Developmental manipulation-induced changes in cognitive functioning

Abstract

Schizophrenia is a complex neurodevelopmental disorder with as-yet no identified cause. The use of animals has been critical to teasing apart the potential individual and intersecting roles of genetic and environmental risk factors in the development of schizophrenia. One way to recreate in animals the cognitive impairments seen in people with schizophrenia is to disrupt the prenatal or neonatal environment of laboratory rodent offspring. This approach can result in congruent perturbations in brain physiology, learning, memory, attention, and sensorimotor domains. Experimental designs utilizing such animal models have led to a greatly improved understanding of the biological mechanisms that could underlie the etiology and symptomology of schizophrenia, although there is still more to be discovered. The implementation of the Research and Domain Criterion (RDoC) has been critical in taking a more comprehensive approach to determining neural mechanisms underlying abnormal behavior in people with schizophrenia through its transdiagnostic approach toward targeting mechanisms rather than focusing on symptoms. Here, we describe several neurodevelopmental animal models of schizophrenia using an RDoC perspective approach. The implementation of animal models, combined with an RDoC framework, will bolster schizophrenia research leading to more targeted and likely effective therapeutic interventions leading to better patient outcomes.

Overview

Diagnostic Considerations

Schizophrenia is a complex neurodevelopmental disorder that is characterized by delusions, hallucinations, disorganized speech, and behavior, as well as negative symptoms such as social withdrawal and/or blunted affect (American Psychiatric Association, 2013; Substance Abuse and Mental Health Services Administration 2016; Patel et al., 2014). Schizophrenia is often first diagnosed in early adulthood and its chronic nature makes it a challenging disease to manage (Patel et al., 2014). The disease has a lifetime prevalence rate of approximately 0.5% (Simeone et al., 2015), and in addition to its high prevalence, schizophrenia is a leading cause of disability around the globe (Disease and Injury Incidence and Prevalence Collaborators, 2016).

The Diagnostic and Statistical Manual of Mental Disorders (DSM) has been the flagship source for the nosology for mental illness; however, the diagnostic criterion of schizophrenia has changed considerably with each edition of the DSM - indicative of the complexity associated with the etiology, symptomology, and treatment of this disorder (Tandon et al., 2013). Despite a broadening of the clinical description of schizophrenia, especially with the advent of the DSM-5, it still sharply contrasts with much of the neurobiological research that demonstrates considerable heterogeneity among patients with schizophrenia, ultimately underscoring the need for a multidimensional approach (Potkin et al., 2020; Cariaga-Martinez, Saiz-Ruiz & Alelú-Paz, 2016).

In favor of this need for a dynamic and comprehensive understanding of schizophrenia and other psychiatric disorders, the National Institute of Mental Health (NIMH) created the Research Domain Criteria (RDoC). This framework, in conjunction with the DSM-5 and similar classification manuals, provides a foundation for simultaneous investigation of pathological behavior and cognition and the subsequent molecular mechanisms that drive them. These biological mechanisms could include aberrations at the genomic, cellular, or neural circuit level (Young et al., 2017).

The goals of RDoC are to establish an understanding of the essence of mental health and illnesses across several different degrees of dysfunction and to study psychological constructs that are relevant to psychopathology (National Institutes of Health, n.d). The six domains of human functioning that it aims to study are negative valence, positive valence, cognitive systems, systems for social processes, arousal/regulatory systems, and sensorimotor systems (Figure 1; National Institutes of Health, n.d.; Cuthbert and Morris 2021). Furthermore, RDoC’s matrix of study is designed to evolve as novel constructs and domains are proposed and integrated (Carcone & Ruocco, 2017; National Institutes of Health, n.d).

Figure 1.

RDoC Matrix relating constructs to schizophrenia.

While there is still a strong preference for the use of DSM-5 in patient diagnoses, the evolution of the RDoC framework has fostered a movement to increase the types of transdiagnostic approaches used by providers (Cuthbert & Morris 2021). This movement has been bolstered by a general agreement within the scientific community that disorder classifications need to be revised (Kapadia, Desai & Parikh, 2020; Cerveri, Gesi, & Menacci, 2019; Cuthbert & Morris 2021). Furthermore, the implementation of RDoC has enabled the use of consistent methodologies in animal studies, allowing a more in-depth focus on the etiology and nature of dysfunction of those behaviors relevant to people with schizophrenia, and the direct assessment of their biological underpinnings (Cuthbert & Insel, 2013; Young et al., 2017).

Developmental Origins

Despite the promising potential of RDoC, an important limitation to this framework is that its initial concept is testing adults, thus it lacks a developmental component where external factors, such as early life adversity, interact with the trajectory of healthy brain development leading to psychiatric disease (Franklin et al., 2015). Indeed, adversity during early life is one major risk factor for developing schizophrenia. For example, epidemiological studies found that exposure to gestational infection increases offspring risk for developing psychiatric illnesses, including schizophrenia (Mahic et al., 2017; Meyer, 2019; Estes & McAllister 2016) and there is an association between birth season and schizophrenia development risk; people born in the early spring have the highest risk (Konrath et al 2016, Wang and Zhang 2017). In addition, exposure to stress after parturition (e.g., emotional, social, physical, or sexual), also increases risk for developing schizophrenia (Fachim et al., 2021; Matheson et al., 2013). This evidence warrants the incorporation of a developmental perspective into the RDoC framework, which could offer tremendous insight on when certain populations are most vulnerable to environmental insults, or when treatment intervention is most effective (McLaughlin & Gabard-Durnam, 2021).

A model is a representation of ideas or processes that are intended to be studied through the use of a manipulation and quantification of the manipulation’s outcomes. Because many of the cognitive and behavioral symptoms seen in people with schizophrenia can be modeled in small laboratory rodents, particularly following exposure to insults experienced during early development, researchers are able to explore the biological mechanisms underlying the intersection of early life adversity and the development of schizophrenia. A good test of an animal model is to evaluate whether it provides construct, face, and predictive validity. Construct validity refers to the mechanistic similarities between the disease and the animal model, face validity is the extent that the animal model looks like it is modeling the disease of interest, and predictive validity is the degree that the response to a therapeutic treatment is similar between the disease and the animal model. Using an RDoC perspective, we will discuss and critique several developmental manipulations used in the generation of animal models for schizophrenia-related behaviors. We will also provide a review of several behavioral and cognitive tests commonly used in the laboratory to test the translational utility of these models.

Commonly Used Animal Models of Schizophrenia-related Behaviors

There are several animal models that are used to understand mechanisms underlying the development of neuropsychiatric disorders such as schizophrenia. These animal models can be broken down into categories based on the timepoint in which the stressor is administrated. These include prenatal, neonatal, and combination models.

Prenatal Models:

Maternal Immune Activation (MIA) Model

Epidemiological research suggests that maternal immune activation (MIA) in pregnant women can result in prolonged and permanent changes in the behavior and brain function of their offspring (Brown et al., 2009; Estes & McAllister., 2016; Kentner et al., 2019a; Mednick et al., 1988). The immune responses that lead to these changes can be caused by a variety of pathogens and bacteria including Taxoplasma gondii, influenza, and rubella (Estes & McAllister 2016; Kentner et al., 2019a). Exposure to these pathogens in utero increases the risk for several disorders including autism and schizophrenia. In the animal laboratory, behaviors associated with schizophrenia and autism can be modelled by challenging pregnant rodents with immunogens. The two most commonly studied immunogens are lipopolysaccharide (LPS) and polyinosinic:polycytidylic acid [poly (I:C)] which are bacterial and viral mimetics, respectively (Brown et al., 2009; Arsenault et al., 2014). These MIA models are excellent for studying the mechanisms underlying the behavioral dysfunctions related to schizophrenia as they provide construct, face, and predictive validity (Estes and McAllister 2016, Kentner et al 2019a). While the MIA manipulation does not recreate the exact profile that occurs in people, it does offer insights into the long-term impact of such MIA that can be compared to changes seen in people with schizophrenia.

People with schizophrenia show abnormal dopamine regulation and prefrontal cortex function and most pharmacological treatments target dopamine receptors (Meyer-Lindenberg et al., 2002; Purves-Tyson et al., 2021). In parallel, several studies have documented alterations in such dopaminergic and serotonin signaling among MIA-infected mouse offspring (Winter et al., 2009; Meyer et al., 2008; Csatlosova et al., 2021). In addition, elevated immune-related markers, such as interleukin (IL)-6, SERPINA3, and IL-8, seen in MIA mice are also observed in people with schizophrenia (Purves-Tyson et al., 2021; Kneusel et al., 2014; Reisinger et al., 2015; Fillman et al., 2013; Borovcanin et al., 2017). Additionally, medial prefrontal cortex functioning is disrupted in adult MIA exposed mice, mediated through altered tyrosine hydroxylase expression, and lowered spontaneous firing rate within this brain region (Purves-Tyson et al 2021). MIA also results in reduced glutamate levels in the hippocampus of rats (Connors et al., 2014), also seen among people with schizophrenia (Howes et al., 2015). In fact, this observation is central to the glutamate hypothesis of schizophrenia and the premise of using glutamatergic-based treatments like bitopertin andminocycline which have shown some promise in clinical trials (Howes et al., 2015; Liu et al., 2014; Umbricht et al., 2014). Thus, despite some potential limitations, the MIA models recreate many of the structural abnormalities seen in people with schizophrenia.

Beyond structural changes, it has been consistently documented that MIA rodent offspring show many behaviors relating to schizophrenia and autism spectrum disorder. For example, offspring of rodents exposed to MIA exhibit decreased sensorimotor gating, increased amphetamine sensitivity, and disruptions in the latent inhibition task. Sensorimotor gating measures the animal’s ability to process/filter out irrelevant information (Braff et al., 2001; Purves-Tyson et al 2021). Prepulse inhibition (PPI; Figure 2), is used to study sensorimotor gating and is conducted through an auditory, visual, or tactile modality (Basu & Ray 2016; Powell et al., 2012). It measures the response of the animal to a stimulus, called the pulse, after the presentation of a smaller pulse, called the prepulse. Frontal dopaminergic pathways are disrupted when decreased sensorimotor gating is observed (Tan et al., 2018; Powell et al 2012). It is hypothesized that the prepulse induces an inhibitory mechanism which blocks the response of the subject to additional stimulation until the prepulse is processed completely in the brain (Basu & Ray, 2016). MIA causes a decrease in PPI as MIA-treated offspring are not able to filter out the extraneous information thus disrupting the response (Scarborough et al., 2020, Shi et al., 2003). Similarly, people with schizophrenia have shown significant deficits in information processing as measured by PPI (Cadenhead et al 2000). Moreover, the antipsychotic clozapine attenuates PPI deficits in rats prenatally exposed to MIA (Basta-Kaim et al., 2012). Attesting to this pharmacological predictive validity, antipsychotic medication also improves PPI in people with schizophrenia. However, PPI may be confounded by the class of antipsychotic prescribed in that typical antipsychotics such as haloperidol don’t improve PPI in humans but the atypical antipsychotics such as clozapine do. Moreover, the duration of pharmacological treatment at the time the PPI task is administered matters as those who have been on antipsychotics longer show more improvement in PPI compared to those who just started taking them (Kumari et al., 2002; Hedberg et al., 2021). Overall, PPI remains an inexpensive and potentially valuable tool to integrate into the RDoC matrix, given its transdiagnostic applicability and the abundance of research identifying its biological constituents (Pineles et al., 2016).

Figure 2. Prepulse inhibition.

An animal’s startle response to a sound (pulse) is reduced by the presentation of a less intense sound (prepulse). Using an acoustic startle chamber, researchers can measure the animal’s ability to inhibit its startle response (measured by average millivolts) during trials that contain a prepulse.

With latent inhibition (Figure 3), previous exposure to a stimulus makes it more difficult to make new associations with that same stimulus (Castagné, Moser, & Porsolt, 2009). MIA is generally associated with impairments in latent inhibition (Bikovsky et al., 2016; Haddad et al., 2020). In one experiment, researchers used white noise as the conditioned stimulus and an electric foot shock as the unconditioned stimulus (Bitanihirwe et al 2010). Half of the rats received a preexposure to the white noise and the other half did not. On conditioning day, after administering the white noise, rats would avoid the shock if they shuttled, which is the process of moving to a different compartment, or area of the chamber to avoid the shock punishment. If animals did not shuttle (reduced response latency), they would immediately receive the shock. MIA-exposed animals showed impairments in social interaction and abnormal latent inhibition was seen in males (Bitanihirwe et al., 2010). There was also a sex-dependent effect on the male offspring from the poly (I:C)-treated mothers. In this two-way active avoidance procedure, the male offspring from poly (I:C)-treated mothers displayed significantly enhanced latent inhibition effects (Bitanihirwe et al 2010). This effect shows heightened cognitive and behavioral inflexibility within the male, but not female, offspring (Bitanihirwe et al 2010). Latent inhibition is a form of associative learning (Lubow & Moore, 1953; Weiner, 1984), although some have described it as a task of attention (Lubow, 2005), making it closely applicable to the declarative memory subconstruct within the cognitive systems domain of RDoC.

Figure 3. Latent inhibition.

Preexposure to the conditioned stimulus (light) before pairing with a conditioned stimulus (footshock) reduces the fear response compared to animals that receive the conditioned and unconditioned stimulus pairing without preexposure to the conditioned stimulus.

Although reductions in PPI and latent inhibition are among the most consistent findings across studies that use the MIA model (Haddad et al., 2020), not all have demonstrated these deficits (Li et al., 2009; Meehan et al., 2017; Bitanihirwe et al 2010). Some important factors that may contribute to this discrepancy include the timing of immune challenge, offspring sex, offspring age, drug dose, the source of immunogen, as well as the animal strain/genetic background. In addition (Table 1), most MIA model studies use nonvirulent agents that activate the immune system (poly (I:C) for example). However, these do not invoke the natural level of immune response caused by infectious pathogens experienced in nature (Mueller et al., 2019). An additional limitation of the MIA model is that it does not capture potential confounds, such as stressful environments, photoperiods, or stress brought on by navigating the world differently because of a triggered immune response. These limitations make it difficult to model the role of broader factors in contributing to altered offspring development. Epidemiological studies in this case, are more robust as the human experience is built-in (Brown & Meyer 2018). Lastly, even with established protocols, immunogens may result in immune responses that are either too weak, too strong, or even potentially fatal which can lead to variability in this model (Kentner et al., 2019a; Roderick & Kentner, 2019; Chow et al., 2016).

Table 1.

Summary of developmental manipulations and cognitive effects.

Developmental manipulation	Cognitive effects	References
Maternal immune activation (MIA)	↓ PPI	(Scarborough et al., 2020; Shi et al., 2003; Cadenhead et al 2000; Haddad et al., 2020)
	↓ PPI (dose dependent)	(Meyer et al., 2005)
	No effect of MIA on PPI	(Li et al., 2009; Meehan et al., 2017)
	↓ Latent inhibition	(Bikovsky et al., 2016; Haddad et al., 2020)
	↑ Latent inhibition (males)	(Bitanihirwe et al., 2010)
Maternal Methylazoxymethanol Acetate Exposure (MAM)	↓ Performance on radial arm maze memory task	(Gourevitch et al., 2004)
	↓ Attention set-shifting	(Featherstone et al., 2007; Mar et al., 2017)
	↓ Spatial navigation of water maze	(Leng et al., 2005)
	↑ Premature responding on 5C-SRT	(Featherstone et al., 2007; Mar et al., 2017)
	↓ PPI	(Chalkiadaki et al., 2019; Hazane et al., 2009)
Prenatal psychosocial stress	↓ Novel object recognition (males)	(Markham et al., 2010)
	↓ PPI	(Koenig et al., 2005; Niu et al., 2020; Burton et al., 2007)
Maternal vitamin D deficiency	↓ Memory in brightness discrimination task	(Becker et al., 2005)
	↓ Performance on 5C-SRT and 5C-CPT	(Harms et al., 2012b; Turner et al., 2013)
Neonatal ventral hippocampal lesion (NVHL)	↓ PPI	(Brady, 2016)
	↓ Spatial navigation of radial arm maze	(Brady, 2016)
	↓ Performance on the continuous spatial delayed alternation task	(Maruki et al., 2001; Brady et al., 2010; Lipska et al., 2002; Kim & Frank, 2009)
	↓ Performance on the discrete paired-trial variable-delay task (time dependent)	(Lipska et al., 2002)
Maternal separation (MS)	↓ PPI	(Ellenbroek & Riva, 2003; Garner et al., 2007)
	↓ Reversal learning in Morris water maze	(Xue et al., 2013)
	↓ Spatial navigation in Morris water maze	(Zhu et al., 2010; Garner et al., 2007)
	No effect of MS on Morris water maze	(Enthoven et al., 2008)
	↑ Reversal learning in Morris water maze	(Mooney-Leber et al., 2020)
Postnatal dopamine agonism	↓ Spatial navigation in Morris water maze, elevated plus maze, and Cincinnati water maze	(Dawirs et al., 1996; Acevedo et al., 2007; Grace et al., 2010; Hrubá et al., 2010; Skelton et al.,2007; Vorhees et al., 2008, 2009)
	↓ Novel location and object recognition	(Siegel et al., 2011; Acevedo et al., 2007)
	↓ PPI	(Acevedo et al., 2007)
Postnatal NMDA receptor antagonism	↓ PPI	(Mouri et al.,2013; Lim et al., 2012; Neill et al., 2010; Plataki et al. 2021)
	↓ Fear conditioning	(Lim et al., 2012; Neill et al., 2010; Plataki et al., 2021)

Maternal Methylazoxymethanol Acetate Exposure (MAM) Model

In the maternal methylazoxymethanol acetate exposure (MAM) model, pregnant female rodents are treated with MAM, producing offspring that demonstrate schizophrenia-relevant behaviors (Flagstad et al., 2004; Featherstone et al 2007). MAM produces such effects in offspring because it is a neurotoxin that selectively targets neuroblasts and is typically administered on gestational day (GD) 16-17. GD17 in rodents coincides with the late second trimester in human gestation (Bitanihirwe et al 2010). MAM treatment results in offspring with cognitive deficits as seen in people with schizophrenia e.g., impairments in attention, long-term memory, working memory, and behavioral flexibility through the use of tests like the radial-arm maze task and water maze experiments (Gourevitch et al 2004, Featherstone et al 2007, Leng et al 2005). In addition to affecting similar cognitive domains as schizophrenia, MAM treatment leads to abnormal hippocampal and cortical morphology and function (Featherstone et al., 2007; Lavin et al., 2005). Importantly, the timing of MAM administration is critical to the effectiveness of this model. If given earlier, such as before GD15, MAM can produce widespread physiological changes in the brain, such as reduction in cortical thickness and overall brain weight (Di Fausto, Fiore, & Aloe, 2007; Gourevitch et al., 2004). These neuroanatomical changes result in damage that is too drastic to serve as a valid model of schizophrenia, highlighting the importance of the timing of the neurodevelopmental manipulation.

To study effects of MAM exposure on the brain and cognitive function, two translationally relevant tests are the attentional set-shifting task and the 5-choice serial reaction time task (5-CSRTT; Pezze et al., 2007; Cope, Powell, & Young, 2016). The attentional set-shifting task (Figure 4) requires the animal to discriminate between a relevant stimulus and an irrelevant stimulus to receive a reward, building the relevance of one dimension by repeated testing within that dimension driving the formation of an ‘attentional set’. Next, animals are required to shift that attentional set, where the previously relevant dimension becomes irrelevant, and a stimulus from the originally irrelevant dimension must be chosen in order to receive the reward (Birrell & Brown, 2000; Featherstone et al., 2007; Heisler et al., 2015). This test was designed to assay cognitive flexibility and held as analogous to the Wisconsin Card Sort Task (WCST; Barceló et al., 2000; Heisler et al., 2015), and more specifically the interdimensional/extradimensional task. The WCST reveals deficits in executive function in people with schizophrenia (Everett et al., 2001; Nieuwenstein, Aleman, & de Haan, 2001; Singh et al., 2017), supporting the translational relevance of attentional set-shifting task in the study of the cognitive mechanisms related to schizophrenia. Offspring of MAM-treated dams require more trials than control, saline-treated groups to learn an extradimensional shift in the attentional set-shifting task (Featherstone et al. 2007). Furthermore, they have difficulties in reversing discriminations that were previously acquired in the experiment (Featherstone et al. 2007). As the attentional set-shifting task is used to study cognitive function and flexibility, naturally it fits into the cognitive systems domain of RDoC.

Figure 4. Attentional set-shifting.

An animal learns to discriminate between two stimuli to receive a food reward. Then the animal must learn to discriminate between two new stimuli to receive the reward. During the test trials, animals are required to ‘shift’ their attention toward previously irrelevant stimuli.

The 5-choice serial reaction time task (5-CSRT; Figure 5) is an analog of Leonard’s (1959) choice reaction time task which was devised to test the impacts of stressors on human performance. Now, the 5-CSRT is widely used to study deficits in attention in animals (Robbins, 2002). In this task, animals receive a visual light stimulus at one of several locations. Animals must then go to the location while the light is on to receive a reward, generally a food reinforcement (Carli et al., 1983). The location of the light will change every trial. The 5-CSRT recruits several parts of the brain including the prefrontal cortex (Liu et al., 2019; Robbins, 2002). Similar to the 5-CSRT in rodent models is the 5-choice continuous performance test (5C-CPT; Figure 6). In the 5C-CPT, animals are first trained in the 5-CSRT test (see Figure 5), as above. When moved to the 5C-CPT, some trials will have all five light stimuli turned on (Bhakta & Young, 2017), wherein the animal needs to inhibit from responding, recorded as a correct rejection (response inhibition), and rewarded. A response to any of the five lights is recorded as a false alarm and punished. Like the 5-CSRT, if the animal responds before any lights appear (no stimulus present), it is recorded as a premature response. Essentially, the 5C-CPT task requires response inhibition, unlike the 5-CSRT (Young et al., 2009; Bhakta & Young, 2017), because it includes non-target stimuli. As a result, researchers can better measure cognitive functions like response inhibition (Bhakta & Young 2017; McKenna et al., 2013; Young et al., 2009), and recruits brain regions consistent with human CPTs like the parietal cortex (Young et al, 2020; McKenna et al, 2013), not needed in the 5-CSRT (Muir et al, 1996), recreating the core aspects of CPTs used to assess attention in people with schizophrenia.

Figure 5. 5-choice serial reaction time task (5C-SRT).

Animals are provided a food reward for entering a hole with a light illuminated. Misses and premature responses trigger a timeout.

Figure 6. 5-choice continuous performance test (5C-CPT).

Animals are first trained in the 5C-SRT test (see Figure 5). In this task, animals must also demonstrate inhibition, by not approaching any of the lights when all of the lights are illuminated. This is recorded as a “correct rejection,” which is rewarded.

In the attention set-shifting task, MAM-treated rats require more attempts to learn an extradimensional shift, have higher errors, and have more difficulty in reversing previously learned associations compared to the control group (Featherstone et al., 2007; Mar et al., 2017). A higher error rate would mean that the animals display reduced cognitive flexibility. On the 5-CSRT, MAM-treated animals show increased premature responding compared to the control group (Featherstone et al., 2007; Mar et al., 2017) suggestive of motor impulsivity and/or poor temporal perception (Cope et al., 2016), but no deficits in attentional measures – therefore assessment in the 5C-CPT is warranted. Given that people with schizophrenia do not show deficits in the 5-CSRT but do so in the 5C-CPT, the latter may provide more finite measurement of schizophrenia-relevant performance. In post-mortem analyses, MAM-treated animals had decreased tissue weight of the prefrontal cortex, dorsal striatum, parietal cortex, and the hippocampus (Featherstone et al., 2007; Haijma et al., 2013; Nelson et al., 1998; Ohtani et al., 2014; Olabi et al., 2011; Veijola et a.,l 2014). These changes are analogous to post-mortem studies in humans with schizophrenia (Arnold et al., 2015; Gaser et al., 2004; Rajarethinam et al., 2007). Lastly, similar to the MIA model, MAM-treated rats display reduced PPI of the startle reflex (Chalkiadaki et al 2019, Hazane et al 2009). The MAM model is valuable for the study of behavioral and cognitive deficits that run in parallel to schizophrenia patients in terms of phenotypic expression (Table 1) and underlying mechanisms. However, to connect the experimental MAM-treated rodent data with human studies, complex effective connective modeling is required and the roles of neurotransmitters (GABA, dopamine, etc.,) can generally only be inferred (Modinos et al., 2015). Moreover, it should be noted that MAM is a potentially potent toxin and carcinogen and so researchers should exercise extreme caution and limit their exposure to MAM.

Prenatal Psychological Stress Models

Research shows that psychological stress experienced during gestation is a risk factor for impaired cognitive functioning and schizophrenia in offspring in later life (Khashan et al., 2008; Malaspina et al., 2008). In fact, exposure to prenatal psychological stressors, like natural disasters during pregnancy, can result in lower cognitive and language skills, increased risk for depression, and may increase risk of developing schizophrenia (Antonelli et al., 2017; Laplante et al., 2008; Watson et al., 1999). In the animal laboratory, prenatal psychological stressors also lead to cognitive impairments in offspring (Lemaire et al., 2000; Markham et al., 2010, Szuran et al., 2000). Examples of prenatal stressors used in animal models include restraint or predator odor which are used to induce both acute (e.g., a short term or single stressor) and chronic (e.g., multiple, or long-term stress exposure) stress across gestation. In chronic stress paradigms, the stressors applied can be homotypic (e.g., using the same stressor repeatedly) or heterotrophic (e.g., using variable or different types of stressors). A commonly used test for evaluating cognitive function following prenatal stress is the novel object recognition task (Figure 7). This task determines whether an animal can distinguish between a novel and familiar object. In this test, a rodent is placed inside of an arena with two different objects equidistant from each other. In the introduction phase, an animal is given a timed duration (e.g., 5 minutes or 10 minutes) to investigate both objects. Following this presentation, the animal is removed from the arena and returned to its home cage. The investigators then replace one of the two familiar objects with a new (novel) one. After a timed delay (e.g., 3 min, 30 min, 1 hour, 24 hours), the animal is placed back into the arena and the amount of time they spend investigating the novel versus the familiar object is recorded, referred to as the recognition phase. When interpreting the results, if the animal spends more time investigating the new object than chance, it is thought to relate to memory and learning (Denninger, Smith, & Kirby, 2018; Markham et al., 2010). Using this task, male rodent offspring exposed to repeated variable stress during the final week of gestation were found to have an impaired ability to recognize objects, even as adults. Prenatal stress did not impair object recognition in female animals (Markham et al., 2010). This result differs from studies which show schizophrenia is present in men and women, however, schizophrenia prevalence is higher among men and merits further research. (Li et al., 2016; Ochoa et al., 2012). While this task more broadly falls into the RDoC domain of cognitive systems (Vengeliene et al., 2017), it is important to acknowledge that the variation in time delays and lack of human counterpart for adults has made the translational validity of this task difficult to determine (Young et al., 2012).

Figure 7. Novel object recognition.

The animal is first presented with two of the same stimuli for the duration of the habituation period (e.g., 5 or 10 minutes). One of these stimuli is then replaced with a novel object after a predetermined delay period (e.g., 3 min, 30 min, 1 hour, 24 hours) and the amount of time the animal spends exploring the novel object during the testing period is measured.

In rodents, prenatal stress also impairs sensorimotor gating and development of the hypothalamic-pituitary-adrenal (HPA) axis, effects which are commonly found among schizophrenia patients (Kinnunen et al., 2003; Markham et al., 2010). The HPA axis is key in our bodies regulation to stress through secretions of corticotropin releasing factor, adrenal corticotropin releasing hormone, cortisol, and other hormones through a negative feedback process (Smith and Vale 2006). Cognitive disruptions in hippocampal processes following prenatal stress have been tied to reductions in hippocampal neurogenesis (Couch et al., 2021; Lemaire et al., 2000; Markham et al., 2010). In addition to damage to the hippocampus, impaired performance on fear conditioning tests (Figure 8) points to damage or altered functioning in the amygdala (LeDoux, 2003; Park & Chung, 2020). These are associative learning tests where animals learn to associate a neutral conditioned stimulus (e.g., a light or a tone) with an aversive unconditioned stimulus (e.g., a shock) to show a conditioned response (e.g., freezing behavior). Fear conditioning would most likely fall under the negative valence systems domain and acute threat construct of RDoC. Lastly, prenatal stress through restraint stress has been shown to decrease the PPI of acoustic startle responses in rodents (Koenig et al., 2005; Niu et al., 2020; Burton et al., 2007). Similarly, there has been a strong connection between HPA axis dysfunction (e.g., elevated cortisol secretion) and hippocampal function with schizophrenia severity (Walder et al 2000). These findings speak to the validity of such prenatal stressors for relevance to schizophrenia.

Figure 8. Fear conditioning.

A conditioned stimulus (light) does not trigger a fear response until after it is paired with an unconditioned stimulus (footshock). After this pairing, an animal learns to associate the light with the footshock and will demonstrate freezing behavior in response to the light only.

Diet/Nutritional Deficiency

Nutrition is one of the most important factors contributing to the health and development of mothers and their offspring. There have been studies indicating that calcium and vitamin D deficiencies (DVD) in the mother may result in central nervous system disorders in offspring, including schizophrenia, multiple sclerosis, and type 1 diabetes (Bordeleau et al., 2021; Hyppönen et al., 2001; Mirzaei et al., 2011). Vitamin D (VD) and calcium are critical across gestation as they are foundational for skeleton formation and calcification (Curtis et al., 2014). In addition to being an essential nutrient for bone development, VD is also crucial for brain, heart, immune, and reproductive health (Xue et al., 2016). Furthermore, VD has anti-inflammatory properties, where the depletion of VD receptor or hydroxylase Cyp27b1 can increase placental inflammation (Liu et al., 2011). Concomitantly, VD receptor agonists reduce the expression of immune cells in vitro (Vojinovic, 2014). Increased schizophrenia prevalence at higher latitudes, combined with the incidence of schizophrenia coinciding with summer and winter patterns (e.g., increased prevalence in births from December to March), provides epidemiological data in support of a DVD link to schizophrenia (McGrath et al., 2011; Cui et al., 2020). As a result of these observations – which include cross-sectional studies and longitudinal cohort studies – researchers have also been able to associate DVD to cognitive function, Parkinson’s disease, multiple sclerosis, and autism (Cui et al., 2015). Directly testing the relationship between VD and schizophrenia in humans is difficult given the ethical issues surrounding VD restriction during pregnancy. Whether or not VD supplementation reduces risk of developing schizophrenia in this population remains to be elucidated.

Animal models of nutritional deficiency have been critical to understanding the underlying mechanisms of DVD to cognitive functioning. For example, developmentally DVD rodents exhibit enlarged lateral ventricles, smaller hippocampi, and altered expression of apoptosis compared to control groups (Harms et al., 2012a; Eyles et al., 2003), which are common physiological patterns seen in psychiatric patients (Del Re et al., 2016; Ershova et al., 2017). Developmental DVD may also lead to disruptions to glutamate and dopamine system functioning (McGrath et al., 2010) as these DVD rats are hypersensitive to the locomotor-stimulating effects of MK-801 and amphetamine, which are both psychotomimetic drugs (Kesby et al., 2006; Kesby et al., 2010). Furthermore, prenatal DVD rats show impairments in learning and memory, as indicated by the brightness discrimination task, (Figure 9), along with damaged latent inhibition and attention, which are all features of schizophrenia (Becker et al., 2005; Harms et al., 2012b). For the brightness discrimination task, animals are trained to enter the illuminated alley of a Y-maze. Once the animal is trained, their ability to relearn this procedure is tested 24 hours later. This measure of memory recall fits into the Declarative Memory construct within the cognitive systems domain of RDoC.

Figure 9. Brightness discrimination task.

Animals are trained to enter the illuminated alley of a Y-maze. Entering an unilluminated alley triggers a foot shock. Once the animal is trained, their ability to relearn to enter the illuminated alley is tested 24 hours later.

One notable experiment tested attentional processing among developmental DVD rats using both the 5-CSRT and 5C-CPT (Turner et al., 2013). While not impaired in the 5-CSRT, adult offspring of DVD dams made more false alarm responses in the 5C-CPT and had more premature responses compared to controls (Turner et al., 2013). This finding suggests that DVD rats have impaired inhibitory control and increased impulsivity resembling the impaired CPT of people with schizophrenia (Turner et al., 2013; Table 1). Concomitantly, 5C-CPT validly translates to human populations, can be fMRI assessed (McKenna et al., 2013). People with schizophrenia exhibit deficits in the human task with EEG biomarker deficits and these neural correlates associated with the task are conserved from mice to humans (Young et al, 2013; Cavanagh et al., 2021). This evidence further supports the integration of 5C-CPT into the attention construct within the cognitive systems domain of RDoC (Cope et al., 2016).

The 5C-CPT requires activation in several brain areas including the inferior frontal cortex, presupplementary motor area, premotor cortex, inferior parietal lobe, basal ganglia, and thalamus (McKenna et al., 2013; Young et al., 2020). Developmentally DVD exposed animals exhibit NMDA receptor and dopamine dysfunction within these brain regions in concert with increased motoric impulsivity and response disinhibition in the 5C-CPT (Kesby et al., 2017; Kesby et al., 2006). Interestingly, DVD rats also exhibited deficits in novel object recognition (Overeem et al., 2019) and spatial navigation (Al-Harbi et al., 2017), but did not display altered PPI or delay-match-to-sample (Burne et al., 2014). Overall, developmental DVD may more reliably model symptoms of schizophrenia associated with deficits in locomotor function compared to those related to working memory or sensorimotor gating.

Early Postnatal Developmental Models

Neonatal Hippocampal Lesion Model

The neonatal hippocampal lesion model, also referred to as the neonatal ventral hippocampal lesion (NVHL) model, is a surgically induced developmental model for studying phenotypes that correlate to schizophrenia symptoms. NVHL recapitulates several neurological and anatomical impairments commonly seen in people with schizophrenia. To induce this model, an excitotoxin, generally ibotenic acid, is injected into the hippocampus of neonatal animals causing several behavioral abnormalities (Brady, 2016). This excitotoxin is generally administered to the rodent hippocampus at the end of the first postnatal week, which is analogous with late third trimester or perinatal period of human development (Tseng et al., 2009). When administered, the excitotoxin disrupts the development of hippocampal pathways affecting behavior. Behavioral abnormalities include reduced PPI and impairments in spatial working memory of the radial arm maze (RAM; Brady et al., 2010; Figure 10). In the RAM task, animals are trained to enter arms of a maze that are baited with a food reward. The un-baited arms are barricaded. After training, the test phase is followed by a predetermined delay (e.g., 1 minute, 60 minutes). During the test, entrances to all of the arms of the maze are opened, and the animal is timed for how long it takes to navigate to all of the food rewards, with working memory errors equated to entering already visited arms (Brown & Giumetti, 2006). NVHL results in decreased GAD67 mRNA expression, decreased NAA (N-acetylasparate) levels, and increased glutamatergic neurotransmission (Lipska et al., 2002; Lipska & Weinberger, 2000; Modinos et al., 2015; Roffman et al., 2000; Schroeder et al., 1999).

Figure 10. Radial arm maze.

During the training phase, animals enter the arms of a maze that are baited with a food reward. The un-baited arms are blocked. The test phase is followed by a predetermined delay (e.g., 1 minute, 60 minutes). During the test, entrances to all of the arms of the maze are opened, and the latency to retrieve food rewards, number of incorrect, and correct entries are recorded.

NVHL rats have impaired performance in the continuous spatial delayed alternation and the discrete paired-trial variable-delay tasks (Brito et al., 1982; Aultman & Moghaddam, 2001; Lipska et al 2002). In the continuous spatial delayed alternation task (Figure 11), researchers employ a “T” shaped maze with food rewards located at the ends of both arms of the T-maze. At the start of this task, an animal is habituated to the maze with full access to the food rewards in each goal arms of the maze. During the test, one arm of the T-maze is blocked, forcing the animal to access the food reward in the opposite arm. After a delay (e.g., 0, 10, or 60 seconds) the animal is given full access to the previously blocked arm. If the animal demonstrates intact spatial working memory, they will enter the previously blocked arm to collect the food. When the delay between rounds is extended past one minute (typically the upper limit of working memory), animals with hippocampus inactivation perform significantly worse indicating impaired spatial-working memory (Maruki et al., 2001). In the continuous spatial delayed alternation task, NVHL rats performed significantly worse compared to sham-operated rats (Brady et al., 2010; Lipska et al 2002; Kim & Frank, 2009). Interestingly, Lipska (2002) found control and NVHL rats had similar performance at the start of training until test day 11, however, the performance of control animals progressively improved over time. The inability for the neonatal lesion group to master this task shows an impaired acquisition and working memory (Lipska et al., 2002). Based on this pattern of deficits, the continuous spatial delayed alternation task would fall into the goal maintenance subdomain of the working memory construct.

Figure 11. Continuous spatial delayed alternation task.

After habituation, one of the arms to the T-maze is closed off, forcing the animal to collect the food reward from the open arm. Next, following a predetermined delay (e.g., 0, 10, or 60 seconds) the closed arm is opened, and the animal must spatially recall that this arm contains the remaining reward. A correct response is recorded when the animal enters the previously closed arm after the forced run trial.

In the discrete paired-trial variable-delay task (Figure 12), the animal is placed into a discrete paired-trial variable-delay T-maze. Next, the animal is put through a randomly chosen forced run - a maze designed to force the animal to turn right or left after being placed inside the maze - following a delay in the home cage. Then, the animal is placed back into the maze and allowed to go through the maze. Every day, the randomly chosen pattern changes but the pattern stays constant the whole day (Right-Right, Left-Right, Right-Left, or Left-Left). The delays are also variable at 1 second, 10 seconds, and 40 seconds (Papaleo et al., 2008; Leggio et al., 2020; Lipska et al 2002). Although, all the animals reached the 80% criterion of choice accuracy, NVHL animals performed worse than sham-operated rats during larger delay periods (Lipska et al., 2002), suggesting their goal-oriented working memory was compromised (Table 1). Both the continuous spatial delayed alternation task and discrete paired-trial variable-delay task fall under the goal maintenance working memory construct of the cognitive systems domain of the RDoC framework, and their mechanistic deconstruction could illuminate the biological constituents of schizophrenia symptomology.

Figure 12. Discrete paired-trial variable-delay task.

This test runs like the continuous spatial delayed alternation task, except experimenters can evaluate the strength of an animal’s memory over time. Animals are first trained to learn specific patterns of forced runs (e.g., left-right). Subjects are then tested on their memories of the patterns following varying lengths of delay (e.g., 1, 10, or 40 seconds).

Many of the anatomical changes related to cognitive disruptions that are induced by NVHL occur in the prefrontal cortex which is a key brain region implicated in schizophrenia (Brady et al., 2010; Placek et al., 2013). Thus, rather than inducing localized cell death, which lacks the ability to assess the role of cross-regional networks, the excitotoxin damages pathways to the prefrontal cortex and forebrain (Brady, 2016). Furthermore, the cognitive effects of the NVHL model temporally matches the onset of schizophrenia in that abnormalities do not generally manifest until the adolescence or early adulthood (Brady, 2016; Gogtay et al., 2011; Häfner, 2019).

There are some disadvantages to the NVHL model. First, only 70% to 80% of all animals may produce a lesion and disorganization or cell loss following exposure to ibotenic acid (Brady, 2016). This form of stereotaxic surgery requires heavy exposure to anesthetics and triggers an immune response which may explain the 70-80% success rate (Kirby et al., 2012). This model may also be influenced by environmental factors (Chambers & Sentir, 2019), which can lead to variability between and within species. Lastly, the construct validity of all lesion models is questionable due to the lack of environmentally induced lesions present in people with schizophrenia (Modinos et al., 2015).

Psychosocial Stress in Neonates

Childhood neglect and abuse are major predictors of schizophrenia in humans (Bennouna-Greene et al., 2011; Kraan, et al., 2015; Rokita et al., 2020; Matheson et al., 2013). The maternal separation (MS) model is a commonly used method to induce early life stress in laboratory animals (Murthy & Gould 2018; Solas et al., 2010). In this model, a mother and her offspring are separated for a specified duration of time (e.g., minutes or hours) daily across several postnatal days or weeks (Kalinichev et al., 2002; Solas et al., 2010). MS impaired PPI and sensory gating (Ellenbroek & Riva, 2003; Table 1) consistent to people with schizophrenia. MS also induced depressive-like behaviors, anxiety-like behaviors, and other cognitive impairments in offspring (Kalinichev et al., 2002; Romeo et al., 2003; Solas et al., 2010), suggesting that MS may not exclusively model schizophrenia and may be an effective model of other disorders. Indeed, some researchers have also used MS to model certain aspects of depression (Solas et al., 2010).

In the Morris Water Maze (MWM; Figure 13), researchers utilize a rodents’ aversion to water to test their spatial memory and learning (Bromley-Brits et al., 2011; Vorhees & Williams, 2006). The MWM is the dominant form of testing to assess allocentric navigation, navigation using external cues and landmark in relation to each other to navigate, in animals (Vorhees and Williams 2016). A small pool is filled with water, made opaque, and a submerged platform is placed in a certain location. Then, the rodent is brought to the environment where the pool is located to acclimate to the room. After acclimation, the animal is placed in the water, requiring a fixed time to locate and mount the platform (escaping the water). If the animal does not do this within the allotted time (around 60 seconds), the researcher places the animal on top of the platform and allows it to stand on the platform for several seconds. After this, the animal is returned to its cage. This training is repeated several times a day, for several days, wherein the animal gets faster at locating the platform. A probe trial is then conducted wherein the platform is removed from the pool and the time the animal spends in the quadrant of where the platform used to be is used as a measure of their memory (Vorhees & Williams, 2006). The initial trials test the animal’s acquisition ability while the probe trial assesses reference memory. For example, if the animal travels to the original location where the platform used to be relatively quickly, it may be presumed that the animal’s reference memory is intact (Vorhees & Williams, 2006). This test is relatively simple, inexpensive, and can provide results in as little as a week. As a result, it is commonly used to study learning and memory deficits in animal disease models including those for schizophrenia (Tian et al., 2019; Ning et al., 2017). There is even a virtual analog of the MWM that has been developed for humans (Fajnerova et al., 2014). The MWM falls into the declarative memory construct within the cognitive systems domain of the RDoC matrix, given the relational representation formed between the allocentric landmark and the location of the platform.

Figure 13. Morris water maze.

After several trials, the animal learns to swim to a platform it cannot see. For the test, the platform is removed and the experimenter times how long it takes the animal to swim to the relative location of where the platform was previously.

Repeated postnatal maternal separation leads to lasting impacts on rat performance on various tests, but initial learning is intact in the MWM (Xue et al., 2013). Impaired spatial reversal learning ability (e.g., when the platform is located in a new position and the rats are retested to determine how long it takes them to learn the new location of the platform) was observed in the MWM as MS-rats spent more time looking for the platform compared to the control group (Xue et al., 2013). There are ambiguous results regarding the impact of MS on spatial learning. Some studies have shown that MS impaired spatial learning and PPI (Garner et al., 2007; Zhu et al., 2010). Other work shows unaffected spatial learning and long-term memory (Enthoven et al., 2008) or even enhanced reversal learning when the platform was placed in a new quadrant, which indicates improved cognitive flexibility (Mooney-Leber et al., 2020). These variations in results may be due to the frequency and length of time involved in the maternal separation, strain differences, variations in the MWM set up and experimental conditions (i.e., varying trial intervals, length of training or timing of probe trials etc.) or other factors. Inter-laboratory differences in experimental MS protocols makes reproducibility a challenge across these studies. Lastly, MS-related deficits in reversal learning may be explained by a reduction in medial prefrontal cortex BDNF – a protein involved in neurodevelopment, neuroplasticity, learning, and memory (Xue et al., 2013; Récamier-Carballo et al., 2017; Roceri et al., 2004). This finding is important as inadequate BDNF may play a role in the neurodevelopmental impairments and molecular mechanisms seen in people with schizophrenia (Nieto et al., 2013). Furthermore, schizophrenia patients also show impaired reversal learning which is similar with the MWM utilized in the animal models.

The cognitive deficits commonly seen among people with neurodevelopmental disorders, like schizophrenia, can be seen using the MS model in conjunction with tests like the MWM, forced swimming test, and object recognition testing (Table 1). While the MS model provides an inexpensive, simple, and effective way to invoke depressive- and cognitive deficit-like symptoms, it does come with a major drawback. As noted, there is not a singular, well-defined MS model. Most MS models have varied times of separation and varied frequencies, standardization across laboratories may help improve reproducibility of findings.

Postnatal Drug Challenges

Dopamine Agonist Models (Amphetamines):

Animal models using postnatal drug administration have the ability to specifically target certain neurotransmitter systems that are implicated in schizophrenia-like phenotypes. Amphetamine-induced models of schizophrenia have been used for over 50 years, given that amphetamine overdose in humans can result in a stimulant-induced psychosis with a variety of behaviors consistent with schizophrenia (Snyder, 1973, Featherstone et al., 2007, Robinson and Becker, 1986). Importantly, we need to distinguish between animal models of adult exposure that typically use chronic low doses to mimic drug-induced psychosis (for review see: Grant et al., 2012, Jones et al., 2011) and models that use early-life developmental exposures, often via a single high dose or brief exposures to disturb the maturation of essential monoamine neurotransmitter systems. Though one may argue that the postnatal administration of high doses of amphetamines has low construct validity, the resulting phenotype does indeed resemble many of the pathological and behavioral changes we observe in people with schizophrenia and thus these models have high face validity. Both amphetamine and methamphetamine (METH) are transporter inhibitors and dopamine releasers that cause an excess release of dopamine into the synaptic cleft, with METH being more potent and faster acting compared to amphetamine (Goodwin et al., 2009). Especially during development, the excessive presence of dopamine results in neurotoxicity and the formation of reactive oxygen species, which in turn cause the degeneration of synaptic terminals (Ricaurte et al., 1982, Cadet and Brannock, 1998, Kita et al., 2003). In a gerbil model using a single high dose of METH on postnatal day 14 (PN14), dopaminergic innervation seems to decrease in prefrontal cortex areas while increasing in mesolimbic structures like the basolateral amygdala (Brummelte et al., 2008, Grund et al., 2007). This is in line with the ‘revised dopamine hypothesis’ that suggests an imbalance in the dopaminergic systems with hyperactive dopamine transmission in the mesolimbic and reduced dopamine transmission in the prefrontal cortex (PFC) in people with schizophrenia (Brisch et al., 2014). Importantly, postnatal METH exposure in this gerbil model changed several other neurotransmitter systems including serotonin, acetylcholine, GABA, and glutamate (Brummelte et al., 2007, Bagorda et al., 2006, Busche et al., 2006, Lehmann et al., 2004, Lehmann et al., 2003). For instance, postnatal METH exposure resulted in a shift in GABAergic innervation in the PFC with a reduction in somatic inhibition of pyramidal cells in layer V, but an increase in dendritic innervations in layers II/III (Brummelte et al., 2007, Nossoll et al., 1997). This pattern resembles reports of reduced GABAergic and parvalbumin-positive synapses on pyramidal cells in the postmortem brains of people with schizophrenia (Blum and Mann, 2002, Jahangir et al., 2021, Nahar et al., 2021, Pierri et al., 1999, Woo et al., 1998). Parvalbumin-positive interneurons are considered an essential RDoC element as they may be important for attention. In line with this, a decrease in parvalbumin-positive innervation and associated desynchronized activity in prefrontal cortices are commonly reported pathologies of schizophrenia, which may underlie issues with attention and related cognitive dysfunction such as working memory deficits (Spencer et al., 2004, Uhlhaas and Singer, 2006, Mathalon and Sohal, 2015).

In accordance with this, animal models of postnatal METH exposure frequently observe deficits in RDoC-relevant cognitive functions such as impairments in spatial learning and working memory in various mazes such as the elevated plus maze and water mazes (Dawirs et al., 1996, Acevedo et al., 2007, Grace et al., 2010, Hrubá et al., 2010, Skelton et al., 2007, Vorhees et al., 2009), novel location and novel object recognition (Siegel et al., 2011, Acevedo et al., 2007), and reduced sensorimotor gating in a pre-PPI tests (Acevedo et al., 2007). For instance, neonatal METH administration (with doses ranging from 10-25mg/kg, given 4/day) on PN6-15 or PN11–20 produced increased errors in the Cincinnati Water Maze at all doses with the earlier exposure having a stronger impact (i.e., more errors) (Vorhees et al., 2008, Vorhees et al., 2009). In conjunction with the behavioral similarities observed in cases of schizophrenia and METH-induced psychosis, MRI evidence has demonstrated comparable losses in hippocampal volume between these two groups (Orikabe et al., 2011). Therefore, METH-induced psychosis may function as a human pharmacological model of certain behavioral and physiological features of schizophrenia (Snyder, 1973).

The Cincinnati water maze (CWM; Figure 14) is a multiple T-maze that requires animals to use route-based egocentric navigation. Egocentric navigation in the CWM depends on striatal- and dopamine-dependent learning as well as internal cues and movement cues to find and remember the correct pathway. This contrasts with allocentric navigation, the main navigation strategy in the MWM, which depends on distal/spatial cues outside the organism and is thus more hippocampus-dependent (Vorhees and Williams, 2016). The procedures for testing animals in the CWM are very similar to those described above for the MWM with some slight modifications. Before being tested in the CWM, rats undergo a pre-maze assessment by performing 4 consecutive trials in a separate straight tube to practice finding the submerged platform. On the test day, rats are placed in the start position in the maze containing room temperature water and are allowed 5 min to find the platform at the end of the maze. Errors and latencies are recorded on each trial, with 2 trials a day for 5-6 days under normal light conditions or for 15-20 days in dark conditions (infrared light) (Vorhees and Williams, 2016). In the CWM, animals do not receive any “help” from the researcher, i.e., they are not placed on the platform if they cannot find it by themselves. Interestingly, opposite sex differences in learning across the MWM and CWM are observed. Females learn to find the correct pathway faster than males in the CWM, while males learn to find the platform faster than females in the MWM (Vorhees and Williams, 2016). This sex difference may reflect a dichotomy in the spatial strategy adopted to learn the tasks. Indeed, spatial learning strategy differs as a result of sex in rodents and gender in humans (Hawley et al., 2012; Yuan et al., 2019). Together this supports the importance of considering sex and gender differences in task learning and performance which could confound results regarding task performance in the broader context of schizophrenia.

Figure 14. Cincinnati water maze.

The animal is given 5 minutes to find the goal platform. The amount of time it takes for the animal to find the platform and number of entries made into the incorrect maze arms are recorded for each trial.

Interestingly, neonatal METH exposure from PN1-10 (i.e., right after birth) did not result in an increase in errors in the CWM (Vorhees et al., 2009), while later METH exposure (i.e. PN6-15 or PN11–20) did cause deficits in long-term path learning and spatial deficits (Vorhees et al., 2009). This difference suggests that the disturbances caused by METH depend on the development of the dopaminergic system and possibly the hippocampus (Siegel et al., 2011), which are slowly maturing during the postnatal period (Brummelte and Teuchert-Noodt, 2006, Dawirs et al., 1993, Shin et al., 2019, Tarazi and Baldessarini, 2000). Importantly, learning deficits in the CWM after postnatal METH-exposure are somewhat preventable with pre-administration of a dopamine D1 receptor antagonist (SCH23390) (Jablonski et al., 2019) or other drugs that preferentially affect dopamine, while drugs that primarily affect serotonin (i.e. MDMA and fenfluramine) did not impact performance in the CWM (Vorhees et al., 2011). This underlines the idea that the dopaminergic system plays a crucial role in METH-induced learning deficits in the CWM. Taken together, the CWM is a complex water maze that tests a different aspect of learning and memory (egocentric navigation) than other spatial navigation tests, which may be particularly relevant for the RDoC framework. Egocentric navigation requires some form of self-recognition and perception (i.e., knowing where oneself is in a certain space) and could relate to the social processes and cognitive systems domain of the RDoC framework. In line with this, egocentric navigation has been used as a tool for the assessment and treatment of cognitive deficits related to self-recognition in patients with schizophrenia (Siemerkus et al., 2012). When people with schizophrenia are put through a virtual maze without topographic markings and are required to use egocentric navigation techniques, they had an impaired performance in comparison to the control group (Siemerkus et al., 2012). Furthermore, schizophrenia symptoms often reflect impairments in social interactions and social context; virtual mazes and environments provide researchers with the ability to test and control social environments (Siemerkus et al., 2012).

N-methyl-D-aspartate (NMDA) Receptor Antagonist Models (Phencyclidine, ketamine, and MK-801)

Like amphetamines, schizophrenia research with N-methyl-D-aspartate receptor (NMDA-R) antagonists began from the observation that acute inhibition of the glutamatergic NMDA-R can exacerbate the positive and cognitive symptoms of schizophrenic patients and induce similar symptoms in healthy volunteers (Gilmour et al., 2012, Murray et al., 2013). Likewise, in rodents both NMDA-R antagonists and amphetamines can induce behaviors such as hyperlocomotion and increased stereotypic behavior which may be analogues of positive symptoms of schizophrenia in humans (Nakazawa et al., 2017). NMDA-R antagonists and amphetamines can promote cognitive-related symptoms, such as impaired PPI, working memory, cognitive flexibility, and attention (Mouri et al., 2013). However, NMDA-R antagonists are better able to model negative symptoms, such as social withdrawal, than amphetamines (Neill et al., 2010, Gilmour et al., 2012, Mouri et al., 2013, Lee and Zhou, 2019). The capacity of NMDA-R antagonists to model the three main categories of schizophrenia symptoms (i.e. positive, cognitive, and negative) suggests that dysregulation of glutamatergic activity may play a major role in the pathology of schizophrenia (Coyle et al., 2012). Notably, the inhibition of NMDA-Rs in neonates alters dopaminergic, serotonergic, noradrenergic, and GABAergic neurotransmitter systems which are all implicated in schizophrenia (as reviewed by Lim et al., 2012, Lee and Zhou, 2019).

In animal models of NMDA-R hypoactivity, subjects are injected with NMDA-R antagonists such as phencyclidine (PCP), ketamine, and MK-801. These compounds are often administered to subjects during the neonatal period (around postnatal days 1-15) to recreate the neurodevelopmental aspects of schizophrenia (Lee and Zhou, 2019). Neonatal exposure to NMDA-R antagonists impairs learning and memory, PPI, fear conditioning, and sociability in addition to decreasing parvalbumin-positive neurons and cortical volume while increasing neuronal apoptosis and stereotypic behaviors (Lim et al., 2012, Neill et al., 2010, Plataki et al., 2021; Table 1).

Although neonatal exposure to NMDA-R antagonists is intended to model the neurodevelopment of schizophrenia, Grayson et al. (2016) state that many outcomes from neonatal PCP exposure do not differ greatly from adult PCP exposure. Consistent with this, other NMDA-R antagonists decrease the number of parvalbumin-positive GABAergic neurons in the prefrontal cortex, alter other neurotransmitter systems, and impair learning and memory regardless of neonatal or adult administration (Lee and Zhou, 2019, Ghotbi Ravandi et al., 2019, Lim et al., 2012, Neill et al., 2010, Bubenikova-Valesova et al., 2008, Plataki et al., 2021). Grayson et al. (2016) argues that this lack of separation shows that current neonatal NMDA-R antagonist models may not adequately reproduce neurodevelopmental aspects of schizophrenia, but all models have their limitations and the postnatal administration of NMDA-R antagonists such as PCP to rodents can still successfully recreate some of the behavioral abnormalities linked to the underlying pathologies of schizophrenia.

There are other limitations to consider though when evaluating the validity and translational value of neonatal drug exposure models. Postnatal exposures to high levels of stimulants or NMDA-R antagonists are unlikely to happen in humans and thus these models have limited construct validity. However, infants may be exposed to amphetamines or other drugs prenatally through maternal transmission and this can lead to cognitive impairments in visual motor integration executive function, lower school achievements and changes in cortical and subcortical brain structures related to cognitive function in exposed children (Chang et al., 2009, Lester and Lagasse, 2010, Cernerud et al., 1996, Sanjari Moghaddam et al., 2021, Roos et al., 2015). Importantly, there is, to our knowledge, no link to an increased risk for schizophrenia (yet), though there is a lack of longitudinal outcome data on children exposed to METH or amphetamines in utero (Li et al., 2021). However, given the limitations of human research in this area (due to polydrug use, foster care, lack of longitudinal data, etc.), postnatal drug challenge animal models are an important part of continuing schizophrenia research as they can mimic many aspects of schizophrenia symptoms.

Combination Models

Animal laboratory researchers can combine multiple early life stress models to make more complex ‘two’ or even ‘three’ or more hit models (Bilbo et al., 2005; Strzelewicz et al., 2021; Strzelewicz et al., 2019). In one classic study, rats displayed memory impairments only if they received both a peripheral challenge with E. Coli on postnatal day 4 and a “second hit” LPS challenge in adulthood (Bilbo et al., 2005). Other models have utilized different combinations of immunogens (e.g., poly (I:C) and LPS administered during the prenatal and postnatal periods respectively; Li et al., 2018), postnatal methamphetamine challenges with altered rearing conditions (Brummelte et al., 2007, Lehmann et al., 2007; Vorhees et al. 2008), or even integrated prenatal immunogen exposure with peripubertal stress (Monte et al., 2017; Giovanoli et al., 2014). There is mounting epidemiological and animal model-based evidence demonstrating that multiple hits may better predict neuropsychiatric disease risk than a single environmental insult alone (Giovanoli et al., 2013; Giovanoli et al., 2014; Mahic et al., 2017; Meyer, 2019; Estes & McCallister, 2016). Ecologically relevant models such as the limited bedding model, which mimics low resource environments by providing rodent dams with reduced nesting and bedding material (Gilles et al., 1996; Ivy et al., 2008), have been combined with diet and other early life stress models to study the developmental impacts of increasing burden on mental health outcomes (Strzelewicz et al., 2021; Strzelewicz et al., 2019). Multidimensional models allow investigators to examine the simultaneous contribution and interaction of several risk factors implicated in the development and etiology of diseases such as schizophrenia. This approach provides support for the importance of studying these combination two- and multi-hit models in the animal laboratory (Maynard et al., 2001).

Translational Approaches: From Bench to Bedside

Pharmacological Interventions

Researchers use animal models to study cognitive, behavioral, and physiological impairments relevant to brain diseases to identify their underlying mechanisms. In addition to this, researchers can evaluate the safety and efficacy of potential treatments for these impairments. These animal models are called preclinical models. If a preclinical model can significantly mimic prominent features of a disease or disorder, it can help inform the development of clinically relevant treatments that can eventually be translated into patient care.

For example, studies make use of the rat MAM model to induce impairments in cognitive function like those seen in people with schizophrenia. Maternally exposed MAM offspring generally show a reduction in volume in various brain regions including the mediodorsal thalamus, hippocampus, prefrontal cortex, occipital cortex, and parahippocampal cortex (Matricon et al., 2010; Moore et al., 2006), see further details above. In addition, they show motor disturbances like dyskinesia, cognitive inflexibility, and sensorimotor gating deficits (Ratajczak et al., 2015; Ciofalo et al., 1971; Moore et al., 2006). This model has provided scientists a way to develop and test treatment methods aimed at stabilizing cortical function and identify biomarkers implicated in the potential preventative treatment of schizophrenia. As a result, researchers can test the efficacy of various types of drugs on MAM-treated rats to see if they alleviate or minimize symptoms induced by MAM. Studies have tested the safety and efficacy of various nootropic drugs (Mar et al., 2017) and mGlu5 positive allosteric modulators (Gastambide et al., 2012), and have found improvements of cognitive impairments related to schizophrenia. Modafinil is one drug that has shown promise in treating ADHD and narcolepsy (Lanni et al., 2008; Outhoff et al., 2016) and it may ameliorate certain cognitive deficits and impulsivity in people with schizophrenia (Ghahremani et al., 2011). Animal models that have tested the efficacy of modafinil have found mixed results. For example, Mar et al (2017) found modafinil improved attention and executive function in rats exposed to MAM when testing on a touchscreen-based CPT, although modafinil also improved performance in sham control rats. In humans, modafinil improved attention set shifting in healthy individuals (Turner et al., 2004) and one study showed that modafinil did not improve cognition or reduce negative symptoms in people with schizophrenia when compared with a placebo group (Freudenreich et al 2009). While this drug has not shown much translational promise for schizophrenia treatment, its prevalent use amongst both healthy and patient populations (Teodorini, 2020) warrant continued research on its potential as a cognition enhancing pharmaceutical.

Another drug receiving attention for the treatment of cognitive impairments in schizophrenia is choline. Choline significantly improved cognitive flexibility in combination with working memory training (Waddell et al., 2020) and activates on-time development of cerebral inhibition in children received choline perinatally (Ross et al., 2013). Choline is an essential nutrient and plays significant roles in the development of the brain. Choline is used in the biosynthesis of acetylcholine and is a part of phospholipids and lipoproteins (Zeisel & Costa, 2009). Further, choline is an important methyl donor, and therefore important for the maintenance of epigenetic marks (Zeisel 2017), and abnormalities in methyl, or one carbon metabolism have been noted in the study of schizophrenia since the 1960’s. In fact, choline is so important that it was designated as an essential nutrient by the Institute of Medicine in the late 1990s.

Sensory gating is also altered in individuals with schizophrenia (Atagun et al., 2020), and can be quantified through P50 response in the dual click paradigm. In this task, two clicks are presented in succession, and the difference in brain response (measured by event-related potential) to the second click is compared against the first click (Adler et al., 1982; Jerger et al., 1992). Generally, people with schizophrenia show less inhibition to the second of the paired stimuli when it is presented compared to healthy patients, indicative of poorer sensory gating (Olincy et al 2010; Freedmen, 2021). A randomized, placebo-controlled study with 100 pregnant women found that when mothers received twice the daily recommended dose of choline, via phosphatidylcholine administered during the second trimester of pregnancy through to the third postnatal month, the extra choline-treated children were able to suppress the P50 EEG response (76%) more than the control group children (43%; Ross et al., 2013). Furthermore, researchers also found the presence of a single nucleotide polymorphism (SNP) of the CHRNA7 gene to lower P50 inhibition among the place-group infants (Ross et al., 2013). Other studies that have implicated SNPs of CHRNA7, the gene that codes for the alpha-7 subunit of nicotinic acetylcholine receptor in P50 inhibition in patients with schizophrenia (Stephens et al., 2009; Clementz et al., 1997), and choline has been shown to ameliorate sensory gating deficits presumably through its binding to the alpha-7 subunit of nicotinic acetylcholine receptor (Knott et al., 2015; Choueiry et al., 2019). CNRA7 is also regulated by stress, suggesting it may be a point of interaction between several models of schizophrenia (Hunter, 2012). Biomarkers, such as the P50 suppression in infants, can help researchers identify successful interventions that help prevent or reduce the onset of various diseases (Ross et al., 2013). Choline supplementation to dams during neurodevelopment in the above-described models could be tested to determine whether it could normalize deficits described.

Environmental Enrichment Interventions

Researchers have also explored the therapeutic potential of complex environments on rodent models for schizophrenia. Environmental enrichment (EE) is a laboratory housing condition that provides more opportunity for stimulation in comparison with standard housing cages, due to increased novelty in alternating toys and increased cage space. As a result, EE promotes more species-typical behaviors in rodents (Kentner et al., 2019b). The link between positive environmental experiences in mitigating neurodevelopmental disruptions is actively being investigated in both humans and animals (Núñez Estevez et al., 2020; Kentner et al., 2019b; Schneider & Przewlocki, 2005; Woo & Leon 2013; Whittingham et al., 2020; Zhao et al., 2021). In animals, such housing environments have shown improvements in social interaction and general cognitive functioning following MIA. These enrichment mediated improvements were generally associated with an upregulation of genes known to be involved in synaptic plasticity (e.g., excitatory amino acid transporters, brain derived neurotrophic factor and NMDA receptors) and stress (e.g., corticotropin releasing hormone and oxytocin associated receptors). Each of these genes had been downregulated within the hypothalamus, hippocampus, prefrontal cortex, and amygdala in a sex-specific manner following MIA (Núñez Estevez et al., 2020; Zhao et al., 2021; Zhao et al., 2020; Kentner et al., 2016). These findings are relevant as they may underlie the pathogenesis of various neurodevelopmental disorders like schizophrenia and point to clinical interventions that may target these developmentally induced changes in brain. Moreover, there may be critical periods for when environmental manipulations may be effective, and interventions such as EE likely need to be administered early in life for best results (Kentner et al., 2019b).

Conclusions

The complex etiology of schizophrenia - consisting of environmental and genetic factors – has made it very difficult to find medications and treatments that are effective among all patients. However, the use of animal models has provided a cost-effective, time-sensitive method to test different hypotheses regarding the underlying mechanisms of schizophrenia. This research is done through the identification of changes in brain structure and genes, and targeted testing of pharmaceutical therapies. Unfortunately, therapeutic interventions identified from animal research often have a high likelihood of not being as effective in humans, partly due to the complex heterogeneity associated with many brain disorders. This complexity has given rise to the National Institute of Mental Health’s RDoC initiative. By integrating neurodevelopmental, behavioral, genetic, and environmental perspectives, we hope to obtain a deeper more translational perspective of the spectrum of brain function ranging from “healthy functioning” to disease.

We have discussed several behavioral tasks through this chapter used to assay cognition in animal models of schizophrenia with relevance to such RDoC domains (Table 1). While operant procedures requiring the simultaneous use of many behavioral tasks within a single experiment has been the preferred practice over the last several decades, these methods can be disadvantageous. The repeated handling of animals can promote stress and affect task performance (Horner et al., 2013). More recently, touchscreen technology has become valuable in testing cognitive functioning, such as reversal learning in rodents (Horner et al., 2013). With the introduction of this technology, experimenters can evaluate several cognitive measures with limited animal handling (Horner et al., 2013). The development of touchscreens involves showing animals visual stimuli and training them to make correct response by touching the correct location on a screen, using an appetitive reward as motivation (e.g., a food pellet or liquid milkshake). These automated touchscreen models provide a high level of standardization as they enable limited investigator interference as well as high translational potential given the consistent use of visual stimuli (Kangas & Bergman, 2017; Horner et al., 2013). Furthermore, touchscreen testing and operant chambers can be used to test multiple domains within the same group of animals.

In the last decade, there has been a drastic increase in the number of neuroimaging and EEG studies investigating patterns of brain function and connectivity that could be utilized as “biomarkers” of schizophrenia. Machine learning has been particularly useful in extrapolating biomarkers identified by a panoply of imaging and EEG studies across diverse samples (Kim, Blumberger, & Daskalakis, 2020; Park et al., 2021). In addition, new evidence has shown some exciting promise, where the physiological mechanisms of certain RDoC domains are conserved from rodents to humans (Cavanagh et al., 2021). The use of animal models is a cost-effective first step in identifying neural circuits of interest, and when used in tandem with MRI or EEG, will provide more clarity regarding the biochemical pathways and mechanisms seen in neurodevelopmental disorders such as schizophrenia, and aid in the development of more translational pharmacological therapies (Young & Markou, 2015; Cavanagh et al., 2021).

Furthermore, there needs to be a better accounting for sex differences in schizophrenia. While the incidence of early onset schizophrenia is higher in males, incidence in later life is higher in females (Pederson et al., 2014; Häfner, 2019). Several animal studies have opted to use only male rodents in their studies (Beery & Zucker, 2011) due to a common misconception that males provide more robust data. This myth has been completely discredited (Prendergast et al., 2014) and the exclusion of female animals hinders our ability to fully understand these disorders.

While no single animal model can replicate schizophrenia precisely, they each offer unique opportunities to identify the contributors involved in the etiology and symptomology of this complex disease. By employing these models, we can identify causal risk factors and map out detailed brain circuits involved in the disease process. Although treatments for schizophrenia are still being investigated, the RDoC approach may help accelerate this process, especially if the future rendition of this framework adopts sensory processing and developmental perspectives (Hirjak et al., 2021; Harrison et al., 2019; McLaughlin & Gabard-Durnam, 2021).