Even for defined syndromic neurodevelopmental disorders, with known biomarkers, introducing a causal mutation and the resulting pathophysiology within neurobiological substrates can be highly varied (see below). Perhaps this should not be surprising, given the clinical heterogeneity of disorders defined by single gene mutations. However, one view of the strength of model systems is the ability to provide phenotypic consistency and reproducibility while deciphering biological mechanisms. Yet there remain practical pressures of defining more biomarkers, particularly for disorders in which such features currently are lacking (psychosis, autism), and testing their validity as disorder-causing in model systems. How, then, can advances in pathogenesis be made via the basic-clinical interface? We suggest that addressing how risk factors, both genetic and environmental, modify the fundamental, evolutionary conserved process of hierarchical circuit and skill development is ripe for discovery. This also has the possibility of leading to the identification of sub-clinical, intermediate endophenotypes which, when translated into clinical populations, may be useful in assessing developmental trajectory and intervention outcomes, either as treatments that redirect development or as preventive measures. For example, in early ASD interventions, targeted environmental enrichment is used to positively influence socially relevant sensory information processing that is critical for later developing social cognition (Vismara and Rogers, 2010).

The example referenced above reflects a strategy of addressing core pathogenic mechanisms of neurodevelopmentally based disorders through investigation of the developmental assembly and integrity of both architectural and ultimately functional hierarchies. This approach acknowledges a fundamental property of the nervous system – that basic circuitry components that underlie the organisms ability to utilize sensory information from its environment to respond appropriately and maintain homeostatic balance are built first, followed by an integration of this basic information into complex circuit hierarchies that process demanding information (Hammock and Levitt, 2006). In fact, this view is supported more by functional studies rather than neurodevelopmental analyses of complex circuitry. Simply stated, we know little about the development of circuit hierarchies from neuroanatomical and molecular perspectives. Instead, data are much more robust regarding functional, fundamental sensory and motor development across vertebrates, and emotional and cognitive development in humans. The hierarchical nature of the emergence of primary sensory modalities, language and then top-down frontal cortical control (executive function) in humans is well reviewed (Fox et al., 2010). In the structural and molecular domains, however, there are few studies that have examined directly the development of key circuitry involved in higher order information processing. For example knowledge regarding the ontogeny of cortico-thalamic (Auladell et al., 2000; Torii and Levitt, 2005), and cortico-striatal connectivity (Sharpe and Tepper, 1998) (not simply axon penetration into targets) is minimal, despite existing techniques that could probe these connections. Yet the cortex provides the vast majority of the synapses to these sub-cortical structures (Jones, 2002). Using pseudorabies virus tracing to monitor the onset of synapse formation within the central circuitry that is involved in integrating autonomic and hypothalamo-pituitary-adrenal (HPA) functions, we showed that brainstem to peripheral structure connectivity, and hypothalamus to brainstem connectivity are assembled in the rodent by birth (Rinaman et al., 2000). While the descending axonal input to brainstem targets from the amygdala and frontal cortex are present during the first few days after birth, the onset of synaptogenesis is delayed by almost one week (Rinaman et al., 2000). This coincides with the period of time when maternal influences on pup physiology and behavior are most profound (Moriceau et al., 2006; Shionoya et al., 2007). In fact, introduction of stressors can change the timing of onset of synapse formation in the higher order, later developing circuits, while having no impact on the basic hypothalamo-brainstem connections (Banihashemi and Rinaman, 2010; Card et al., 2005). Molecular physiology methods can be adapted for examining developmental trajectory of circuitry. For example, lentivirustransduced channelrhodopsin could be used to selectively activate anatomically defined circuits. This would allow monitoring of postsynaptic responsiveness, and perhaps even adaptations following particular interventions in animal models (Cruikshank et al., 2010). The improved resolution of methods used for live imaging provides opportunities to examine circuit function and plasticity over developmental epochs (Cruz-Martin et al., 2010; Trachtenberg et al., 2002). These approaches provide opportunities to reveal the importance of timing in terms of disrupting the hierarchical nature of circuit development, and the analysis of the basic elements that comprise more complex behavioral dimensions such as social cognition and executive function (Hammock and Levitt, 2006). Early alterations to the circuitry underlying a specific functional modality can lead to more widespread disruptions (e.g. autonomic nervous system disruption and systemic dysautonomia (Axelrod et al., 2002), whereas later perturbations may lead to more specific disruptions within a particular skill set, but may not necessarily have a broad impact (e.g. postnatal toxic stressors and emotional dysregulation (Caspi et al., 2003; Plotsky and Meaney, 1993)). Thus, the same insult, but presented at different times during development, is likely to lead to very different outcomes (Figure 1) based on the relative degree of maturation of neural circuits (i.e. the breadth of connections) and the particular role played by a molecular component present during development compared to the adult. In fact, in animal studies examining structure-function effects of prenatal cocaine exposure, timing of exposure was one of several key variables in explaining variance across models (Stanwood and Levitt, 2008). Similarly, deletion of 5HT1_aR only during development results in very distinct adult-onset emotion regulation outcomes compared to deletion of the gene encoding the receptor in the adult mouse (Gross et al., 2002). How this temporally circumscribed manipulation results in long-term changes remains unknown. One hypothesis is that this receptor modulates the assembly of relevant circuits during development, and then switches its role as a traditional biogenic aminergic neurotransmitter in the adult. Therefore, disrupting this receptor during development could directly alter neural circuitry. There is precedence for 5-HT serving this novel developmental role (Bonnin et al., 2007), but relevant studies examining the ontogeny of amygdala and related circuits in the 5HT1_aR conditional null mouse should assess this directly.

The developmental impact of allostatic load, however, is under-investigated, and may be quite different from what is reported in the adult (Lupien et al., 2009; McEwen, 2001). Chronic high levels of glucocorticoids in the adult rodent or non-human primate can permanently damage prefrontal cortical and hippocampal neurons (Radley et al., 2004; Sapolsky et al., 1985; Uno et al., 1989). During development, this same challenge may also alter the fundamental wiring of circuits that mediate the stress response (Plotsky and Meaney, 1993; Wei et al., 2007). Thus, in the context of adult-onset disorders that originate in development, allostatic overload may result in an altered developmental trajectory (Thompson and Levitt, 2010). This changed trajectory alone, or in combination with other environmental challenges, could possibly lead to a diseased/disordered system, for example schizophrenia. As such, behavioral and etiological heterogeneity within each disorder is easier to understand, as disorder onset would require multiple events (environmental, genetic, interactions) to occur, and does not necessitate identical causality, nor the identical trajectory of biological development even within a single clinical population. Mechanistically, this has emerged through the long-held concept of ‘early life programming’, with the growth of epigenetic studies examining the permanent and transient nature of genomic modifications (Bale et al., 2010; Weaver et al., 2004). The concept of early allostatic overload is not without precedence. For example, allostatic overload is likely to occur in premature onset of adult hypertension, where reduction in the number of kidney glomeruli, established prenatally, is accompanied by a maladaptive pathological increase in glomerular size, correlating with disease vulnerability (Keller et al., 2003). Rather than focusing on end state, examining developmental changes following genetic and/or environmental perturbations in animal models will provide fundamental information regarding the maladaptive nature of the fundamental molecular and structural architecture in the brain that underlies a particular disorder pathophysiology, especially for childhood-onset disorders. This also provides opportunities to examine normal system interactions as skills are built, thus introducing more naturalistic, integrated strategies.

Adding to the complexity of disorder heterogeneity based on clinical observation, is clear evidence for multiple genetic mechanisms (Geschwind, 2009) contributing to idiopathic disorders such as ASD (see review by State, this volume) and schizophrenia (Owen et al., 2010). Both very small or modest contributors to disorders, like common polymorphisms (Campbell et al., 2006) (Anney et al., 2010; Stefansson et al., 2009; Wang et al., 2009), and potentially major contributors, such as rare coding mutations and copy number variations (Christian et al., 2008; Glessner et al., 2009; Merikangas et al., 2009; Sebat et al., 2007; Weiss et al., 2008) are themselves heterogeneous in disorder impact. Common risk alleles are, by themselves, only moderators of physiological states. Moreover, there are now several descriptions in which well-defined mutations present in neurotypical individuals within a family that also include an affected person with the mutation (Sebat et al., 2007). Emerging from these genetic studies is a perspective emphasizing the importance of defining the molecular and neurobiological pathways that are disrupted (Akil et al., 2010; Bill and Geschwind, 2009; El-Fishawy and State, 2010; Levitt and Campbell, 2009). This parallels an initial conceptualization, put forth in previous gene microarray studies (Mirnics et al., 2000; Mirnics et al., 2001), of targeting defined molecular networks that are specific to a particular disorder, though achievable through different genetic and environmental mechanisms across subjects, that therefore introduces clinical heterogeneity.

Modeling a biological process that is disrupted in the disorder, rather than producing ‘the’ clinical model is achievable. Nine factors, not exhaustive in nature, are presented here as considerations for the design and utilization of experimental animal models of neurodevelopmental disorders (Figure 3). This is presented in the spirit of moving away from the need to design models that mimic the diagnostic criteria of an early- or late-onset disorder, to instead, approaches that may provide more relevant experimental data, and thus, inform better intervention designs. Moreover, this allows for individual behaviors to be studied, which when disrupted, are often shared across multiple neurodevelopmental disorders. Understanding the developmental disruptions that lead to aberrant behavior provides insight into common circuit problems and possible treatment strategies that can be shared across these disorders. The nine considerations: 1) ethologically appropriate behavior. Insight can be gained either from probing functions that are conserved across species boundaries (e.g. sensorimotor gating (Powell et al., 2009)) or common functions that are expressed in distinct, species-relevant ways. Regarding the latter, for example, unique aspects of maternal care, use of particular sensory modalities, or unique predator-prey relationships are features that can be used to an advantage in designing models. The stressors introduced by intermittent removal of offspring from the dam reflect careful design of an environmental factor that takes advantage of species-unique characteristics (Mendoza et al., 1978; Plotsky and Meaney, 1993; Suchecki et al., 1993). These studies in turn have led to common themes of epigenetic modulation of the impact of early stress on brain development and function in animal models (Weaver et al., 2004) and humans (McGowan et al., 2009); 2) the presence of species-specific genetic modifiers. The emergence of more detailed information regarding genetic modulatory factors will lead to better model design to probe similarities and differences between a clinical dimension of a disorder and phenotypes in models using the same mutations. For example, the rare coding mutation in the human serotonin transporter (SERT) associated with ASD and co-morbid with obsessive-compulsive behavior is an intrinsic variant to the C57BL6^J mouse (Carneiro et al., 2009), which may impact specific behavioral traits of the wild type mouse. However, while this strain is used in the majority of gene targeting studies, and the presence of the SERT variant could impact phenotypes, the wild type strain is not used as a model of either ASD or OCD; 3) species-specific timing of histogenic events. Cortical synaptogenesis, for example, begins in earnest during the third trimester in primate species, but postnatally in rodents (Thompson et al., 2009). This detail impacts the design of environmental manipulations for perturbing this developmental process; 4) species dependent regulation and patterning of gene expression. Information on transcriptional regulation of risk genes, detailed spatio-temporal developmental expression patterns, and their genetic modifiers provides a foundation for relating genetic manipulations in animal models to human. However, the limited data to date poses problems for designing appropriate studies. Constitutive or conditional introduction of gene mutations, disorder-specific rare variants, manipulation of expression via RNAi or shRNA transfer, or viral expression vectors all promote the examination of genetic mechanisms that may underlie a pathophysiological state. Each technique has strengths and weaknesses, and none, even those that introduce a clinically-defined mutation, need to phenocopy the disorder to make valuable contributions to understanding neurodevelopmental pathogenic processes; 5) species-specific representation of circuits and their individual components. The debate regarding the presence or absence of prefrontal cortex in rodents notwithstanding, consideration of structure-function relations from a developmental perspective enhances animal model studies. Recent history however reflects oversight of this consideration, for genes are introduced, behaviors measured, and drugs given to reverse one or two elements of a phenotype. Studies of neurodevelopmental disorders incorporating relationships between neuroanatomical substrates may provide more comprehensive intervention strategies, specifically targeting these circuits and substrates using pharmacological and experience-based approaches; 6) species-relevant environmental impact. Similar to selecting ethologically appropriate behavioral tasks putative environmental factors that impact developing systems should also be considered; 7) species-unique metabolic and homeostatic systems. Allostatic load and developmental homeostatic challenges are impacted by species-unique mechanisms responsible for controlling metabolic and physiologic states. This also can be used to an advantage in designing animal model studies; 8) species-unique neural and non-neural system interactions. Animal models could be designed to explore these relationships, which may be useful for understanding pathogenesis and treatment approaches. For example, there is substantial co-occurrence of seizures in multiple neurodevelopmental disorders (Gaitatzis et al., 2004; Matsuo et al.), though this particular pathophysiology may not be represented or focused upon in model systems designed to mirror the clinical disorder. Additionally, gastrointestinal disorders are prevalent among individuals with ASD (Buie et al., 2010; Campbell et al., 2009), but are not currently examined in animal models of ASD; 9) species-specific onset and regulation of puberty. This is specifically relevant for addressing sensitive and critical periods of development that transition through puberty, which, when perturbed can lead to adult-onset psychiatric disorders. Mechanistically, experimental approaches in animal models will assist in clarifying potential factors that influence post-pubertal pathophysiology emergence.

Trying not to cast too broad a net, once identified as particular risk or disorder-causing factors, we would argue that models in which specific genetic or environmental manipulations are introduced will be clinically relevant. The specificity and sensitivity of a particular model relies on defining ethologically relevant strategies for probing specific molecular and neural elements that contribute to phenotype. For example, deciphering the impact of a particular allelic variant in ASD on basic associative learning skills integral to more complex social and emotional regulation may provide novel insight that will lead to understanding the trajectory that a developing system may endure as a result of the disorder. The resulting long-term challenges to the circuits and behavioral hierarchies, discovered through animal model studies, can ultimately become targets of therapeutic strategies that engage the system at its most readily responsive time for clinical treatment or even prevention.