) and parallel amplification of ␤-actin (positive control) (Ibrahim et al., ’98). GenBank sequence data were retrieved from the National Center of Biotechnology Information (http:// ncbi.nlm.nih.gov) and the Basic Local Alignment Search Tool (BLAST) algorithm was used for sequence comparison. Based on sequence data in GenBank-specific oligo(d)nucleotide primers were designed using OLIGO Primer Software (version 5.0, National Biosciences). The upper primer (5⬘-CTTTCTAAGTCGGCTTCGAG-3⬘) was designed to the mitochondrial leader sequence of human dGK (the murine dGK sequence was unavailable at the time this experiment was performed) and the lower primer (5⬘-AGATATACCTGTCACTGTAC-3⬘) to a DRH motif conserved among many cellular kinases that use ATP as a highenergy phosphate donor (Balasubramaniam et al., ’90). PCR amplification consisted of 30 cycles (95°C for 1 min, 55°C for 2 min, and 72°C for 1 min). The PCR reaction mixture was 10 mM Tris-HCl, pH 9.0, containing 50 mM KCl, 0.2 ␮M primers, 0.2 mM each dNTP, and 1.5 U Taq polymerase. For T/A cloning, PCR product was purified (Qiagen) for use in a 10-␮l ligation reaction with T4 DNA ligase (1 ␮l, 1⫻) and pGEM-T Easy vector (1 ␮l, 50 ng) overnight at 4°C. Positive (Control Insert DNA provided by Promega) and negative (omission of insert DNA) controls were performed concurrently. In this study, 2 ␮l of each ligation reaction was used to transform JM 109 high-efficiency competent cells (50 ng; Promega). Duplicates of agar plates containing ampicillin (100 ␮g/ml), isopropyl-␤-D-thiogalactoside (IPTG; 0.5 mM) and 5-bromo-4-chloro-3indolyl-␤-D-galactoside (X-Gal; 80 ␮g/ml) for blue/white screening were each plated with 100 ␮l of the transformation culture and incubated for 16 –24 hr at 37°C. Individual white colonies were picked and grown overnight at room temperature with vigorous shaking in 2 ml of Luria-Bertani (LB) medium containing ampicillin (100 ␮g/ml); 500 ␮l of the overnight culture was used (the remainder was frozen at ⫺20°C in 60% glycerol) in Qiaprep plasmid DNA extraction (Qiagen). To confirm the presence of the PCR product, a restriction digest using EcoRI was done. The reaction was incubated for 2 hr at 37°C and the enzyme was inactivated after a 15-min incubation at 65°C. An aliquot of the digested sample was run alongside an equal volume of undi-

158

WUBAH ET AL.

gested sample on 8% nondenaturing polyacrylamide gel at 100 V for 1 hr. Samples were sequenced using PRISM Ready Reaction DyeDeoxy Terminator Sequencing kit on an Applied Biosystems model 377 DNA Sequencer (Applied Biosystems, Foster, CA) at the Nucleic Acid Facility of the Kimmel Cancer Center of Thomas Jefferson University. The sequencing primers were T7 and SP6 RNA polymerases, which allowed sequencing to be performed in both the forward and reverse directions. We used Marathon RACE (Clontech) to obtain the full-length cDNA sequence of murine dGK. Gene-specific primers were: GSP1 (5⬘-ACACAGACCTCTCAAAGACCCGCACAGA-3⬘) and GSP2 (5⬘-CCCAAAAAGATGGCACTTCCAAACGTCT-3⬘). These primers corresponded to antisense and sense sequence information for reverse (5⬘-RACE) and forward (3⬘-RACE) products, respectively. The cDNA was used in an adaptor ligation reaction with a Marathon cDNA adaptor and T4 DNA ligase. The reaction was incubated for 4 hr at room temperature and heated for 5 min at 70°C to inactivate the ligase. This library of adaptor-ligated doublestranded cDNA was diluted to 1:50 (Tricine-EDTA buffer), heated for 2 min at 94°C to denature the double strands and used in the next step. 5⬘ RACE used 1 ␮l of GSP1 (0.2 ␮M) and 1 ␮l of AP1 adapter primer (0.2 ␮M). 3⬘ RACE used 1 ␮l of GSP2 (0.2 ␮M) with 1 ␮l AP1. All necessary positive and negative controls were performed in parallel. PCR conditions were as follows: 94°C for 1 min, 5 cycles (94°C for 30 sec, 72°C for 5 min), 5 cycles (94°C for 30 sec, 70°C for 5 min), and 25 cycles (94°C for 30 sec, 68°C for 5 min). Polyacrylamide gel electrophoresis (3.5%) identified the most robust bands; these bands were excised, extracted, and cloned into pGEM-T Easy vector for bidirectional sequencing. RESULTS Teratological evaluation Dosing pregnant mice on GD8 generated the 2CdA dose-response curve. Each treatment group and an untreated control group consisted of 10 litters. Results confirmed the specificity of this exposure for microphthalmia and anophthalmia (Fig. 1). Exencephaly and microcephaly were produced in a minority (3–11%) of fetuses presenting with micro-/anophthalmia (not shown). Complete data for the effect of 2CdA treatment on resorptions, fetal weight, and malformations are shown in Table 1. Concerning eye malformations, 1 in 10 litters showed an affected fetus for the control group and each of the low-dose groups (0.5 and 1.5 mg/kg), whereas 9 in 10 litters for the 5.0-mg/kg dose group, and 10 of the 11 litters surviving the combined high-dose treatments (7.5, 10.0 mg/kg), showed at least one affected fetus. Therefore, 1.5 mg/kg was the highest dosage having no observable adverse effect on the quantal (litter) parameter. For the continuous (individual) parameter, the mean incidence of fetuses displaying micro-/anophthalmia, 0.9% in the control population and 0.5-mg/kg

Fig. 1. Eye malformations in CD-1 mouse fetuses induced by 2-chloro-2⬘-deoxyadenosine (2CdA) exposure on day 8 of gestation. Fetuses were fixed in formalin to make the lens apparent. (Left) control fetus displaying normal eye (normophthalmia). (Middle) microphthalmia with 1.5 mg/kg 2CdA injected on GD8. (Right) anophthalmia caused by 10 mg/kg 2CdA.

treatment group, increased with 2CdA dosage from 1.8% of fetuses at 1.5 mg/kg 2CdA to 85% of fetuses surviving 10.0 mg/kg 2CdA. The effect on the developing eye became significant at the 5.0-mg/kg dose level (24.4% malformed). Therefore, 1.5 mg/kg was the highest dosage having no observable adverse effect on the continuous (individual) parameter. Significant effects on fetal weight and resorption rate did not appear until the 5.0-mg/kg and 7.5-mg/kg dose levels, respectively. Dichotomous (quantal) data from the percentage of malformed fetuses (Table 1) were plotted as a function of dosage (Fig. 2), choosing 5.9% as the Benchmark Response (BMR) for an increased 5% risk of microphthalmia over background level of 0.9%. Nonlinear regression used BMDS software (www.epa.gov/ncea, US EPA, ’95) to fit the data to a probit model and resulted in a smooth curve that models the dose-response curve for percentage of fetuses with microphthalmia. The function describing this exposure– disease relationship was f ⫽ CumNorm共a ⫹ b ⴱ dose兲 where CumNorm is the standard normal distribution function, constant a is the intercept (estimate ⫺2.485, standard error 0.247), and constant b is the slope (estimate 0.356, standard error 0.041). The chi-square for goodness of fit was 0.23 with 3 degrees of freedom (df). The P-value for the test of fit is 0.97. Thus, the model describes the data quite well. The 2CdA dose that produced an extra 5% risk of microphthalmia (BMD5) can be estimated as 2.52 and the corresponding BMDL for the lower 95% confidence limit is 1.99 mg/kg. To determine whether useful information for lowdose extrapolation could be gathered by integrating the “graded” with “individual” dose-response data, gross observable reduction in size of the optic globe was used as the developmental outcome variable (Table 1). Eye malformations became more severe at higher 2CdA doses. For example, microphthalmia prevailed at 5.0 mg/kg whereas anophthalmia prevailed at 10.0 mg/kg. Eyes were individually scored as 0, 2, or 3, depending on whether they were judged to be normal, microph-

† Mean (95% CI). Statistical comparison by one-way analysis of variance (ANOVA) used unweighted litter means (n ⫽ 10); differences were localized by comparison to the control group, using the unpaired t-test, *P ⱕ 0.05; **P ⱕ 0.01; ***P ⱕ 0.001. ERD grade was calculated by assigning a value to the eye of 0 if normophthalmic, 2 if microphthalmic, and 3 if anophthalmic, and then dividing the cumulative value for the entire population by the number of eyes in the nested data set. CI, 95% confidence interval; ERD, eye reduction defects; 2-CdA, 2-chloro-2⬘-deoxyadenosine; GD8, gestational day 8.

ERD grade (CI)

0.0 0.0 0.0 2.1 23.0 61.1 — 0.4 0.5 0.8 16.8 26.3 18.5 — 5.7 (2.0–13.8) 7.0 (1.8–20.6) 10.4 (5.8–17.6) 9.2 (5.1–15.7) 40.2* (15.2–70.7) 78.7*** (50.5–93.9) P ⬍ 0.001 12.3 (10.5–14.1) 11.4 (10.2–12.6) 12.5 (11.4–13.6) 13.1 (12.4–13.8) 12.7 (11.8–13.6) 12.7 (12.0–13.4) P ⫽ 0.437 0 0.5 1.5 5.0 7.5 10.0 one-way ANOVA

1.00 (0.95–1.04) 0.95 (0.84–1.06) 0.96 (0.92–1.01) 0.85*** (0.81–0.90) 0.88** (0.83–0.94) 0.82* (0.74–0.89) P ⫽ 0.008

0.9 (0.0–5.8) 0.9 (0.0–6.3) 1.8 (0.0–10.7) 24.4*** (16.3–34.6) 57.9*** (34.3–78.7) 85.2** (65.6–95.1) P ⬍ 0.001

Missing eye (%) Small eye (%) Malformed fetuses (%) (CI) Fetal weight (g) (CI) Resorptions (%) (CI) Implants/litter (CI) Dose (mg/kg) GD8

TABLE 1. Developmental toxicity of 2CdA in CD-1 mice†

0.014 (0.00–0.028) 0.014 (0.00–0.028) 0.031 (0.00–0.091) 0.502** (0.263–0.741) 1.156** (0.625–1.687) 1.925** (1.185–2.665) P ⬍ 0.001

DOSIMETRY OF 2CdA-INDUCED MICROPHTHALMIA

159

Fig. 2. Benchmark dose calculation. Experimental data points and 95% confidence intervals were plotted for micro-/anophthalmia rates (% malformed fetuses in Table 1) and fit to a probit model using BMDS software (www.epa.gov/ncea/bmds). The BMR (benchmark response) and BMD (benchmark dose) are shown as an extra 5% risk above the control level; the BMD5 was estimated as 2.52 mg/kg. BMDL was estimated as 1.99 mg/kg.

thalmic, or anophthalmic, respectively. An ERD grade was then calculated by dividing the cumulative values for the entire population by the total number of eyes in the nested data set. The graded response revealed the increasing effect of 2CdA on malformation incidence and a trend toward higher ERD grade below the significance level (e.g., 1.5 mg/kg). Subtle defects, such as coloboma and narrowing of the pupillary ring previously recorded in pigmented TSG-p53 mice and graded 1 (Wubah et al., ’96), were not included in this evaluation of the nonpigmented CD-1 mice. Therefore, we recalculated ERD grades for TSG-p53 mice, reported in the previous study with 2CdA, using microphthalmia (grade 2) and anophthalmia (grade 3) only. With this scale, the effect of the p53 genotype variable on a constant exposure gave ERD grades as follows: 0.965 for p53⫹/⫹ fetuses, 0.633 for p53⫹/⫺ fetuses, and 0.000 for p53⫺/⫺ fetuses. Corresponding malformation rates for these fetuses were 57%, 56%, and 0%, respectively. Thus, a graded response revealed the intermediate effect of a heterozygous p53-deficient condition that was not seen in the malformation incidence. p53 protein induction Induction of nuclear accumulation of p53 was previously observed in GD8 mouse embryos 4.5 hr postexposure to 10.0 mg/kg 2CdA (Wubah et al., ’96). To determine the dosimetry of nuclear p53 accumulation, pregnant CD-1 mice were dosed with 2CdA on GD8 and the embryos were procured for immunohistochemical staining 4.5 hr postexposure. Control (untreated) embryos were also obtained at the same time. The normal pattern of p53 immunoreactivity observed in control embryos was uniformly weak cytoplasmic–nuclear staining. Omission of the primary antibody showed no residual staining to suggest the control pattern was

160

WUBAH ET AL.

specific for the p53 protein (not shown). A minority of cells showed enhanced immunostaining in the cytoplasm and/or nucleus, particularly in some cranial neural crest cells. Embryos exposed to 0.5 or 1.5 mg/kg 2CdA displayed a normal distribution of p53-immunoreactivity (Fig. 3B–D). In contrast, embryos exposed to 5.0 mg/kg 2CdA displayed many cells with nuclear ⬎ cytoplasmic phenotypes (Fig. 3E–G). This means that doses at or below the developmental NOAEL failed to elicit demonstrable p53 induction. Patterns further shifted toward nuclear ⬎ cytoplasmic p53 staining after exposures to 7.5- and 10.0-mg/kg (Fig. 3H–J) dose levels. The different grades of p53 protein induction are demonstrated in Figure 3 for the neural ectoderm, neural crest, and foregut endoderm. No response was observed in the primitive heart at any dosage of 2CdA examined in this report (not shown). Furthermore, dosing scenarios invoking low-grade induction in the headfold had no demonstrable effect at 1.5-hr or 3.0-hr postexposure intervals, whereas dosing scenarios invoking high-grade induction had an early effect at 3.0 hr. Pharmacokinetics of 2CdA exposure Pregnant mice were dosed with 0.5, 1.5, 5.0, or 10.0 mg/kg 2CdA on day 8. Extracts of the antimesometrial compartment were analyzed at postexposure times of 0 min (untreated), 5 min, 15 min, 30 min, 45 min, 60 min, 90 min, and 180 min. Because a day 8 mouse embryo, still nourished by yolk sac diffusion, has a potential for unacceptable loss of extracellular metabolites (hence 2CdA) during microdissection of the implantation chamber, the embryo was extracted as a unit with the antimesometrium for this analysis. Co-retention and spectral congruency versus authentic standard easily identified the 2CdA peak in biological samples (Fig. 4). 2CdA was measured in the antimesometrium at each dose ⫻ time combination. Data for the mean pmol/mg protein estimated by the linear mixed effects models are presented in Table 2, along with the uncertainty in the estimates of the means expressed as 95% confidence intervals. The among-litter standard deviation of log[pmol/mg protein] with approximate 95% confidence interval was 0.42 (0.30 – 0.59). The within-litter standard deviation with 95% confidence interval was 0.23 (0.19 – 0.28). Interpretation of pharmacokinetic parameters using empirical data gives Tmax for the disposition phase ⫽ 5 min for the highest dose (10.0 mg/kg), 15 min for the two intermediate doses (1.5, 5.0 mg/kg), and 30 min for the lowest dose (0.5 mg/kg). 2CdA decayed to or below the detection limits within 180 minutes and the half-lives of the elimination phase were approximately 30 min at each dose (Table 2). To visualize the disposition and elimination of 2CdA from the antimesometrium a smooth curve was fit to the empirical data at each dose level (Fig. 5). These results were consistent with the two-compartment model for the plasma of adult mice treated subcutaneously with 42.0 mg/kg 2CdA showing half-lives of 11.4 and 150 min for the two phases (Reichelova et al., ’95).

AUC estimates were derived from the empirical data using the trapezoidal rule. Resulting Cmax and AUC values revealed a direct relationship between the applied dosage and the amount of 2CdA in the antimesometrium (Table 3). 2CdA was converted from units of pmol per mg protein to micromolar (␮M) concentration by two different methods. Method 1 used empirical characteristics of the antimesometrium on day 8 (4.2 mg protein, 15-␮l vol, 20 mg wet wt). It assumed an even distribution of 2CdA in the tissue sample. Method 2 was based on plasma data reported by Reichelova et al., ’95 (Cmax ⫽ 54.80 ␮M) for 42.0 mg/kg 2CdA, which is considerably higher than the embryotoxic range (0.5–10.0 mg/kg). That plasma value was extrapolated to the doses tested here and assumed 2CdA levels in the antimesometrium were similar to plasma levels in a nonpregnant mouse. Both methods gave similar results (Table 3). Thus, we obtain similar results when pmol/mg protein levels were transformed to micromolar concentration by two different methods depending on parameters and assumptions from the antimesometrium (Method 1) or published maternal plasma (Method 2). This suggests good agreement between antimesometrium and maternal plasma kinetics. An average of the two methods estimated Cmax and AUCs for 2CdA at 0.9 ␮M and 31.5 ␮M-min, respectively, for the no observable adverse effect level (e.g., 1.5 mg/kg), and 2.7 ␮M and 97.8 ␮M-min for the BMD5 (e.g., 2.5 mg/kg). Therefore, the threshold for ocular teratogenicity in CD-1 mouse embryos falls within a range of 0.9 –2.7 ␮M peak exposure and 31.5–97.8 total integrated exposure on day 8 of gestation. The comparison with the day 8 mouse embryo proper is more difficult. Some of the reasons are the same as those mentioned above pertaining to compartmentation of the extracellular fluids. A study of 2CdA in the rat showed similar kinetics in the embryo proper and maternal plasma (albeit relatively later in gestation) and will be reported elsewhere. 5ⴕ/3ⴕ RACE amplification and sequence analysis of murine deoxyguanosine kinase The rate-limiting step in 2CdA cytotoxicity is phosphorylation to 2CdAMP. Two enzymes may catalyze this physiological reaction in mammalian cells: deoxycytidine kinase (dCK) or deoxyguanosine kinase (dGK). Previous studies found no evidence for a contribution of dCK deoxyadenylate stress in early mouse embryos (Gao et al., ’94) unlike human cell lines (Gao et al., ’95). Therefore, we sought to characterize the dGK transcript in early mouse embryos using a two-step strategy. Step 1, conventional RT-PCR, amplified a transcript using relevant sequence information from human dGK. An upper primer was designed to the cleavable mitochondrial leader sequence and a lower primer to the DRH motif conserved among many cellular kinases that utilize ATP as a high-energy phosphate donor. These primers amplified a ⬃0.3 kilobase pair (kb) PCR product from the prosencephalon and primitive heart (Fig. 6A). The resulting sequence

DOSIMETRY OF 2CdA-INDUCED MICROPHTHALMIA

Fig. 3. p53-immunoperoxidase staining in the headfold region of day 8 mouse embryos procured 4.5 hr after in utero exposure to 2-chloro2⬘-deoxyadenosine (2CdA). A: Low scan of a head-fold section to orient enlargements of cranial neural crest region (B,E,H), foregut pocket (C,F,I), and cranial neuroepithelium (D,G,J). Embryos were from dams dosed with 0.5 mg/kg (B–D); 5.0 mg/kg (E–G); or 10.0 mg/kg (H–J). Specific p53 immunoperoxidase staining in cytoplasm (e.g., C) and nucleus (e.g., I) was lost with omission of the primary antibody. Staining of the 0.5-mg/kg embryo was indistinguishable from the

161

negative-control (untreated) embryos and adequately represented the distribution of p53 protein in an unstimulated embryo; these cells were phenotyped as no response (cytoplasm ⬎ nucleus). Staining of the 10.0 mg/kg embryo was equivalent to the positive-control human breast tumor (not shown); these cells were phenotyped as a highgrade response (many cells with nucleus ⬎ cytoplasm). Staining of the 5.0-mg/kg embryo represented an intermediate, low-grade response (some cells with nucleus ⬎ cytoplasm).

162

WUBAH ET AL.

Fig. 4. High-performance liquid chromatography (HPLC) detection of 2-chloro-2⬘-deoxyadenosine (2CdA) in nucleoside extracts of the antimesometrium. A: Sample hybrid calibration curve generated with authentic 2CdA standard. The curve was generated with replicates at each level and differences were smaller than the diameter of plotting points. B–D: Chromatograms of antimesometrial samples monitored

at 264 nm absorbance. B: Tissue sample extracted 30 min after 5.0-mg/kg 2CdA treatment. C: Untreated (control) tissue sample showing baseline in the 2CdA region of the chromatogram. D: Baseline sample spiked with 50 pmol 2CdA standard. Arrows denote retention time of the 2CdA peak, which was confirmed by spectral congruency with authentic standard.

TABLE 2. Dose ⴛ time distribution of 2CdA in the antimesometrium of pregnant CD-1 mice* Time (min)

0.5 mg/kg

1.5 mg/kg

5.0 mg/kg

10.0 mg/kg

5 15 30 45 60 90 180

0.909 (0.470–1.760) 0.903 (0.467–1.750) 0.989 (0.511–1.910) 0.235 (0.121–0.454) — — —

2.270 (1.17–4.38) 3.11 (1.61–6.02) 2.12 (1.09–4.09) 0.458 (0.237–0.886) 0.514 (0.266–0.994) 0.408 (0.211–0.790) —

9.87 (5.10–19.10) 14.20 (7.34–27.50) 7.56 (3.91–14.6) 6.53 (3.38–12.6) 5.29 (2.73–10.2) 1.83 (0.944–3.53) 0.275 (0.108–0.700)

33.7 (17.4–65.2) 9.99 (5.16–19.3) 17.7 (9.16–34.3) 7.78 (4.02–15.1) 6.50 (3.36–12.6) 3.25 (1.68–6.29) 2.04 (1.06–3.95)

*Means and 95% confidence intervals (CI) for 2CdA measured in pmol/mg protein.

showed the highest probability match to dGK in a blast search of GenBank. This PCR product was ⬃0.1 kb shorter than predicted from human or murine dGK sequence information due to the specific absence of region A (Johansson and Karlsson, ’96). In step 2, forward and reverse gene-specific primers designed for 5⬘/3⬘-RACE yielded 0.5 kb (5⬘) and 0.85 kb (3⬘) bands (Fig. 6B) that had the correct overlapping sequence. The amalgamated cDNA sequence (Fig. 7) revealed a 982-bp cDNA with 57% homology to the human dGK cDNA. There was 10 bp of 5⬘-untranslated region (UTR). An open reading frame of 711 bp showed all five highly conserved substrate binding domains (S1–S5) predicted for nucleoside kinases (Brown et al., ’95). The deduced 236 amino acid sequence was 76% homologous to human dGK (Johansson and Karlsson, ’96) and 100% homologous to murine dGK (Petrakis et al., ’99). However, it lacked the last 35 (mouse) or 41 (human) C-terminal amino acids. Petrakis et al. (’99) reported

two splice variants of murine dGK referred to as dGK-1 and dGK-2. The embryonic splice variant was designated dGK-3. DISCUSSION Within the context of structural birth defects, a biological threshold can be defined as the minimal stimulus that begins to produce an observable adverse effect on the fetus. A priori, a threshold would manifest as the “critical effect malformation,” usually defined in terms of the first malformation to appear in a doseresponse curve. Microphthalmia appeared first in the dose-response curve with 2CdA and thus was designated the critical effect malformation for 2CdA exposure on GD8. Susceptibility of the eye to diverse teratogenic exposure scenarios is well known. Maternal folic acid deficiency (Armstrong and Monie, ’66), prenatal irradiation (Strange and Murphree ’72), zinc defi-

DOSIMETRY OF 2CdA-INDUCED MICROPHTHALMIA

Fig. 5. Dose means and approximate 95% confidence intervals, with the fitted concentration ⫻ time curves superimposed. Each data point used four measured samples (duplicate within-litter sets of five-pooled antimesometrial halves of embryodecidua from two different litters). A smooth curve was fit to the estimated means for each dose group

163

implemented in the function nlme (S-Plus, version 3.4). The function used was biexponential: y ⫽ dose 䡠 eCdose (e⫺At ⫺ e⫺Bt) Parameters A, B, and Cdose were estimated to fit the data. In addition, it was assumed that exponent Cdose not only varied systematically with dose, but an additional random component varied among litters.

TABLE 3. Pharmacokinetic parameters of 2CdA in the antimesometrium C max (␮M)a

Dose (mg/kg) GD8

C max pmol/mg protein (CI)

Method 1

Method 2

AUC pmol-min/mg protein (CI)

Total exposure (␮M/min)

0.5 1.5 5.0 10.0

1.0 (0.5–1.9) 3.1 (1.6–6.0) 14.2 (7.3–27.5) 33.7 (17.4–65.2)

0.28 0.87 3.98 9.45

0.34 1.06 4.86 11.50

34.7 (22.2–47.2) 112.2 (77.8–146.6) 703.9 (520.6–887.2) 1193.3 (892.5–1492.2)

9.7 31.5 197.3 334.6

*Estimates of peak concentration (C max) and area under the Cxt curve (AUC) with 95% confidence intervals (CI). Predicted ␮M concentration based on pmol 2CdA per mg protein data. Two methods were used for this transformation. Method 1 used empirical characteristics of the antimesometrial half of the GD8 embryodecidual unit (4.2 mg protein, 0.015-ml volume, 20 mg wet weight). Method 2 derived values by regressing C max ⫽ 54.8 ␮M at 42.0 mg/kg 2CdA (Reichelova et al., ’95) to the specific doses tested here (r 2 ⫽ 0.9997).

a

ciency (Rogers and Hurley, ’87), cyclophosphamide (Peiffer et al., ’91), hypervitaminosis A (Sulik et al., ’95), methylmercury (Chambers and Klein, ’93), and trichloroethylene (Narotsky et al., ’95) all cause microphthalmia and anophthalmia in rodents. In fact, the eye ranks among the most sensitive structures in the embryo to many agents at a time when susceptibility to teratogens is at a peak. Microphthalmia is also part of the syndrome of human malformations linked to thalidomide (Miller and Stromland, ’99). 2CdA, although not an environmental hazard, invokes a common disease pathway in the developing embryo, justifying its use as a tool compound for BBDR model development.

Risk assessment methods commonly used to evaluate the developmental toxicity of environmental exposures have traditionally focused on identification of the no observable adverse effect level (NOAEL) (Faustman et al., ’99). The present study identified 1.5 mg/kg 2CdA as the NOAEL defined by the highest dosage having no observable adverse effect on mouse fetuses evaluated by the quantal (Q) or continuous (C) teratological parameters. The former is based on the percentage of litters showing an effect within the study group and the latter on the mean incidence of affected implants within litters (Kavlock and Setzer, ’96). One approach to low-dose extrapolation in risk assessment uses the

164

WUBAH ET AL. Fig. 6. 5⬘/3⬘ rapid amplification of cDNA ends (RACE) of murine deoxyguanosine kinase. Embryonic poly(A⫹)RNA was isolated from the prosencephalon (PN) or primitive heart (HT) on GD8 and reverse-transcribed to cDNA. 5⬘- and 3⬘-RACE was performed in two steps. A: Step 1 used straight PCR to amplify cDNA with upper primer to the mitochondrial import signal sequence and lower primer to the conserved DHR kinase domain; a 282-bp product was amplified (arrow). B: Step 2 developed 5⬘ and 3⬘ RACE primers from the 282-bp sequence and generated 500-bp and 850-bp RACE products (arrows), which were then cloned and sequenced. The final gel on the right is EcoRI digest of the appropriate clones.

critical NOAEL to calculate a reference dose (RfD), which is an estimate of daily exposure that is assumed to be without appreciable risk of deleterious effects (Kimmel and Kimmel, ’94). However, NOAEL has only operational meaning within the context of a specific experimental design. Apart from weaknesses in statistical power and endpoint sensitivity, reliance on NOAEL as the entrance point for dose extrapolation fails to account for the shape and variability of the dose-response curve (Kavlock and Setzer, ’96). Another approach to derive RfD, the benchmark dose (BMD), uses measurements across a range of exposures and estimates risk by fitting a mathematical model to the available data and deriving a lower confidence limit on the effective dose associated with a defined level of effect (Kimmel, ’98; Krewski and Zhu, ’95). The 2CdA dose that produced an extra 5% risk of microphthalmia (BMD5) was estimated at 2.5 mg/kg. Proximity of the BMR to the predicted dose level that induced a 5% increase in malformations placed the risk levels for BMD5 (2.5 mg/kg) and NOAEL (1.5 mg/kg) within a factor of 2. Most conventional developmental toxicity studies find proximity between the NOAEL and the BMD5 (Kavlock and Setzer, ’96). Thus, we get essentially the same information from low dose extrapolation of BMD5, quantal (Q-NOAEL), and continuous (C-NOAEL) data. BMDL (replaces NOAEL) at approximately 2.0 mg/kg fell exactly midway between NOAEL and BMD5. Although BMD5 is the more accurate derivation of RfD within the context of an exposure– disease continuum for a mechanistic study model, the present descriptive study reveals the conservative nature of dose-response models based on quantal endpoints that reduce the data to whether a litter contained at least one malformed fetus (Kavlock and Setzer, ’96). Kavlock and Setzer (’96) state that it is “important not to confuse the concept of NOAEL with that of threshold for biological effects, as the former merely reflects the statistical power of an experiment to see an

effect when in fact one does exist.” Our abilities to assess subtle effects on the size of the optic globe were limited with CD-1 mice, which are nonpigmented, versus fetuses on a C57BL/6J genetic background. Pigmentation of the retina permitted visual detection of subtle malformations such as narrowing of the pupillary ring and colobomas of the iris and retina even before microphthalmia was observable (Wubah et al., ’96). These subtle malformations were difficult to detect in CD-1 fetuses, so the minimal stimulus that begins to produce an observable adverse effect on the eye might actually have been lower than 1.5 mg/kg. Hence, our experience with 2CdA supports the aforementioned caveat (Kavlock and Setzer, ’96) and corroborates the need for quantitative information about critical biomolecular changes in the exposed embryo and the resulting shape of the low end of the dose-response curve for these sensitive parameters (Lau et al., ’01). Delineating the exposure– disease continuum requires first-hand knowledge of the pharmacokinetic profiles in a pregnant dam. For 2CdA, a determination of peak levels (Cmax) and concentration versus time (C ⫻ t) profiles in the antimesometrium provides direct analysis of the embryonic environment while preserving the integrity of embryonic nucleoside pools (Knudsen et al., ’92). Similarity between concentration ⫻ time profiles of the antimesometrium and adult mouse plasma (Reichelova et al., ’95) makes it possible to predict exposure levels in the very low dose range (e.g., ⬍0.5 mg/kg) by limited sampling of the maternal plasma. However, the ability of 2CdA to enter intermediary metabolism makes it difficult to determine the actual 2CdA exposure at the level of the target organ. BBDR models can provide a framework for defining the dose response for subtle changes associated with lower, environmentally relevant exposures (Conolly and Andersen, ’91; Shuey et al., ’94; Kavlock and Setzer, ’96; Faustman et al., ’99; Lau et al., ’00). Lau et al. (’00) outlined three assumptions to render construction of BBDR models tenable: (1) a major, or critical,

DOSIMETRY OF 2CdA-INDUCED MICROPHTHALMIA

165

Fig. 7. Full-length cDNA sequence of murine deoxyguanosine kinase (dGK-3). The 982-bp cDNA is compared with the deduced amino acid sequence. Features included (a) 10-bp 5⬘-UTR and 251-bp 3⬘-UTR reading with polyadenylation signal (displayed in upper case text); (b) 711-bp open reading frame; (c) an asterisk (*) indicates putative mitochondrial peptide cleavage site; (d) coordinates for step 1 primers (28 – 48 upper, 444 – 461 lower in italics) and step 2 primers (416 – 443

GSP1, 258 –286 GSP2 in boldface); (e) 5-boxes correspond to five conserved regions for nucleoside kinases: S1, glycine loop; S2, aids in substrate binding; S3, DRH motif; S4, C(Y/F)P motif; and S5, arginine-rich site described for herpesviral deoxythymidine kinases (Balasubramaniam et al., ’95). The embryonic d6K-3 sequence was deposited into GenBank. (Accession number AY037862).

mechanistic pathway is directly responsible for the adverse developmental outcomes; (2) perturbation of this mechanistic pathway can be evaluated quantitatively relative to dose and time of exposure; and (3) homologous mechanisms of toxicity via a specific pathway exist between test animal species and humans. Concerning 2CdA-induced micro-/anophthalmia, the first assumption is supported by evidence for the involvement of p53 (Wubah et al., ’96) and dose-dependent effects on p53 protein induction described here. The dosimetry of 2CdA-induced microphthalmia, also provided in the present study, adds evidence for an empirical link between exposure (pharmacokinetics) and disease (malformations) to support at least part of the second assumption. Since p53 function appears to be a critical determinant in developmental toxicity, it is important to understand how this pathway contributes to the mode of action of toxicant-induced malformations. p53 is a transcription factor that is frequently mutated in human neoplasm, with most of the oncogenic mutations mapping to or near sites of DNA contact (Cho et al., ’94). Wild-type p53 protein under normal conditions

has rapid intracellular turnover (Rogel et al., ’85). The protein is stabilized after cellular stress such as DNA damage (Kastan et al., ’91), hypoxia (Graeber et al., ’94), and ribonucleotide pool imbalance (Linke et al., ’96). Any of these imbalances are plausible initiating mechanisms for 2CdA. Downstream changes associated with p53 protein induction include cell cycle arrest and programmed cell death (apoptosis) (Choi and Donehower, ’99). These changes have been observed in association with p53-dependent developmental toxicity (Wubah et al., ’96; Moallem and Hales, ’98; Torchnisky et al., ’99; Wang et al., ’00). In the present study, we observed that 2CdA doses below the BMDL had no demonstrable effect on the immunohistochemical distribution of p53 protein in the mouse embryo. Lowgrade p53 induction was evident at a moderately teratogenic dose (5.0 mg/kg), whereas a high-grade induction followed strongly embryotoxic dosage (e.g., 10.0 mg/kg). These findings support a direct connection between p53 protein induction and developmental toxicity (with 2CdA). The direct correlation between Trp53 genotype and teratogenesis has now been observed for 2CdA-induced

166

WUBAH ET AL.

microphthalmia (Wubah et al., ’96) and radiation-induced ectrodactyly (Wang et al., ’00). Both studies implicated excessive amounts of p53-dependent apoptosis as the mode of action in teratogenesis. The present study extends these findings to support a direct link between 2CdA exposure, p53 protein induction, and the risk of disease. However, it is important to note the contrasting observation of increased risk of teratogeninduced malformation with p53 deficiency (Nicol et al., ’95; Moallem and Hales, ’98; Kato et al., ’01). This suggests that the role of p53-dependent apoptosis as a cellular proofreading process may shift the threshold for teratogenicity upward or downward, depending on factors such as dose rate of exposure (Kato et al., ’01). Consistent with this notion, p53 protein induction in response to 2CdA exposure was not confined to the optic pit nor did it prescribe a clearly defined embryonic precursor target cell population. Therefore, p53 protein induction may be part of the mode of action of 2CdA but cannot account for the mechanism of ocular specificity. A contribution of p53-independent cellular processes may be possible (Pettitt et al., ’99) especially at higher dosages of 2CdA leading to resorptions (Wubah et al., ’96). Analysis of the pharmacokinetic profile of 2CdA in the antimesometrium showed that disposition-elimination phases (e.g., 0.25–3.0 hr postexposure) precedes p53 protein induction in the embryo proper (e.g., 3.0 – 4.5 hr postexposure). This latency is consistent with the notion of a pro-drug requiring metabolic activation via the nucleoside “salvage” pathway (Carson et al., ’80, ’83). Additional evidence derives from the relationship between the pro-apoptotic activity of 2CdA and accumulation of 2CdATP (Hirota et al., ’89; Gao et al., ’95; Leoni et al., ’98). A teratogenic dose of 5.0 mg/kg 2CdA corresponded with an estimated pro-drug exposure of 4 ␮M (Cmax) in the antimesometrium. This value is consistent with the in vitro potency for 2CdAinduced apoptosis in day 8 mouse embryos (e.g., 5 ␮M) reported in Wubah et al. (’96) (see publisher’s erratum with that reference). A study on the growth of human leukemic cell lines showed an accumulation of 2CdATP to 20 ␮M after 6-hr continuous exposure to 1 ␮M 2CdA (Wilson et al., ’98). This accumulation was accompanied by a decline in cellular deoxynucleoside 5⬘-triphosphates (dNTPs) between 0 – 4 hr. The quantitative pharmacodynamic model of Guchelaar et al. (’98) included the time and dose-dependent cell killing curves of 2CdA and other nucleoside analogues in human leukemic cells. These investigators reported dose-dependent apoptosis upon continuous incubation with 0.1– 10.0 ␮M 2CdA, with no effect at 0.01 ␮M and a maximal rate of apoptosis was achieved at 6 hr of continuous exposure. If the threshold for 2CdA teratogenicity is assumed between NOAEL (1.5 mg/kg) and BMD5 (2.5 mg/kg), a threshold exposure can then be modeled from the pharmacokinetic parameters in the antimesometrium (Table 3) using a three-parameter Hill equation (r2 ⫽ 0.999). The threshold exposure for 2CdA teratogenicity corresponded to a Cmax of 1.0 –1.8

␮M, a range somewhat higher than Guchelaar’s model (0.1 ␮M). However, in terms of total integrated exposure the threshold exposure corresponded to AUC 30 – 80 ␮M-min, similar to Guchelaar’s model (0.1 ␮M ⫻ 360 min ⫽ 36.0 ␮M-min). Hence, pharmacokinetic analysis of the potency of 2CdA in vivo compares with the published in vitro potency for various biochemical and cellular events. 2CdATP levels are related to the rate of 2CdAMP formation, which in turn is rate-limiting in 2CdA cytotoxicity (Carson et al., ’80, ’83). Mammalian cells contain four main salvage enzymes: thymidine kinase 1 (TK1), thymidine kinase 2 (TK2), deoxycytidine kinase (dCK), and deoxyguanosine kinase (dGK) (Arner and Erikkson, ’95; Johansson and Eriksson, ’96). Both dCK and dGK exhibit 2CdA phosphorylating activity under physiological conditions; however, they differ from one another in their subcellular localization: dCK resides principally in the nucleus, whereas dGK resides in the mitochondrion (Wang et al., ’93, ’96; Zhu et al., ’98; Jullig and Eriksson ’00). In some cells, dCK accounts for observed tissue and species differences of 2CdA toxicity (Reichelova et al., ’95). This finding is consistent with the ability of 2⬘-deoxycytidine to protect against 2CdATP accumulation in human leukemia cell lines (Gao et al., ’95). Predictably, exogenous 2⬘-deoxycytidine did not protect mouse embryos from deoxyadenylate stress (Gao et al., ’94). The present studied identified dGK-3 as an embryonic variant of dGK. Although further studies are needed to confirm dGK-3 as important in 2CdA embryotoxicity its sequence had several distinctive features pertaining to the N-terminal peptide. Only one ATG translation start codon was evident in murine dGK-3, as for murine dGK-1 (Petrakis et al., ’99). The first 40 amino acid residues of dGK-3 perfectly matched dGK-1 and shared 51% homology with human dGK. In each case the N-terminal peptide fit biophysical criteria for proteins destined for mitochondrial import: (1) a nearly complete absence of acidic amino acid residues; (2) a preponderance of arginine, serine, and leucine residues relative to a sample of N-terminal peptides of cytosolic proteins; and (3) a segment with a large predicted helical hydrophobic moment (Hendrick et al., ’89). Biochemical and immunochemical studies confirm that a substantial fraction of cellular dGK activity, perhaps all of it, resides in the mitochondrion (Wang et al., ’93; Johansson et al., ’97; Zhu et al., ’98; Jullig and Eriksson ’00). These findings suggest the early mouse embryo harbors significant capacity to bioactivate 2CdA in the mitochondrion. Because dGK-3 was detected in the headfold and heart, we cannot attribute differential susceptibility of these structures to pro-drug bioactivation. A more plausible explanation pertains to differences in mitochondrial genetic activity between the headfold and heart, and between p53 genotypes, in day 8 mouse embryos (Ibrahim et al., ’98). Precursor deoxynucleotide pools for mitochondrial DNA (mtDNA) synthesis are maintained by the enzymatic activities of

DOSIMETRY OF 2CdA-INDUCED MICROPHTHALMIA TK2 and dGK in the mitochondrial matrix, most likely making mitochondria independent of de novo precursor synthesis and cell cycle control (Arner and Erikkson, ’95; Johansson and Eriksson, ’96). The same property can explain why 2CdA and other 2⬘-deoxyadenosine analogues exhibit unique activity toward both cycling and quiescent cells (Carson et al., ’83; Seto et al., ’85; Kizaki et al., ’88; Benveniste and Cohen, ’95), an activity that renders 2CdA (Cladribine, Leustatin) an effective anticancer agent toward slow-growing, chronic lymphoid malignancies (Rai, ’98). Contamination of mitochondrial dATP pools (with 2CdATP) can perturb mtDNA replication (Chunduru et al., ’93; Hentosh and Grippo, ’94) and mitochondrial function in target cells (Hentosh and Tiduban, ’97). Bioactivation of 2CdA in the mitochondrion (Zhu et al., ’98), and the fact that mitochondrial biogenesis in general and mtDNA replication in particular more tightly couple to cellular bioenergetic demand than to progression of the cell cycle (Wiesner, ’97), point to the mitochondrion as the subcellular target for 2CdA embryotoxicity. Perhaps focus on the mitochondrion can guide basic research aimed at uncovering basic disease mechanisms in developmental toxicity and more specifically informed judgments about the shape of the dose-response curve in regions below experimental observation. ACKNOWLEDGMENTS The authors gratefully acknowledge scientific input from Drs. Carole Kimmel, Gary Kimmel, Michael Narotsky, and Robert Kavlock of the U.S. Environmental Protection Agency. J.A.W. was a predoctoral fellow on Training Grant T32 ES07282 from the National Institute of Environmental Health Sciences and NRSA Grant F31 HD08167 from the National Institute of Child Health and Human Development. Although the research described in this article has been supported by the United Stated Environmental Protection Agency through assistance agreement EPA-CR 824 445-01 to Jefferson Medical College, it has not been subjected to Agency review and therefore does not necessarily reflect the views of the Agency and no official endorsement should be inferred. Mention of trade names or commercial products does not constitute endorsement or recommendation for use. LITERATURE CITED Airhart MJ, Robbins CM, Knudsen TB, Church JK, Skalko RG. 1993. Occurrence of embryotoxicity in mouse embryos following in utero exposure to 2⬘-deoxycoformycin (pentostatin). Teratology 47:17–27. Armstrong C, Monie IW. 1966. Congenital eye defects in rats following maternal folic-acid deficiency during pregnancy. J Embryol Exp Morphol 16:531–542. Arner ESJ, Eriksson S. 1995. Mammalian deoxyribonucleoside kinases. Pharmacol Ther 67:155–186. Balasubramaniam NK, Veerisetty V, Gentry GA. 1990. Herpes viral deoxythymidine kinases contain a site analogous to the phosphorylbinding arginine-rich region of porcine adenylate kinase; comparison of secondary structure predictions and conservation. J Gen Virol 71:2979 –2987. Benveniste P, Cohen A. 1995. p53 expression is required for thymocyte apoptosis induced by adenosine deaminase deficiency. Proc Natl Acad Sci USA 92:8373– 8377.

167

Blackburn MR, Gao X, Airhart MJ, Skalko RG, Thompson LF, Knudsen TB. 1992. Adenosine levels in the early postimplantation mouse uterus. Quantitative analysis by HPLC-fluorometric detection and spatio-temporal regulation by 5⬘-nucleotidase and adenosine deaminase. Dev Dynam 194:155–168. Blackburn MR, Knudsen TB, Kellems RE. 1997. Genetically engineered mice demonstrate that adenosine deaminase is essential for early postimplantation development. Development 124:3089 –3097. Brown DG, Visse R, Sandhu G, Davies A, Rizkallah PJ, Melitz C, Summers WC, Sanderson MR. 1995. Crystal structures of the thymidine kinase from herpes simplex virus type-1 in complex with deoxythymidine and ganciclovir. Nature Struct Biol 2:876 – 881. Carson DA, Wasson DB, Kaye J, Ullman B, Martin DW Jr, Robins RK, Montgomery JA. 1980. Deoxycytidine kinase-mediated toxicity of deoxyadenosine analogs toward malignant human lymphocyblasts in vitro and toward murine L1210 leukemia in vivo. Proc Natl Acad Sci USA 77:6865– 6869. Carson DA, Wasson DB, Taetle R, Yu A. 1983. Specific toxicity of 2-chlorodeoxyadenosine toward resting and proliferating human lymphocytes. Blood 62:737–743. Chambers BJ, Klein NW. 1993. Role of laminin autoantibodies on the embryo toxicity of sera from mercuric chloride treated Brown Norway rats. Reprod Toxicol 7:333–341. Cho Y, Gorina S, Jeffrey PD, Pavletich NP. 1994. Crystal structure of a p53 tumor suppressor-DNA complex: understanding tumorigenic mutations. Science 265:346 –355. Choi J, Donehower LA. 1999. p53 in embryonic development: maintaining a fine balance. Cell Mol Life Sci 55:38 – 47. Chunduru SK, Appleman JR, Blakley RL. 1993. Activity of humand DNA polymerases ␣ and ␤ with 2-chloro-2⬘-deoxyadenosine 5⬘triphosphate as a substrate and quantitative effects of incorporation on chain extension. Arch Biochem Biophys 302:19 –30. Cohen A, Hirschhorn R, Horowitz SD, Rubinstein A, Palmar SH, Hong R, Martin DW Jr. 1978. Deoxyadenosine triphosphate as a potentially toxic metabolite in adenosine deaminase deficiency. Proc Natl Acad Sci USA 75:472– 476. Conolly RB, Andersen ME. 1991. Biologically based pharmacodynamic models: tools for toxicological research and risk assessment. Annu Rev Pharmacol Toxicol 31:503–523. Davidian M, Giltinan D. 1995. Nonlinear models for repeated measurement data. London: Chapman & Hall. Faustman EM, Lewandowski TA, Ponce RA, Bartell SM. 1999. Biologically based dose-response models for developmental toxicants: lessons learned from methylmercury. Inhal Toxicol 11:559 –572. Gao X, Blackburn MR, Knudsen TB. 1994. Activation of apoptosis in early mouse embryos by 2⬘-deoxyadenosine exposure. Teratology 49:1–12. Gao X, Knudsen TB, Ibrahim MM, Haldar S. 1995. Bcl-2 relieves deoxyadenylate stress and suppress apoptosis in pre-B leukemia cells. Cell Death Diff 2:69 –78. Gaylor D, Razzaghi M. 1992. Process of building biologically based dose-response models for developmental defects. Teratology 46:573–581. Graeber TG, Peterson M, Tsai M, Monica K, Fornace AJ Jr, Giacca AJ. 1994. Hypoxia induces accumulation of p53 protein, but activation of a G1-phase checkpoint by low oxygen conditions is independent of p53 status. Mol Cell Biol 14:6264 – 6277 Guchelaar H-J, Vermes I, Koopmans RP, Reutelingsperger CPM, Haanen C. 1998. Apoptosis- and necrosis-inducing potential of cladribine, cytarabine, cisplatin, and 5-fluorouracil in vitro: a quantitative pharmacodynamic model. Cancer Chemother Pharmacol 42:77– 83. Hendrick JP, Hodges PE, Rosenberg LE. 1989. Survey of aminoterminal proteolytic cleavage sites in mitochondrial precursor proteins: leader peptides cleaved by two matrix proteases share a three-amino acid motif. Proc Natl Acad Sci USA 86:4056 – 4060. Hentosh P, Grippo P. 1994. Template 2-chloro-2⬘-deoxyadenosine monophosphate inhibits in vitro DNA synthesis. Mol Pharmacol 45:955–961. Hentosh P, Tiduban M. 1997. 2-Chloro-2⬘-deoxyadenosine, an antileukemic drug, has an early effect on cellular mitochondrial function. Mol Pharmacol 51:613– 619.

168

WUBAH ET AL.

Hershfield MS, Fetter JE, Small WC, Bagnara AS, Williams SR, Ullman B, Martin DW Jr, Wasson DB, Carson DA. 1982. Effects of mutational loss of adenosine kinase and deoxycytidine kinase on deoxyATP accumulation and deoxyadenosine toxicity in cultured CEM human T-lymphoblastoid cells. J Biol Chem 257:6380 – 6386. Hirota Y, Yoshioka A, Tanaka S, Watanabe K, Otani T, Minowada J, Matsuda A, Ueda T, Wataya Y. 1989. Imbalance of deoxyribonucleoside triphosphates, DNA double-strand breaks, and cell death caused by 2-chlorodeoxyadenosine in mouse FM3A cells. Cancer Res 49:915–919. Hoever M, Clement JH, Wedlich D, Montenarh M, Knochel W. 1994. Overexpression of wild-type p53 interferes with normal development in Xenopus laevis embryos. Oncogene 9:109 –120. Ibrahim MM, Razmara M, Nguyen D, Donahue RJ, Wubah JA, Knudsen TB. 1998. Altered expression of mitochondrial 16S ribosomal RNA in p53-deficient mouse embryos revealed by differential display. Biochim Biophys Acta 1403:254 –264. Ikeda S, Hawes NL, Chang B, Avery CS, Smith RS, Nishina PM. 1999. Severe ocular abnormalities in C57BL/6 but not in 129/Sv p53-deficient mice. Invest Ophthalmol Vis Sci 40:1874 –1878. Johansson M, Eriksson S. 1996. Structure-activity relationships for phosphorylation of nucleoside analogs to monophosphates by nucleoside kinases. Acta Biochim Polon 43:143–160. Johansson M, Karlsson A. 1996. Cloning and expression of human deoxyguanosine kinase cDNA. Proc Natl Acad Sci USA 93:7258 – 7262. Johansson M, Brismar S, Karlsson A. 1997. Human deoxycytidine kinase is located in the cell nucleus. Proc Natl Acad Sci USA 94:11941–11945. Jullig M, Eriksson S. 2000. Mitochondrial and submitochondrial localization of human deoxyguanosine kinase. Eur J Biochem 267: 5466 –5472. Kastan MB, Onyekwere O, Sidransky D, Vogelstein B, Craig RW. 1991. Participation of p53 protein in the cellular response to DNA damage. Cancer Res 51:6304 – 6311. Kato F, Ootsuyama A, Nomoto S, Kondo S, Norimura T. 2001. Threshold effect for teratogenic risk of radiation depends on dose-rate and p53-dependent apoptosis. Int J Radiat Biol 77:13–19. Kavlock RJ, Setzer RW. 1996. The road to embryologically based dose-response models. Environ Health Persp 104:107–121. Kimmel CA. 1998. Current approaches to risk assessment for developmental neurotoxicity. In: Handbook of developmental neurotoxicity. Slikker W, Chang LW, editors. San Diego, CA: Academic Press. p 675– 685. Kimmel CA, Kimmel GL. 1994. Risk assessment for developmental toxicity. In: Kimmel CA, Buelke-Sam J, editors. Developmental toxicity. 2nd ed. New York: Raven Press. p 429 – 453. Kizaki H, Shimada H, Oshaka F, Sakurada T. 1988. Adenosine, deoxyadenosine, and deoxyguanosine induce DNA cleavage in mouse thymocytes. J Immunol 141:1652–1657. Knudsen TB, Gray MK, Church JK, Blackburn MR, Airhart MJ, Kellems RE, Skalko RG. 1989. Early postimplantation embryolethality in mice following in utero inhibition of adenosine deaminase with 2⬘-deoxycoformycin. Teratology 40:615– 626. Knudsen, TB, Winters RS, Otey SK, Blackburn MR, Airhart MJ, Church JK, Skalko RG. 1992. Effects of (R)-deoxycoformycin (pentostatin) on intrauterine nucleoside catabolism and embryo viability in the pregnant mouse. Teratology 45:91–103. Krewski D, Zhu Y. 1995. A simple data transformation for estimating benchmark doses in developmental toxicity experiments. Risk Anal 15:29 –39. Laird NM, Ware JH. 1982. Random-effects models for longitudinal data. Biometrics 38:963–974. Lau C, Andersen ME, Crawford-Brown DJ, Kavlock RJ, Kimmel CA, Knudsen TB, Muneoka K, Rogers JM, Setzer RW, Smith G, Tyl R. 2000. Evaluation of biologically based dose-response modeling for developmental toxicity: a workshop report. Reg Tox Pharmacol 31:190 –199. Lau C, Mole ML, Copeland MF, Rogers JM, Kavlock RJ, Shuey DL, Cameron AM, Ellis DH, Logsdon TR, Merriman J, Setzer RW. 2001. Toward building a biologically based dose-response model for developmental toxicity of 5-fluorouracil in the rat: acquisition of experimental data. Toxicol Sci 59:37– 48.

Leoni LM, Chao Q, Cottam HB, Genini D, Rosenbach M, Carrera CJ, Budihardjo I, Wang X, Carson DA. 1998. Induction of an apoptotic program in cell-free extracts by 2-chloro-2⬘-deoxyadenosine 5⬘triphosphate and cytochrome c. Proc Natl Acad Sci USA 95:9567– 9571. Liliemark J. 1997. The clinical pharmacokinetics of cladribine. Clin Pharmacokinet 32:120 –131. Lindstrom MJ, Bates DM. 1988. Newton-Raphson and EM algorithms for linear mixed-effects models for repeated-measures data. J Am Stat Assoc 83:1014 –1022. Linke SP, Clarkin KC, Di Leonardo A, Tsou A, Wahl GM. 1996. A reversible, p53-dependent G0/G1 cell cycle arrest induced by ribonucleotide depletion in the absence of detectable DNA damage. Genes Dev 10:934 –947. Miller MT, Stromland K. 1999. Teratogen update: thalidomide: a review, with a focus on ocular findings and new potential uses. Teratology 60:306 –321. Moallem SA, Hales BF. 1998. The role of p53 and cell death by apoptosis and necrosis in 4-hydroperoxycyclophosphamide-induced limb malformations. Development 125:3225–3234. Narotsky MG, Weller EA, Chinchilli VM, Kavlock RJ. 1995. Nonadditive developmental toxicity in mixtures of trichloroethylene, di(2thylhexyl)phthalate, and hepaochlor in a 5 ⫻ 5 ⫻ 5 design. Fundam Appl Toxicol 27:203–216. Nicol CJ, Harrison ML, Laposa RR, Gimelshtein IL, Wells P. 1995. A teratological suppressor role for p53 in benzo[a]pyrene-treated transgenic p53-deficient mice. Nature Genet 10:181–187. O’Flaherty EJ, Clarke DO. 1994. Pharmacokinetic/pharmacodynamic approaches for developmental toxicity. In: Kimmel CA, Buelke-Sam J, editors. Developmental toxicity. 2nd ed. New York: Raven Press. p 215–244. Peiffer RL, McCullen R, Alles A, Sulik KK. 1991. Relationship of cell death to cyclophosphamide-induced ocular malformations. Teratogenesis Carcinog Mutagen 11:203–212. Petrakis TG, Ktistaki E, Wang L, Erikkson S, Talianidis I. 1999. Cloning and characterization of mouse deoxyguanosine kinase. Evidence for a cytoplasmic isoform. J Biol Chem 274:24726 –24730. Pettitt AR, Sherrington PD, Cawley JC. 1999. The effect of p53 dysfunction on purine analogue cytotoxicity in chronic lymphocytic leukaemia. Br J Haematol 106:1049 –1051. Rai KR 1998. Cladribine for the treatment of hairy cell leukemia and chronic lymphocytic leukemia. Semin Oncol 25:19 –22. Rao JNK, Scott AJ. 1992 A simple method for analysis of clustered binary data. Biometrics 48:577–585. Reichel MB, Ali RR, D’Esposito F, Clarke AR, Luthert PJ, Bhattacharya SS, Hunt DM. 1998. High frequency of persistent hyperplastic primary vitreous and cataracts in p53-deficient mice. Cell Death Diff 5:156 –162. Reichelova V, Juliusson G, Spasokoukotskaja T, Eriksson S, Liliemark J. 1995. Interspecies differences in the kinetic properties of deoxycytidine kinase elucidate the poor utility of a phase I pharmacologically directed dose-escalation concept for 2-chloro-2⬘-deoxyadenosine. Cancer Chemother Pharmacol 36:524 –529. Rogel A, Popliker M, Webb CG, Oren M. 1985. p53 cellular tumor antigen: analysis of mRNA levels in normal adult tissues, embryos, and tumors. Mol Cell Biol 5:2851–2855. Rogers JM, Hurley LS. 1987. Effects of zinc deficiency on morphogenesis of the fetal eye. Development 99:231–288. Sah VP, Attardi LD, Mulligan CJ, Williams BO, Bronson RT, Jacks T. 1995. A subset of p53-deficient embryos exhibits exencephaly. Nature Genet 10:175–180. Seto S, Carrera CJ, Kubota M, Wasson DB, Carson DA. 1985. Mechanism of deoxyadenosine and 2-chlorodeoxyadenosine toxicity to nondividing human lymphocytes. J Clin Invest 75:377–383. Setzer RW, Lau C, Mole ML, Copeland MF, Rogers JM, Kavlock RJ. 2001. Toward building a biologically based dose-response model for developmental toxicity of 5-fluorouracil in the rat: a mathematical construct. Toxicol Sci 59:49 –58. Shuey DL, Lau C, Logsdon TR, Zucker RM, Elstein KH, Narotsky MG, Setzer RW, Kavlock RJ, Rogers JM. 1994. Biologically based dose-response modeling in developmental toxicology: biochemical and cellular sequelae of 5-fluorouracil exposure in the developing rat. Toxicol Appl Pharmacol 126:129 –144

DOSIMETRY OF 2CdA-INDUCED MICROPHTHALMIA Spaeth GL, Nelson LB, Beaudoin AR. 1982. Experimental teratogenesis. In: Jakobiec FA, editor. Ocular anatomy, embryology, and teratology. Philadelphia: Harper & Row. p 968 –981. Strange R, Murphree RL. 1972. Exposure-rate responses in the prenatally irradiated rat: effect of 100 R on day 11 of gestation to the developing eye. Radiat Res 51:674 – 684. Sulik KK, Dehart DR, Rogers JM, Chernoff N. 1995. Teratogenicity of low doses of all-trans retinoic acid in presomite mouse embryos. Teratology 51:398 – 403. Torchinsky A, Ivnitsky I, Savion S, Shepshelovich J, Gorivodsky M, Fein A, Carp H, Schwartz D, Frankel J, Rotter V, Toder V. 1999. Cellular events and the pattern of p53 protein expression following cyclophosphamide-initiated cell death in various organs of developing embryo. Teratogenesis Carcinog Mutagen 19:353–367. US EPA. 1995. The use of the benchmark dose approach in health risk assessment. Office of Research and Development, Washington, DC: EPA/630/R-94/007. Wang L, Karlsson A, Arner ES, Eriksson S. 1993. Substrate specificity of mitochondrial 2⬘-deoxyadenosine kinase: efficient phosphorylation of 2-chlorodeoxyadenosine. J Biol Chem 268:22847–22852.

169

Wang L, Hellman U, Eriksson S. 1996. Cloning and expression of human mitochondrial deoxyguanosine kinase cDNA. FEBS Lett 390:39–43. Wang B, Ohyama H, Haginoya K, Odaka T, Yamada T, Hayata I. 2000. Prenatal radiation-induced limb defects mediated by Trp53dependent apoptosis in mice. Radiat Res 154:673– 679. Weisner RJ. 1997. Adaptation of mitochondrial gene expression to changing cellular energy demands. New Physiol Sci 12:178 –184. Wilson PK, Szabados E, Mulligan SP, Christopherson RI. 1998. Comparative effects of cladribine, fludarabine and pentostatin on nucleotide metabolism in T- and B-cell lines. Int J Biochem Cell Biol 30:833– 842. Wubah JA, Ibrahim MM, Gao X, Nguyen D, Pisano MM, Knudsen TB. 1996. Teratogen-induced eye defects mediated by p53-dependent apoptosis. Curr Biol 6:60 – 69 (publisher’s erratum appears in Curr Biol 6:753). Zhu C, Johansson M, Permert J, Karlsson A. 1998. Enhanced cytotoxicity of nucleoside analogs by overexpression of mitochondrial deoxyguanosine kinase in cancer cell lines. J Biol Chem 273:14707– 14711.