Diminished experience-dependent neuroanatomical plasticity: evidence for an improved biomarker of subtle neurotoxic damage to the developing rat brain.

Millions of children are exposed to low levels of environmental neurotoxicants as their brains are developing. Conventional laboratory methods of neurotoxicology can detect maldevelopment of brain structure but are not designed to detect maldevelopment of the brain's capacity for plasticity that could impair learning throughout life. The environmental complexity (EC) paradigm has become classic for demonstrating the modifications in brain structure that occur in response to experience and thus provides a set of indices for plasticity in the healthy brain. In this study, we have tested the hypothesis that if degradation of experience-dependent cortical plasticity is used as a biomarker, then developmental neurotoxic effects will be detected at doses below those that alter cortical morphogenesis overtly. Pregnant Long-Evans hooded rats received a single injection of either saline vehicle or 1, 5, 10, or 25 mg/kg of the well-characterized developmental neurotoxicant methylazoxymethanol acetate (MAM) on the 16th or 17th day of gestation. On postnatal days 35-39, male offspring were assigned to either a complex environment (EC) or an individual cage (IC) for 28 days to stimulate neuroanatomical plasticity. This response was measured as the difference between the thickness of visual cortex of IC and EC littermates at a given dose. The threshold dose for significant reduction of cortical thickness was 25 mg/kg, but the threshold dose for failure of plasticity was much lower and could be detected at 1 mg/kg, the lowest dose used. No other method of assessment has detected lasting effects of prenatal exposure to MAM at such a low dose. These data suggest that this simple test of plasticity could be an efficient way to detect subtle neurotoxic damage to the developing brain.

The development of the human nervous system is an extended process. It begins during early embryogenesis with the specification of neural epithelium; it requires that complex synaptic circuits be established [reviewed by Pomeroy and Kim (2000); Volpe (2000)], and it may never truly end in that synaptic connections continue to be modified throughout life, presumably as a mechanism of brain information storage (Black et al. 1998). Therefore, intrusion of a neurotoxicant could produce different defects depending on the timing and level of exposure (Spencer 2000) because it might disrupt populations of neurons engaged in cellular proliferation, migration, differentiation, or maturation (Jensen and Catalano 1998;Rodier 1995). This timing issue alone makes assessing the risk of even a single compound at a single dose a complex task, even discounting individual differences in response. Ideally, a biomarker would assess an important brain function with broad sensitivity. In this study, we questioned whether measuring experience-dependent brain plasticity late in development can reveal subtle lasting deficits caused by exposure to a neurotoxicant early in development.
By late embryogenesis, cortical neurons have been intrinsically programmed to form dendrites of the appropriate architecture (Banker and Waxman 1988). However, dendrites continue to grow into the postnatal period (Becker et al. 1984;Juraska 1982) and remain "plastic," such that extrinsic factors, notably, the unique behavioral experience of the individual, can induce readily quantifiable modifications in dendritic structure throughout life [reviewed by Kolb and Whishaw (1998)]. The therapeutic potential of behavioral stimulation is well established (e.g., Walsh and Greenough 1976), and recent animal studies demonstrate positive plastic changes in brain structure or physiology following housing in a broadly stimulating environment (e.g., Rema and Ebner 1999) or regimens of training designed to address specific functional deficits (Klintsova et al. 1999). The fact that plasticity is often reduced (e.g., Rema and Ebner 1999), or even eliminated after these manipulations (e.g., Berman et al. 1996), however, suggests that measurements of plasticity per se could provide an improved biomarker for detecting latent neurotoxic damage relevant to lasting learning and memory deficits.
The environmental complexity (EC) paradigm, originally devised by Hebb (1947), has been a key tool in documenting that behavioral experience produces enduring modifications of brain circuitry and behavior throughout life. It has also been used to identify cellular and molecular changes associated with the longterm storage of information in the mammalian brain [reviewed by Greenough (1976); Rosenzweig et al. (1972);van Praag et al. (2000)]. In this model, litters of rats are divided into conditions of differential environmental stimulation. Rats housed individually in standard cages (IC) are compared with their siblings that have been housed as a group in a complex environment that provides continuing opportunity for exploration, play, and spatial learning. Experience in the EC induces a robust and coordinated growth response in neurons (Kempermann et al. 1997;Turner and Greenough 1985;Volkmar and Greenough 1972), as well as in supporting cells (Black et al. 1987;Jones et al. 1996;Sirevaag and Greenough 1991). This array of cellular changes accumulates in some regions of neocortex to the extent that the thickness of cortex increases significantly (Diamond et al. 1964).
In the present study, we have selected this classic behavioral paradigm to test the hypothesis that cortical plasticity will begin to fail at a lower neurotoxicant dose than will the gross maturation of cortex. Our goal was to establish a laboratory assay for neuroanatomical plasticity using a summary measure such as cortical thickness that would be efficient yet sensitive. Such an assay could then be used to screen a variety of substances, including mixtures such as those found in groundwater surrounding Superfund sites.
As an initial test of this model, we selected a compound with well-characterized effects on brain development in rats, methylazoxymethanol acetate (MAM), a derivative of the cycad plant . MAM causes DNA damage, and through this mechanism it is thought to kill actively dividing neurons (Esclaire et al. 1999). When administered during rat cortical neurogenesis [gestational days (GDs) 14-18], MAM produces a dose-dependent forebrain microcephaly that is most pronounced at GDs 14-15 (Fischer et al. 1972;Haddad et al. 1972;Tamaru et al. 1988). MAM-induced microcephaly has become a common experimental model for studying cortical dysgenesis (Berger-Sweeney and Hohmann 1997; Colacitti et al. 1999). In addition to killing some neurons as they are being generated, prenatal MAM treatment causes detectable elevations in DNA adducts that remain detectable at postnatal day (PD) 30 ) and generates aberrant patterns of connectivity in the remaining cells (Colacitti et al. 1999). We administered MAM on GD 16 or 17 because this exposure a) coincides with the generation of association neurons in the superficial layers of cortex (Bayer and Altman 1995) that express robust plasticity (Juraska et al. 1980;Wallace et al. 1992) and b) produces milder morphogenic effects than do exposures at GDs 14-15 (e.g., Balduini et al. 1991). We chose a range of doses based upon data from GD-15 exposures in which the reported threshold dose for obvious reduction of cortical volume by MAM is > 10 mg/kg (Fischer et al. 1972;Haddad et al. 1972;Tamaru et al. 1988) and the reported threshold dose for behavioral deficits (e.g., open field, maze training and avoidance conditioning) is > 15 mg/kg (e.g., Balduini et al. 1991;Ohta et al. 2000). A high dose (25 mg/kg) was included as a positive control for morphogenic effects, but we focused on relatively low doses expected to produce a grossly normal brain (i.e., 1, 5, and 10 mg/kg).

Materials and Methods
Embryonic exposure to MAM. Litters of Long-Evans hooded rats (Simonsen Laboratories, Gilroy, CA) were exposed in utero to the neurotoxin MAM (Midwest Research Institute, Kansas City, MO) on GD 16 or 17 via a single intraperitoneal injection to the pregnant dam. The day of sperm plug detection was recorded as GD 1.
Timed-pregnant dams were selected randomly to receive 0 (saline vehicle), 1, 5, 10, or 25 mg/kg MAM in saline (12 dams per GD). For all doses, 2 dams were injected, except 3 dams each received saline or 5 mg/kg MAM injections on GD 16 and 3 dams each were injected with 5 mg/kg or 10 mg/kg treatments on GD 17. Offspring were housed with the dam until weaning on PD 28 and then housed in groups of two or three until the start of differential housing. Because handling per se can apparently induce neuroanatomical changes (Horner et al. 1991), we left rats undisturbed between weaning and assignment to differential housing to maximize EC-IC differences. All animals were handled according to the guidelines of the National Institutes of Health Office of Laboratory Animal Welfare (2002).
Differential housing. At PDs 35-39, male littermates were selected randomly for assignment to differential housing conditions, and these conditions were then balanced for body weight. Two littermate pairs from each dose were assigned to differential housing, with the exception of two litters that produced fewer than four males (one litter treated with saline, one litter treated with 25 mg/kg MAM). Rats were housed either individually in standard laboratory cages (IC) or as a cohort of 12 with the opportunity to play and explore in a large cage filled with toys that were changed daily (EC). As an additional opportunity for exploration, EC animals were placed in an open-field play cage with a novel arrangement of toys for 1 hr daily. All doses of MAM-treated and saline-treated rats were represented in a single EC cage, such that two parallel EC environments were required to accommodate the number of animals involved. Differential housing was continued in this manner for 28 days. Food and water in both conditions were administered ad libitum. EC and IC conditions were maintained in the same room under a 12-hr light/12-hr dark cycle to control for ambient conditions.
Cortical thickness measures and quantitative analyses. Cortical thickness was used as a simple, quantitative measure to determine both the effects of MAM treatment and the effects of differential housing. At the end of environmental exposure, rats were euthanized [sodium pentobarbital (Nembutal), 150 mg/kg, intraperitoneally], and their brains were removed, fresh-frozen using isopentane chilled over dry ice, sectioned into 20 µm coronal sections with a cryostat, and stained with a Nissl stain. A standard sample section of occipital cortex (OC) was approximately -4.3 mm from bregma and identified by the presence of both dorsal and ventral hippocampal formation with a narrowing in the middle [comparable with Figure 21 from Zilles (1985)]. The section was scanned to produce a digital image [Polaroid SprintScan 35+ with scan path enabler (Polaroid Corporation, Waltham, MA) using Photoshop 5.5 software (Adobe Systems Inc., San Jose, CA)]. To correct for overall differences in brain size that might arise from toxicant treatment, standard atlas coordinates for cortical subdivisions within the sample section (Zilles 1985) and the spacing of triplicate measures were normalized to the crosssectional area of each brain. Cortical thickness was measured using NIH Image 1.62 software (National Institutes of Health, Bethesda, MD) as the distance between the lower boundary of layer I and the beginning of the white matter beneath layer VI. All histological processing and data collection were carried out blind to the experimental history of the cases.
The average of the three measurements was analyzed using a general linear model (GLM), allowing for unequal numbers of subjects, with dose, housing condition, and GD of dose (GD 16 vs. GD 17) as main effects. Design and number of animals per condition are reported in Table 1. Unequal numbers arose because of the a priori decision to use additional animals in some conditions (particularly intermediate doses), low numbers of males in some litters, and tissue loss during processing. The influences of dose and housing condition on cortical thickness were assessed using t-statistics for individual effects estimated from the GLM (Netter et al. 1989).

Results
Experimental design and main effects. We exposed the developing rats to a range of MAM doses via injections to timed-pregnant For each dose, subjects were assigned as littermate pairs, and housing conditions were balanced for body weight.

Figure 1.
Dose-response analysis of MAM on brain size. The relative effects of prenatal exposure to 0, 1, 5, 10, and 25 mg/kg of MAM are shown in a comparison of brains from rats exposed at GD 16 and photographed at weaning. An enduring reduction in cortical volume was obvious at the highest dose, 25 mg/kg, but was not apparent at lower doses. These defects are most apparent in the extensive exposure of the subcortical colliculi in the posterior cortex (arrow), suggesting that the overlying cortex is diminished in size. dams on GD 16 or 17, a time when the superficial cortex is undergoing significant development. We then screened for enduring defects when these animals were in late adolescence (PDs 60-64) using two types of biomarkers. The first biomarker was evidence of persistent cortical maldevelopment (i.e., an index of deficient morphogenesis). The second biomarker was the capacity of MAM-exposed cortex to grow thicker in response to behavioral stimulation (i.e., an index of deficient plasticity).
Because MAM was delivered in utero, it was possible to compare the relative sensitivities of these two biomarkers in littermates exposed to MAM under the same conditions and assigned randomly to the EC or IC condition. Statistical analysis revealed a significant main effect of MAM upon cortical thickness (p < 0.0001), a significant main effect of housing condition upon cortical thickness (p < 0.0001), and a significant main effect of GD of exposure (p < 0.0001). Significant interactions were detected between MAM dose and housing condition (p < 0.05) and MAM dose and GD of dose (p < 0.01) but not between housing condition and GD of dose (p = 0.51).

Disruption of cortical morphogenesis as a biomarker of prenatal MAM exposure.
Upon weaning, representative brains from litters exposed at GD 16 were compared (Figure 1). Visual inspection of these brains confirmed that the volume of cerebral cortex was reduced dramatically by a dose of 25 mg/kg, as indicated by the exposed dorsal surface of both inferior and superior colliculi (Figure 1). Using these criteria, however, the brains of rats receiving MAM doses of 1, 5, and 10 mg/kg were not obviously different from those exposed to saline vehicle only (0 mg/kg). Brains of rats exposed to 25 mg/kg MAM on GD 17 also showed obvious reductions of cortical mass (data not shown). The observation of microcephaly at the 25 mg/kg dose served as a positive control for the action of MAM at lower doses where exposed brains showed no overt signs of cortical maldevelopment.
For a more fine-grained assessment of morphogenic defects, we measured the overall thickness of cerebral cortex (Figure 2A). We sampled thickness in the OC because this region is appropriate for assessing both defective morphogenesis ( Figure 1) and defective experience-dependent plasticity. Rats (PDs 60-64) housed in the IC were used to assess toxicant effects on thickness of cortex ( Figure 2B). A decrease in cortex was observed in animals exposed to 25 mg/kg MAM on GD 16 but was not detected in animals exposed during GD 17. Taken together with our observations of overall brain size (Figure 1), these data suggest that animals exposed to MAM on GD 17 may have less cortical volume, but the thickness of the cortex that survives appears to be relatively normal. Nonetheless, the sensitivity of cortical mass as a biomarker appears limited to relatively high dose effects, with a detection threshold in the present experiment similar to that of previous reports (Fischer et al. 1972;Haddad et al. 1972;Tamaru et al. 1988).
The capacity for robust plasticity emerges from healthy development. The usefulness of experience-dependent plasticity as a biomarker hinges on its reliable and robust induction. Figure 3 demonstrates the stimulatory effects of the EC condition ( Figure 3A) compared with housing in the IC condition ( Figure 3B) and shows that the thickness of the OC is significantly increased after 1 month of exposure to the EC condition. The mean thickness of cortex was increased by 18% (p < 0.001; Figure 3C).

Disruption of cortical plasticity as a biomarker of prenatal MAM exposure.
To determine if exposure to MAM interfered with this growth response, we compared differences in mean cortical thickness between EC and IC at each dose of MAM (0, 1, 5, 10, and 25 mg/kg). Although significant differences were detected for both main effects of dose and housing condition with the GLM analysis of variance, the significant interaction between environment and MAM dose revealed that MAM during cortical development altered the Article | Wallace et al. 1296 VOLUME 111 | NUMBER 10 | August 2003 • Environmental Health Perspectives  Figure 4B, GD 17). Rats exposed to the highest dose of MAM (25 mg/kg) failed to exhibit significant increases in cortical thickness in response to the EC condition. Rats exposed at GD 16, however, showed both a reduction in baseline thickness of cortex (indicating a morphogenic defect; Figure 2B) and a reduction in plasticity, whereas rats exposed at GD 17 showed only reduced plasticity. A critical test of experience-dependent plasticity as a biomarker of subtle neurotoxic insult is the comparison between the effects of EC and IC conditions at low doses of MAM. At the two lowest doses (1 and 5 mg/kg of MAM), no basal morphogenic defect was detected ( Figure 2B), but the effect of the EC condition was blunted or lost (Figure 4). At 1 mg/kg, the response to the EC condition differed with GD of exposure; at GD 16 the magnitude of response was reduced but still significant (p < 0.05; Figure 4A), whereas at GD 17 there was no significant effect of EC (p = 0.23; Figure 4B). At 5 mg/kg, neither the GD 16 nor the GD 17 rats exhibited a significant EC effect (GD 16, p = 0.45; GD 17, p = 0.69). Rats exposed to 10 mg/kg MAM also failed to show a significant effect of EC (GD 16, p = 0.15; GD 17, p = 0.13). The trend for EC rats to have numerically thicker cortices than their IC siblings across the 1-10 mg/kg MAM range suggests that, although reduced significantly, some capacity for growth may have been spared.
To determine if exposure to MAM reduced plasticity in a dose-dependent manner, we tested the hypothesis that the EC cortices would decrease in thickness with increased dose using Jonckheere's test (Jonckheere 1954). With this test, the null hypothesis was rejected both in rats exposed at GD 16 (p < 0.001) and at GD 17 (p < 0.05). Because these analyses examined only the differences between EC groups across doses, we also determined the correlation of the difference in EC/IC cortical thickness and dose ( Figure 5). For rats treated with MAM on GD 16, this analysis produced a correlation coefficient of -0.87 ( Figure 5A); for those exposed on GD 17, the correlation was -0.63 ( Figure 5B). Together, these results indicate that as the dose of MAM increased, the magnitude of the EC response was diminished.

Discussion
The answer to the question of whether the brain develops normally in the presence of a toxicant may depend upon on how development is measured. One way is use a variety of sensitive neuropathological techniques to assess the integrity of neural circuits and associated metabolic support at the close of a key stage (Garman et al. 2001). Here, as an alternative way to define healthy development, we evaluate experience-dependent plasticity, a late component of brain development that emerges after basic circuitry has formed and remains active into maturity (Black and Greenough 1986). We hypothesized that plasticity would be a more sensitive biomarker because it appears to be an emergent integrative property of a properly assembled brain. The induction of plasticity depends on a rat's active exploration of its environment (Ferchmin et al. 1975), and its expression involves changes in the structure of neurons (Volkmar and Greenough 1972), glia (Jones et al. 1996;Sirevaag and Greenough 1991), and blood vessels (Black et al. 1987). The major finding of this study was that young adult rats could appear to have normal cortical thickness after prenatal exposure to low doses of MAM, yet exhibit significantly diminished experiencedependent neuroanatomical plasticity.
The threshold dose of MAM required to significantly reduce cortical thickness was clearly > 10 mg/kg, but doses as low as 1 mg/kg impaired plasticity. This initial assessment of plasticity as a biomarker demonstrates deficits at lower prenatal exposures of MAM than have been documented previously (Ohta et al. 2000). The potential of this model is in its ability to test the integrity of cellular plasticity directly under conditions relevant to learning and memory. Its strength lies in the ability to detect the expression of experience-dependent plasticity, as well as the degradation of plasticity by neurotoxic insult, through the use of within-litter comparisons. In the present study we did not sample a sufficient number of subjects per dose to fully document the absolute threshold for disruption of plasticity, nor was that its intent. Instead, the study was designed to test the relative sensitivity of this measure using a well-characterized neurotoxicant. Although the present data suggest that prenatal MAM dose and postweaning capacity for experience-dependent plasticity may be inversely related (Figure 5), confirmation of this trend will also require a larger data sample.
Based on this initial study with MAM, we would not predict that exposure to neurtoxicants at the low doses of greatest interest would necessarily eliminate plasticity. In fact, residual plasticity in the face of toxic insult Figure 5. The capacity for plasticity is inversely correlated with dose of MAM; the difference between effects of the EC and IC conditions is plotted by dose for rats exposed to MAM on GD 16 (r = -0.87) (A) and on GD 17 (r = -0.63) (B). CT, cortical thickness. These plots show a consistent, systematic decline in plasticity, most notably across the low doses (0-5 mg/kg). They also show an unexpected, but consistent, break in trend at 10 mg/kg. Any exposure to MAM resulted in a diminished plasticity response compared with that of saline-treated control rats.  Figure 4. Prenatal MAM exposure impairs postweaning cortical plasticity. OC thickness was measured in rats exposed to either saline vehicle (0 mg/kg) or MAM (1, 5, 10, or 25 mg/kg) at GD 16 (A) or GD 17 (B) and housed postweaning in either the EC or the IC condition for 30 days. Plasticity failed at MAM doses below those required to reduce baseline cortical thickness (compare with Figure 2). Data are reported as mean thickness ± SE. could play an important role in compensating for brain damage during development and therefore holds therapeutic promise . Remarkable levels of cognitive function have been reported in individuals with profound defects in brain morphogenesis (Lewin 1980), indicating that the compensatory capacity of human developmental plasticity can be profound under certain conditions. Accordingly, these data suggest that subtle deficits are likely to show up as quantitative differences in the magnitude of plasticity or in the rate at which it is expressed.
The generality of this model awaits assessment in future studies. The EC paradigm has proven quite reliable in inducing robust neuroanatomical plasticity in a variety of species (Rosenzweig and Bennett 1996) and appears to represent a fundamental mechanism of brain information storage relevant to humans (Bertenthal et al. 1984;Karni et al. 1995;Maguire et al. 2000). MAM is a compound that affects the development of cerebral cortex directly . We would predict, however, that even if cortical cells are not damaged directly, cortical plasticity is likely to be vulnerable to disruptions in the integrity of the extrinsic afferent systems that trigger it or modulate its expression.
We observed relatively large increases in cortical thickness of untreated rats after 28 days of the EC condition [e.g., Diamond (1967)]. However, this analysis may not be sufficiently sensitive in all species or strains (e.g., Wainwright et al. 1993), and more refined assessments of plasticity based specifically on the growth response of cells may be required. In species in which both neuropathology can be modeled appropriately and experience-dependent plasticity induced robustly, however, the efficiency of a simple measure of plasticity such as cortical thickness has tremendous advantages in screening a large set of toxicants. A testable prediction is that this paradigm can be applied to a variety of neurotoxicants to reveal deficits at lower doses than can routine morphological examination. If so, this novel biomarker based on experience-dependent neuroanatomical plasticity could become an important tool in establishing limits for acceptable levels of toxicants present in the environment.