Decision-making results from a complex system of corticolimbic brain structures: the mesolimbic pathway is responsible for encoding motivation and promoting actions (Goto and Grace, Reference Goto and Grace2005), while the hippocampus, the amygdala and the prefrontal cortex (PFC) assign an affective value to life experiences (Calhoon and Tye, Reference Calhoon and Tye2015). These areas integrate the dopaminergic activity of the mesolimbic pathway, ultimately promoting or inhibiting the action (Adhikari et al., Reference Adhikari, Topiwala and Gordon2010; Abela et al., Reference Abela, Dougherty, Fagen, Hill and Chudasama2013; Abela and Chudasama, Reference Abela and Chudasama2014). Thus, a healthy individual undertakes goal-directed actions as a result of an inner conflict among corticolimbic brain areas, which ultimately drives the behavioural choice to approach or avoid a life paradigm, following a continuous gradient of integration between needs, rewards and risks (Cornwell et al., Reference Cornwell, Franks and Higgins2014). This ‘conflict’ is usually referred to as the approach-avoidance conflict (AAC) and its disbalance, in both directions, represents a highly shared symptom of several psychiatric disorders (Aupperle and Paulus, Reference Aupperle and Paulus2010; Aupperle et al., Reference Aupperle, Melrose, Francisco, Paulus and Stein2015; Kirlic et al., Reference Kirlic, Young and Aupperle2017; Loijen et al., Reference Loijen, Vrijsen, Egger, Becker and Rinck2020). According to Gray's Reinforcement Sensitivity Theory (RST), approach-avoidance behaviour arises from a balanced interaction between the Behavioural Approach System (BAS) and the Behavioural Inhibition System (BIS) (Gray, Reference Gray1970; Gray and McNaughton, Reference Gray and McNaughton2003; Bijttebier et al., Reference Bijttebier, Beck, Claes and Vandereycken2009). In mammals, BAS promotes behavioural responses to appetitive and rewarding stimuli, while BIS organises individuals' responses to punishing or threatening situations (Gray and McNaughton, Reference Gray and McNaughton2003). BIS and BAS are informed by implicit and explicit information processing, involved in positive or negative valence attribution to specific cues (Loijen et al., Reference Loijen, Vrijsen, Egger, Becker and Rinck2020). Implicit information processing is fast and emotion-driven and depends on subcortical brain regions involved in affective processing (i.e. mesolimbic areas including the hippocampus and amygdala). Conversely, explicit information processing is relatively slow, requires intentional reasoning and attentional efforts, and seems to depend on frontal brain regions associated with cognitive control and emotional reappraisal (Wager et al., Reference Wager, Davidson, Hughes, Lindquist and Ochsner2008; Aupperle and Paulus, Reference Aupperle and Paulus2010). Gray's RST assumes that individuals with an unbalanced BIS/BAS ratio are at risk of psychiatric drift (Loijen et al., Reference Loijen, Vrijsen, Egger, Becker and Rinck2020). Consistently, dysfunctional approach-avoidance tendencies have been implicated in the development and progression of several mental health disorders such as anxiety, depression, eating and addictive disorders and schizophrenia (Bijttebier et al., Reference Bijttebier, Beck, Claes and Vandereycken2009; Struijs et al., Reference Struijs, Lamers, Vroling, Roelofs, Spinhoven and Penninx2017; Loijen et al., Reference Loijen, Vrijsen, Egger, Becker and Rinck2020). We aim to emphasise the translational role of AAC paradigms in clinical psychiatric research, by discussing preclinical and clinical evidence of the brain circuits associated with AAC.
Cross-species paradigms to assess AAC
Cross-species behavioural tests aim at resolving shared molecular and circuital determinants between humans and rodents that underlie homologous disease-relevant behaviours in the AAC domain (Bach, Reference Bach2021). This approach should build a more solid basis to develop novel tools to help clinical diagnosis and maximise the predictive validity of AAC preclinical paradigms used to test pharmacological interventions for humans (Belzung and Lemoine, Reference Belzung and Lemoine2011). Cross-species behavioural tests to assess AAC mainly exploit paradigms of exploration (Gromer et al., Reference Gromer, Kiser and Pauli2021). Mammals' functioning is indeed based on an innate exploratory drive for territorial recognition, food supply and reproduction, which are in turn associated with a significant reward (Kidd and Hayden, Reference Kidd and Hayden2015). However, exploration naturally entails a conflict between its reward component, and the objective risk (Arzate-Mejía et al., Reference Arzate-Mejía, Lottenbach, Schindler, Jawaid and Mansuy2020; Italia et al., Reference Italia, Forastieri, Longaretti, Battaglioli and Rusconi2020). Thus, explorative engagement results from a fine balance between motivation (driven by reward) and refusal (supported by fear) (Blanco et al., Reference Blanco, Otto, Maddox, Beevers and Love2013; La-Vu et al., Reference La-Vu, Tobias, Schuette and Adhikari2020). Exploratory avoidance (which also includes to some extent social avoidance) is a pathological status that can either be caused by a loss of the natural desire to explore and socialise (anhedonia) or can be the result of a pathological fear (anxiety) (Kim and Kirkpatrick, Reference Kim and Kirkpatrick1996; Arzate-Mejía et al., Reference Arzate-Mejía, Lottenbach, Schindler, Jawaid and Mansuy2020). The first case is more evocative of a depressive-like drift, the second is associated with an aberrant state of anxiety which prevents an individual from indulging in social and exploratory needs, even though social and exploratory desire can be intact (Bijttebier et al., Reference Bijttebier, Beck, Claes and Vandereycken2009). In the following sections, we describe those AAC paradigms that are suitable for both humans and rodents.
In rodents, one of the most used exploratory paradigms to assess AAC is the Open Field (OF) test. It consists of a simple observation of animal deambulatory behaviour (Prut and Belzung, Reference Prut and Belzung2003). Modern software digitally traces the distance moved by the rodent from the perimetric walls to the centre of an arena. The longer the time spent and the distance walked in the centre, the more a rodent is prone to ‘approach’ exploration (Noldus et al., Reference Noldus, Spink and Tegelenbosch2001). Conversely, animals walking near the perimetric walls of the cage (also known as thigmotaxis) adopt an instinctive behaviour aimed at protecting against a perceived novelty-related threat (Prut and Belzung, Reference Prut and Belzung2003). In literature, the OF test is almost often described as an anxiety-assessing paradigm. However, it is important to underline that decreased exploration is also due to decreased exploration-related reward sensitivity (Blanco et al., Reference Blanco, Otto, Maddox, Beevers and Love2013; Cornwell et al., Reference Cornwell, Franks and Higgins2014; La-Vu et al., Reference La-Vu, Tobias, Schuette and Adhikari2020).
Human versions of the OF test have been developed and administered to clinical/non-clinical groups. Kallai et al., measured human thigmotaxis during the exploration of virtual and physical spaces, showing that thigmotaxis positively correlated with fear and avoidance bias for fear-mobilising situations during early trials of both tasks, but not with self-reported trait anxiety (Kallai et al., Reference Kallai, Makany, Csatho, Karadi, Horvath, Kovacs-Labadi, Jarai, Nadel and Jacobs2007). Walz et al., instructed patients with agoraphobia and healthy controls to perform a 15 min solitary walk on a 146 × 79 m soccer field. Patients with agoraphobia and participants with high self-reported anxiety sensitivity exhibited enhanced thigmotaxis (Walz et al., Reference Walz, Mühlberger and Pauli2016). In an additional virtual reality OF test performed on 141 individuals, the participants – like rodents in animal studies – preferred to stay in the outer region of the open field but there was no consistent association between thigmotaxis and self-report scales of anxiety and fear (Gromer et al., Reference Gromer, Kiser and Pauli2021). Overall, human OF test demonstrated cross-species validity, although, the modulatory effects of anxiety on human open-field behaviour should be further examined (Bach, Reference Bach2021).
Another exploratory paradigm frequently used in rodents is the elevated plus maze (EPM) test (Walf and Frye, Reference Walf and Frye2007). The maze, a cross-like structure mounted on elevated strut, is made up of two open and two closed arms where opaque plexiglass walls protect the rodent path (Rusconi et al., Reference Rusconi, Grillo, Ponzoni, Bassani, Toffolo, Paganini, Mallei, Braida, Passafaro, Popoli, Sala and Battaglioli2016). Closed arms are more comfortable for rodents and represent the preferred part of the maze. However, instinctive propensity for exploration prompts the animals to abandon the closed arms for the open ones for a short time. The percentage of time spent and the frequency of open arms entries represent a reliable readout of the rodents' propensity to approach. EPM test is largely used as an anxiety-assessment test that, however, does not take into consideration the reward-related drive to explore. Thus, these two tests specifically measure the AAC, as the result of the contribution of two components: anxiety and reward sensitivity (Cornwell et al., Reference Cornwell, Franks and Higgins2014; Bryant and Barker, Reference Bryant and Barker2020).
In humans, the EPM corresponding paradigm is the Mixed Reality version of EPM. Biedermann et al. (Reference Biedermann, Biedermann, Wenzlaff, Kurjak, Nouri, Auer, Wiedemann, Briken, Haaker, Lonsdorf and Fuss2017) translated the rodent EPM test to humans using a combination of real-world and virtual elements namely, a real-world wooden maze combined with a representation of this maze in virtual reality. Briefly, participants were instructed to step into the maze and walk slowly towards its centre, and wait for the scene to change before exploring the environment of the maze. After 90 s, the scenario switched and, instead of being in a virtual room, the maze was placed on a virtual rocky mountain surrounded by water. The subjects were allowed to explore the EPM for 300 s and, reporting higher anxiety about open arms, they preferentially avoided them. This tendency increased or decreased when they were given the anxiogenic yohimbine and the benzodiazepine lorazepam, respectively.
Other cross-species approaches featuring operant conflict tests were developed to emphasise both anxiety and reward sensitivity components thus further enhancing the conflict load of the choice (Bach, Reference Bach2021). These tests are based on the association of a specific reward or punishment to a given action. In rodents, the Vogel conflict test (VCT) represents one of the best constructs of AAC (Millan and Brocco, Reference Millan and Brocco2003). In this paradigm, within a habituation phase, a thirsty animal learns to drink from a metallic gauge. During the trial phase, after a few licks, the animal receives a mild electric shock. Depending on the relative balance between the motivation to seek the drinking reward and facing the punishing shock, the rodent will stop or keep on drinking. The number of shocks the animal decides to stand, directly correlates with its approach behaviour. Another similar test, the Geller-Seifter Conflict Test (GSCT) (Geller et al., Reference Geller, Demarco and Seifter1960) exploits a food-related reward instead of water. Although the human counterparts of the VCT and GSCT (described below) are less similar to the rodent variants than those of the OF and EPM, evidence suggests that these tests are equally valuable within cross-species approaches (Bach, Reference Bach2021). Aupperle et al. (Reference Aupperle, Sullivan, Melrose, Paulus and Stein2011) developed a third-person view computer task, named ACC conflict task, in which human participants move an avatar on a runway to decide between their chances of receiving a conflict outcome (negative affective image/sound combined with point rewards) v. non-conflict outcome (affective image/sound coupled with no points). The trials were designed to elicit the simultaneous desire to approach the reward and avoid the negative affective punishment. A limitation of this paradigm is that the reward offered during the conflicting conditions can vary while the affective threat remains stable. As such, the task allows to study conflict-related neuronal activations that are associated with the higher salience of the reward, but not those that might be elicited by increased salience of the negative outcome. Within a similar rationale, two additional human-designed conflict tests, based on third-person view computer tasks, were developed by Bach et al. (Reference Bach, Guitart-Masip, Packard, Miró, Falip, Fuentemilla and Dolan2014). In these tasks, the player is instructed to press a key (Bach, Reference Bach2015) or move on a rectangular grid (Bach et al., Reference Bach, Guitart-Masip, Packard, Miró, Falip, Fuentemilla and Dolan2014) to collect virtual tokens under the threat of being caught by a virtual ‘predator’ and losing all previously collected tokens. Threat probability corresponds to the wake-up rate of the predator, and the magnitude of potential loss corresponds to the number of already collected tokens. The wake-up rate is signalled by different colours and tailored to result in 3 different wake-up probabilities if the player stays outside the safe place for 100 ms. The player cannot escape once the predator is active. When participants have to press the key, they tend to collect fewer tokens when the potential loss is higher (Bach, Reference Bach2015). When participants have to move on the screen, they tend to explore and collect tokens early on, but as time progresses, the subjects retreat more to the safe place.
The inhibitory role of the hippocampus in approach-related behaviours
In the ‘80 Jeffrey Gray and Neil McNaughton suggested that the mammalian hippocampus may represent a central component of the BIS, hence sustaining avoidance within the AAC (McNaughton and Gray, Reference McNaughton and Gray2000; Gray and McNaughton, Reference Gray and McNaughton2003). This interesting theory initially accounted for robust data showing how partial or total surgical hippocampus ablation in rodents leads to increased approach behaviours. Similarly, it was shown that local hippocampal infusion of anxiolytic drugs that inhibit excitatory hippocampal neurotransmission enhances approach-related actions (Gray and McNaughton, Reference Gray and McNaughton2003). Lately, many studies described hippocampal functional polarisation, showing how the hippocampus is grossly divided into two portions, dorsal and ventral hippocampus (dHIP; vHIP) in rodents, respectively involved in spatial memory consolidation and affective/emotional processing (Kheirbek et al., Reference Kheirbek, Drew, Burghardt, Costantini, Tannenholz, Ahmari, Zeng, Fenton and Hen2013; Jimenez et al., Reference Jimenez, Su, Goldberg, Luna, Biane, Ordek, Zhou, Ong, Wright, Zweifel, Paninski, Hen and Kheirbek2018). Interestingly, anatomical segregation of hippocampal circuits has also been described in humans, being the anterior hippocampus homologous to the rodent vHIP and the posterior to the dHIP (Clark and Squire, Reference Clark and Squire2013). Thus, these studies better address the vHIP in rodents and the anterior in humans as a relevant seat of emotional information processing possibly related to behavioural avoidance. In the following sections, we examine the latest experimental evidence, in particular optogenetic-mediated surgical circuitry characterisation in rodents, and fMRI in humans, supporting the role of the hippocampus in behavioural avoidance, and further endowing Gray's RST with spatial, molecular and metabolic determinants.
Rodents optogenetic studies
Optogenetics refers to a biological technique to control the activity of genetically labelled neurons with light, an approach that significantly contributed in the last years to map brain functional connectivity (Adamantidis et al., Reference Adamantidis, Arber, Bains, Bamberg, Bonci, Buzsáki, Cardin, Costa, Dan, Goda, Graybiel, Häusser, Hegemann, Huguenard, Insel, Janak, Johnston, Josselyn, Koch, Kreitzer, Lüscher, Malenka, Miesenböck, Nagel, Roska, Schnitzer, Shenoy, Soltesz, Sternson, Tsien, Tsien, Turrigiano, Tye and Wilson2015).
One of the first optogenetic evidence of hippocampal involvement in the approach-avoidance outcome showed that specific inhibition of glutamatergic neurons of the basolateral amygdala projecting to the vHIP promoted exploratory approach measured by the EPM test (Felix-Ortiz et al., Reference Felix-Ortiz, Beyeler, Seo, Leppla, Wildes and Tye2013), while optogenetic activation of the same circuit limited exploration of the EPM open arms, increasing avoidance (Felix-Ortiz et al., Reference Felix-Ortiz, Beyeler, Seo, Leppla, Wildes and Tye2013). Another interesting study outlined the role of an additional vHIP efferent pathway, directed to the medial PFC in the modulation of approach-avoidance (Padilla-Coreano et al., Reference Padilla-Coreano, Bolkan, Pierce, Blackman, Hardin, Garcia-Garcia, Spellman and Gordon2016). In particular, optogenetic inhibition of vHIP axon terminals projecting to the medial PFC biases the AAC towards approach behaviours measured as increased exploration in the EPM test, a profile that was further validated by the Novelty Suppressed Feeding test (Padilla-Coreano et al., Reference Padilla-Coreano, Bolkan, Pierce, Blackman, Hardin, Garcia-Garcia, Spellman and Gordon2016).
Recently, it was also described that optogenetic enhancement of the excitatory vHIP afferents that project to the lateral hypothalamus increases anxiety, shifting the AAC conflict toward exploratory avoidance as scored by the OF and EPM tests (Jimenez et al., Reference Jimenez, Su, Goldberg, Luna, Biane, Ordek, Zhou, Ong, Wright, Zweifel, Paninski, Hen and Kheirbek2018).
Unexpectedly, a similar positive optogenetic manipulation that was performed over those vHIP afferents that innervate basal amygdala (BA), limited fear memory encoding and retrieval in the contextual fear conditioning test but displayed no effect in OF test readouts (Jimenez et al., Reference Jimenez, Su, Goldberg, Luna, Biane, Ordek, Zhou, Ong, Wright, Zweifel, Paninski, Hen and Kheirbek2018). The authors concluded that positive manipulation of these two vHIP glutamatergic afferents affects different emotional domains in rodents. However, the inhibitory activity of the vHIP-BA circuit toward fear memory consolidation contrasts with the potential involvement of the hippocampus in behavioural avoidance. In general, the stronger the fear memory the less an animal will be engaged in approach behaviours. It is possible that, within the complexity of limbic circuitry, inner homoeostatic needs leave a minority of the hippocampal circuitry (including vHIP-BA) free to contribute to approach, while the majority, as reviewed here, contributes to behavioural avoidance; therefore, balanced functioning of these circuitries would serve to support adaptive behaviours. The aforementioned evidence suggests the hippocampus as a brain area involved in the discrimination of those advantages and potential threats that have to be weighted in decision making.
Chronic environmental stress including psychosocial trauma has been shown to modulate AAC towards avoidance in vulnerable mice (Toth and Neumann, Reference Toth and Neumann2013; Anacker et al., Reference Anacker, Luna, Stevens, Millette, Shores, Jimenez, Chen and Hen2018). Thus, a question raises about whether the hippocampus contributes to translating stress into avoidance. In mice, chronic social defeat stress diminishes, in a subset of susceptible animals, the willingness to socially explore conspecific animals and exploratory behaviour. Such susceptibility is promptly reverted to resiliency by negatively regulating hippocampus excitability (Anacker et al., Reference Anacker, Luna, Stevens, Millette, Shores, Jimenez, Chen and Hen2018). Interestingly, has also been shown that chronic treatment with imipramine, a tricyclic antidepressant drug increasing serotonin levels in the synaptic cleft, is able to restore a normal approach-avoidance balance in psychosocial stress susceptible mice previously evaluated as social avoidants (Tsankova et al., Reference Tsankova, Berton, Renthal, Kumar, Neve and Nestler2006). This effect is again mediated by an overall decrease of hippocampal excitability, in accordance with the inhibitory effect of the hippocampal serotonin receptors, 5HT1A, highly abundant in this area (Tsankova et al., Reference Tsankova, Berton, Renthal, Kumar, Neve and Nestler2006).
Humans' studies
Recent functional magnetic resonance imaging (fMRI), and pharmacological and brain lesions studies confirm a relevant role for the hippocampus in the AAC in humans (Kheirbek et al., Reference Kheirbek, Drew, Burghardt, Costantini, Tannenholz, Ahmari, Zeng, Fenton and Hen2013; Bach et al., Reference Bach, Guitart-Masip, Packard, Miró, Falip, Fuentemilla and Dolan2014; Weeden et al., Reference Weeden, Roberts, Kamm and Kesner2015; Ito and Lee, Reference Ito and Lee2016; Schumacher et al., Reference Schumacher, Villaruel, Ussling, Riaz, Lee and Ito2018; Bach et al., Reference Bach, Hoffmann, Finke, Hurlemann and Ploner2019; Abivardi et al., Reference Abivardi, Khemka and Bach2020; Bryant and Barker, Reference Bryant and Barker2020; La-Vu et al., Reference La-Vu, Tobias, Schuette and Adhikari2020; Yeates et al., Reference Yeates, Ussling, Lee and Ito2020). For instance, Bach et al., conducted a fMRI study to investigate the role of the hippocampus in arbitrating ACC under various levels of potential threat, comparing neurologically healthy controls and patients with selective hippocampal lesions (Bach et al., Reference Bach, Guitart-Masip, Packard, Miró, Falip, Fuentemilla and Dolan2014). Bold signal in the anterior hippocampus increased with the probability of predator attack, supporting the putative role of the hippocampus as a negative regulator of approach in AAC paradigms in humans. Importantly, the threat levels much less influenced the behaviour of patients with selective hippocampal lesions, which showed reduced passive avoidance behaviour and inhibition across all threat levels (Bach et al., Reference Bach, Guitart-Masip, Packard, Miró, Falip, Fuentemilla and Dolan2014). The same AAC computer game was then used to investigate the impact of benzodiazepines and amygdala lesions on putative human anxiety-like behaviour (Korn et al., Reference Korn, Vunder, Miró, Fuentemilla, Hurlemann and Bach2017). The task was administered to (i) a group of healthy controls after a single dose of lorazepam v. placebo and (ii) two patients with bilateral amygdala lesions v. age- and gender-matched healthy controls. Lorazepam and amygdala lesions reduced loss adaptation, decreasing patients' anxiety-related avoidance behaviour.
A more recent study from the same group (Bach et al., Reference Bach, Hoffmann, Finke, Hurlemann and Ploner2019) confirmed that, in humans, hippocampal lesions increase approach under conflict whereas amygdalar lesions impair the return to safety.
In summary, also human studies report a role of the hippocampus and amygdala in AAC under threat, linking the integrity of these regions to conditioned fear expression and inhibitory avoidance (Ito and Lee, Reference Ito and Lee2016). Such new knowledge, however, warrants further inherent neuroimaging studies, based in particular on fMRI to better dissect inherent circuitry.
Conclusion and future directions
Deepening the knowledge of AAC circuitry and mechanisms in rodents and humans holds a huge translational potential as it may help to unravel psychopathological elements of several psychiatric disorders featuring unbalanced AAC. Further studies combining hippocampus-focused functional brain imaging using the described AAC cross-species paradigms with clinical (i.e. questionnaire-based) evaluation of AAC and anxiety, should be performed to validate preliminary observation of increased hippocampal activity as a biomarker of threat or punishment sensitivity and avoidance, ultimately helping to refine psychiatric patient stratification and diagnosis along with treatment options and prognosis.
Data
All data used to write this paper is in the reference list.
Acknowledgements
We acknowledge Paolo Brambilla for stimulating discussions and critical reading of the manuscript. We also acknowledge Cariplo Foundation Grant 2016-0908 to EB, Competitive Research Grant KAUST Grant 2019 to EB, PSR_2019 to FR and SEED Seal of Excellence (University of Milan) to FR.
Conflict of interest
None.