Dopamine (DA) participates in locomotor activity, reward, motivation and feeding behavior and its dysfunction is implicated in several neurobehavioral disorders including ADHD and addiction. The central DA system consists of neurons that originate in substantia nigra pars compacta (SNpc), ventral tegmental area (VTA), and project through nigrostriatal (locomotor activity), mesolimbic (reward, feeding), and mesocortical (motivation, emotional) pathways. Dopamine-related gene expression is also prominent within the hypothalamus. A broad range of prenatal insults, including malnutrition, maternal stress or infection and environmental toxins, have been found to alter dopamine function (Wang et al., 2009; Zhou et al., 2009; Palmer et al., 2008; McArthur et al., 2007; Son et al., 2007; Stanwood and Levitt, 2007; Valdomero et al., 2005).

The molecular link between adverse neurobehavioral outcomes and a suboptimal prenatal environment may involve epigenetic changes (DNA methylation, histone methylation or acetylation (Jirtle and Skinner, 2007; Weaver et al., 2004; Szyf, 2009)). Imprinted genes, due to their monoallelic and epigenetically regulated expression, are particularly susceptible to dysregulation as a result of suboptimal prenatal conditions (Jirtle and Skinner, 2007; Kwong et al., 2006). Similarly, disruption of the imprinting status of several imprinted genes in the CNS leads to neurodevelopmental disorders (e.g, Prader-Willi Syndrome, Angelman syndrome) (Davies et al., 2005). Cyclin-dependent kinase inhibitor 1C (Cdkn1c/p57) is an imprinted gene located in an imprinted region on mouse chromosome 7 encoding a cyclin-dependant kinase inhibitor that acts to negatively regulate cell proliferation and, in some tissues, to actively direct differentiation (Smith et al., 2007; Ye et al., 2009). In the brain, Cdkn1c is expressed in dopaminergic neurons where it exerts a crucial role in dopamine neuron differentiation (Joseph et al., 2003; Freed et al., 2008). Furthermore, overexpression of Cdkn1c in a transgenic mouse model leads to embryonic growth retardation and low birth weight, a characteristic IUGR phenotype (Andrews et al., 2007). Igf2, another imprinted gene in this locus, has also been linked to fetal growth (Zeisel, 2009). Therefore, gene expression and differential methylation was determined for genes within the Cdkn1c locus.

IUGR infants are at significant risk for long term neurobehavioral disabilities, including ADHD, depression, and addiction. In utero malnutrition does not lead to major structural or anatomical deficits within the CNS, but rather permanent suboptimal development (Ranade et al., 2008). Experiments in the present manuscript were designed to examine dopaminergic dysregulation at the molecular and behavioral level in a mouse model of IUGR. Vulnerability to epigenetic modifications was determined for dopamine-related genes as well as potentially more vulnerable imprinted genes within the Cdkn1c locus.

Early life exposure to LP diet increased expression of TH in all regions examined, while upregulation of DAT and DARPP-32 was evident in only select brain regions. Importantly, we also observed an increase in TH-ir neurons within the VTA, and an increase in DA within the PFC. Dopamine levels within the striatum were not altered, suggesting that VTA to PFC dopaminergic projections may be uniquely sensitive to this perinatal challenge or that the increased expression of DAT in the NAc (which was not elevated in the PFC) participated in normalizing dopamine levels. These data complement reports of increased whole brain DA levels (Chen et al., 1997; Marichich et al., 1979), TH activity (Marichich et al., 1979), and altered dopamine receptor binding in the striatum in rats from protein-restricted pregnancies (Palmer et al., 2008; Marichich et al., 1979). TH-producing neurons within the PFC are likely non-classical TH-producing interneurons (Asmus et al., 2008) although little is known about these cells, including their final transmitter phenotype. These neurons lack additional downstream enzymes necessary for catecholamine synthesis and these neurons have been termed “dopaergic” (Ugrumov et al., 2004). Interestingly, both classical and non-classical catecholaminergic cell groups responded similarly to the early life protein restriction.

Altered DA tone has significant implications for human health, given the broad range of normal and pathological behaviors that rely on dopamine. Studies in infants have shown that reduced fetal growth represents a consistent risk for ADHD (Hultman et al., 2007; Lahti et al., 2006; Mick et al., 2002). Although the relative importance of dopamine in ADHD pathology has been debated (Gonon, 2009), there is a wealth of data that support a role for dopamine dysfunction in ADHD pathology (van der Kooij and Glennon, 2007; Madras et al., 2005; DiMaio et al., 2003). Overexpression of DAT in the NAc (similar to the increased accumbens DAT expression observed in the IUGR mice) was shown to increase impulsivity and risk taking behavior (Adriani et al., 2009), prominent behavioral components of ADHD, and methylphenidate administration was shown to decrease DAT expression in the striatum (Moll et al., 2001), which may contribute to its therapeutic effectiveness. However, decreases in DAT (and D2/D3) in the left midbrain and accumbens have been linked to measures of inattention in ADHD patients (Volkow et al., 2009), highlighting the difficulty of replicating the totality and heterogeneity of ADHD symptoms in any one animal model. Hyperactivity (Wilson, 2000), a component of ADHD, was also present in the IUGR mice, but was only evident in animals fed a high fat diet. While the explanation for this is currently unknown, consumption of a high fat diet is known to alter dopaminergic activity (Lee et al., 2009). It is possible that the dopaminergic alterations due to IUGR establish an underlying vulnerability to hyperactivity and when coupled with altered dopamine signaling in response to consumption of high fat diet, hyperactivity emerged. The present data suggest that altered DA in response to a suboptimal perinatal environment may contribute to some of the behavioral components of ADHD observed in IUGR infants.

Altered reward processing (Haenlein and Caul, 1987) is also a component of ADHD. IUGR mice demonstrated a decrease in sucrose preference, reflecting a decreased responsiveness to rewarding stimuli. Individual differences in sucrose consumption (high preference versus low preference) positively predict amphetamine self-administration (DeSousa et al., 2000) and correlate with D2 binding in the striatum (Tonissaar et al., 2006). Decreased sucrose preference in response to stress has also been associated with increased dopamine in the hypothalamus, striatum and PFC (Bekris et al., 2005), which is similar to our observation of increased DA in the PFC and increased TH in the hypothalamus. It is important to note, however that dopamine deficient mice have been shown to have an intact sucrose preference (Cannon and Palmiter, 2003), and additional neurotransmitter systems are clearly involved. An extensive literature has documented decreased sucrose preference in response to chronic mild stress, commonly interpreted as an animal model of “anhedonia”. Accordingly, this decrease can be ameliorated by antidepressant treatment (i.e., broad acting tricyclics such as imiprimine (Bekris et al., 2005), as well as selective serotonin reuptake inhibitors like citalopram (Rygula et al., 2006)). In addition to dopamine, decreased activity of serotonin in the PFC has also been linked to reductions in sucrose preference (Bekris et al., 2005) and citalopram can reverse both the behavior and the neurochemical changes (Rygula et al., 2006). Of note, however, citalopram also decreases striatal dopamine and increases D2 binding (Dewey et al., 1995), indicating that an interaction between serotonin and dopamine systems is likely to be important in mediating sucrose preference. Activation of the reward system by drugs of abuse (Martinez et al., 2007; Ron and Jurd, 2005), coupled with underactivation of reward circuitry in response to natural rewards (i.e., sucrose) may act synergistically in the development of addiction (Koob and Le Moal, 2008). These two behaviors were noted in IUGR animals (exaggerated response to cocaine and decreased sucrose preference), suggesting that the IUGR animals may represent an animal model of increased addiction risk as a result of suboptimal prenatal environment. These findings are consistent with other observations in the literature linking malnutrition and addictive behavior in rodents (Palmer et al., 2008; Valdomero et al., 2007; Valdomero et al., 2006; Shultz et al., 1999) and humans (Franzek et al., 2008).

In an effort to define potential mechanisms driving the observed gene expression changes, we investigated the methylation status of specific target genes. A recent report examined genome-scale methylation status of CpG islands in the liver in response to maternal malnutrition (van Straten et al., 2010), and found that less than 0.5% of CpG islands were differentially methylated. Further, they observed equal amounts of hyper- and hypomethylation, suggesting that limited substrate (e.g., methyl donors) due to the low protein diet was not likely to be responsible for the differential methylation patterns. Further, CpG islands within promoter regions are known to demonstrate varying degrees of sensitivity to aberrant methylation (Feltus et al., 2003), and here we report that TH, DAT and Igf2 are not susceptible to differential methylation, but Cdkn1c is vulnerable to differential methylation in response to early life protein restriction. Hypomethylation and subsequent overexpression of Cdkn1c in IUGR animals is notable for three primary reasons: (1) Cdkn1c is involved in DA neuron differentiation, (2) neural overexpression of Cdkn1c results in growth retardation (Andrews et al., 2007) and (3) as an imprinted locus, this region may be more susceptible to epigenetic dysregulation as a result of suboptimal prenatal conditions. Cdkn1c is important for neuronal migration and coordination of the timing of progenitor cell cycle exit in developing mouse neocortex (Ye et al., 2009; Itoh et al., 2007) and most likely partners with Nurr1 to promote terminal differentiation of DA neurons (Joseph et al., 2003; Freed et al., 2008). Interestingly, Igf2, overexpressed in the PFC and hypothalamus of IUGR mice, has also been shown to play a role in DA neuron development (Vazin et al., 2009). Counts of TH-ir neurons were made within the VTA, as this is the primary source of dopaminergic innervation of the PFC where an increase in DA was detected, even though we did not detect an increase in Cdkn1c in this region. It is possible that elevations in Cdkn1c were not detected using qRT-PCR in a mixed (VTA/SN) dissection or it is possible that Cdkn1c levels were elevated at an earlier timepoint in development. At this time, it is premature to define Cdkn1c overexpression as causative in the dopaminergic changes observed in the IUGR animals. However, given its dysregulation in this model of IUGR, the importance of Cdkn1c to dopamine neuron development, and the link between neuronal overexpression of Cdkn1c and embryonic growth retardation, our data highlight Cdkn1c as an important molecule for future research within this context. Only a subset of genes in Cdkn1c-Kcnq1 locus was differentially expressed in IUGR animals, which implies that the methylation status of IC1 (see Fig 4a), normally inherited from gametes, is not responsible for observed gene expression changes. Instead the Cdkn1c promoter is hypomethylated in IUGR offspring, which is established during the post-implantation period (Bhogal et al., 2004), identifying this as a vulnerable timeframe for long-lasting epigenetic dysregulation of Cdkn1c gene.

In these studies, maternal exposure to the low protein diet extended through the entirety of breeding, pregnancy and lactation. At birth, rodents are considered more “immature” than humans, with significant brain development occurring postnatally (e.g., postnatal days 3–10 are thought to mirror the third trimester of a human pregnancy (Livy et al., 2003) with precise pairing of developmental time periods dependent upon brain region (Clancy et al., 2007). Therefore, to model the entirety of human gestation, we chose to maintain the low protein diet through pregnancy and lactation. A number of factors may participate in driving the observed neurodevelopmental phenotype, including the prenatal protein restriction of the pup or the maternal endocrine response (Langley-Evans et al., 1996). Postnatal maternal behavior has been shown to modify the methylation status of the glucocorticoid receptor (Weaver et al., 2004) and estrogen receptor (Champagne et al., 2006), and differences in maternal care between the groups are possible (i.e., while the direct effect of protein malnutrition on the maternal behavior has not been reported, protein-deprived rat pups nursed by control dams experienced an increase in active nursing and grooming (Massaro et al., 1974)). Defining the causative factor(s) for the observed phenotypes remains an area of active research. As these experiments represented an initial evaluation of maternal low protein diet on neurobehavioral outcomes in the offspring, experiments were necessarily limited with regard to age of assessment (adults) and sex (males). Future work can now be directed at defining other key variables. For example, an examination of the critical period for exposure to low protein diet (e.g., early versus late gestation, inclusion or exclusion of lactation) will provide important information on specific windows of vulnerability. Further, future studies may extend the analyses to female offspring, as certain developmental programming effects have been shown to be sex-specific (Skinner et al., 2008) or other assessment timepoints that represent particularly vulnerable ages (e.g., adolescence).

01