Chapter 10: Developmental Toxicology

Developmental toxicology encompasses the study of developmental exposures, pharmacokinetics, mechanisms, pathogenesis, and outcomes potentially leading to adverse health effects. Manifestations of developmental toxicity include structural malformations, growth retardation, functional or metabolic impairment, and/or death of the organism. Developmental exposures may also alter the risk of diseases in adulthood. Developmental toxicology defined as such is a relatively new science, but teratology, the study of structural birth defects, as a descriptive science preceded written language. For example, a marble sculpture from southern Turkey dating to 6500 bc depicts conjoined twins (Warkany, 1983), and Egyptian wall paintings of human conditions such as cleft palate and achondroplasia date to as long as 5000 years ago. Conjecture has it that mythological figures such as the Cyclops and sirens took their origin in the birth of malformed infants (Thompson, 1930; Warkany, 1977). Ancient Babylonians, Greeks, and Romans believed that abnormal infants were reflections of celestial events and were considered to be portents of the future. Indeed, the Latin word monstrum, from monstrare (to show) or monere (to warn), connotes an ability to foretell the future. In turn, derivation of the word teratology is from the Greek word for monster, teras.

Hippocrates and Aristotle thought that abnormal development could originate in physical causes such as uterine trauma or pressure, but Aristotle also held a widespread belief that maternal impressions and emotions could influence the development of the child. He advised pregnant women to gaze at beautiful statuary to increase their child’s beauty. Although this theory sounds fanciful, it is present in diverse cultures throughout recorded history. Indeed, we now know that maternal stress, depression, and anxiety during pregnancy can be deleterious to the developing conceptus and child (Dunkel Schetter and Tanner, 2012).

In 1649, the French surgeon Ambrois Paré expounded upon the theory of Aristotle and Hippocrates by writing that birth defects could result from narrowness of the uterus, faulty posture of the pregnant woman, or physical trauma such as a fall. Fetal limb amputations were thought to result from amniotic bands, adhesions, or twisting of the umbilical cord. This conjecture has proven to be accurate. With the blossoming of biology in the 16th and 17th centuries, scientific theories of causation of birth defects began to emerge. In 1651, William Harvey put forth the theory of developmental arrest, which stated that malformations resulted from incomplete development of an organ or structure. An example given by Harvey was harelip in humans, a congenital malformation that represents a normal developmental stage. Much later, the theory of developmental arrest was supported by the experiments of Stockard (1921) using eggs of the minnow, Fundulus heteroclitus. By manipulating the chemical constituents and temperature of growth media, he produced malformations in the embryos, the nature of which depended on the developmental stage at the time of the insult. He concluded that developmental arrest explained all malformations except those of hereditary origin (Barrow, 1971).

With the advent of the germplasm theory elucidated by Weissmann in the 1880s and the rediscovery of Mendel’s laws in 1900, genetics as the basis for some birth defects was accepted. In 1894, Bateson published his treatise on the study of variations in animals as a tool for understanding evolution, inferring that inheritance of such variations could be a basis for speciation (Bateson, 1894). His study contains detailed descriptions and illustrations of such human birth defects as polydactyly and syndactyly, supernumerary cervical and thoracic ribs, duplicated appendages, and horseshoe (fused) kidneys. Bateson coined the term homeosis to denote morphological alterations in which one structure takes on the likeness of another. Studies of such alterations in mutants of the fruit fly Drosophila melanogaster and, more recently, vertebrates, have served as the basis for much of our present knowledge of the genetic control of development. Homeobox genes are found throughout the animal and plant kingdoms and direct embryonic pattern formation (Graham et al., 1989; Deschamps and van Nes, 2005; Mallo et al., 2010). Acceptance of a genetic basis of birth defects was furthered with studies of human inborn errors of metabolism in the first decade of the 20th century.

Modern experimental teratology began in the early 19th century with the work of Etienne Geoffrey Saint-Hilaire. Saint-Hilaire produced malformed chick embryos by subjecting eggs to various environmental conditions including physical trauma (jarring, inversion, pricking) and toxic exposures. In the latter part of the 19th century, Camille Dareste experimented extensively with chick embryos, producing a wide variety of malformations by administering noxious stimuli, physical trauma, or heat shock at various times during development. He found that timing was more important than the nature of the insult in determining the type of malformation produced. Among the malformations described and beautifully illustrated by Dareste (1877, 1891) were the neural tube defects (NTDs) anencephaly and spina bifida, cyclopia, heart defects, situs inversus, and conjoined twins. Many of the great embryologists of the 19th and 20th centuries, including Loeb, Morgan, Driesch, Wilson, Spemann, and Hertwig, performed teratological manipulations employing various physical and chemical probes to deduce principles of normal development.

In the early 20th century, a variety of environmental conditions (temperature, microbial toxins, drugs) were shown to perturb development in avian, reptilian, fish, and amphibian species. In contrast, mammalian embryos were at that time thought to be resistant to induction of malformations, protected from adverse environmental conditions by the maternal system. The first reports to the contrary were published in the 1930s and were the result of experimental maternal nutritional deficiencies. Hale (1935) produced malformations including anophthalmia and cleft palate in offspring of sows fed a diet deficient in vitamin A. Beginning in 1940, Josef Warkany and co-workers began a series of experiments in which they demonstrated that maternal dietary deficiencies and other environmental factors could affect intrauterine development in rats (Warkany and Nelson, 1940; Warkany, 1945; Warkany and Schraffenberger, 1944; Wilson et al., 1953). These experiments were followed by many other studies in which chemical and physical agents, for example, nitrogen mustard, trypan blue, hormones, antimetabolites, alkylating agents, hypoxia, and x-rays, to name a few, were clearly shown to cause malformations in mammals (see Warkany, 1965).

The first recognized human epidemic of malformations induced by an environmental agent was reported by Gregg (1941), who linked an epidemic of rubella virus infection (German measles) in Australia to an elevation in the incidence of eye, heart, and ear defects, and mental retardation. The triad of deafness, cataracts, and cardiac disease is now recognized as the clinical signature of congenital rubella syndrome. Heart and eye defects predominated with maternal infection in the first or second months of pregnancy, whereas hearing and speech defects and mental retardation were most commonly associated with infection in the third month. Later, the risk of congenital anomalies associated with rubella infection in the first four weeks of pregnancy was estimated to be 61%; in weeks five to eight, 26%; and in weeks nine to 12, 8% (Sever, 1967). It has been estimated that in the United States alone approximately 20,000 children have been impaired as a consequence of prenatal rubella infections (Cooper and Krugman, 1966). Although maternal rubella is now uncommon in developed countries due to vaccination programs, there are still rubella outbreaks in developing countries (De Santis et al., 2006; Chandy et al., 2011), and some cases in developed countries as well (Vauloup-Fellous et al., 2010).

Although embryos of mammals, including humans, were found to be susceptible to common external influences such as nutritional deficiencies and intrauterine infections, the impact of these findings was not fully appreciated at the time (Wilson, 1973). That changed, however, in 1961, when the association between thalidomide ingestion by pregnant women and the birth of severely malformed infants was established (see below).

The overall miscarriage rate in humans is estimated to be 15% to 20% of recognized pregnancies. However, with the development of highly sensitive tests, pregnancies can now be detected shortly after fertilization. When these tests are used early, recognized pregnancy loss increases to about 60% to 70%. Estimates of adverse outcomes include postimplantation pregnancy loss, 31%; major birth defects, 2% to 3% at birth and increasing to 6% to 7% at one year as more manifestations are diagnosed; minor birth defects, 14%; low birth weight, 7%; infant mortality (prior to one year of age), 1.4%; and abnormal neurological function, 16% to 17% (Schardein, 2000). Reasons for these adverse outcomes are largely unknown. Brent and Beckman (1990) attributed 15% to 25% of human birth defects to genetic causes, 4% to maternal conditions, 3% to maternal infections, 1% to 2% to deformations (eg, mechanical problems such as umbilical cord limb amputations), <1% to chemicals and other environmental influences, and 65% to unknown etiologies. These estimates are similar to those suggested by Wilson (1977). Regardless of etiology, the sum total represents a significant health burden in light of the over four million births annually in the United States.

It has been estimated that well over 4000 chemicals have been tested in animals for teratogenicity, with approximately 66% shown to be nonteratogenic, 7% teratogenic in more than one species, 18% teratogenic in most species tested, and 9% producing equivocal experimental results (Schardein, 2000). In contrast, only about 35 to 40 chemicals, chemical classes, or conditions (Table 10-1) have been documented to alter prenatal development in humans (Schardein and Keller, 1989; Shepard and Lemire, 2004). A brief review of selected human developmental toxicants provides both a historical view of developmental toxicology and an illustration of some of the key principles of the science.

Thalidomide

In 1960, a striking increase in newborns with rare limb malformations was recorded in West Germany. The affected individuals had amelia (absence of the limbs) or various degrees of phocomelia (reduction of the long bones of the limbs), usually affecting the arms more than the legs and usually involving both left and right sides, although to differing degrees. Congenital cardiac, ocular, otic, intestinal, and renal anomalies were also involved. However, the limb defects were pathognomonic, as limb reduction anomalies of this nature are exceedingly rare. At the university clinic in Hamburg, for example, no cases of phocomelia were reported between 1940 and 1959. In 1959 there was a single case; in 1960, there were 30 cases; and in 1961, a total of 154 cases (Taussig, 1962). The unusual nature of the malformations was a key factor in unraveling the epidemic. In 1961, Lenz and McBride, working independently in Germany and Australia, respectively, identified the sedative thalidomide as the causative agent (McBride, 1961; Lenz, 1961, 1963). Thalidomide had been introduced in 1956 by Chemie Grunenthal as a sedative/hypnotic and was used throughout much of the world as a sleep aid and to treat nausea and vomiting in pregnancy. The drug was widely prescribed at an oral dose of 50 to 200 mg/day. There were a few reports of peripheral neuritis attributed to thalidomide, but only in patients with long-term use for up to 18 months (Fullerton and Kermer, 1961). Following the association with birth defects, thalidomide was withdrawn from the market by Grunenthal in November 1961, and case reports ended in mid-1962 as exposed pregnancies were completed. Estimates of the number of infants malformed by thalidomide worldwide during this period range from 5840 (Lenz, 1988) to over 10,000 in 46 countries (Bartlett et al., 2004). Quantitative estimates of malformation risks from exposure have been difficult to compile but are believed to be in the range of one in two to one in 10 (Newman, 1985). Due to concerns regarding the severity of the peripheral neuritis and subsequent questions with regard to safety in pregnancy, thalidomide did not receive marketing approval by the US Food and Drug Administration (FDA) prior to its removal from the world market following the epidemic. Much credit is due to Dr Frances Kelsey, who was a new FDA staff member at the time. She later received the highest honor that can be bestowed upon a US civilian, the medal for Distinguished Federal Civilian Service, from President John F. Kennedy, for her diligence in protecting the country.

As a result of the thalidomide catastrophe, regulatory agencies in many countries began developing animal testing requirements for evaluating the effects of drugs during pregnancy (Stirling et al., 1997). In the United States, the discussions ultimately led to the development of the Segment I, II, and III testing protocols of the FDA (Kelsey, 1988). Current regulatory testing guidelines for developmental toxicity are discussed later in this chapter.

It is both ironic and telling that the chemical largely responsible for the advent of modern regulation of potential developmental toxicants presents a complex pattern of effects in different animal species. It has been tested for prenatal toxicity in at least 19 laboratory species. Malformations and increased embryonic loss have been observed in some studies in rats, whereas generally no effects were reported in studies with hamsters or most mouse strains. Effects more or less similar to those observed in humans have been reported for several rabbit strains and in eight of nine primate species. The teratogenic potency of thalidomide ranges from approximately 1 to 100 mg/kg among sensitive species. In this ranking the human sensitivity was estimated to be 1 mg/kg (Schardein, 1993).

Studies of the relationship between timing of drug use and type of malformation revealed that thalidomide was teratogenic between 20 and 36 days after fertilization (Lenz and Knapp, 1962). Because of its short half-life, teratogenic potency, and good records/recall of prescribed drug use, fairly concise timetables of susceptibility have been constructed (Lenz and Knapp, 1962; Nowack, 1965; Neubert and Neubert, 1997; Miller and Stromland, 1999). During the susceptible period of 20 to 36 days postfertilization, anotia (missing ear) was the defect induced earliest, followed by thumb, upper extremity, lower extremity, and triphalangeal thumb (Miller and Stromland, 1999). Despite extensive effort, research to understand the species and strain differences in response to thalidomide has met with limited success until recently. Extensive structure–activity studies involving analogs of thalidomide found strict structural requirements (eg, an intact phthalimide or phthalimidine group) but shed little light on potential mechanisms (Jonsson, 1972; Schumacher, 1975; Helm et al., 1981). Stephens (1988) reviewed 24 proposed mechanisms, including biochemical alterations involving vitamin B, glutamic acid, acylation, nucleic acids, and oxidative phosphorylation; cellular mechanisms including cell death and cell–cell interactions; and tissue level mechanisms including inhibition of nerve and blood vessel outgrowth. None was considered sufficient by that reviewer. More recent hypotheses concerning the mechanism of thalidomide teratogenesis include effects on angiogenesis (D’Amato et al., 1994; Joussen et al., 1999; Sauer et al., 2000; Stephens et al., 2000; Therapontos et al., 2009), integrin regulation (Neubert et al., 1996), oxidative DNA damage (Parman et al., 1999), TNF-α inhibition (Argiles et al., 1998), growth factor antagonism (Stephens et al., 1998; Stephens and Fillmore, 2000), effects on glutathione and redox status (Hansen et al., 1999), and inhibition of ubiquitin-mediated protein degradation (Ito et al., 2011; Ito and Handa, 2012).

The idea that thalidomide exerts its teratogenicity in sensitive species by altering cellular redox status has been explored by Wells and colleagues (Parman et al., 1999), who compared the effects of thalidomide in the pregnant mouse (an insensitive species) and the pregnant rabbit (a sensitive species). First, they demonstrated a 380% increase in DNA oxidation in rabbit embryos. Pretreatment of pregnant rabbits with the spin-trapping (free-radical scavenging) agent α-phenyl-N-t-butylnitrone (PBN) reduced thalidomide-induced embryonal DNA oxidation by 73%. Thalidomide was teratogenic and embryotoxic in the rabbit, inducing characteristic limb defects as well as other malformations, embryo/fetal mortality, and reduced fetal weight. Almost all of these effects on the rabbit embryos were abolished by maternal pretreatment with PBN. These results indicate that, in the rabbit, thalidomide is activated to a free-radical intermediate that initiates the formation of reactive oxygen species that appear to be a key element of the mechanism of teratogenicity because both oxidation and teratogenicity are largely reduced by PBN. This group has recently demonstrated limb bud embryopathies in rabbit embryos cultured in medium containing thalidomide or two of its hydrolysis products, with concomitant DNA oxidation (Lee et al., 2011). Although oxidation of DNA may be part of the teratogenic mechanism of thalidomide, Hansen and Harris (2004) have provided strong evidence to support their hypothesis that the limb defects induced by thalidomide are due to misregulation of the expression of genes critical for the outgrowth of the limb. They too have shown that thalidomide causes oxidative stress in sensitive (rabbit) but not in resistant (rat) pregnant animals and their embryos, by comparing glutathione depletion following administration of thalidomide (Hansen et al., 2002). They further demonstrated that this shift in redox potential reduces binding of a key transcription factor, NF-κB, to its binding sites on DNA. Binding of NF-κB is required to turn on the expression of the genes twist and FGF-10 in the mesenchyme of the developing limb. Lack of expression of these two genes resulted in loss of FGF-8 expression in the apical ectodermal ridge of the developing limb bud. Given our understanding of the importance of these genes for normal limb development, ablation of their expression by thalidomide could be predicted to cause the limb defects seen in sensitive species. That these genes are under the control of the redox-sensitive NF-κB may be the basis for the differences in teratogenicity of thalidomide in the sensitive rabbit (and presumably human) and the resistant mouse and rat.

Research on alterations in immune function and angiogenesis has opened the possibility of expanded use of thalidomide in diseases including dermatological disorders, arthritis, multiple myeloma, diabetic retinopathy, and macular degeneration (Adler, 1994; Calabrese and Fleischer, 2000; Schwab and Jagannath, 2006; Chen et al., 2010). Thalidomide has been approved by the FDA for erythema nodosum leprosum, an inflammatory complication of Hansen disease (leprosy) and multiple myeloma. In addition, it is used off-label for a number of disorders. An unprecedented level of safeguards, embodied in the STEPS program (System of Thalidomide Education and Prescribing Safety: http://www.thalomid.com/steps_program.aspx), surrounds thalidomide use to prevent accidental exposure during pregnancy, including required registration of all prescribers, pharmacies, and patients, required use of contraception, and periodic pregnancy testing for patients of childbearing ability (Lary et al., 1999).

Diethylstilbestrol

Diethylstilbestrol (DES), a synthetic nonsteroidal estrogen, was widely used from the 1940s to the 1970s in the United States to prevent miscarriage by stimulating synthesis of estrogen and progesterone in the placenta. Between 1966 and 1969, seven young women between the ages of 15 and 22 were seen at Massachusetts General Hospital with clear cell adenocarcinoma of the vagina. This tumor had not previously been seen in patients younger than 30. An epidemiological case–control study subsequently found an association with first trimester DES exposure (reviewed in Poskanzer and Herbst, 1977). The Registry of Clear Cell Adenocarcinoma of the Genital Tract of Young Females was established in 1971 to track affected offspring. Maternal use of DES prior to the 18th week of gestation appeared to be necessary for induction of the genital tract anomalies in offspring. The incidence of genital tract tumors peaked at age 19 and declined through age 22, with absolute risk of clear cell adenocarcinoma of the vagina and cervix estimated to be 0.14 to 1.4 per 1000 exposed pregnancies (Herbst et al., 1977). However, the overall incidence of noncancerous alterations in the vagina, cervix, and uterus was estimated to be as high as 75% (Poskanzer and Herbst, 1977). Effects of developmental exposure on the female reproductive tract may be due to alterations in genetic pathways governing uterine differentiation (Huang et al., 2005). In male offspring of exposed pregnancies, epididymal cysts, hypotrophic testes, and capsular induration along with low ejaculated semen volume and poor semen quality have been observed (Bibbo et al., 1977). The realization of the latent and devastating manifestations of prenatal DES exposure has broadened our concept of the magnitude and scope of potential adverse outcomes of intrauterine exposures. Yet, recent studies suggest even broader developmental targets; a study in mice by Newbold et al. (2006) showed that the increased susceptibility to tumors and reproductive tract abnormalities conferred by DES exposure may be passed on to future generations (both males and females) of exposed mothers. Developmental exposure to DES has also been shown to cause obesity in mice (Newbold et al., 2009).

Ethanol

The developmental toxicity of ethanol has been a recurrent concern throughout history and can be traced to biblical times (eg, Judges 13: 3–4). Yet, only since the description of the Fetal Alcohol Syndrome (FAS) by Jones and Smith in the early 1970s (Jones and Smith, 1973; Jones et al., 1973) has there been a clear recognition and acceptance of alcohol’s developmental toxicity. Since that time, there have been hundreds of clinical, epidemiological, and experimental studies of the effects of ethanol exposure during gestation, revealing an ever-widening array of effects of alcohol on development (for review, see Jones, 2011).

The FAS comprises craniofacial dysmorphology, intrauterine and postnatal growth retardation, retarded psychomotor and intellectual development, and other nonspecific major and minor abnormalities (Abel, 1982, 2006; Jones, 2011). The average IQ of FAS children has been reported to be 68 (Streissguth et al., 1991a) and changes little over time (Streissguth et al., 1991b). The craniofacial malformations as well as other morphological and anatomical effects of prenatal ethanol exposure may be due, at least in part, to interference with retinol metabolism in the early embryo, at a time when the anterior part of the embryo, including the brain, is just beginning to form (Yelin et al., 2005). Oxidation of retinol to the signaling molecule, retinoic acid, is crucial for normal development of the embryo. However, the molecular mechanisms underlying FAS are still poorly understood, and the possible molecular and cellular targets are many (eg, Sulik, 2005; Smith, 2010; Zeisel, 2011).

Full-blown FAS has been observed primarily in children born to alcoholic mothers, and among alcoholics the incidence of FAS has been estimated at 25 per 1000 (Abel, 1984). Numerous methodological difficulties are involved in attempting to estimate the level of maternal ethanol consumption associated with FAS, but estimates of a minimum of 3 to 4 oz of alcohol per day have been made (Clarren et al., 1987; Ernhart et al., 1987).

Decades of research has demonstrated that FAS represents an extreme of the range of effects of developmental alcohol exposure, now termed fetal alcohol spectrum disorders (FASDs) (Koren et al., 2003; Sokol et al., 2003; Manning and Hoyme, 2007; Hellemans et al., 2010). In utero exposure to lower levels of ethanol has been associated with a wide range of effects, including isolated components of FAS and effects on the developing brain leading to cognitive and behavioral deficits. Little (1977) studied 800 women prospectively to evaluate the effects of drinking on birth weight. After adjusting for smoking, gestational age, maternal height, age, parity, and sex of the child, it was found that for each ounce of absolute ethanol consumed per day during late pregnancy there was a 160-g decrease in birth weight. Effects of maternal alcohol consumption during pregnancy on attention, short-term memory, and performance on standardized tests have been noted in a longitudinal prospective study of 462 children (Streissguth et al., 1994a,b). Alcohol intake was related to decrements in these measures, the number of drinks per drinking occasion being the strongest predictor. Streissguth and co-workers (Baer et al., 2003) have also reported that prenatal alcohol exposure was significantly associated with alcohol drinking problems at age 21, independent of family history or other environmental factors. Later life vulnerability to stress, depression, and anxiety disorders after in utero exposure to alcohol are reviewed by Hellemans et al. (2010). For reviews of the neurotoxic effects of developmental ethanol exposure, see the special issue of Neuropsychology Review (June 2011;21(2)) on this topic.

One animal model of FAS in which pathogenesis of the craniofacial effects has been extensively studied involves intraperitoneal injection of ethanol to pregnant C57Bl/6J mice in early pregnancy when embryos are undergoing gastrulation (Sulik et al., 1981; Sulik and Johnston, 1983). Following such exposures, term fetuses exhibit many of the features of FAS, including microcephaly, microphthalmia, short palpebral fissures, deficiencies of the philtral region, and a long upper lip. The specific set of craniofacial malformations produced in offspring depends on the time of exposure. Excess cell death in sensitive cell populations is a common finding (Kotch and Sulik, 1992; Sulik, 2005; Smith, 2010).

Tobacco Smoke

Prenatal and early postnatal exposure to tobacco smoke or its constituents may well represent the leading cause of environmentally induced developmental disease and morbidity today (Rogers, 2008) (Fig. 10-1). Approximately 25% of women in the United States continue to smoke during pregnancy, despite public health programs aimed at curbing this behavior. Because of the high number of pregnant smokers and the relative accuracy of assessing smoking during pregnancy, results of epidemiological studies provide a well-characterized picture of the consequences of developmental tobacco smoke exposure. These include spontaneous abortions, perinatal deaths, sudden infant death syndrome (SIDS), learning, behavioral, and attention disorders, and lower birth weight (Slotkin, 1998; Fried et al., 1998; Tuthill et al., 1999; Haug et al., 2000; Markussen Linnet et al., 2006; Rogers, 2008). In a review of smoking and SIDS (Mitchell and Milerad, 2006), the relative risk associated with maternal smoking was 2.86 (95% confidence interval 2.77–2.95) before prone sleep-position intervention (putting infants on their backs for sleep reduces the overall risk of SIDS) and 3.93 (3.78–4.08) after intervention programs were started. The authors of the review estimated that about one-third of SIDS deaths might have been prevented if no fetuses had been exposed in utero to maternal smoking. Recent studies have found that offspring of smoking mothers have a higher risk of obesity, hypertension, and type 2 diabetes than offspring of nonsmokers (Oken et al., 2008; Suzuki et al., 2009; Bakker and Jaddoe, 2011; Cupul-Uicab et al., 2012).

Maternal smoking has not been associated with a high risk of structural malformations. However, Hackshaw et al. (2011) performed a systematic review and meta-analysis of observational studies reporting odds ratios for having a nonchromosomal birth defect for women who smoked versus nonsmokers. There were 172 published articles comprising 173,687 malformed cases and 11.7 million controls considered in the review. Significant positive associations with maternal smoking were found for these defects [odds ratios]: cardiovascular [1.09], musculoskeletal [1.16], limb reduction [1.26], missing/extra digits [1.18], clubfoot [1.28], craniosynostosis [1.33], facial [1.19], eye [1.25], orofacial clefts [1.28], gastrointestinal [1.27], gastroschisis [1.5], anal atresia [1.20], hernia [1.40], and undescended testes [1.13]. Lammer and co-workers (Lammer et al., 2005) found that maternal smoking was a risk factor for orofacial clefts, and increased risk is associated with particular genotypes indicating lack of enzymes involved in detoxification of smoking-derived chemicals. Gene–environment interactions may also be the basis for increased risk of other birth defects with maternal smoking (Shi et al., 2008). Perinatal exposure to tobacco smoke can also affect branching morphogenesis and maturation of the lung, leading to altered physiological function (Pinkerton and Joad, 2000; Gilliland et al., 2000; Maritz and Harding, 2011). In a genome-wide association study, two genes, TBK1 on chromosome 12 and ZNF236 on chromosome 18, showed multiple single nucleotide polymorphisms associated with higher risk of nonsyndromic cleft palate in the presence of maternal smoking (Beaty et al., 2011). One component of tobacco smoke, nicotine, is a known neuroteratogen in experimental animals and can by itself produce many of the adverse developmental outcomes associated with tobacco smoke, including impaired fertility, neurobehavioral effects, respiratory dysfunction, obesity, hypertension, and type 2 diabetes (Slotkin, 1998; Bruin et al., 2010).

It is important to state unequivocally that exposure to environmental (secondhand) tobacco smoke also represents a significant risk to the pregnant nonsmoker and her baby, and exposure to secondhand smoke has been associated with many of the effects caused by active maternal smoking (eg, Windham et al., 2000; George et al., 2006; Mitchell and Milerad, 2006). In a literature review with meta-analysis, Leonardi-Bee and co-workers (2008) found that exposure of nonsmoking pregnant women to secondhand smoke was associated with reduction of birth weight by 33 g or more, as well as a 22% increased risk of being of low birth weight (<2500 g). In a subsequent review and meta-analysis (Leonardi-Bee et al., 2011), they found that nonsmoking pregnant women exposed to secondhand tobacco smoke were 23% more likely to experience stillbirth and 13% more likely to have a child with a congenital malformation.

Cocaine

Cocaine, a plant alkaloid derived from coca, is a local anesthetic with vasoconstrictor properties. Effects on the fetus are complicated and controversial and demonstrate the difficulty of monitoring the human population for adverse reproductive outcomes (reviewed in Scanlon, 1991; Volfe, 1992; Addis et al., 2001; Ackerman et al., 2010). Accurate exposure ascertainment is difficult, as many confounding factors, including socioeconomic status and concurrent use of cigarettes, alcohol, and other drugs of abuse, may be involved. In addition, reported effects on the fetus and infant (neurological and behavioral changes) are difficult to identify and quantify. Cocaine exposure during pregnancy in humans appears to be associated with abruptio placentae (Addis et al., 2001), and preterm birth, low birth weight, and small-for-gestational age (SGA) infants (Gouin et al., 2011). Congenital malformations of the genitourinary tract have also been reported (Lutiger et al., 1991), and kidney and bladder function is diminished in fetuses of pregnant women using cocaine (Mitra, 1999). Moreover, fetal cocaine exposure has been associated with impaired neonatal auditory processing (Potter et al., 2000). Fetal cocaine exposure can be estimated by chemical analysis of fetal meconium, which can provide a measure of developmental exposure to xenobiotic agents ranging from food additives to over-the-counter medications to drugs of abuse (Ostrea et al., 1998).

Retinoids

The ability of excess vitamin A (retinol) to induce malformations has been known for at least 50 years (Cohlan, 1954). Effects on the developing embryo include malformations of the face, limbs, heart, central nervous system, and skeleton. Similar malformations were later shown to be induced by retinoic acid (RA) administration in the mouse (Kochhar, 1967) and hamster (Shenefelt, 1972). Since those observations, knowledge relating to the effects of retinol, RA, and structurally related chemicals that bind to and activate specific nuclear receptors has expanded rapidly (Chambon, 1994; Lohnes et al., 1994; Mendelsohn et al., 1994; Collins and Mao, 1999; Arafa et al., 2000). The RXR-α receptor appears to play an important role in cleft palate induced by RA (Nugent et al., 1999). The teratogenic effects of vitamin A and retinoids have been reviewed (Nau et al., 1994; Collins and Mao, 1999).

Beginning in 1982, one retinoid, 13-cis-RA (isotretinoin or Accutane), was marketed as an effective treatment of cystic acne. Despite clear warnings against use in pregnancy on the label of this prescription drug (FDA pregnancy category X: proven fetal risks clearly outweigh any benefit), an extensive physician and patient education program, and restrictive requirements for prescription to women of childbearing potential, infants with pathognomonic malformations involving the ears, heart, brain, and thymus were reported as early as 1983 (Rosa, 1983; Lammer et al., 1985). Among 115 exposed pregnancies not electively terminated, 18% ended in spontaneous abortion and 28% of live-born infants had at least one major malformation (Dai et al., 1992). In another prospective study, there was nearly a doubling of the risk for premature delivery after first trimester exposure, and about 50% of the exposed children had full-scale IQ scores below 85 at age five (Lammer, 1992). Robertson et al. (2005) reported that almost 25% of women with an isotretinoin-exposed pregnancy did not recall having contraception counseling before starting their medication. Rates of recommended birth control and monthly pregnancy testing were low. Thus, voluntary pregnancy prevention programs for women receiving prescriptions for isotretinoin were not highly effective, resulting in fetal exposures to isotretinoin. Adams (2010) has reviewed the neurobehavioral teratology of retinoids over the last 50 years.

Antiepileptic Drugs

Clinical management of women of childbearing age who have epilepsy is difficult. Although control of seizures during pregnancy is crucial, most current antiepileptic drugs (AEDs) have been shown to carry risk of developmental toxicity including birth defects, cognitive impairment, and fetal death. As a class, including phenytoin, carbamazepine and valproic acid, AEDs are considered human teratogens. These drugs can be used as monotherapy or in combination. In general, the incidence of birth defects increases with the number of drugs used in combination. Newer AEDs such as gabapentin, lamotrigine, oxcarbazone, topiramate, and zonizamide have now been used for over a decade, and studies to date suggest that these alternatives may be safer than the older AEDs. The interested reader is directed to several recent reviews (Tatum, 2006; Tomson and Battino, 2005; Ornoy, 2006; Hansen, 2010).

Phenytoin (diphenylhydantoin) has been used as an AED for over 70 years and its antiepileptic effect occurs through inhibition of voltage-gated sodium channels. It is metabolized by the cytochrome P450 system to a reactive intermediate and then to an active metabolite that is excreted in the urine. The fetal hydantoin syndrome, described by Hanson and Smith (1975), includes craniofacial abnormalities, limb defects, growth deficiency, and mild-to-moderate mental retardation. This syndrome has been reported to occur in 5% to 10% of exposed offspring (Schardein, 2000). Proposed mechanisms involved in the teratogenicity of phenytoin include folic acid deficiency and the formation of reactive intermediates that damage tissue macromolecules (for review see Hansen, 2010).

Carbamazepine has been reported to be teratogenic in rats and mice (Schardein, 2000), and a report from the FDA (Rosa, 1991) linked the drug to NTDs in humans. A meta-analysis of 16 studies reported a 2.89-fold increased risk of a major birth defect in offspring of pregnant women treated with carbamazepine (Matalon et al., 2002). An epoxide metabolite of carbamazepine was found to be teratogenic in mice in vivo (Bennett et al., 1996) but not in mice or rats in vitro (Hansen et al., 1996). The teratogenicity observed in mice may have been related to maternal toxicity or to a different metabolite.

Valproic acid, or 2-propylpentanoic acid, was first marketed in Europe in 1967 and in the United States in 1978. In 1982, Elizabeth Robert reported that, of 146 cases of spina bifida aperta recorded in Lyon, France, nine of the mothers had taken valproic acid during the first trimester of pregnancy. The odds ratio in a case–control study was 20.6, and the estimated risk of a valproate-exposed woman having a child with spina bifida was 1.2%, a risk similar to that for women with a previous child with an NTD (Centers for Disease Control and Prevention [CDC], 1982). The report was quickly confirmed in other areas of the world through the efforts of the International Clearinghouse of Birth Defect Registries (CDC, 1983). It was fortunate that several events came together to allow the determination of valproate as a human teratogen. These included the active birth defects registry, an interest by Robert in the genetics of spina bifida, a question on epilepsy and anticonvulsant use in a Robert survey, and the prevalence of valproate monotherapy for epilepsy in that region (Lammer et al., 1987). These findings spurred a great deal of research on the effects of valproate in multiple species, including the effects of enantiomers of valproate analogs. Yet, the mechanism of action, as for most developmental toxicants, remains elusive (Nau et al., 1991; Ehlers et al., 1992a,b; Hauck and Nau, 1992). Use of inbred mouse strains differing in their sensitivity to valproate-induced teratogenesis has revealed several candidate genes conferring sensitivity (Finnell et al., 1997, 2000; Craig et al., 2000; Bennett et al., 2000). Similar to phenytoin, some studies have indicated a possible role of interference with folate metabolism (Hansen, 2010). Valproic acid is also a histone deacetylase inhibitor, and this may play a role in its teratogenicity (Menegola et al., 2006; Eikel et al., 2006).

Angiotensin-Converting Enzyme Inhibitors and Angiotensin Receptor Antagonists

The renin–angiotensin system is a key controller of blood pressure. The active signaling messenger of this system is angiotensin II, which binds to angiotensin II (AT1) receptors to cause vasoconstriction and fluid retention, resulting in elevation of blood pressure. Antihypertensive agents acting on this system include the angiotensin-converting enzyme (ACE) inhibitors and the AT1 receptor blockers (collectively known as “sartans”). ACE inhibitors are widely prescribed and, when used in the second half of pregnancy, are known to cause oligohydramnios (low amniotic fluid volume), fetal growth retardation, pulmonary hypoplasia, joint contractures, hypocalvaria, neonatal renal failure, hypotension, and death. Angiotensin II receptor antagonists taken during the same period (second or third trimester) in pregnancy cause a similar spectrum of developmental toxicity (Friedman, 2006; Quan, 2006). These findings in the fetus can be related to the reduced amniotic fluid volume, a consequence of impaired fetal renal function (Tabacova, 2005). The developmental toxicity of first trimester use of ACE inhibitors and AT1 receptor antagonists has been somewhat controversial (Schaefer, 2003), although a recent study found an increased risk of major congenital malformations following first trimester exposure to ACE inhibitors (Cooper et al., 2006). These authors studied a cohort of 29,507 infants born between 1985 and 2000. Of these, 209 infants had exposure to ACE inhibitors in the first trimester alone. These infants had an increased risk of major congenital malformations (relative risk = 2.71, 95% confidence interval 1.72–4.27). The risk ratio for malformations of the cardiovascular system and the central nervous system were 3.72 and 4.39, respectively. These authors concluded that exposure to ACE inhibitors during the first trimester should be avoided. However, a recent literature review and meta-analysis of studies of first trimester exposures to ACE inhibitors and AT1 blockers found that while there was a significant increase in risk of major malformations compared to healthy controls, the risk was similar to that of patients on other antihypertensives (Walfisch et al., 2011). These results suggest that a first trimester exposure to antihypertensives in general may be teratogenic.

Principles of teratology were put forth by James Wilson in 1959 and in his watershed monograph Environment and Birth Defects (Wilson, 1973) (Table 10-2). Much progress has been made in the ensuing decades, yet these principles have withstood the test of time and remain basic to developmental toxicology.

Critical Periods of Susceptibility and Endpoints of Toxicity

Familiarity with principles of normal development is a prerequisite to understanding abnormal development. Development is characterized by change: change in size, changes in biochemistry and physiology, and changes in form and functionality. These changes are orchestrated by a cascade of factors regulating gene expression, the first of which are maternally inherited and present in the egg prior to fertilization. In turn, these maternal factors activate regulatory genes in the embryonic genome, and sequential gene activation continues throughout development. Intercellular and intracellular signaling pathways essential for normal development have been elucidated and rely on transcriptional, translational, and posttranslational controls.

Because of the rapid changes occurring during development, the nature of the embryo/fetus as a target for toxicity is also changing. Although the basic tenets of toxicology discussed elsewhere in this text also apply during development, the principle of critical periods of sensitivity based on developmental stage of the conceptus at the time of insult is a primary and somewhat unique consideration. In this section, normal developmental stages are discussed in the context of their known and potential susceptibility to toxicants. It should be emphasized, however, that development is a continuum; these stages are used for descriptive purposes and do not necessarily represent discrete developmental events. Timing of some key developmental events in humans and experimental animal species is presented in Table 10-3.

As a logical starting point, gametogenesis is the process of forming the haploid germ cells or gametes, the egg and sperm. These gametes fuse in the process of fertilization to form the diploid zygote, or one-celled embryo. Gametogenesis and fertilization are vulnerable to toxicants, but this is the topic of another chapter in this text. It is now known that the maternal and paternal genomes are not equivalent in their contributions to the zygotic genome. The process of imprinting occurs during gametogenesis, conferring to certain allelic genes a differential expressivity depending on whether they are of maternal or paternal origin (Latham, 1999; Li and Sasaki, 2011). Because imprinting involves cytosine methylation and changes in chromatin conformation, this process may be susceptible to toxicants that affect these targets (Murphy and Jirtle, 2000). Jirtle and colleagues have demonstrated that both early nutrition (Waterland and Jirtle, 2003, 2004) and the estrogenic compound genistein (Dolinoy et al., 2006) can modify the methylation patterns of specific genes and cause altered phenotypes in mice. Anway et al. (2006) demonstrated a transgenerational effect of the fungicide vinclozolin in rats, in which male offspring (F1) of treated F0 females had reduced spermatogenic capacity, which was transmitted to subsequent generations (F2–F4). For a review of transgenerational effects of environmental agents, the reader is referred to Skinner et al. (2011).

Epigenetics refers to the biochemical changes in chromatin that lead to changes in conformation and gene expression. Epigenetic changes include DNA methylation, histone modifications, and expression of microRNAs. Although epigenetics provides the basis for imprinting of genes, the role of epigenetics is much broader (Rogers et al., 2010; Grace et al., 2011). There are specific stages of the life cycle during which epigenetic marks may be erased and reestablished (Sasaki and Matsui, 2008), including two periods of development during which large-scale demethylations of the genome are known to occur. One is during migration and proliferation of the primordial gem cells (between embryonic days 10.5 and 12.5 in the mouse). It is during this period that imprinted genes are demethylated, with remethylation occurring in a gender-specific manner during gametogenesis in the offspring. The other period of widespread epigenetic reprogramming occurs shortly after formation of the zygote and in the early embryo, with total genomic methylation reaching a nadir at the blastocyst stage (Hales et al., 2011). This process of epigenetic reprogramming represents a putative target for environmental chemicals, and examples of environmentally mediated epigenetic changes are emerging (Skinner et al., 2011; Rogers et al., 2010; Perera and Herbstman, 2011).

Following fertilization, the embryo moves down the fallopian tube (oviduct) and attaches to the wall of the uterus. The preimplantation period comprises mainly an increase in cell number through a rapid series of cell divisions with little growth in size (cleavage of the zygote) and cavitation of the embryo to form a fluid-filled blastocoele. This stage, termed the blastocyst, consisting of about a thousand cells, may contain as few as three cells destined to give rise to the embryo proper (Markert and Petters, 1978), and these cells are within a region called the inner cell mass. The remainder of the blastocyst cells gives rise to extraembryonic membranes and support structures (eg, trophoblast and placenta). However, the fates of the cells in the early embryo are not completely determined at this stage. The relatively undifferentiated preimplantation embryo has great restorative (regulative) growth potential (Snow and Tam, 1979). For example, experiments by Moore et al. (1968) demonstrated that single cells from eight-celled rabbit embryos are capable of producing normal offspring.

Toxicity during the preimplantation period is generally thought to result in either no or slight effect on growth (because of regulative growth) or in death (through overwhelming damage or failure to implant). Preimplantation exposure to dichlorodiphenyltrichloroethane (DDT), nicotine, or methyl-methane sulfonate results in body and/or brain weight deficits and embryo lethality, but not malformations (Fabro, 1973; Fabro et al., 1984). However, there are a few examples of toxic exposures during the preimplantation period leading to fetal malformations. Exposure to toxicants during a brief period (approximately six hours) immediately following fertilization has been demonstrated to result in malformed fetuses for a few chemicals including ethylene oxide (Generoso et al., 1987), ethylmethane sulfonate, ethylnitrosourea, and triethylene melamine (Generoso et al., 1988). The mechanisms underlying these unexpected findings have not been elucidated but probably do not involve point mutations. Epigenetic changes were not examined, but given the time of the exposure, this is a plausible target. Treatment of pregnant mice with methylnitrosourea on days 2.5, 3.5, and 4.5 of gestation resulted in NTDs and cleft palate (Takeuchi, 1984). Cyproterone acetate and medroxyprogesterone acetate are capable of producing malformations when administered on day two of gestation (Eibs et al., 1982). Rutledge and co-workers (Rutledge et al., 1994) produced hind-limb and lower body duplications by treating pregnant mice with all-trans RA on gestation day 4.5 to 5.5, at which time the embryos are at the late blastocyst and proamniotic stages. Because of the rapid mitoses occurring during the preimplantation period, chemicals affecting DNA synthesis or integrity, or those affecting microtubule assembly, would be expected to be particularly toxic.

Following implantation, the embryo undergoes gastrulation. Gastrulation is the process of formation of the three primary germ layers—the ectoderm, mesoderm, and endoderm. During gastrulation, cells migrate through a midline structure called the primitive streak, and their movements set up basic morphogenetic fields in the embryo (Smith et al., 1994). A prelude to organogenesis, the period of gastrulation is highly susceptible to teratogenesis. It is common for teratogens administered during gastrulation to produce malformations of the eye, brain, and face. These malformations are indicative of damage to the anterior neural plate, one of the regions defined by the cellular movements of gastrulation.

The formation of the neural plate in the ectoderm marks the onset of organogenesis, during which the rudiments of most bodily structures are established. This is a period of heightened susceptibility to malformations and extends from approximately the third to the eighth weeks of gestation in humans. Within this short period, the embryo undergoes rapid and dramatic changes. At three weeks of gestation, the human conceptus is in most ways indistinguishable from other mammalian and indeed other vertebrate embryos, consisting of only a few cell types in a trilaminar arrangement. By eight weeks, the conceptus, which can now be termed a fetus, has a form clearly recognizable as human. The rapid changes during organogenesis require cell proliferation and migration, cell–cell interactions, and morphogenetic tissue remodeling. These processes can be exemplified by the ontogeny of the neural crest cells. These cells originate at the border of the neural plate and migrate to form a wide variety of structures throughout the embryo. For example, neural crest cells derived from segments of the hindbrain (rhombomeres) migrate to form bone and connective tissues in the head (Krumlauf, 1993; Vaglia and Hall, 1999; Trainor and Krumlauf, 2001), while other neural crest cells form diverse structures including melanocytes, enteric ganglia, and adrenal medulla, and contribute to the cardiac outflow tract.

Within organogenesis, there are periods of peak susceptibility for each forming structure. This is nicely illustrated by the work of Shenefelt (1972), who studied the malformations caused by carefully timed doses of RA in pregnant hamsters. The incidences of some of the defects seen after RA administration at different times in pregnancy are shown in Fig. 10-2. The peak incidence of each malformation coincides with the timing of key developmental events in the affected structure. Thus, the specification of developmental fields for the eyes is established quite early, and microphthalmia has an early critical period. Establishment of rudiments of the long bones of the limbs occurs later, as does susceptibility to shortened limbs. The palate has two separate peaks of susceptibility, the first corresponding to the early establishment of the palatal folds and the second to the later events leading to palatal closure. Notice also that the total incidence of malformations is lower prior to organogenesis but increases to 100% by gestation day 7¾. The processes underlying the development of normal structures are incompletely understood but involve a number of key events. A given toxicant may affect one or several developmental events, so the pattern of sensitivity of a structure can change depending on the nature of the toxic insult. Cleft palate can be induced in mouse fetuses following maternal exposure to methanol as early as day five of gestation, with a peak sensitivity at day seven and little or no sensitivity after day nine (Rogers and Mole, 1997). In contrast, the typical peak critical period for induction of cleft palate in the mouse for most teratogens is between gestation days 11 and 13, during the stages of growth, elevation, and fusion of the palatal shelves. In a large series of experiments in naval medical research institute (NMRI) mice, the day of peak sensitivity to the induction of cleft palate was day 11 for 2,3,7,8-tetrachlorodibenzodioxin (TCDD), day 12 for 2,4,5-trichlorophenoxyacetic acid, and day 13 for dexamethasone (Neubert et al., 1973). Detection of unexpected critical periods like the early one for induction of cleft palate by methanol can provide clues to normal developmental processes not presently understood.

The end of organogenesis marks the beginning of the fetal period (from day ~57 to birth in humans), characterized primarily by tissue differentiation, growth, and physiological maturation. This is not to say that formation of the organs is complete, but rather that almost all organs are present and grossly recognizable. Further development of organs proceeds during the fetal period to attain requisite functionality prior to birth, including fine structural morphogenesis (eg, neural outgrowth and synaptogenesis, branching morphogenesis of the bronchial tree and renal cortical tubules) as well as biochemical maturation (eg, induction of tissue-specific enzymes and structural proteins). One of the latest organogenetic events is closure of the urethral groove in the male, which occurs at about gestation day 90. Failure of this event produces hypospadias, a ventral clefting of the penis.

Exposure during the fetal period is most likely to result in effects on growth and functional maturation. Functional anomalies of the central nervous system and reproductive organs—including behavioral, cognitive, and motor deficits as well as decreases in fertility—are among the possible adverse outcomes. These manifestations are not apparent prenatally and may require careful postnatal observation and testing. Such postnatal functional manifestations can be sensitive indicators of in utero toxicity, and reviews of postnatal functional deficits of the central nervous system (Rodier et al., 1994; Miodovnik, 2011), immune system (Holladay and Luster, 1994; Dietert, 2008), and heart, lung, and kidneys (Lau and Kavlock, 1994) are available. Major structural alterations can occur during the fetal period, but these generally result from deformations (disruption of previously normal structures) rather than malformations. The extremities may be affected by amniotic bands, wrapping of the umbilical cord, or vascular disruptions, leading to loss of distal structures. Over the past two decades, the concept of “developmental programming” has emerged, in which the developmental environment is thought to influence the metabolic parameters of the offspring that will persist throughout life and may affect lifelong risk of disease (for review, see McMillen and Robinson, 2005; Gluckman et al., 2011; Barouki et al., 2012). Much of the work on fetal programming has focused on the role of maternal nutrition, and there is a paucity of data concerning the long-term effects of chemical exposure during the fetal and early postnatal periods (Lau and Rogers, 2004). Some effects could require years to become apparent (such as those noted above for DES), and others may even result in the premature onset of senescence and/or organ failure late in life. In rats, prenatal exposure to high dosages of ethanol during the second half of pregnancy shortens life span of the offspring, by about 20 weeks in females and 2.5 to seven weeks in males (Abel et al., 1987).

Dose–Response Patterns and the Threshold Concept

The major effects of prenatal exposure, observed in fetuses near term in developmental toxicity studies, are growth retardation, malformations, and embryo death. The relationships among these effects are complex and vary with the chemical, the time of exposure, and the dose. For some chemicals these endpoints may represent a continuum of increasing toxicity, with low dosages producing growth retardation and increasing dosages producing malformations and then death. Malformations and/or death can occur without a concomitant effect on intrauterine growth, but this is unusual. Likewise, growth retardation and embryo lethality can occur without malformations. Chemicals producing the latter pattern of response would be considered embryotoxic or embryolethal but not teratogenic (unless it was established that death was due to a structural malformation). Effects on function in viable fetuses will usually require postnatal assessment. Transition to postnatal life is a rigorous test in itself, and severe functional effects may manifest as neonatal mortality.

Another key element of the dose–response relationship is the shape of the dose–response curve at low exposure levels. Because of the high restorative growth potential of the mammalian embryo, cellular homeostatic mechanisms, and maternal metabolic defenses, mammalian developmental toxicity has generally been considered to be a threshold phenomenon. Assumption of a threshold means that there is a maternal dosage below which no increase in an adverse outcome is elicited. Daston (1993) summarized two approaches for establishing the existence of a threshold. The first, exemplified by a large teratology study on 2,4,5-T (Nelson and Holson, 1978), suggests that no study is adequately capable of evaluating the dose–response at low response rates (eg, in this study it was calculated that 805 litters per dose would be necessary to detect a 5% increase in resorptions of embryos). The second approach is to determine whether a threshold exists for the molecular mechanism responsible for the observed effect. Although relatively few mechanisms of abnormal development have been thoroughly studied, it is clear that cellular and embryonic repair mechanisms and dose-dependent kinetics both support the plausibility of a mechanistic threshold. Lack of a threshold implies that exposure to any amount of a toxic chemical, even one molecule, has the potential to cause developmental toxicity. Hypothetically, a point mutation in a critical gene could be induced by a single hit or single molecule, leading to a deleterious change in a gene product and consequent abnormal development. This, of course, carries the large assumption that the molecule could traverse the maternal system and the placenta and enter a critical progenitor cell in the embryo. An effect on a single cell might result in abnormal development at the zygote (one-cell) stage, the blastocyst stage (when only a few cells in the inner cell mass are embryo progenitors), or during organogenesis, when organ rudiments may consist of only a few cells.

An apparent threshold for developmental toxicity based at least in part on cellular homeostatic mechanisms is demonstrated in studies of mechanisms underlying the teratogenicity of 5-fluorouracil (Shuey et al., 1994; see also “Safety Assessment” later in this chapter). This chemical inhibits the enzyme thymidylate synthetase (TS), thus interfering with DNA synthesis and cell proliferation. Significant embryonal TS inhibition is evident at maternal dosages an order of magnitude lower than those required to produce malformations and about fivefold below those affecting fetal growth (Fig. 10-3). The lack of developmental toxicity despite significant TS inhibition probably reflects an ability of the embryo to compensate for imbalances in cellular nucleotide pool sizes.

For human health risk assessment, it is also important to consider the distinction between individual thresholds and population thresholds. There is wide variability in the human population, and a threshold for a population can be defined as the threshold of the most sensitive individual in the population (Gaylor et al., 1988). Indeed, even though the biological target of a developmental toxicant may exhibit a threshold, background factors such as health status or concomitant exposures may render some individuals already at or even beyond the threshold for failure of that biological process. Any further toxic impact on that process, even one molecule, would theoretically increase risk. The concept of thresholds of adversity for reproductive toxicants has recently been reviewed by Piersma et al. (2011a).

The term mechanisms is used here to refer to cellular events that initiate a process leading to abnormal development. Pathogenesis comprises the cell-, tissue-, and organ-level sequelae that ultimately lead to abnormality. Mechanisms of teratogenesis listed by Wilson (1977) include mutations, chromosomal breaks, altered mitosis, altered nucleic acid integrity or function, diminished supplies of precursors or substrates, decreased energy supplies, altered membrane characteristics, osmolar imbalance, and enzyme inhibition. Although these cellular insults are not unique to development, they may relatively quickly trigger unique pathogenic responses in the embryo, such as reduced cell proliferation, cell death, altered cell–cell interactions, reduced biosynthesis, inhibition of morphogenetic movements, or mechanical disruption of developing structures.

Elucidating the mechanism(s) and pathogenesis of chemically induced developmental toxicity can be extremely challenging. Cyclophosphamide (CP), a chemotherapeutic and immunosuppressive drug, has been studied for its teratogenicity for 50 years, and is a good case study (Mirkes, 1985a; Ozolins, 2010). Because of its use as an immunosuppressive drug to treat lupus, at least nine cases of teratogenesis induced by inadvertent exposure to CP during early pregnancy have been recorded (Paladini et al., 2004). Efforts to understand the teratogenicity of this drug provide an example of approaches to understanding metabolic activation, teratogenic mechanisms, and pathogenesis. Much of this and other mechanistic work was made possible by the advent of rodent whole embryo culture (WEC), which involves removing rodent or rabbit embryos from the uterus at the beginning of organogenesis and growing them in serum-containing culture media (New, 1978; Sadler and Warner, 1984; Ellis-Hutchings and Carney, 2010; Robinson et al., 2012a,b,c,d; Lee et al., 2012). Embryos will grow normally for about 48 hours, completing most of organogenesis. The ability to grow embryos in isolation allows direct exposure, manipulation, and observation. The use of rodent WEC as a screening system for developmental toxicity will be discussed later in this chapter.

Using WEC, Fantel et al. (1979) and Sanyal et al. (1979) showed that hepatic S9 fractions and cofactors were required to elicit abnormal development by CP, demonstrating that it must be metabolically activated to be teratogenic. Activation of CP was inhibited by metyrapone or carbon monoxide, indicating involvement of P450 monooxygenases. Of the CP metabolites, 4-hydroxycyclophosphamide (4OHCP) and aldophosphamide (AP) are unstable. A stable derivative of 4OHCP, 4-hydroperoxy-cyclophosphamide (4OOHCP) was tested in vivo (Hales, 1982) and in WEC (Mirkes, 1987). In the latter study, the morphology of the treated embryos was indistinguishable from that of embryos cultured with CP plus an activating system. Spontaneous conversion of 4OOHCP to 4OHCP and then to phosphoramide mustard (PM) and acrolein (AC) occurs rapidly, and these further metabolites, as well as 4-ketocyclophosphamide (4-ketoCP) and carboxyphosphamide (CaP), have also been studied for their teratogenicity. It appears that 4OHCP is not teratogenic (Hales, 1983) and toxicity elicited by 4-ketoCP is dissimilar to that of activated CP (Mirkes et al., 1981). Subsequent work centered on the two remaining metabolites, PM and AC. Mirkes et al. (1981) demonstrated that the effects of PM on cultured rat embryos were indistinguishable from those of activated CP. Hales (1982) administered CP, PM, or AC to gestation day 13 rat embryos by intra-amniotic injection. CP and AC caused hydrocephaly, open eyes, cleft palate, micrognathia, omphalocele, and tail and limb defects, whereas PM produced only hydrocephaly and tail and limb defects. Thus, both PM and AC appear to be teratogenic metabolites of CP.

What are the cell and molecular targets of activated CP, and what is the nature of the interactions? Experiments with (³H)CP showed that approximately 87% of bound radioactivity was associated with protein, 5% with DNA, and 8% with RNA (Mirkes, 1985b). Using alkaline elution, it was demonstrated that CP and PM produce single-strand DNA breaks and DNA–DNA and DNA–protein cross-linking. To determine whether DNA cross-linking is essential for teratogenicity, a monofunctional derivative of PM, capable of producing single-strand breaks but not cross-links in DNA, was tested. Although higher concentrations were needed, this derivative produced the same spectrum of effects as PM (Mirkes et al., 1985). Later, Little and Mirkes (1990) showed that 4-hydroperoxydechlorocyclophosphamide, a CP analog that yields AC and a nonalkylating derivative of PM, did not produce DNA damage when embryos were exposed in serum-containing medium, but this analog retained teratogenicity. Using radiolabeled CP, they further found that AC preferentially binds to protein and shows high incorporation into the yolk sac, whereas PM binds preferentially to DNA. Hales (1989) showed that PM and AC have strikingly different effects on limb buds in culture. These results indicate that PM and AC have different targets in the embryo and that PM is responsible for CP-induced DNA damage.

How do chemical insults at the cell and molecular level translate to a birth defect? To illustrate pathogenesis, we will consider cell cycle perturbations and cell death, continuing our example of CP. Cell death plays a critical role in normal morphogenesis. The term programmed cell death refers to a specific type of cell death, apoptosis, under genetic control in the embryo (Lavin and Watters, 1993). Apoptosis is necessary for sculpting the digits from the hand plate, for instance (Hernandez-Martinez and Covarrubias, 2011), and for assuring appropriate functional connections between the central nervous system and distal structures. Cell proliferation is obviously essential for development. Cells within the primitive streak of the gastrula-stage rat embryo have the shortest known cell cycle time of any mammalian cell, three to 3.5 hours (MacAuley et al., 1993). Cell proliferation rates change both spatially and temporally during ontogenesis, as can be demonstrated by examining the proportion of cells in S-phase over time in various tissues during mid-to late gestation (Fig. 10-4). There is a delicate balance between cell proliferation, cell differentiation, and apoptosis in the embryo, and one molecular mechanism discussed above, DNA damage, might lead to the cell cycle perturbations and cell death induced in specific cell populations by CP.

Maternal CP treatment on gestation day 10 in the rat causes an S-phase cell cycle block as well as widespread cell death in the embryo (Fig. 10-5). In agreement with the S-phase cell cycle block, cell death is observed in areas of rapid cell proliferation (Chernoff et al., 1989; Francis et al., 1990). Similar embryonal cell cycle blockage and cell death were observed using activated CP in WEC (Little and Mirkes, 1992). The embryonal neuroepithelium is quite sensitive to CP-induced cell death, whereas the heart is resistant; differences in cell cycle length may, in part, underlie this differential sensitivity. The neuroepithelium of the day 10 rat embryo has a cell cycle time of approximately 9.5 hours, whereas the cell cycle length in the heart was estimated to be 13.4 hours. This difference is due to a longer G0/G1 phase in the heart cells compared to the neuroepithelium (Mirkes et al., 1989). Damage to DNA by PM occurs predominately during S-phase (Little and Mirkes, 1992), which constitutes a relatively smaller proportion of the cell cycle in the heart than in the neuroepithelium. Damage to DNA can inhibit cell cycle progression at the G1/S transition, through the S phase, and at the G2/M transition. If DNA damage is repaired, the cell cycle can return to normal, but if damage is too extensive or cell cycle arrest too long, apoptosis may be triggered. An increasing number of genes are being identified that play a role in apoptosis (White, 1993; Jurisicova and Acton, 2004). The p53 gene, which functions as a tumor suppressor, can promote apoptosis or growth arrest. Apoptosis occurring during normal development does not require this gene, as p53-deficient embryos develop normally. However, p53 may be critical for induction of growth arrest or apoptosis in response to DNA damage. The incidence of benzo[a]pyrene-induced resorptions were increased 2.6-fold and 3.6-fold, respectively, in heterozygous and homozygous p53-deficient embryos compared to normal homozygous controls (Nicol et al., 1995). Growth factors and some cytokines (eg, IL-3, IL-6) can prevent p53-dependent apoptosis. Expression of c-myc produces continued DNA synthesis, which may precipitate apoptosis in the face of DNA damage. Bcl-2 functions as a repressor of apoptosis and functions in conjunction with Bax, a homolog that dimerizes with itself or with Bcl-2. Bax homodimers favor cell death whereas Bcl-2/Bax heterodimers inhibit cell death (Oltvai and Korsmeyer, 1994; Boumela et al., 2011).

A common approach to test a hypothesized mechanism of teratogenesis is to administer chemicals expected to ameliorate or exacerbate the effects based on the putative mechanism of action. Several investigators have taken this approach in studying CP. Because bioactivation of CP was hypothesized to be necessary for its teratogenicity, Gomes-Carneiro et al. (2003) treated rats with β-ionone, a known inhibitor of CYP2B1, the enzyme that activates CP, four minutes prior to CP injection. Severity and incidence of teratogenicity was reduced, supporting the hypothesis. Conversely, licorice extract, an upregulator of CYP2B1 expression, exacerbated the teratogenicity of CP (Park et al., 2011). Ginseng extract inhibits the CP-detoxifying enzyme CYP3A, and treatment with this extract also enhances the teratogenicity of CP. One hypothesis for the downstream teratogenic mechanism of the active CP metabolite(s) involves alkylation of cellular macromolecules. Thiol-containing molecules protect against alkylating agents, and the thiol molecules cysteine, glutathione, and Na-2-mercaptoethane sulfonate have been shown to ameliorate CP teratogenicity (Ashby et al., 1976; Hales, 1981; Slott and Hales, 1986). However, these studies did not measure the degree of CP alkylation. Treatment with the antioxidant ascorbic acid attenuates CP-induced embryonic death and chromosomal damage (Kola et al., 1989), whereas another antioxidant, butylated hydroxyanisole, protects against CP teratogenicity in rats (Kang et al., 2003) and mice (Kim et al., 2003).

Advances in the Molecular Basis of Dysmorphogenesis

Rapid advances in molecular biology, genomics, proteomics, and bioinformatics have brought new understanding of mechanisms of normal and abnormal development. A key approach, targeted gene disruption by homologous recombination (gene “knockout”; KO), has been used to study the function of many gene products in developing and adult animals. An example is the retinoic acid receptor (RAR) family of nuclear ligand-inducible transcription factors that bind RA and 9-cis RA. Chambon and colleagues have produced mice lacking several of these receptors either singly or as multiple KOs. Single-receptor isoform mutants were often unaffected, suggesting functional redundancy, whereas double mutants presented with widespread malformations of the skeleton and viscera (Lohnes et al., 1994; Mendelsohn et al., 1994). The compound RARγ–RARβ null mouse exhibits syndactyly, indicating that RA plays a role in interdigital cell death (Dupe et al., 1999). The rexinoid receptors (RXRα, β, and γ) bind 9-cis RA exclusively, and it is now understood that RAR/RXR heterodimers are involved in RA signaling via transcriptional control in the embryo (Mark et al., 2006, 2009). A wide variety of spontaneous and genetically altered mouse strains are commercially available, and expanding compendia of mutants of other animals including zebrafish and frogs are available.

Synthetic antisense oligonucleotides have been used to ablate gene products in a temporally and spatially restricted manner. In this technique, 15 to 25 mer oligonucleotides are synthesized that are complimentary to the mRNA to be disrupted (Helene et al., 1990). These probes can enter embryonal cells, and hybridization with cellular mRNA causes disruption of native message, such that gene function can be turned off at specific times. The proto-oncogenes Wnt-1 and Wnt-3a have been implicated in the development of the midbrain and hindbrain. Augustine et al. (1993) attenuated Wnt-1 expression using antisense oligonucleotide inhibition in mouse embryos developing in culture. Exposure during neurulation produced mid- and hindbrain malformations similar to those seen in Wnt-1 null mutant mice, as well as cardiac anomalies not observed in Wnt-1 KOs created by homologous recombination. Antisense attenuation of Wnt-3a caused anomalies of the forebrain, midbrain, and spinal cord. Simultaneously attenuating both Wnt-1 and Wnt-3a targeted all brain regions and worsened the effect on the spinal cord, suggesting that these genes may serve a complementary function in the development of the central nervous system. Morpholino antisense oligonucleotides contain nucleic acid analogs that resist breakdown, and morpholinos have been used extensively in the study of zebrafish biology (Bedell et al., 2011). RNA interference is a more recent gene knockdown technique, exploiting the discovery of the RNA interference pathway. Small interference (Si)RNA, plasmid-, and virus-encoded small RNAs can be used to down regulate the expression of specific genes posttranscriptionally (Shan, 2010).

Gain of gene function can also be studied by engineering genetic constructs with an inducible promoter attached to the gene of interest. Ectopic gene expression can be made ubiquitous or site-specific depending on the choice of promoter to drive expression. Ectopic expression of the Hoxa-7 gene induced in mouse embryos by attaching it to the chicken β-actin promoter resulted in a phenotype exhibiting multiple craniofacial and cervical vertebral malformations (Balling et al., 1989; Kessel et al., 1990). Transient overexpression of specific genes can be accomplished by adding extra copies using adenoviral transduction. In proof-of-concept, Hartig and Hunter (1998) injected an adenoviral vector containing either the bacterial β-galactosidase or green fluorescent protein reporter gene under the control of the human cytomegalovirus early gene promoter into the intra-amniotic space of neurulation-stage mouse embryos and achieved intense gene expression in the neuroepithelium.

Advances in gene targeting and transgenic strategies now allow modification of gene expression at specific points in development and in specific cell types. Conditional KOs or knockins, inducible gene expression and other techniques are being used to study the effects of specific gene products on development in great detail (Mikkola and Orkin, 2005).

Reporter transgenes contain a gene with a readily detectable product fused downstream of a selected regulatory region. The Escherichia coli lacZ (β-galactosidase) gene is commonly used for this purpose. Cell lineage studies can be carried out by fusing lacZ to a constitutive regulatory sequence and introducing the construct into a somatic cell early in ontogenesis. The reporter gene will then be expressed in and mark all progeny of the transfected cell. This method has been used to study postimplantation development in the mouse embryo (Sanes et al., 1986), although intracellular injection of fluorescent dyes has also proven highly reliable for cell lineage studies (eg, Smith et al., 1994). The pattern of expression of a particular gene of interest can be discriminated by fusing upstream regulatory elements of the gene to lacZ, which will then be transcribed under control of those upstream elements (Zakany et al., 1990). RA can activate hox genes in vitro, and some hox genes have multiple RA response elements (RAREs). Evidence that RA-induced malformations in mouse embryos are related to changes in hox expression was first provided by staining of hoxlacZ transgenic embryos (Marshall et al., 1992). Within a few hours of RA treatment, hoxb-1 expression extended anteriorly, suggesting that hox genes could be direct targets of RA induction. Regions of altered hox expression could be manifest as abnormal cell fate and morphogenesis (Marshall et al., 1996; Collins and Mao, 1999).

Reporter genes have recently been used to great advantage to study gene expression in zebrafish (eg, Saydmohammed et al., 2011).

The use of modern biological techniques to elucidate a mechanism of developmental toxicity is exemplified by work on the perfluorinated alkyl acids (PFAAs), perfluorooctanoic acid (PFOA), and perfluorooctane sulfonate (PFOS). The PFAAs are a family of industrial chemicals that has been widely used as surfactants, hydrophobic and oleophobic treatments, and nonstick coatings. These chemicals are highly resistant to environmental degradation and are therefore persistent in the environment. Studies in experimental animals have demonstrated that developmental toxicity is among the most sensitive toxic effects of these chemicals (Lau et al., 2004). Treatment of pregnant mice with either PFOS or PFOA resulted in dose-related increases in mortality of offspring in the first few days of life. Although the toxic endpoint is similar for PFOA and PFOS, their mechanisms of action may be different. PFAAs have varied affinity for the peroxisome proliferator-activated receptor alpha (PPARα), and activation of PPARα is a putative mechanism of action for the developmental toxicity of these chemicals. To explore the role of PPARα activation in the developmental toxicity of PFOA and PFOS, Abbott and colleagues tested the developmental toxicity of these chemicals in PPARα null mice. Exposure of pregnant wild-type and PPARα KO mice to PFOA revealed that neonatal mortality, the signature effect of developmental exposure to PFOA, was dependent on PPARα expression. The survival of PPAR null offspring was not affected by maternal PFOA exposure, whereas survival of wild-type offspring was significantly reduced (Abbott et al., 2007). Somewhat surprisingly, when the same experiments were carried out with PFOS exposure, the PPARα null offsprings were not protected from neonatal mortality (Abbott et al., 2009). Studies were undertaken to compare the ability of PFAAs to activate mouse and human PPARα in COS-1 cells (a commonly used cell line derived from kidney cells of the African green monkey) transiently transfected with either a mouse or human PPARα-luciferase reporter plasmid, such that activation of the receptor caused light emission that could be quantified on a luminometer (Wolf et al., 2008). In agreement with the KO mouse studies, PFOA activated both the mouse and human PPARα at lower concentrations than did PFOS. Further evidence that PPARα activation was required for PFOA but not PFOS developmental toxicity came from studies of gene expression changes induced in fetal tissues by maternal treatment with PFOA or PFOS (Rosen et al., 2007, 2009). Although both PFOA and PFOS activated PPARα downstream genes, PFOA did so in a more consistent and robust fashion. This could also be seen in livers from adult mice treated with PFOA or PFOS. Although PPARα KO mice did not upregulate certain PPARα–related genes upon PFOA or PFOS treatment, the expression of other genes was changed by exposure to either of these PFAAs, indicating that the influence of either PFOA or PFOS may extend beyond activation of PPARα (Rosen et al., 2010). This elegant set of experiments on the developmental toxicity of the PFAAs is illustrated in Fig. 10-6. It remains to be determined which of the pathways downstream of PPARα are involved in the developmental toxicity of PFOA, and which non-PPARα–related pathways might be perturbed by PFOS.

The manner in which chemicals are absorbed during pregnancy and the extent to and form in which they reach the conceptus are important determinants of whether the chemical can impact development. The maternal, placental, and embryonic compartments are distinct but interacting systems that undergo profound changes during the course of pregnancy. Maternal adaptations to pregnancy include changes in hepatic metabolism, the gastrointestinal tract, cardiovascular system, excretory system, and the respiratory system (Hytten, 1984; Krauer, 1987; Mattison et al., 1991). Although these changes are necessary to support the growing needs of the conceptus in terms of energy supply and waste elimination, they can have significant impact on the uptake, distribution, metabolism, and elimination of xenobiotics. For example, decreases in intestinal motility and increases in gastric emptying time result in longer retention of ingested chemicals in the upper gastrointestinal tract. Cardiac output in humans increases by 30% to 40% by the 27th week of pregnancy, whereas blood volume increases and plasma proteins and peripheral vascular resistance decrease. The relative increase in blood volume over red cell volume leads to borderline anemia and a generalized edema with a 70% elevation of extracellular space. Thus, the volume of distribution of a chemical and the amount bound by plasma proteins may change considerably during pregnancy. Renal blood flow and glomerular filtration rate also increase during pregnancy. Respiratory changes including increases in tidal volume, minute ventilation, and minute O₂ uptake can result in increased pulmonary distribution of gases and decreased time to reach alveolar steady state.

In addition to changes in maternal physiology, limited evidence suggests that relative rates of drug metabolizing enzymes also change during pregnancy (Juchau, 1981; Juchau and Faustman-Watts, 1983). Maternal hepatic P450 enzymes increase during pregnancy in humans, whereas decreased hepatic monooxygenase activity has been observed during pregnancy in rats, attributed to decreased enzyme levels and competitive inhibition by circulating steroids (Neims, 1976). Another factor that contributes to lower monooxygenase activity may be changes in inducibility—pregnant rats appear to be less responsive to induction of hepatic monooxygenases by phenobarbital (but not 3-methylcholanthrene) than are nonpregnant females (Guenther and Mannering, 1977). Maternal handling of a chemical bears considerable weight in determining the extent of embryotoxicity. In one of the few studies of its type, Kimmel and Young (1983) found that a linear combination of the 45-minute and the 24-hour maternal blood concentrations was able to predict the litter response rate for pregnant rats dosed with 500 mg/kg sodium salicylate on gestation day 11. These two kinetic parameters probably reflect the influence of peak drug concentration as well as cumulative area under the concentration–time curve on developmental toxicity.

The placenta plays a central role in influencing embryonic exposure by helping to regulate blood flow, by offering a transport barrier, and by metabolizing chemicals (Slikker and Miller, 1994). Functionally, the placenta acts as a lipid membrane that permits bidirectional transfer of substances between maternal and fetal compartments. The extent of transfer depends on three major elements: the type of placentation, the physicochemical properties of the chemical, and rates of placental metabolism. Although there are marked species differences in types of placentas, orientation of blood vessels, and numbers of exchanging layers (Carney et al., 2004), these do not seem to play a dominant role in placental transfer of most chemicals. It is important to note that virtually any substance present in the maternal plasma will cross the placenta to some extent. The passage of most drugs across the placenta seems to occur by simple passive diffusion, the rate of which is proportional to the diffusion constant of the drug, the concentration gradient between the maternal and embryonic plasma, the area of exchange, and the inverse of the membrane thickness (Nau, 1992). Important modifying factors to the rate and extent of transfer include lipid solubility, molecular weight, protein binding, the type of transfer (passive diffusion, facilitated or active transport), the degree of ionization, and placental metabolism. Weak acids appear to be rapidly transferred across the placenta during early pregnancy in rats and mice, due in part to the pH gradient between the maternal and embryonic plasma which can trap ionized forms of the drug in the slightly more basic embryonic compartment (Nau and Scott, 1986). In humans and rabbits, maternal serum is either similar to or slightly higher in pH than is the yolk sac fluid (Carney et al., 2005).

Quantifying the form, amount, and timing of chemical delivery to the embryonic compartment in the context of concurrent developmental processes is an important component of understanding mechanisms of embryotoxicity and species differences in embryonic sensitivity (Nau, 1986). The small size of the conceptus during organogenesis and the fact that the embryo is changing at a rapid rate during this period make assessment of toxicokinetics difficult. Nevertheless, there has been considerable progress in this area (Nau and Scott, 1987; Clark, 1993; Corley et al., 2003; Corley, 2010). Increasingly sensitive analytical methods are now providing evidence to challenge the historical view, particularly for cytochrome P450-dependent monooxygenases, that the early embryo has low metabolic capabilities (Juchau et al., 1992). Using the WEC system, Juchau and co-workers demonstrated that the rat conceptus was able to generate sufficient amounts of metabolites of the proteratogen 2-acetylaminofluorene (2-AAF) to induce dysmorphogenesis, and that the proximate toxicant, the 7-hydroxy metabolite, was different from the metabolite responsible for 2-AAF mutagenesis and carcinogenesis. Prior exposure of the dams to 3-methylcholanthrene increased the sensitivity of the cultured embryos to 2-AAF, thus demonstrating the inducibility of at least some cytochromes in the conceptus. These investigators later showed that embryos could further metabolize the 7-hydroxy metabolite to an even more toxic catechol. No previous induction was necessary for this activation step, demonstrating the presence of constitutive metabolizing enzymes in the embryo. Although the rates of metabolism for these activation steps may be low relative to the maternal liver, they occur close to the target site of the embryo or even within it, and thus are significant in terms of inducing embryotoxicity.

Physiologically based pharmacokinetic models provide the framework to integrate what is known about physiological changes during pregnancy, both within and between species, with aspects of drug metabolism and embryonic development into a quantitative description of the events. Olanoff and Anderson (1980) described the kinetics of the antibiotic tetracycline in the pregnant rat and her fetuses. Importantly, growth of the fetus and placenta was incorporated into the model. Gabrielson and co-workers (Gabrielson and Paalkow, 1983; Gabrielson and Larsson, 1990) were among the early investigators to develop physiologically based models of pregnancy, and others (Fisher et al., 1989; O’Flaherty et al., 1992; Clark et al., 1993; Luecke et al., 1994, 1997; Young, 1998) have added to their comprehensiveness. The Fisher model included growth of several maternal tissues during pregnancy. The model of O’Flaherty and co-workers describes the entire period of gestation, and includes the uterus, mammary tissue, maternal fat, kidney, liver, other well-perfused maternal tissues, embryo/fetal tissues and yolk sac, and chorioallantoic placenta. It takes into account the growth of various compartments during pregnancy (including the embryo itself), as well as changes in blood flow and the stage-dependent pH gradients between maternal and embryonic plasma. Transfer across the placenta is diffusion limited in this model. The utility of the model was evaluated using 5,5-dimethyloxazolidine-2,4-dione (DMO), a weak acid that is not appreciably bound to plasma proteins and is eliminated by excretion in the urine. The model demonstrated that the whole body disposition of DMO, including distribution to the embryo, can be accounted for solely on the basis of its pK_a and of the pH and volumes of body fluid spaces. Differences between the disposition of DMO by the pregnant mouse and rat are consistent simply with differences in fluid pH. Luecke and co-workers developed quantitative models for human embryofetal growth for use in models of human pregnancy (Luecke et al., 1994, 1997, 1999).

Chemical-specific PBPK models for pregnancy can range from simple to complex, with the degree of complexity often limited by lack of knowledge about mechanism of action or limited biological or chemical-specific data. One example is the solvent 2-methoxyethanol found to be embryotoxic and teratogenic in all species tested to date. The proximate teratogen appears to be the metabolite 2-methoxyacetic, acid (2-MAA). A physiologically based pharmacokinetic model has been developed for the pregnant mouse (Terry et al., 1995). Pharmacokinetics and tissue partition coefficients for 2-MAA were determined at different stages of embryonal development, and various models were tested based on the alternative hypotheses involving (1) blood flow limited delivery of 2-MAA to model compartments, (2) pH trapping of ionized 2-MAA within compartments, (3) active transport of 2-MAA into compartments, and (4) reversible binding of 2-MAA within compartments. Although the blood flow limited model best predicted gestation day eight dosimetry, the active transport models better described dosimetry on gestation days 11 and 13. Using published data on biotransformation of 2-methoxyethanol to ethylene glycol and 2-MAA in rats, Hays et al. (2000) adapted the pregnant mouse PBPK model to the pregnant rat and successfully predicted tissue levels of 2-MAA following oral or intravenous administration of 2-methoxyethanol. The next step was to extrapolate this model to the inhalation route of exposure, and to model both rats and humans (Gargas et al., 2000). The extrapolation of the model enabled predictions of the exposures needed for pregnant women to reach blood concentrations (C_max or area under the curve [AUC]) equivalent to those in pregnant rats exposed to the no observed adverse effect level (NOAEL) or lowest observed adverse effect level (LOAEL) for developmental toxicity. The body of work on PBPK modeling of 2-methoxyethanol is exemplary of the power of these techniques for extrapolating across dose, developmental stage, route, and species. For a review of chemical-specific PBPK models, see Corley et al. (2003) and Corley (2010).

Although all developmental toxicity must ultimately result from an insult to the conceptus at the cellular level, in mammals the insult may occur through a direct effect on the embryo/fetus, indirectly through toxicity of the agent to the mother and/or the placenta, or a combination of direct and indirect effects. Maternal physiological conditions capable of adversely affecting the developing organism include decreased uterine blood flow, maternal anemia, under or malnutrition, toxemia, altered organ function, autoimmune states, diabetes, electrolyte or acid–base disturbances, decreased milk quantity or quality, elevated stress hormones, and abnormal behavior (Chernoff et al., 1989; Daston, 1994). Induction or exacerbation of such maternal conditions by chemicals and the degree to which they manifest in abnormal development are dependent on maternal genetic background, age, parity, size, nutrition, disease, stress, and other health parameters and exposures (DeSesso, 1987; Chernoff et al., 1989; Rogers et al., 2005). These relationships are depicted in Fig. 10-7. In this section, we will discuss maternal conditions known to adversely affect the conceptus, as well as examples of xenobiotics whose developmental toxicity results completely or in large part from maternal or placental toxicity.

The distinction between direct and indirect developmental toxicity is important for interpreting results of standard developmental toxicity bioassays in pregnant animals, as the highest dosage levels in these experiments are deliberately chosen to produce measurable maternal toxicity (eg, decreased food or water intake, weight loss, clinical signs). However, maternal toxicity defined only by such overt manifestations provides little insight into the toxic actions of a xenobiotic. When developmental toxicity is observed only in the presence of maternal toxicity, the developmental effects may be indirect; however, an understanding of the physiological changes underlying the observed maternal toxicity and elucidation of their association with developmental effects is needed before one can begin to address the relevance of the observations to human safety assessment. Many known human developmental toxicants, including ethanol and cocaine, adversely affect the embryo/fetus predominately at maternally toxic levels, and part of their developmental toxicity may be ascribed to secondary effects of maternal physiological disturbances. For example, the nutritional status of alcoholics is generally poor, and effects on the conceptus may be exacerbated by effects of alcohol on placental transfer of nutrients. Chronic alcohol abuse alters maternal folate and zinc metabolism, which may play a role in the induction of fetal alcohol syndrome (Dreosti, 1993). Carney (2010) has constructed a useful decision tree for interpreting developmental toxicity observed concomitantly with maternal toxicity (Fig. 10-8).

Genetics

The genetic makeup of the pregnant female has been well documented as a determinant of developmental outcome in both humans and animals. The incidence of cleft lip and/or palate [CL(P)], which occurs more frequently in whites than in blacks, has been investigated in offspring of interracial couples in the United States (Khoury et al., 1983). Offspring of white mothers had a higher incidence of CL(P) than offspring of black mothers after correcting for paternal race, whereas offspring of white fathers did not have a higher incidence of CL(P) than offspring of black fathers after correcting for maternal race.

Among experimental animals, two related mouse strains, A/J and CL/Fr, produced spontaneous CL(P) at 8% to 10% and 18% to 26% frequencies, respectively. The incidence of CL(P) in offspring depends on the genotype of the mother rather than that of the embryo (Juriloff and Fraser, 1980). The response to vitamin A of mouse embryos heterozygous for the curly-tail mutation also depends on the genotype of the mother (Seller et al., 1983). The teratogenicity of phenytoin has been compared in several inbred strains of mice. The susceptibility of offspring of crosses between susceptible A/J mice and resistant C57BL/6J mice was determined by the maternal, but not the embryonic genome (Hansen and Hodes, 1983).

Recent work has begun to identify genes associated with differential susceptibility of mouse strains to valproic acid (Finnell et al., 1997; Craig et al., 2000; Bennett et al., 2000; Faiella et al., 2000; Okada et al., 2004, 2005). A recent study utilized inbred C57BL/6J (B6) and DBA/2J (D2) mice to examine the role of maternal genotype in sensitivity to valproic acid teratogenesis (Downing et al., 2010). The B6 mice are more susceptible to valproate-induced digital and vertebral malformations, whereas D2 mice are more susceptible to rib malformations. In a reciprocal cross experiment between B6 and D2 mice, genetically identical F1 offspring responded in a manner consistent with the genotype of the mother: F1 carried in a B6 mother exhibited more vertebral and digital malformations, whereas F1 carried in a D2 mother exhibited more rib malformations. Such maternal effects clearly demonstrate the importance of the maternal milieu in determining outcomes of developmental exposures.

Disease

A number of maternal diseases can exert untoward effects on the conceptus. Chronic hypertension is a risk factor for the development of preeclampsia, eclampsia, and toxemia of pregnancy, and hypertension is a leading cause of pregnancy-associated maternal deaths. Uncontrolled maternal diabetes mellitus is a significant cause of prenatal morbidity. Certain maternal infections can adversely affect the conceptus (eg, rubella virus, discussed earlier), either through indirect disease-related maternal alterations or direct transplacental infection. Cytomegalovirus infection is associated with fetal death, microcephaly, mental retardation, blindness, and deafness (MacDonald and Tobin, 1978) and maternal infection with Toxoplasma gondii is known to induce hydrocephaly and chorioretinitis in infants (Alford et al., 1974). One factor common to many disease states is hyperthermia. Hyperthermia is a potent experimental animal teratogen (Edwards, 1986), and there is a body of evidence associating maternal febrile illness during the first trimester of pregnancy with birth defects in humans, most notably malformations of the central nervous system (Warkany, 1986; Milunsky et al., 1992; Edwards, 2006).

Nutrition

A wide spectrum of dietary insufficiencies ranging from protein-calorie malnutrition to deficiencies of vitamins, trace elements, and/or enzyme cofactors is known to adversely affect pregnancy (Keen et al., 1993). Among the most significant findings related to human nutrition and pregnancy outcome in recent years are results of studies in which pregnant women at risk for having infants with NTDs were supplemented with folate (Wald, 1993; Czeizel, 2011). The largest and most convincing study is the Medical Research Council (MRC) Vitamin Study, in which supplementation with 4 mg of folic acid reduced NTD recurrence by over 70% (MRC, 1991; Bendich, 1993). Results of these studies prompted the US CDC to recommend folate supplementation for women of childbearing age and folate supplementation of some foodstuffs. In 1996, FDA mandated enrichment of cereal grain products with folate by 1998. One study has demonstrated more than a doubling of mean serum folate concentrations across all sex and age groups since then (Dietrich et al., 2005). Yet, in the same population, less than 10% of women of childbearing age reached the recommended erythrocyte folate concentration of greater than 906 nmol/L that has been shown to be associated with a reduction in the risk of NTDs. The CDC (2004) estimated that the number of NTD-affected pregnancies in the United States declined from 4000 in 1995–1996 to 3000 in 1999–2000. Although this survey indicates partial success of the folate fortification program, women capable of becoming pregnant were urged to continue to follow the US Public Health Service recommendation to consume 400 μg folate daily. Mills and Signore (2004) compared the rates of antenatal detection of NTDs in the United States (incomplete ascertainment and underprediction of NTDs) and Canada (more complete antenatal ascertainment) and determined that the studies that best identify cases of NTDs show that folic acid fortification is preventing around 50% of NTDs. These authors estimate that the percentage of folate-preventable NTDs in the United States is about 50% to 60%, suggesting that we may be close to achieving optimum protection. However, recent studies using multivitamins containing folate have achieved even better results (Czeizel et al., 2011).

Stress

Diverse forms of maternal toxicity may have in common the induction of a physiological stress response. Thus, understanding the potential effects of maternal stress on development may help interpret developmental toxicity observed in experimental animals at maternally toxic dosages. Various forms of physical stress have been applied to pregnant animals in attempts to isolate the developmental effects of stress. Subjecting pregnant rats or mice to noise stress throughout gestation can produce developmental toxicity (Kimmel et al., 1976; Nawrot et al., 1980, 1981). Restraint stress produces increased fetal death in rats (Euker and Riegle, 1973) and cleft palate (Barlow et al., 1975), fused and supernumerary ribs, and encephaloceles in mice (Beyer and Chernoff, 1986).

Objective data on effects of stress in humans are difficult to obtain. Nevertheless, studies investigating the relationship of maternal stress and pregnancy outcome have indicated a positive correlation between stress and adverse developmental effects, including low birth weight and adverse mental health outcomes in offspring (Stott, 1973; Gorsuch and Key, 1974; Beydoun and Saftlas, 2008; Entringer et al., 2010; Dunkel Schetter and Tanner, 2012).

Placental Toxicity

The placenta is the interface between the mother and the conceptus, providing attachment, an immune interface, nutrition, gas exchange, and waste removal. The placenta also produces peptide and steroid hormones critical to the maintenance of pregnancy and the regulation of fetal and maternal metabolism. The placenta can also metabolize or store xenobiotics. Placental toxicity may compromise these functions and produce or contribute to adverse effects on the conceptus. Slikker and Miller (1994) listed 46 substances known to be toxic to the yolk sac or chorioallantoic placenta, including metals such as cadmium (Cd), arsenic or mercury, cigarette smoke, ethanol, cocaine, endotoxin, and sodium salicylate. Cd is among the best studied of these, and it appears that the developmental toxicity of Cd during mid-to-late gestation involves both placental toxicity (necrosis, reduced blood flow) and inhibition of nutrient transport across the placenta. Maternal injection of Cd during late gestation resulted in fetal death in rats, despite little cadmium entering the fetus (Parizek, 1964; Levin and Miller, 1980). Fetal deaths occurred concomitantly with reduced uteroplacental blood flow within 10 hours (Levin and Miller, 1980). The authors’ conclusion that fetal death was caused by placental toxicity was supported by experiments in which fetuses were directly injected with Cd. Despite fetal Cd burdens almost 10-fold higher than those following maternal administration, only a slight increase in fetal death was observed.

Cd is a transition metal similar in physicochemical properties to the essential metal zinc (Zn). Cd interferes with Zn transfer across the placenta (Ahokas et al., 1980; Sorrell and Graziano, 1990), possibly via metallothionein (MT), a metal-binding protein induced in the placenta by Cd. Because of its high affinity for Zn, MT may sequester Zn in the placenta, impeding transfer to the conceptus (induction of maternal hepatic MT by Cd or other agents can also induce fetal Zn deficiency, as discussed below). Cadmium inhibits Zn uptake by human placental microvesicles (Page et al., 1992) suggesting that Cd may also compete directly with Zn for membrane transport. Cadmium may also competitively inhibit other Zn-dependent processes in the placenta. Coadministration of Zn ameliorates the developmental toxicity of administered Cd, further indicating that interference of Cd with Zn metabolism is a key to its developmental toxicity (Ferm and Carpenter, 1967, 1968; Daston, 1982).

In the last decade it has become evident that the placenta can undergo adaptive responses to the intrauterine milieu that can influence the metabolic “programming” of the fetus (Myatt, 2006). Adverse intrauterine conditions including hypoxia and nutritional deficiencies can lead to alterations in placental structure and function, including altered expression or activity of transporters and receptors. Placental morphometry has been proposed as a biomarker of fetal programming (Thornburg et al., 2010). One role of the placenta that is key for fetal programming is as a regulator of glucocorticoid transfer between the mother and her fetus(es) (Seckl and Holmes, 2007). The enzyme 11β-hydroxysteroid dehydrogenase-2 (11β-HSD2) inactivates glucocorticoids; its activity in the placenta protects the fetus from over exposure to maternal stress hormone. In experimental animals, lowering the activity of 11β-HSD2 by nutritional deficiency or gene ablation leads to lower birth weight, and causes increases in adult blood pressure, circulating glucose concentrations and hypothalamic-pituitary-adrenal axis activity in offspring. In humans, babies with lower birth weight have higher levels of plasma cortisol throughout life.

Maternal Toxicity

A retrospective analysis of relationships between maternal toxicity and specific types of prenatal effects found species-specific associations. Yet, among rat, rabbit, and hamster studies, 22% failed to show any developmental toxicity in the presence of significant maternal toxicity (Khera, 1984, 1985). In a study designed to test the potential of maternal toxicity to affect development, Kavlock et al. (1985) administered 10 structurally unrelated compounds to pregnant mice at maternotoxic dosages. Developmental effects were agent-specific, ranging from complete resorption to lack of effect. An exception was an increased incidence of supernumerary ribs (ribs on the first lumbar vertebra), which occurred with seven of the 10 compounds. Chernoff et al. (1990) dosed pregnant rats for 10 days with a series of compounds chosen because they exhibited little or no developmental toxicity in previous studies. When these compounds were administered at higher dosages producing maternal toxicity (weight loss or a low incidence of death), varied adverse developmental outcomes were noted, including increased intrauterine death (two compounds), decreased fetal weight (two compounds), supernumerary ribs (two compounds), and enlarged renal pelves (two compounds). Two of the compounds produced no developmental toxicity despite substantial maternal toxicity. These diverse developmental responses led the authors to conclude that maternal toxicity characterized by weight loss or mortality was not associated with any consistent syndrome of developmental effects in the rat.

There have been a number of studies relating specific forms of maternal toxicity to developmental toxicity, including those in which the test chemical caused maternal effects that exacerbated the developmental toxicity of the chemical, and instances in which developmental toxicity was thought to be secondary to maternal effects. However, clearly delineating the relative role(s) of indirect maternal and direct embryo/fetal toxicity is difficult. What follows are examples of the kinds of experiments that may allow one to do so.

Acetazolamide inhibits carbonic anhydrase and is teratogenic in mice (Hirsch and Scott, 1983). Although maternal weight loss was not correlated with malformation frequency, maternal hypercapnia potentiated the teratogenicity of acetazolamide. In C57Bl/6J mice, maternal hypercapnia alone resulted in right forelimb ectrodactyly, the characteristic malformation induced by acetazolamide. Correction of maternal acidosis failed to reduce developmental toxicity, suggesting that the primary teratogenic factor was elevated maternal plasma CO₂ tension (Weaver and Scott, 1984a,b).

Diflunisal, an analgesic and anti-inflammatory drug, causes axial skeletal defects in rabbits. Developmentally toxic dosages resulted in severe maternal anemia (hematocrit = 20%–24% vs 37% in controls) and depletion of erythrocyte ATP levels (Clark et al., 1984). Teratogenicity, anemia, and ATP depletion were unique to the rabbit among the species studied. A single dose of diflunisal on day five of gestation produced maternal anemia that lasted through day 15. Concentration of the drug in the embryo was less than 5% of the peak maternal blood level, and diflunisal was cleared from maternal blood before day nine, the critical day for induction of similar axial skeletal defects by hypoxia. Thus, the teratogenicity of diflunisal in the rabbit was probably due to hypoxia resulting from maternal anemia.

Phenytoin, an anticonvulsant, can affect maternal folate metabolism in experimental animals, and these alterations may play a role in the teratogenicity of this drug (Hansen and Billings, 1985). Further, maternal heart rates were monitored on gestation day 10 after administration to susceptible A/J mice and resistant C57Bl/6J mice (Watkinson and Millikovsky, 1983). Heart rates were depressed by phenytoin in a dose-related manner in the A/J mice but not in C57Bl/6J mice. A mechanism of teratogenesis was proposed relating depressed maternal heart rate and embryonic hypoxia. Supporting studies have demonstrated that hyperoxia reduced the teratogenicity of phenytoin in mice (Millicovsky and Johnston, 1981).

Reduced uterine blood flow has been proposed as a mechanism of teratogenicity caused by hydroxyurea, which produces elevated systolic blood pressure, altered heart rate, decreased cardiac output, severely decreased uterine blood flow, and increased vascular resistance in pregnant rabbits (Millicovsky et al., 1981). Embryos exhibited craniofacial and pericardial hemorrhages immediately after treatment (Millicovsky and DeSesso, 1980a) and identical embryopathies were achieved by clamping the uterine vessels of pregnant rabbits for 10 minutes (Millicovsky and DeSesso, 1980b).

MT synthesis is induced by a wide variety of chemical and physical agents including metals, alcohols, urethane, endotoxin, alkylating agents, hyper- or hypothermia, and ionizing radiation (Daston, 1994). MT synthesis is also induced by endogenous mediators such as glucocorticoids and certain of the cytokines (Klaassen and Lehman-McKeeman, 1989). A mechanism common to the developmental toxicity of these diverse agents may be Zn deficiency of the conceptus secondary to induction of maternal MT. Induction of synthesis can produce hepatic MT concentrations over an order of magnitude higher than normal, leading to substantial sequestration of circulating Zn in maternal liver, lowered plasma Zn concentrations, and reduced Zn availability to the conceptus. Embryofetal zinc deficiency secondary to maternal hepatic MT induction has been demonstrated for diverse chemicals including valproic acid (Keen et al., 1989), 6-mercaptopurine (Amemiya et al., 1985, 1989), urethane (Daston et al., 1991), ethanol, and α-hederin (Taubeneck et al., 1994). In a study combining data for many of these compounds, Taubeneck and co-workers (1994) found a strong positive relationship between maternal hepatic MT induction and maternal hepatic ⁶⁵Zn retention, and a negative relationship between maternal MT induction and ⁶⁵Zn distribution to the litter. This is an example of a maternally mediated mechanism of exacerbating or directly causing developmental toxicity, one that is applicable to diverse maternal insults and for which direct exposure of the conceptus to the causative agent is not required.

Maternal ethanol exposure and the induction of FASD represent an excellent example of the interplay of indirect (maternally mediated) and direct toxic mechanisms of action (Fig. 10-9). Alcoholic mothers exhibit tolerance to alcohol such that their blood alcohol levels can be quite high and no doubt exert multiple direct effects on the embryo, including induction of cell death, altered neural crest cell migration, mitochondrial dysfunction, and effects on membrane integrity. At the same time, effects of chronic alcohol exposure on the mother include MT induction and secondary embrofetal zinc deficiency, as described above. Zn deficiency is teratogenic and exerts multiple adverse effects on the developing conceptus as shown in Fig. 10-9. Additional maternal effects that may exacerbate teratogenicity include hypoferremia, reduced circulating folate, reduced circulating vitamin A, and hypercupremia. Genetic polymorphisms in both the mother and the fetus likely play roles in manifestation of FASD as well, as evidenced by studies in humans and in animal models (Warren and Li, 2005). Thus, despite literally thousands of epidemiological, clinical, and laboratory studies, the complex interplay of factors involved in the teratogenicity of ethanol is still being elucidated.

An “endocrine disruptor” has been broadly defined as “an exogenous chemical that interferes with the production, release, transport, metabolism, binding, action, or elimination of natural hormones responsible for the maintenance of homeostasis and the regulation of developmental processes” (Kavlock et al., 1996). More recently, the Endocrine Society has stated: “an endocrine-disrupting substance is a compound, either natural or synthetic, which, through environmental or inappropriate developmental exposures, alters the hormonal and homeostatic systems that enable the organism to communicate with and respond to its environment” (Diamanti-Kandarakis et al., 2009). Due to the critical role of hormones in directing differentiation in many tissues, the developing organism is particularly vulnerable to fluctuations in the timing or intensity of exposure to chemicals with hormonal (or antihormonal) activity. A wide variety of chemical classes (eg, pesticides, herbicides, fungicides, plasticizers, surfactants, organometals, halogenated polyaromatic hydrocarbons, phytoestrogens) have been shown to cause developmental toxicity via at least three modes of action involving the endocrine system: (1) by acting as hormone receptor ligands; (2) by altering steroid hormone synthetic and metabolic enzymes; and (3) by perturbing hypothalamic-pituitary release of tropic hormones. Interactions with the functions of estrogens, androgens, and thyroid hormones have been the most studied, but the underlying principles apply to other hormones as well.

Laboratory Animal Evidence

Chemicals with estrogenic activity are a well-described class of developmental toxicants based on standard criteria of causing specific malformations during critical developmental periods of relatively short duration (Schardein, 2000). Estrogens induce pleiotropic effects, acting on diverse cell types with estrogen receptors, and can display cell and organ-specific agonist and antagonist actions. The pattern of outcomes is generally similar across different estrogens. DES provides one of the most well-characterized examples of the effects of a xenoestrogen on development. Manifestations include malformations and functional alterations of the male and female reproductive tract and brain. In the CD1 mouse, effective exposures are in the range of 0.1 to 100 μg/kg on gestation day (GD) nine to 16 (Newbold, 1995). At the higher end of the exposure range (10–100 μg/kg), total sterility of female offspring is noted, due in part to structural abnormalities of the oviduct, uterus, cervix, and vagina, and to depletion and abnormalities of ovarian follicles. In adulthood, male offspring show hypospadias, whereas females exhibit excessive vaginal keratinization and epidermoid tumors of the vagina. Vaginal adenocarcinoma is seen at dosages as low as 2.5 μg/kg maternal body weight, and benign uterine tumors (leiomyomas) are seen at as low as 0.1 μg/kg. In male offspring, sterility is observed at high doses, the result of retained rete testes and Mullerian duct remnants, abnormal sperm morphology and motility, cryptorchidism, abnormal reproductive tract secretions, and inflammation (Newbold, 1995). DES exposure of pregnant mice also causes obesity in female offspring (Newbold et al., 2009). Other estrogenic (or antiestrogenic) developmental toxicants include estradiol (Biegel et al., 1998; Cook et al., 1998), ethinyl estradiol, antiestrogenic drugs such as tamoxifen and clomiphene citrate (Branham et al., 1998), and pesticides and industrial chemicals such as methoxychlor (Gray et al., 1989; Hall et al., 1997; Amstislavsky et al., 2003, 2004, 2006), o,p-DDT (Heinrichs et al., 1971), kepone (Gellert, 1978; Guzelian, 1982), dioxins (Mably et al., 1992; Gray et al., 1997a,b), bisphenol A (Nagel et al., 1997), and phytoestrogens such as genistein and coumestrol (Medlock et al., 1995; Rozman et al., 2006). Female offspring are generally more sensitive than males and altered pubertal development, reduced fertility, and reproductive tract anomalies are common findings.

Although most studies on estrogens have indicated traditional dose–response patterns, vom Saal and co-workers (vom Saal et al., 1997; Nagel et al., 1997) have reported that unusual dose– response patterns may occur for endocrine effects on some endpoints. In their studies, a 50% elevation in fetal serum estradiol concentration resulting from implantation of estradiol-containing Silastic capsules on days 13 to 19 of gestation in mice caused a 30% increase in adult prostate weight in male offspring, whereas higher maternal serum concentrations were associated with decreased adult prostate weight. A similar U-shaped dose–response pattern was observed for DES given on days 11 to 17 of gestation, as increased adult prostate weights were seen between 0.02 and 20 ng/kg/day, whereas 200 ng/kg/day resulted in smaller prostates. Bisphenol A (2 or 20 μg/kg/day on gestation days 11–17) also increased adult prostate weight in these mice. However, the issue is controversial, as other researchers using similar testing paradigms have not seen this pattern (eg, Cagen et al., 1999). Vom Saal and colleagues (vom Saal and Hughes, 2005; Welshons et al., 2006) argue that bisphenol A exerts estrogenic effects at low doses and exhibits an inverted U-shaped dose response. Naciff et al. (2005) assessed changes in gene expression in response to in utero exposure to three estrogen receptor agonists, 17α-ethinyl estradiol, genistein, or bisphenol A. Expression of 50 genes was changed by all three chemicals, suggesting that these chemicals have a consistent mode of action. Further, the dose–response for these gene expression changes was monotonic, arguing against the existence of a U-shaped dose–response for these chemicals. Vandenberg and colleagues (Vandenberg et al., 2012) have recently reviewed the issue of low dose effects and nonmonotonic dose–response curves for endocrine disruptors.

Antiandrogens are another class of endocrine-disrupting chemicals. Effects of developmental exposure to an antiandrogen are generally restricted to males, and include hypospadias, retained nipples, reduced testes and accessory sex gland weights, and decreased sperm production. Examples of chemicals known to affect development via an antiandrogenic mechanism include pharmaceuticals such as the androgen receptor antagonist flutamide (Imperato-McGinley et al., 1992) and the 5α-reductase inhibitor finasteride (Clark et al., 1990), the fungicide vinclozolin (Gray et al., 1994) and the DDT metabolite p,p-dichlorodiphenyldichloroethylene (DDE) (Kelce et al., 1995; You et al., 1998); the latter two are androgen receptor antagonists. Dibutylphthalate has been shown to induce an antiandrogen phenotype in developing rats, but the effect does not appear to be mediated by direct interaction with the androgen receptor (Mylchreest et al., 1998, 1999). The critical developmental window for these effects of dibutylphthalate was gestation days 16 to 18 in the rat (Carruthers and Foster, 2005). Dietary exposure of pregnant rats to dibutylphthalate from gestation days 12 to 19 resulted in dose-related decreases in androgen responsive gene expression and testicular testosterone content in male fetuses, and an antiandrogen phenotype in male offspring (Struve et al., 2009). A similar phenotype has been observed in offspring of dibutylphthalate treated rabbits (Higuchi et al., 2003). Importantly, recent studies have demonstrated that exposure of pregnant rats to mixtures of antiandrogenic chemicals can have a cumulative effect on development of the male offspring (Rider et al., 2009, 2010; Hotchkiss et al., 2010; Hannas et al., 2011).

Hypothyroidism during pregnancy and early postnatal development causes growth retardation, cognitive deficits, delayed eye opening, hyperactivity, and auditory defects in rodents. Polychlorinated biphenyls (PCBs) may act at several sites to lower thyroid hormone levels during development and cause body weight and auditory deficits (Goldey et al., 1995; Goldey and Crofton, 1998). PCBs also cause learning deficits and alter locomotor activity patterns in rodents (Eriksson et al., 1991; Schantz et al., 1995) and monkeys (Bowman, 1982; Schantz et al., 1991). The thyroid hormone synthesis inhibitor propylthiouracil (PTU) has been used to study the effects of chemically induced hypothyroidism on development of the brain in rats (Gilbert, 2011). Administration of 1 to 3 ppm PTU in the drinking water of female rats during pregnancy and lactation produced deficits of thyroxine (T4) in the offspring, but only modest reductions in triiodothyronine (T3, the active thyroid hormone) at the high dose. Excitatory postsynaptic transmission was reduced and long-term potentiation was impaired in adult offspring despite the modest and transient effect on T3 during treatment.

Human Evidence

Despite the biological plausibility of effects demonstrated in numerous laboratory studies, the extent to which human health is being adversely impacted from exposures to environmental endocrine disruptors is unclear. It is difficult to demonstrate cause and effect relationships in epidemiological studies where the signals may be weak, the effects may be evident only long after an exposure, and the endpoints are sensitive to other factors. Reports in humans which are or may be relevant are of two types: (1) Observations of adverse effects on reproductive system development and function following exposure to chemicals with known endocrine activities that are present in medicines, contaminated food, or the workplace. These have tended to involve relatively high exposures to chemicals with known endocrine effects (eg, DES). (2) Epidemiological evidence of increasing trends in adverse reproductive and developmental outcomes having an endocrine basis. With the exception of DES, evidence is either lacking to support a definitive link to an exposure, or is variable across studies. Secular trends have been reported for cryptorchidism (Toppari et al., 1996); hypospadias (Toppari et al., 1996; Paulozzi et al., 1997; Paulozzi, 1999); semen quality (Carlsen et al., 1992; Skakkebaek and Keiding, 1994; Olsen et al., 1995; Swan et al., 1997; Auger et al., 1995; de Mouzon et al., 1996; Irvine et al., 1996; Vierula et al., 1996; Bujan et al., 1996; Fisch and Goluboff, 1996), and testicular cancer (Toppari et al., 1996). Recently, connections between environmental exposures and cryptorchidism and hypospadias (Toppari et al., 2010), and early onset of puberty in girls (Mouritsen et al., 2010) have been suggested.

The most convincing evidence for effects of endocrine-disrupting chemicals in humans comes from reports of neurobehavioral changes and learning deficits in children exposed to PCBs in utero or lactationally, either through their mothers’ consumption of PCB-contaminated fish (Jacobson et al., 1990; Jacobson and Jacobson, 1996) or through exposure to background levels of PCBs in the United States (Rogan and Gladen, 1991) or the Netherlands (Koopman-Esseboom et al., 1996). In addition, there have been two occurrences of high-level exposure to contaminated rice oil (in Japan in 1968 and in Taiwan in 1979) in which alterations in development of ectodermal tissues and delays in neurological development were seen (Hsu et al., 1985; Yu et al., 1991; Guo et al., 1994; Schecter et al., 1994). In these cases, there was coexposure to polychlorinated dibenzofurans and PCBs. The mode of action of the developmental neurotoxicity of PCBs is, however, not yet understood.

Impact on Screening and Testing Programs

The findings of altered reproductive development following early life stage exposures to endocrine-disrupting chemicals prompted revisions of traditional safety evaluation tests such as those issued by the EPA (USEPA, 1997). These include assessments of female estrous cyclicity, sperm parameters (total number, percent progressively motile and sperm morphology in both the parental and F1 generations), the age at puberty in the F1 (vaginal opening in the female, preputial separation in the males); an expanded list of organs for pathology and/or histopathology to identify and characterize effects at the target organ; as well as some triggered endpoints including anogenital distance in the F2 and primordial follicular counts in the parental and F1 generations. An important modification of prenatal developmental toxicity test guidelines aimed at improved detection of endocrine disruptors is the extension of the period of dosing to near the end of pregnancy in order to include the developmental period of urogenital tract differentiation.

Experience with chemicals that have the potential to induce developmental toxicity in humans indicates that laboratory animal testing, surveillance of the human population (ie, epidemiological studies), and alert clinical awareness (eg, Friedman, 2011) are all necessary to provide adequate public health protection. Laboratory animal investigations are guided by both regulatory requirements for drug or chemical marketing and the need to understand mechanisms of toxicity.

Regulatory Guidelines for In Vivo Testing

Prior to the thalidomide tragedy, safety evaluations for reproductive effects were limited in the types of chemicals evaluated and the sophistication of the endpoints. Subsequently, the FDA issued more extensive testing protocols (termed Segments I, II, and III) for application to a broader range of chemicals (USFDA, 1966). These testing protocols, with minor variations, were adopted by regulatory agencies around the world and remained similar for nearly 30 years. Several factors including the historical experience of testing thousands of chemicals, increased knowledge of basic reproductive processes, the ever-rising cost of testing, the acknowledged redundancy and overlap of required protocols, a growing divergence in study design requirements of various countries, and the expanding international presence of the pharmaceutical industry succeeded in producing new and streamlined testing protocols that have been accepted internationally (USFDA, 1994). These guidelines, the result of the International Conference of Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH), allow considerable flexibility in implementation depending on the particular circumstances of the agent under evaluation. Rather than specify study and technical details, they rely on the investigator to meet the primary goal of detecting and bringing to light any indication of toxicity to reproduction. Key elements of the FDA Segment I, II, and III studies, the ICH protocols, and the organisation for economic cooperation and development (OECD) equivalent of the FDA Segment II test are provided in Table 10-4. In each protocol, guidance is provided on species/strain selection, route of administration, number and spacing of dosage levels, exposure duration, experimental sample size, observational techniques, statistical analysis, and reporting requirements. Details are available in the original publications as well as in several reviews (eg, Manson, 1994; Claudio et al., 1999; Reuter et al., 2003). Variations of these protocols exist that include extensions of exposure to early or later time points in development and extensions of observations to postnatal ages with more sophisticated endpoints. For example, the Environmental Protection Agency’s Developmental Neurotoxicity Protocol for the rat includes exposure from gestation day six though lactation day 10, and observation of postnatal growth, developmental landmarks of puberty (preputial separation, vaginal opening), motor activity, auditory startle, learning and memory, and neuropathology at various ages through postnatal day 60 (USEPA, 1998a,b).

In 2006, the Regulation on Registration, Evaluation, Authorization and restriction of Chemicals (REACH, 2006) was published. REACH requires that all chemicals marketed or manufactured in the European Union above one ton per year must be characterized as to their toxicity. The requirements are categorized by the annual tonnage of the chemical and are more extensive as the tonnage increases (Rovida et al., 2011). Requirements for reproductive and developmental toxicity are included and are expected to be the largest user of live animals under REACH, which, although it encourages use of alternative toxicity tests to reduce animal use, does not currently deem screening tests for developmental toxicity sufficient to replace required whole animal tests (Scialli, 2008; Rovida et al., 2011).

In part because of the development of new pharmaceuticals for use in children, a Workshop on Testing Strategies and Design of Juvenile Animal Studies was held in 2003 (Hurtt et al., 2004). Subsequently, the FDA and the European Medicines Agency (EMA) have promulgated guidelines for determining the need for and the design of nonclinical studies in support of registering such drugs (USFDA, 2006; EMA, 2008). The design of such studies is flexible and would depend in part on the intended use of the drug (eg, age range, duration of treatment, drug target). To assess the value of such tests, juvenile animal data were compiled from over 200 studies and reviewed in a workshop (Bailey and Marien, 2011). The consensus was that the survey demonstrated the value of juvenile studies for the development of safe pediatric drugs.

The general goal of regulatory studies is to identify the NOAEL, which is the highest dosage level that does not produce a significant increase in adverse effects in the offspring or juvenile animals. These NOAELs are then used in the risk assessment process to assess the likelihood of effects in humans given certain exposure conditions. A preferable alternative, the Benchmark Dose approach, is discussed below.

Multigeneration Tests

Information pertaining to developmental toxicity can also be obtained from studies in which animals are exposed to the test substance continuously over one or more generations (eg, OECD Test Guideline 416 or USEPA 870.3800). For additional information on this approach, see Chap. 20. A recent study (Piersma et al., 2011b) has called into question the value of including a second generation on overall outcome of the assessment. These investigators conducted a retrospective analysis of 498 rat mulitgeneratonal studies, and found that the second-generation mating and offspring rarely provided critical information. These authors instead recommended adopting a proposed OECD extended one-generation test (Cooper, 2009; Vogel et al., 2010) for risk assessment purposes. The extended one-generation test is flexible and could include a second generation if certain triggers were reached in the first generation. Others have argued that the triggers may be too lenient and will too frequently result in invoking production of the second generation (Beekhuijzen et al., 2009).

Children’s Health and the Food Quality Protection Act

In 1993, the National Academy of Sciences published a report entitled “Pesticides in the Diets of Infants and Children,” which brought to light the fact that infants and children differ both qualitatively and quantitatively from adults in their exposure to pesticide residues in food because of different dietary composition and intake patterns and different activities (National Research Council [NRC], 1993). This report, along with the report from the International Life Sciences Institute entitled “Similarities and Differences Between Children and Adults” (Guzelian et al., 1992) provided background and impetus for passage of the Food Quality Protection Act (FQPA) of 1996. The FQPA incorporates an additional 10-fold safety factor for children, cumulative effects of toxicants acting through a common mode of action, aggregate exposure (ie, same toxicant from different sources), and endocrine disruption. The inclusion (at the discretion of the EPA) of the 10-fold factor for calculating allowable intakes for children affects most strongly the pesticide industry, whose products can appear as residues in food. The application of this safety factor is controversial in part because its opponents’ claim that developmental susceptibility is already considered in other tests, such as the Segment II test for prenatal toxicity, the two-generation test, and the developmental neurotoxicity test. On the other hand, proponents applaud the measure and point to the numerous factors that may increase the exposure of infants and children to environmental toxicants and their susceptibility to harm from these exposures. Children have different diets than adults and also have activity patterns that change their exposure profile compared to adults, such as crawling on the floor or ground, putting their hands and foreign objects in their mouths, and raising dust and dirt during play. Even the level of their activity (ie, closer to the ground) can affect their exposure to some toxicants. In addition to exposure differences, children are growing and developing, which makes them more susceptible to some types of insults. Effects of early childhood exposure, including neurobehavioral effects, obesity, diabetes, and cancer, may not be apparent until later in life. Debate continues over the approach to be used for risk assessment in consideration of infants and children.

A large prospective study on children’s health in the United States, the National Children’s Study (NCS) (Branum et al., 2003; Needham et al., 2005; Kimmel et al., 2005; Landrigan and Miodovnic, 2011) was launched in 2009. For this longitudinal birth cohort, 100,000 children will be followed from before birth through 21 years of age. Among many parameters and endpoints to be assessed is exposure to environmental contaminants, including those present in breast milk. In this unique study, because the children will be followed to adulthood, the opportunity exists to assess the full range of potential adverse developmental consequences of environmental exposures. In addition to the National Children’s Study, the US Environmental Protection Agency and the US National Institutes of Health support 14 Centers for Children’s Environmental Health and Disease Prevention Research (Landrigan and Miodovnic, 2011).

Alternative Testing Strategies

A variety of alternative test systems have been proposed over several decades to refine, reduce, or replace the standard mammalian regulatory tests for assessing developmental toxicity. These assays are based on cell cultures, cultures of embryos in vitro, including those of worms, flies, frogs, fish, and mammals (Chapin et al., 2008; Piersma et al., 2010), and a short-term in vivo test (Chernoff and Kavlock, 1982). Daston (1996) has discussed the theoretical and empirical underpinnings supporting the use of a number of these systems. Yet, validation of these alternative tests was a major issue (Neubert, 1989; Welsch, 1990). Although it was initially hoped that alternative approaches would become generally applicable to all chemicals, and help prioritize full-scale testing, this has not yet been accomplished. Given the complexity of embryogenesis and the multiple mechanisms and target sites of potential teratogens, it was perhaps unrealistic to have expected a single test, or even a small battery of tests, to accurately prescreen the activity of chemicals in general. Over the past decade, a validation study of three in vitro embryotoxicity assays, the rat embryo limb bud micromass assay, the mouse embryonic stem cell test (mEST), and the rat WEC test, has been carried out (Genschow et al., 2000, 2002, 2004; Brown, 2002; Piersma et al., 2004; Seiler et al., 2004; Spielmann et al., 2004). This study involved interlaboratory blind trials to validate these assays, and the approach involved the development of “prediction models,” which mathematically combine assay endpoints to determine the combinations and formulations that are most predictive of mammalian in vivo results. More recently, the mEST and other approaches to using mouse embryonic stem cells have been further developed and tested, and the zebrafish embryo has gained prominence as a short-term test for developmental toxicity. Currently, developmental toxicity assays based on rodent WEC, mEST, and zebrafish embryos are receiving the bulk of research attention for use as teratogen screens (Fig. 10-10).

The rodent WEC has been discussed above as a tool for studies of chemical mechanism of action. For WEC, rodent or rabbit embryos are excised from the mother at an early stage of organogenesis with their yolk sacs intact and cultured for up to 48 hours (Ellis-Hutchings and Carney, 2010). During this time they continue to develop in a manner similar to in utero development. At the end of culture the embryos are scored for a number of growth and developmental parameters and assigned a developmental score (eg, rodents: Brown and Fabro, 1981; Van Maele-Fabry et al., 1990; rabbits: Carney et al., 2007). As a screen for developmental toxicity, the WEC has the advantages of using mammalian embryos and being able to divide the many embryos in a litter into different treatment groups. Disadvantages include the technical difficult and time-consuming process of excising the embryos and scoring them at the end of culture, and the cost in animals for obtaining the embryos and the rat serum used as culture medium. As an alternative or in addition to morphological scoring, effects of chemicals on gene expression in the embryos have been studied to discern common patterns (“signatures”) of transcriptomic effects (Luijten et al., 2010; Robinson et al., 2010, 2011a,b, 2012a,b,c,d).

The mEST uses established stem cell lines and examines the effects of test chemicals on viability and differentiation of the cells. The cells aggregate and form “embryoid bodies” when grown in hanging drop cultures. Cardiomyocyte differentiation in the cultures can be quantified by examining the cultures for beating cardiomyocytes. In a different version of the mEST, mouse embryonic stem cells are cultured to adherency in 96-well plates and test chemical applied for up to eight days (Barrier et al., 2011; Chandler et al., 2011). At the end of culture, cardiomyocyte differentiation is assessed by performing in-cell Westerns for cardiac myosin heavy chain and viability is assessed using a fluorescent dye technique. The mEST has the advantage of using established cell lines, therefore totally avoiding the use of live animals. It is also fairly rapid and inexpensive. A disadvantage is the limited developmental repertoire of the cells under the conditions of these tests. Human embryonic stem cell lines or induced pluripotent stem cells are also being developed for toxicity testing, with the obvious advantage of being human cells (Krtolica et al., 2009; Vojnits and Bremer, 2010). However, at present these cells are less well characterized than are mouse embryonic stem cells, and they are generally less robust in culture.

Submammalian species have been used for many years in the study of normal developmental biology, and among these animal models, the African clawed frog, Xenopus laevis or X tropicalis (Bantle, 1995; Fort et al., 2000, 2004; Fort and Paul, 2002), and the zebrafish, Danio rerio (Fraysse et al., 2006; Love et al., 2004; Ton et al., 2006; Yang et al., 2009; Brannen et al., 2010; Sipes et al., 2011a,b; McCollum et al., 2011; Selderslaghs et al., 2012), have been used for developmental toxicology to a number of advantages. Chief among the features of these species is the rapid external development of the embryos, the large historical and recent literature on their normal development, and the availability of genetic mutants and molecular biological tools for studying these embryos. In addition, these species can be bred to produce large numbers of embryos in a relatively short period and are easy and inexpensive to maintain.

The zebrafish has gained increased usage in a number of fields including developmental biology and cancer research. These small fish exhibit a great deal of homology to higher vertebrates in their development, anatomy, physiology, and behavior. Their rapidly developing embryos are transparent, allowing visualization of internal anatomy. Chemical toxicity studies in zebrafish have been comprehensively compiled and reviewed (McCollum et al., 2011). Zebrafish developmental toxicity testing can be carried out in 96-well plates, an approach used for assessment in zebrafish of the Phase I chemical library in the US Environmental Protection Agency’s ToxCast program (Padilla et al., 2012). For a review of zebrafish development and use of zebrafish as a disease model, see Birth Defects Research Part C: Embryo Today: Reviews (2011;93(2)), an issue dedicated to this topic.

The ToxCast program at EPA (Kavlock et al., 2012; http://actor.epa.gov/actor/faces/ToxCastDB/Home.jsp) includes hundreds of high-throughput assays directed at molecular and cellular targets of toxicity, a number of which are relevant to developmental toxicity. Sipes et al. (2011a,b) have evaluated statistical associations of ToxCast assay data with the mammalian developmental toxicity data in EPA’s Toxicity Reference Database, ToxRefDB (http://actor.epa.gov/toxrefdb/faces/Home.jsp). A high number of associations between the rat and rabbit in vivo data and the Toxcast data were discerned, and species-specific predictive models based on the ToxCast assays exhibited greater than 70% balanced accuracy.

A streamlined alternate in vivo test using rodents for prescreening for developmental toxicity was developed by Chernoff and Kavlock (1982), and is known as the Chernoff–Kavlock assay. Pregnant females are exposed during the period of major organogenesis to a limited number of dosage levels near those inducing maternal toxicity, and offspring are evaluated over a brief neonatal period for external malformations, growth, and viability. It has proven reliable over a large number of chemical agents and classes (Hardin et al., 1987), and a regulatory testing guideline is available (USEPA, 1985).

Epidemiology

Reproductive epidemiology is the study of associations between specific exposures of the father or pregnant woman and her conceptus and the outcome of pregnancy (Chambers and Scialli, 2010). In rare situations, such as rubella, thalidomide, and isotretinoin, where a relatively high risk exists and the outcome is a rare event, formal studies may not be needed to identify causes of abnormal birth outcomes. The likelihood of linking a particular exposure with a series of case reports increases with the rarity of the defect, the rarity of the exposure in the population, a small source population, a short time span for study, and biological plausibility for the association (Khoury et al., 1991). In other situations, such as those with ethanol and valproic acid, associations are sought through either a case–control or a cohort approach. Both approaches require accurate ascertainment of abnormal outcomes and exposures, and a large enough effect and study population to detect an elevated risk. Therein lies one of the difficulties for epidemiologists studying abnormal reproductive outcomes. For example, it has been estimated that the monitoring of more than one million births would have been necessary to detect a statistically significant increase in the frequency of spina bifida following the introduction of valproic acid in the United States, where the frequency of exposure was less than one in 1000 pregnancies and the risk was only a doubling over the background incidence (Khoury and Holtzman, 1987). Another challenge to epidemiologists is the high percentage of human pregnancy wastage, perhaps as much as 31% in the peri-implantation period (Wilcox et al., 1988) and an additional 15% that are clinically recognized. Therefore, pregnancy failures related to a particular exposure may go undetected in the general population. Furthermore, with the availability of prenatal diagnostic procedures, some pregnancies with malformed embryos (particularly NTDs) are electively aborted. Thus, the incidence of abnormal outcomes at birth may not reflect the true rate of abnormalities, and the term prevalence, rather than incidence, is preferred when the denominator is the number of live births rather than total pregnancies.

Epidemiological studies of abnormal reproductive outcomes are usually undertaken with three objectives in mind: the first is scientific research into the causes of abnormal birth outcomes and usually involves analysis of case reports or clusters; a second aim is prevention and is targeted at broader surveillance of trends by birth defect registries around the world; and the last objective is informing the public and providing assurance. In this regard, it is informative to consider the review by Schardein (1993) of the method and year by which humans teratogens were detected. For 23 of 28 chemicals reviewed (including nine cancer therapeutics, androgenic hormones, antithyroid drugs, aminoglycoside antibiotics, coumarin anticoagulants, DES, methylmercury, hydantoins, primidone, penicillamine, lithium, vitamin A, and RA), case reports presented the first evidence in humans. For two of these (DES and lithium), the case reports were soon followed by registries that provided confirmation, while for two others (methyl mercury and hydantoins) follow-up epidemiology studies added support. For only four chemicals, alcohol, PCBs, carbamazepine, and cocaine, did an analytical epidemiological study provide the first human evidence. Evidence for one chemical, valproic acid, was first obtained by analysis of a birth defect registry. For the 28 chemicals in that review, human evidence of developmental toxicity preceded published animal evidence in eleven instances. Cohort studies, with their prospective exposure assessment and ability to monitor both adverse and beneficial outcomes, may be the most methodologically robust approach to identifying human developmental toxicants. The lack of cohort studies demonstrating risk for pregnancy may be in part due to the difficulty in making such associations, but may also reflect the fact that use in pregnancy is not associated with increased risk for the majority of drugs (Irl and Hasford, 2000).

Assessment of the developmental toxicity of tobacco smoke has benefited greatly from the many epidemiological studies that have been carried out (see tobacco smoke under Scope of the Problem: The Human Experience). Reasons for this success include the prevalence of smoking and secondhand smoke exposure (allowing large studies), good recall of exposure, and biomarkers of exposure (eg, cotinine), and comprehensive assessment of effects in offspring. The large number of studies has allowed meta-analyses for selected effects.

Concordance of Data

There have been several reviews of the similarity of responses of laboratory mammals and humans to developmental toxicants. In general, these studies support the assumption that results from laboratory tests are predictive of potential human effects. Concordance is strongest when there are positive data from more than one test species. In a quantitative sense, the few comparisons that have been made suggest that humans tend to be more sensitive to developmental toxicants than is the most sensitive test species. Although concordance among species for chemicals reported as positive is high, often special steps must be taken retrospectively to produce an animal model that reflects the nature of outcome in humans (eg, valproic acid, Ehlers et al., 1992 a,b).

Frankos (1985) reviewed data for 38 compounds having demonstrated or suspect developmental toxicity in humans; all except tobramycin, which caused otologic defects, were positive in at least one test species and 76% were positive in more than one test species. Predictiveness was highest in the mouse (85%) and rat (80%), with lower rates for rabbits (60%), and hamsters (40%). Frankos identified 165 chemicals with no evidence of human effects; only 29% were negative in all species tested while 51% were negative in more than one species. Schardein and Keller (1989) examined concordance by species and developmental manifestation for 51 potential human developmental toxicants that had adequate animal data (three human developmental toxicants did not). Thalidomide received the widest testing, with data from 19 species; 53% had data from three species, whereas 18% had data from four or five species. Across all chemicals, the most common findings in humans, rabbits, and monkeys were spontaneous abortion and fetal/neonatal death, followed by malformations and then growth retardation. In the rat, prenatal death, growth retardation, and then malformations was the typical pattern. All species showed at least one positive response for 64% of the human developmental toxicants and, with only one exception, all of the potential human developmental toxicants showed a positive response in at least one species. Overall, the match to the human, regardless of the nature of the developmental response, was rat, 98%; mouse, 91%; hamster, 85%; monkey, 82%; and rabbit, 77%. Jelovsek et al. (1989) reviewed the predictiveness of animal data for 84 negative human developmental toxicants, 33 with unknown activity, 26 considered suspicious, and 32 considered positive. Variables considered included the response of each species, the number of positive and negative species, percent positive and negative species, and mutagenicity and carcinogenicity. The compounds were correctly classified 63% to 91% of the time based on animal data, depending upon how the suspect and unknown human toxicants were considered. The various models had a sensitivity of 62% to 75%, a positive predictive value of 75% to 100%, and a negative predictive value of 64% to 91%. A summary comparison of rodent and rabbit results for 22 human developmental toxicants has recently been compiled by Daston and Knudsen (2010) (Table 10-5).

Elements of Risk Assessment

The extrapolation of animal test data to predict human risk of developmental toxicity follows two basic directions, one for drugs where exposure is voluntary and usually to high dosages, the other for environmental chemicals where exposure is generally involuntary and to low levels. For drugs, a use-in-pregnancy rating has been utilized (USFDA, 1979). In this system the letters A, B, C, D, and X are used to classify the evidence that a chemical poses a risk to the human conceptus. For example, drugs are placed in category A if adequate, well-controlled studies in pregnant humans have failed to demonstrate a risk, and in category X (contraindicated for pregnancy) if studies in animals or humans, or investigational or postmarketing reports have shown fetal risk, which clearly outweighs any possible benefit to the patient. The default category is C (risks cannot be ruled out), assigned when there is a lack of human studies and animal studies are either lacking or are positive for fetal risk, but the benefits may justify the potential risk. Categories B and D represent areas of relatively lesser, or greater concern for risk, respectively. Manson (1994) reviewed the 1992 Physicians’ Desk Reference and found 7% of the 1033 drugs belonged to category X, 66% to category C, and only 0.7% to category A. The FDA categorization procedure has been criticized (Teratology Society, 1994) as being too reliant on risk/benefit comparisons, especially given that the magnitude of risk is often unknown, or the benefits are not an issue (eg, after the drug in question has been taken during early pregnancy, the question is then directed to the management of the exposed pregnancy). The FDA system has also been criticized for demanding an unrealistically high quality of data for assignment to category A (negative controlled studies in pregnant women) and overuse of category C, interpreted as “risks cannot be ruled out” (Sannerstedt et al., 1996). This is an important issue because presently the perception of teratogenic risk is strong among both patients and prescribers even for safe drugs (Pole et al., 2000). The often necessary reliance on studies conducted in animals has also been criticized, and acceptance by clinicians of these studies as reflective of human risks/benefits may be limited (Doering et al., 2002). The approach of FDA for informing physicians and patients about the risk of drugs for use in pregnancy is currently undergoing a complete overhaul. The proposed new system eliminates the letter categories and includes a risk summary and information about available data for the drug during pregnancy.

For environmental chemicals, the purpose of the risk assessment process for noncancer endpoints such as developmental toxicity is generally to define the dose, route, timing, and duration of exposure that induces effects at the lowest level in the most relevant laboratory animal model (USEPA, 1991a,b). The exposure associated with this “critical effect” is then subjected to a variety of safety or uncertainty factors in order to derive an exposure level for humans that is presumed to be relatively safe (see Chap. 4). The principal uncertainty factors include one for interspecies extrapolation and one for variability in the human population. The default value for each of these factors is 10. In the absence of firm evidence upon which to base decisions on whether or not to extrapolate animal test data, certain default assumptions are generally made. They include (1) a chemical that produces an adverse developmental effect in experimental animals will potentially pose a hazard to humans following sufficient exposure during development; (2) all four manifestations of developmental toxicity (death, structural abnormalities, growth alterations, and functional deficits) are of concern; (3) the specific types of developmental effects seen in animal studies are not necessarily the same as those that may be produced in humans; (4) the most appropriate species is used to estimate human risk when data are available (in the absence of such data, the most sensitive species is appropriate); and (5) in general, a threshold is assumed for the dose–response curve for agents that produce developmental toxicity.

New Approaches

The Benchmark-Dose Approach

The use of safety or uncertainty factors applied to an experimentally derived NOAEL to arrive at a presumed safe level of human exposure is predicated on the risk assessment assumption that a threshold for developmental toxicity exists (see “Principles of Developmental Toxicology”). A threshold should not be confused with the NOAEL, as the NOAEL is dependent entirely on the power of the study and is associated with risks perhaps on the order of 5% over the control incidence in typical studies. Also, the value obtained by the application of uncertainty factors to the NOAEL should not be confused with a threshold, as this exposure is only assumed to be without appreciable added risk.

The use of the NOAEL in the risk assessment process has been criticized for several reasons. Because it is dependent on statistical power to detect pairwise differences between a treated and a control group, the use of larger sample sizes and more dose groups (which might better characterize the dose–response relationship) can only yield lower NOAELs, and thus better experimental designs are actually penalized by this approach. In addition, the NOAEL is limited to an experimental dose level, and an experiment might need to be repeated to develop a NOAEL for risk assessment. A final point relates to the fact that, given varying experimental designs and variability of control values, NOAELs actually represent different levels of risk in different studies.

Crump (1984) proposed using a mathematical model to estimate the lower confidence bounds on a predetermined level of risk [the “benchmark dose” (BMD)] as a means of avoiding many of the disadvantages of the NOAEL. The application of this approach to a large compilation of Segment II type data sets (Faustman et al., 1994; Allen et al., 1994a,b; Kavlock et al., 1995) demonstrated that a variety of mathematical models, including those that incorporate developmental-specific features such as litter size and intralitter correlations, can be readily applied to standard test results. On average, benchmark doses based on a 5% added risk of effect calculated on quantal endpoints (eg, whether an implant was affected or not) were approximately equivalent to traditionally determined NOAELs. When the litter was used as the unit of response (did it contain at least one affected implant?), benchmarks calculated for a 10% added risk were most similar to the correspondingly determined NOAEL. Discrepancies between the benchmark dose and the NOAEL were most pronounced when one or more of the following conditions were present: a shallow dose–response, small sample sizes, wide spacing of experimental dosage levels, or more than the typical number of dose levels. These features tend to make determination of the NOAEL more problematic (usually higher) and the confidence limits around the maximum likelihood estimate broader (resulting in lower BMDs). Software programs, including the USEPA’s Benchmark Dose Software (BMDS), are available, helping to make the BMD approach the method of choice for many risk assessment organizations worldwide (Davis et al., 2011).

Biologically Based Dose–Response Modeling

The introduction of statistical dose–response models for noncancer endpoints is the first step in developing quantitative, mechanistic models that will help reduce the major uncertainties of high-to-low dose and species-to-species extrapolation of experimental data. These biologically based dose–response models integrate pharmacokinetic information on target tissue dosimetry with molecular/biochemical responses, cellular/tissue responses, and developmental toxicity (O’Flaherty, 1997; Lau and Setzer, 2000; Lau et al., 2001; Setzer et al., 2001). Gaylor and Razzaghi (1992) proposed a model that related induction of cleft palate to fetal growth inhibition, and Gaylor and Chen (1993) proposed a model relating fetal weight and the probability of fetal abnormality. Shuey et al. (1994) presented a model using the cancer chemotherapeutic 5-fluorouracil (Fig. 10-11). They postulated that the developmental toxicity observed in the term fetus was due to an active metabolite (FdUMP) inhibiting the enzyme TS, with subsequent depletion of thymidine, decreased DNA synthetic rates, reduced cell proliferation, and ultimately, reduced tissue growth and differentiation. Each step in the process was determined experimentally and the relationships were described by Hill equations. The individual equations were then linked in an integrated model to describe the entire relationship between administered dose and the incidence of hind-limb defects. Although this is still an empirically based model, the process clearly demonstrated the utility of the approach in understanding the relative importance of various pathways of abnormal development and in providing a basis for models that incorporate species-specific response parameters. Ultimately, biologically based dose-response (BBDR) models will need to be generalizable across dose, route of exposure, species, and perhaps even chemicals of similar mechanistic classes. Understanding the mode of action for any toxicant, including developmental toxicants, can greatly inform the human risk assessment process. Although the construction of BBDR models exemplifies the extent of sophistication and quantification to which such exercises can be pursued, even more qualitative information on modes of action can have a great impact on the interpretation of developmental toxicity test results. For excellent examples of the utility of mode of action information, the reader is referred to an issue of Critical Reviews in Toxicology (September–October, 2005;35) devoted to this topic. Developmental toxicants for which mode of action is discussed include ethylene glycol (Corley et al., 2005), nicotine (Slikker et al., 2005), phthalate esters (Foster, 2005), vinclozolin (Kavlock and Cummings, 2005), valproic acid (Wiltse, 2005), hemoglobin-based oxygen carriers (Holson et al., 2005), ACE inhibitors (Tabacova, 2005), and PTU (Zoeller and Crofton, 2005).

In 2000, a committee assembled by the NRC released its report, “Scientific Frontiers in Developmental Toxicology and Risk Assessment” (NRC, 2000), which concluded that major discoveries have been made about mechanisms of normal development that have been conserved in diverse animals. Many of these animals have been used extensively in developmental biology and genetics, including the fruit fly, roundworm, zebrafish, frog, chick, and mouse. That these key pathways have been preserved over millions of years of evolution and speciation provides a strong scientific rationale for using these animal models for developmental toxicology. These organisms are advantageous for developmental toxicity studies due to their well-known genetics and embryology, their rapid generation time, and the ability to manipulate their genetics to probe specific developmental pathways or to incorporate human genes, such as those of drug metabolizing enzymes, to answer questions of interspecies extrapolation. As discussed above, there has been a great deal of progress over the past decade in using alternative approaches for evaluating developmental toxicity. A more recent NRC report entitled “Toxicity Testing in the 21st Century: A Vision and Strategy” (NRC, 2007) lays out an approach toward revolutionizing the way the toxicity is assessed, moving from expensive and time-consuming animal studies to high-throughput in vitro assays based on cellular toxicity pathways and analyzing outputs using a systems biology approach. This is the basis for EPA’s ToxCast program discussed above, including assays relevant to developmental toxicity. Although full implementation of such approaches is probably years away, in the interim these assays should at least provide data for prioritizing chemicals for further study.

As our knowledge of the genetic control of normal development has advanced immensely in the past few years, so has the technology to examine gene expression and its control, networks of interrelated gene products, and changes in gene expression induced by alterations in the developmental environment. Advances in genomics, proteomics, metabolomics, and bioinformatics are being used to advance our understanding of health and disease, including normal and abnormal development. The ability to gather and analyze large amounts of biological data using databases and models of biological systems allows a more comprehensive approach to understanding pathways to developmental toxicity. Systems modeling and virtual tissues are among the nascent cutting-edge approaches being developed (Knudsen and Daston, 2010). An excellent example is the modeling of a proposed adverse outcome pathway for embryonic vascular disruption (Kleinstreuer et al., 2011; Knudsen and Kleinstreuer, 2011). The model includes a number of potentially responsive gene networks involved in responses to hypoxia, the angiogenic switch, extracellular matrix interactions and vessel remodeling, as well as predicted downstream consequences of disrupting these networks (Fig. 10-12).

Public sharing of data is a trend that is increasing in biology and toxicology. The ToxRefDB database at EPA mentioned above is a searchable public database of chemical toxicity results of regulatory guideline studies, and it includes data for developmental and reproductive toxicity. Knudsen et al. (2009) have extracted and analyzed rodent and rabbit maternal and developmental toxicity endpoints from ToxRefDB, providing a useful data set for comparison to tests in alternative species and other purposes. Another database, simply termed “Developmental Toxicity Database,” (http://www.ilsi.org/ResearchFoundation/Pages/DevelopmentalToxicityDatabase.aspx) has recently been launched by the International Life Science Institute (ILSI) and includes maternal and developmental toxicity data from studies in rats, mice, rabbits, and hamsters. Databases such as these, along with advances in biology discussed above, will facilitate progress toward improving our ability to predict developmental toxicity reliably, quickly, and more economically.