Laboratory models of posttraumatic stress disorder: the elusive bridge to translation

Summary

Posttraumatic stress disorder (PTSD) is a debilitating mental illness composed of a heterogeneous collection of symptom clusters. The unique nature of PTSD as arising from a precipitating traumatic event helps simplify cross-species translational research modeling the neurobehavioral effects of stress and fear. However, neurobiological progress on these complex neural circuits informed by animal models has yet to produce novel evidence-based clinical treatment for PTSD. Here, we provide a comprehensive overview of popular laboratory models of PTSD and provide concrete ideas for improving validity and clinical translational value of basic research efforts in humans. We detail modifications to simplified animal paradigms to account for myriad cognitive factors affected in PTSD, which may contribute to abnormalities in regulating fear. We further describe new avenues for integrating different areas of psychological research underserved by animal models of PTSD. This includes incorporating emerging trends in the cognitive neuroscience of episodic memory, emotion regulation, social-emotional processes, and PTSD subtyping to provide a more comprehensive recapitulation of the human experience to trauma in laboratory research.

In brief :

Dunsmoor et al. detail challenges to cross-translation of laboratory PTSD models from laboratory animals to humans, and describe approaches to enrich laboratory paradigms in humans to account for myriad cognitive, social, and emotional factors affected in PTSD.

Posttraumatic stress disorder (PTSD) is a categorical psychiatric syndromal diagnosis characterized by a heterogeneous collection of symptoms that eludes a single laboratory model. A diagnosis of PTSD can involve a host of anxious, depressive, and cognitive symptoms that fall within a multitude of symptom clusters—including intrusive re-experiencing, avoidance, changes in mood and cognition, and arousal-based symptoms—and comorbidity with other mental illnesses is extremely common (Shalev et al., 2017). Although distinguishing PTSD as a distinct disorder is complicated by symptom heterogeneity and comorbidity, PTSD is truly unique among all DSM 5 diagnoses by necessitating exposure to a precipitating traumatic event (American Psychiatric Association, 2013; Stein et al., 2014). This particular feature benefits translational research efforts, as there are several tractable laboratory paradigms that evaluate the psychobiological consequences of aversive events on learning, memory, and behavior. As such, rodent models of PTSD have focused largely on the behavioral and neurophysiological effects of stress and fear (Pitman et al., 2012; Siegmund and Wotjak, 2006; Yamamoto et al., 2009). Today there is substantial and detailed knowledge of the neurocircuitry underlying normative and dysregulated responses to stressful and fearful events (Krabbe et al., 2017; Mahan and Ressler, 2012; Maren and Holmes, 2015). The promise of these advances is to provide a deeper understanding of how psychopathology can develop after trauma and identify putative targets for intervention.

However, despite significant achievements in neurobiologically-driven approaches to PTSD research, there remains a stubborn translational gap towards the development of innovative and effective new clinical treatments (Bienvenu et al., 2021; Flores et al., 2018; Richter-Levin et al., 2019). The development of novel treatments is of paramount importance. Currently available evidence-based psychotherapies such as trauma-focused cognitive-behavior therapy (CBT), prolonged exposure (PE), and cognitive processing therapy (CPT), as well as FDA approved pharmacological agents such as sertraline and paroxetine, and their combination, are insufficient to attain remission in many patients with PTSD (Merz et al., 2019; Steenkamp et al., 2015). This lack of progress is perhaps due, in part, to an overreliance on neurobiological fear-based models, developed and refined in rodents, to inform clinical translational research efforts in humans. The dependence on animal models to drive PTSD research has resulted in a dominant focus on selective phenotypes measurable by experimental protocols optimized for rodent research and the neural circuits that support these behaviors. Nowhere is this more evident than Pavlovian fear conditioning (Bienvenu et al., 2021), wherein the basic laboratory protocol is easily adapted for important cross-species experimentation (LeDoux, 2000; Lonsdorf et al., 2017).

However, progress in animal neurobiology research is starting to create a bottleneck to human research, which borrows similar experimental approaches but lacks equivalent technical tools to probe the dynamic neurocircuitry of fear and anxiety. Indeed, translation of PTSD-related processes into animal models necessarily requires simplification (Deslauriers et al., 2018; Verbitsky et al., 2020). Though some important features of PTSD inevitably get lost in translation, the degree of experimental control, and use of sophisticated invasive technology (e.g., optogenetics and postmortem studies of single cell transcriptomics), in animal models affords immense potential for discovery. By contrast, translating animal models of PTSD back to human subjects research retains the simplification and information loss while often facing challenges in retaining the technological sophistication and experimental control that characterize elegant neurobiological experiments. Further, animal models necessarily omit human-specific psychological functions that play a prominent role in dictating trauma symptomatology, such as verbal-linguistic functions, subjective self-report of internal emotional/cognitive phenomena, and complex social and behavioral components of these experiences (LeDoux and Pine, 2016), leaving critical aspects of PTSD psychopathology unrepresented. Indeed, this fundamentally raises the question of the construct validity of the available animal models of PTSD. We are not, of course, suggesting that these models have not provided important information about the neurobiology of the stress response that have led to important leads in clinical research. However, if emerging neuroscience discoveries in rodents aren’t adequately adapted to drive innovation or inform human research, then these discoveries will ultimately be of little utility to improve treatment outcomes.

With advancements in neurobiological characterization of PTSD-related phenotypes in laboratory animals (Ressler, 2020), it is a good time to take a deep look at areas that have and have not been adequately addressed by conventional animal models of PTSD. In particular, as compared to strong interest in the neural mechanisms of fear, there has been relatively less emphasis on basic laboratory-based research covering fundamental cognitive and psychosocial factors underlying the etiology, treatment, and prevention of PTSD as a psychiatric disorder in humans. As a consequence, an integrated neuroscience-informed model that assimilates known dysregulated stress and fear learning processes with disturbances in higher-order cognitive function and social-emotional processing has remained elusive.

The goal of this monograph is to explore avenues for building on animal models of PTSD to develop integrated approaches for human translational research that accounts for relevant cognitive and social-emotional factors. We describe some of the dominant existing approaches to PTSD research in humans that are built on animal models, with a particular focus on associative fear learning and extinction. Rather than provide a comprehensive review of the PTSD conditioning literature, we instead discuss areas where basic research in this domain can be elaborated to account for more dynamic cognitive processes known to contribute to associative fear learning. We also review an important area of research that is not well-captured by animal models, episodic memory. Many features of PTSD can be conceptualized as emerging from disordered / dysfunctional memory (Brewin, 2011; Elzinga and Bremner, 2002). Animal models of PTSD are exceptional for detailing the neurobiology underlying implicit fear memory, but are less informative for investigating disturbances of explicit memory. We propose a path forward for basic laboratory research on well-known memory disturbances in PTSD built on emerging trends in the cognitive neuroscience of episodic memory. Finally, we discuss avenues to encourage further basic research on social-emotional symptoms in PTSD to integrate with known dysregulation in other symptom clusters.

Efforts to integrate cognitive factors into models of PTSD is not new (Brewin and Holmes, 2003; Creamer et al., 1992; Ehlers and Clark, 2000; Foa et al., 1989; Liberzon and Abelson, 2016; Litz and Keane, 1989). Many paradigmatic approaches and specific protocols we discuss have been proposed in the past as key to understanding the dynamic nature of psychiatric disorders. Nevertheless, the general prominence of neural data over behavior (Niv, 2021)—with the particular focus on the neural circuits of fear and anxiety in murine neuroscience—has in many ways overshadowed the ultimate goal of clinical translational research efforts to improve outcomes in humans. As a consequence, several of the inventive experimental techniques we discuss could have direct relevance to our understanding of the neuropathophysiology of PTSD but remain underutilized in the field.

Altogether, our goal is to discuss new ideas (and dust off some old ones) that could yield fruitful avenues for translational research in humans. Incorporating advances in the neuroscience of fear and anxiety with empirical and theoretical approaches from other domains of psychological research should provide more validity to laboratory models of PTSD.

1. PTSD through the lens of learning theory

An animal’s response to uncontrollable and unpredictable aversive events had long been proposed to mimic aspects of mental health disorders (Wolpe, 1952). Many of these behaviors resemble core fear-related symptoms of PTSD (Blanchard et al., 1982; Foa et al., 1989; Kolb, 1987), such as arousal, reactivity, aggression, and avoidance, thereby providing high face validity for animal models of PTSD. Renewed attention to the neuroscience of Pavlovian fear conditioning throughout the 1990s paralleled the rise of PTSD as a psychiatric diagnostic category (Bienvenu et al., 2021). Breakthroughs in the neural basis of conditioned fear, centered on the amygdala, informed experimental paradigms in human research, and continue to drive experimental approaches to PTSD research.

Associative learning provides a strong theoretical basis for understanding mental health disorders with an acknowledged learning and memory component. Fear conditioning, a form of associative learning, has long served an important role as a mechanistic model for the etiology, maintenance, and treatment of PTSD and anxiety-related disorders (Foa et al., 1992; Grillon et al., 1996; Keane et al., 1985; Kolb and Mutalipassi, 1982). Conditioning refers to both a learning process and an experimental protocol. On both fronts, conditioning provides a number of crucial advantages for translational research. One of the foremost advantages is many basic elements of conditioned learning are evolutionarily conserved across animal species, from fruit flies, to sea slugs, to mammals. As a result, technical tools available for discovery in non-human species (e.g., optogenetics and calcium imaging) can be used to make inferences on homologous processes in the human brain (Anderson and Adolphs, 2014). Conditioning is also widely utilized across the fields of psychology and neuroscience to investigate fundamental learning and memory processes, allowing translational mental health research to benefit from discoveries in laboratories without this explicit fundamental goal. These experimental paradigms are popular because they induce rapid learning that persists as long-term memory (up to the lifetime of the animal), produce an objective behavioral correlate (e.g., freezing), and behavior conforms to predictions from powerful computational learning models (e.g., error-correction learning). Finally, conditioning rests on over a century of rigorous experimentation and theoretical insights, and is perhaps the most replicated and validated finding in psychological research. It therefore provides a deep well of empirical research and theory to draw from for contemporary mental health research.

Notably, conditioning is an associative learning process, and conditioning protocols therefore probe the learning capacity of the animal. By contrast, some of the most popular animal models of PTSD have used stress protocols that are similar to animal models of depression (Krishnan and Nestler, 2011) and induce non-associative sensitization, such as immobilization, predator odor, inescapable shock, forced swim, isolation, and social-defeat, or a combination of stressors built into a single prolonged stress model (Yamamoto et al., 2009). These protocols are appropriately designed to induce behavioral and neurophysiological responses to intense stress, and correspond well to human PTSD research demonstrating that cumulative trauma exposure, rather than single incident trauma exposure, better predicts PTSD development (Breslau et al., 1999; Cisler et al., 2012; Cougle et al., 2009; Neuner et al., 2004). However, stress models lack a strong learning component. Fear conditioning has proved to be the more prevailing cross-species experimental framework for contemporary translational research on PTSD-related phenotypes (Bienvenu et al., 2021), likely because it captures aspects of learning and memory that are not addressed by stress models alone. Critically, many studies have combined multiple stress protocols with learning paradigms to show, for instance, that stress impairs extinction processes (Souza et al., 2017).

In the simplified conditioning model of PTSD, the myriad sensory or contextual cues in the environment during or around the time of trauma serve as conditioned stimuli, while the trauma serves as the aversive unconditioned stimulus. The CSs later act as reminders to reactive the trauma memory, and can generalize to allow a wider array of stimuli to elicit the trauma memory beyond the original stimuli present during the trauma. Abnormalities in forming or retrieving inhibitory associations during or following extinction maintains or exacerbates learned fear behavior, demonstrated through heightened post-extinction recovery of learned defensive responses, i.e., the return of fear. This can be conceptualized as a failure to properly enact a contextually-adaptive implicit (non-deliberate) regulation of fear responses (Liberzon and Abelson, 2016), which is often viewed to be one active component of effective trauma-focused psychotherapy (Milad et al., 2014b) These basic principles of extinction can be applied to understand difficulties in recovering from trauma, including the return of symptoms following exposure-based treatment (Milad and Quirk, 2012).

The elegant nature of this simplified model provides sufficient explanatory power for how some fears develop following a negative emotional experience, and how fear persists despite countervailing experiences of safety. However, the fear conditioning model has been criticized for being overly simplistic to account for complex mental health disorders and lacks clear relevance to non-fear related symptoms. In particular, it doesn’t account for the role of social-emotional factors that predict recovery after trauma, underlie impairment across PTSD symptom clusters, and seem to influence treatment response. It also does not properly account for additional forms of fear or emotional regulatory processes that are specific to humans due to our cognitive-linguistic capabilities (e.g., cognitive reappraisal, modification of belief systems) and which also form a key component of particular forms of evidence-based trauma-focused psychotherapy (e.g., cognitive processing therapy (Resick et al., 2017)). There also remains a debate on whether the term “fear” is too subjective to be applied to paradigms that evaluate the defensive response systems of non-human animals (LeDoux, 2020; Mobbs et al., 2019). Finally, the plethora of complex human emotional responses to trauma, including guilt, shame, and moral injury (Shi et al., 2021) likewise render the fear conditioning model of PTSD an incomplete recapitulation of the human trauma response.

1.1. Fear acquisition.

Defensive behavior is an ordinary response to stimuli that reliably predict danger. The question arises whether heightened fear acquisition is a core pathophysiological feature that should be incorporated into conditioning-based PTSD models. Despite early claims that PTSD was associated with stronger conditioning (Orr et al., 2000), numerous studies using discriminative conditioning procedures have found initial fear acquisition in PTSD patients and comparison groups to be equivalent (Hennings et al., 2020; Milad et al., 2008; Morey et al., 2015; Morey et al., 2020; Pöhlchen et al., 2020), with evidence of elevated responding only to the unpaired CS− (Duits et al., 2015). Equivalent fear acquisition should not necessarily be a surprise given that a behavioral response to cues that predict danger is part of a normative learning process (Beckers et al., 2013; Richter-Levin et al., 2019). Indeed, a lack of conditioned fear acquisition is indicative of poor and maladaptive learning, and associated with abnormalities in socialization (Gao et al., 2010) and psychopathy (Birbaumer et al., 2005). Further, it is not obvious what constitutes an abnormally elevated response to an imminent threat. That is, what distinguishes normal from abnormally elevated defensive responding when threat is actually present, as during fear acquisition?

Conceptually, whether fear acquisition is dysfunctional in PTSD is also somewhat orthogonal to the question of whether fear conditioning is a model for the development of PTSD; as indicated above, dominant animal models that produce phenotypes consistent with PTSD are prolonged stress / sensitization models that do not include clear learning components. While the magnitude of a defensive response to imminent threat is unlikely to be useful to a conditioning model of PTSD, acquisition is a necessary methodological step for probing potential abnormalities in other learning processes following acquisition when the potential for threat is more ambiguous (Bouton, 2002), such as the ability to extinguish defensive responding once the aversive outcome is omitted.

1.2. Fear extinction.

PTSD is often conceptualized as a failure to recover from heightened fear responding following trauma (Yehuda and LeDoux, 2007). In one example, 95 women were followed prospectively shortly after a sexual assault (Rothbaum et al., 1992). At the first assessment in the acute trauma period, 94% of women met criteria for PTSD. After 3 months, 47% met criteria for PTSD. This example highlights that a decrease in baseline functioning following trauma is normative, and what differentiates PTSD is the persistence of decreased functioning. As such, much animal and human work has focused on characterizing deficits in PTSD that might explain failure to recover. One candidate mechanism is a deficit in extinguishing learned fear responses. This mechanism has strong face validity, given the prominence of fear conditioning as an explanation for the acquisition of fear following trauma and the observation of lack of attenuation of fearful responding in PTSD. While there are numerous examples in the literature of PTSD patients demonstrating slowed extinction learning (Orr et al., 2000; Peri et al., 2000; Wessa and Flor, 2007) as well as a deficit in retrieving fear extinction memories (Garfinkel et al., 2014; Milad et al., 2009), the robustness of this effect has recently been questioned (Pöhlchen et al., 2020).

The inconsistency in demonstrating extinction learning or recall deficits in PTSD could reflect either a weak effect and/or sensitivity to specific methodological variations in testing the effect (Lonsdorf et al., 2017). Such inconsistencies may also reflect effects of other higher-order cognitive factors likely to influence or interact with the basic acquisition and extinction processes (Hofmann, 2008; Lovibond, 2004; Rachman, 1977; Vervliet et al., 2013), even for simple experiments derived from animal models. Cognitive factors that affect acquisition and extinction involve causal reasoning (Gershman et al., 2010; Redish et al., 2007), conscious expectancy or contingency beliefs (Grady et al., 2016; Grings, 1973; Lovibond and Shanks, 2002), verbal instructions or social learning (Bandura and Menlove, 1968; Luck and Lipp, 2016; Olsson and Phelps, 2007), propositional analysis (Mitchell et al., 2009), shifting of attention (Craske et al., 2018; Ma et al., 2017), or ongoing cognitive manifestations of anxiety such as worry. These cognitive factors likely play some role in differentiating patterns of acquisition and extinction between PTSD and controls, but some (though certainly not all) are challenging to adequately translate to an animal learning model.

Another notable inconsistency across laboratory animals and humans involves the presumed role of the amygdala and medial prefrontal cortex in acquisition and extinction. Collections of subnuclei in the amygdala are properly considered crucial for acquisition and expression of conditioned fear, and are also necessary for extinction learning (Falls et al., 1992; Pape and Paré, 2010). Yet evidence of amygdala involvement in human fMRI of fear conditioning and extinction is mixed (Visser et al., 2021), and meta-analyses pointedly show the amygdala is not among a collection of regions consistently observed when contrasting activity between the CS+ and CS− (Fullana et al., 2015). Likewise, numerous neurobiological studies in rodents show the infralimbic cortex (homologous to the ventromedial PFC in humans) is important for forming, storing, and retrieving extinction memory (Milad and Quirk, 2012). Although early fMRI evidence seemed to translate this finding to healthy human adults (Milad et al., 2007; Phelps et al., 2004), mounting contemporary evidence questions the robustness of vmPFC activity during extinction learning and recall (Fullana et al., 2018b). As such, reports of differences in brain activity in the amygdala and vmPFC during acquisition and extinction between PTSD and comparison groups should be interpreted in the framework that the comparison group may not necessarily evince distinctive fMRI activity to CSs in these regions either. These challenges may in part explain why fMRI studies comparing PTSD and controls have yielded mixed evidence for involvement of a number of brain regions during acquisition and extinction (Lazarov et al., 2020; Neria, 2021), including whether the amygdala and vmPFC are hyper- or hypo- active relative to controls (Diener et al., 2016; Hennings et al., 2020; Milad et al., 2009; Shvil et al., 2014).

One approach that might help clarify the role of the amygdala, vmPFC, and other core regions associated with threat learning and regulation, is to incorporate machine learning and multivariate analysis methods into functional neuroimaging (Lewis-Peacock and Norman, 2014). MVPA is more sensitive to information content coded across collections of voxels, making it possible to non-invasively measure information content in patterns of spatially distributed brain regions (i.e., across multiple voxels) at relatively small spatial scales. This can be compared to traditional univariate approaches, which are better understand at identifying engagement of particular brain regions through averaged activity that is often spatially smoothed across voxels to improve signal-to-noise. For example, a multi-voxel pattern analysis (MVPA) approach was used to reveal distinct patterns in the amygdala associated with the CS+ and CS− (Bach et al., 2011). Recently, MVPA approaches were used to show engagement of the vmPFC for coding long-term representations of extinction memory in healthy adults versus individuals with PTSD symptoms (Hennings et al., 2022; Hennings et al., 2020).

In summary, the standard extinction learning and recall paradigm is extremely popular, and is even considered a model paradigm in NIMH’s Research Domain Criteria Initiative. These tasks, which tend to be based on simplified animal paradigms, have been used for several decades to compare anxiety-related disorder groups versus control groups (Milad et al., 2014a). Notably, while acquisition and extinction protocols tend to be simplified, humans incorporate higher-order cognitive factors to both acquire and extinguish conditioned fear. Importantly, cognitive processes are affected in PTSD, which may therefore contribute to impaired aspects of extinction learning, retention, and recall. Some higher-order cognitive factors (e.g., verbal instructions, propositional analysis) may be difficult to model in laboratory animals; although contemporary theoretical frameworks can successfully account for many cognitive factors in rodent experiments (e.g., inferring causal relationships; (Blaisdell et al., 2006)). Neuroimaging evidence of abnormal acquisition and extinction in PTSD has also been mixed, perhaps driven by failures to fully translate the role of amygdala and mPFC from rodent neurobiology to human functional neuroimaging. Advances in machine learning and multivariate pattern analysis of neuroimaging data may help bridge rodent neuroscience. Specifically, further advances in high-resolution multivariate neuroimaging methods will help translate findings of distinct neural populations within and across brain regions that process different valence-specific behavior (Ghosh and Chattarji, 2015; Herry et al., 2008; Lacagnina et al., 2019; Namburi et al., 2015), sometimes referred to as engrams (Josselyn et al., 2015). The innovative use of real-time closed-loop MVPA decoded fMRI neurofeedback (Cortese et al., 2021) holds future promise for enhancing self-regulation by targeting activity patterns associated with the inhibition of fear (Taschereau-Dumouchel et al., 2018).

2. Introducing more complexity to the conditioning model of PTSD

While no single laboratory model can account for a complex psychiatric disorder, the typical animal fear conditioning protocol is purposefully reduced to its basic elements. This often involves a simplified design with a unimodal sensory CS or context paired, and then unpaired, with foot shocks. While conditioning is an optimal paradigm for characterizing myriad fear learning processes, it fails to capture important cognitive features of PTSD, such as disturbances in declarative memory, intrusive memories, problems in psychosocial functioning, negative and distorted beliefs, and complex emotions such as guilt and shame. Contemporary views of conditioning have long incorporated cognition into myriad factors important for acquisition, generalization (Figure 1), and inhibition of learned behavior (Mineka and Zinbarg, 2006; Rescorla, 1988). As a result, a number of validated associative learning paradigms reveal complex aspects of learning and memory. These may address conceptual limitations of more simplified conditioning protocols frequently used in animal models of PTSD.

Figure 1.

A schematic comparing how fear transfers to a wide variety of stimuli in a simplified laboratory-based fear learning and a real-world example aversive experience. The complex nature of how fear associations spread to idiosyncratic details associated with trauma complicates efforts to map simplified experimental protocols of direct CS-US learning to real-world manifestations of learned fear.

2.1. Fear Generalization.

One of the hallmark features of PTSD is the tendency to overgeneralize negative emotional associations to stimuli, people, situations, or contexts that are similar to trauma-related details (see Figure 1). This expands the number of trauma-reminders in the environment capable of triggering PTSD symptoms (Foa et al., 1989; Keane et al., 1985). Overgeneralization presents a theoretical challenge for exposure-based treatments, as it is nearly impossible to extinguish every stimulus or situation that could act as a potential reminder.

Empirical research on stimulus generalization dates back to the earliest research from Pavlov’s laboratory (Pavlov, 1927). Fear generalization was also a major focus of the first demonstration of human fear conditioning (i.e., the infamous Little Albert experiment) (Watson and Rayner, 1920). There is now well over a century of research on the experimental factors that promote or reduce generalization of learned behavior following conditioning (Ghirlanda and Enquist, 2003). The past several years has seen a growth of interest in empirical research of fear generalization in humans (Dunsmoor and Paz, 2015; Dymond et al., 2015; Webler et al., 2021). Generalization research in PTSD groups largely confirms broader generalization of fear-related behavior and neural activity to a wider range of stimuli that approximate a learned threat (Kaczkurkin et al., 2017; Morey et al., 2015; Morey et al., 2020). Overall, however, the extinction deficit model remains far more popular, both in rodents and in humans, than protocols that test generalization following fear learning. Thus, it still remains an important question whether overgeneralization is a pathogenic marker of PTSD, and whether the breadth of generalization can distinguish PTSD from other anxiety syndromes such as generalized anxiety disorder and panic disorder.

Humans are also particularly adept at extracting the semantic and conceptual regularities from an emotional learning event, and later generalize defensive behavior based on these abstract properties (Dunsmoor and Murphy, 2015). This can include semantically-related details not directly linked through a purely perceptual resemblance to details from the trauma event. For example, a military veteran with PTSD may regard the anniversary of a traumatic event, holidays associated with wartime (e.g., Veterans Day), or various associations to the foreign country where they were deployed (e.g., the country’s flag, cuisine, weather, symbols of the culture or majority religion) as reminders that trigger PTSD symptoms. Understanding dysregulated fear behavior in PTSD requires accounting for concept-based generalization, as many trauma-related cues are conceptual in nature. Fear generalization that draws on conceptual knowledge structures is clearly a challenge to model in rodents. Early research shows PTSD is associated with broader concept-based fear generalization in fear-related neurocircuitry than controls (Morey et al., 2020).

2.2. Extinction Generalization.

Fear extinction is hypothesized to be a primary means by which exposure therapy enacts symptom change; as such, fear extinction is used as a laboratory model for exposure therapy. Another perspective on fear extinction is informed by work in the field of emotion regulation in humans, wherein this process is viewed as a form of regulating emotional responses through an automatic and implicit process (Braunstein et al., 2017a)—in this case, experiential learning of new associations between a stimulus and an outcome. One methodological feature of exposure therapy, which is often lost in basic laboratory models of fear extinction, is that exposure therapy rarely, if ever, uses the exact stimuli from an individual’s actual learning history. In the case of PTSD, the exact stimulus conditions surrounding the trauma are either difficult to reproduce or are dangerous and contraindicated to use during exposure therapy. Instead, exposure therapy for PTSD largely focuses on generalization stimuli (GS); that is, stimuli that are perceptually and/or conceptually related to the actual traumatic event. Additionally, the goal in exposure therapy for PTSD is often to distinguish between actual dangerous events (e.g., sexual assault) and situations / contexts / stimuli that remind one of the dangerous event (e.g., being around crowds, unfamiliar men, certain TV shows, etc.). In this case, it is explicitly not the goal to reduce fear responding to the original traumatic stimuli (which might remain an actual threat), but rather to reduce fear to non-dangerous stimuli to which the fear has generalized.

Interestingly, the framework of exposure therapy arose out of an emotional processing framework, which hypothesized that novel corrective information needs to be incorporated into an individual’s “fear structure”, which is a cognitive-associative mental structure containing information about the feared stimulus/situation’s characteristics, the individual’s response to the feared stimulus/situation, and aspects of its meaning to the individual (Foa and Kozak, 1986). This concept of a “fear structure” is clearly more complex than that of either a CS or GS, though encompassing elements of both, and it highlights the top-down and bottom-up convergence between human clinical research and basic experimental research in animals.

Distinctions between the use of GSs vs the original CS in exposure therapy is not a trivial methodological detail. Basic laboratory research on fear extinction using GSs, versus the original CS, demonstrates greater return of fear both to the original CS as well as to novel GSs (Lipp et al., 2020; Vervliet et al., 2004; Vervoort et al., 2014; Zbozinek and Craske, 2018). Although extinction to the original CS is the standard in laboratory protocols, exposure to GSs in the absence of an aversive outcome is a better analogue of clinical exposure therapy. But strong fear relapse following extinction to GSs, compared to the original CS, raises a significant question: how does extinction research using original CSs ultimately inform mechanisms relevant to exposure therapy? As one example, laboratory manipulations designed to enhance fear extinction developed using the original CS (Dunsmoor et al., 2015b; Fitzgerald et al., 2014) may only weakly translate to clinical exposure therapy.

Emerging basic laboratory research among humans has investigated the impact of interesting variations of extinction with GSs. One study compared extinction with multiple GSs, extinction with a single GS, and extinction with the original CS (Zbozinek and Craske, 2018). Results demonstrated that while both extinction procedures using GSs resulted in greater fear responding to novel GSs compared to extinction with the original CS, extinction using multiple GSs resulted in less return of fear than extinction with a single GS. Another study compared extinction using a GS that elicited a strong fear response vs a GS that elicited a weak fear response, and results demonstrated that extinction using a GS that elicited a strong fear response resulted in better fear reduction (Struyf et al., 2018). These two studies highlight methodological variations of extinction that are readily translatable to clinical exposure therapy: use of multiple / varied exposure situations and selecting exposure situations that maximize fear responding (Craske et al., 2014; Rowe and Craske, 1998). Further work along these lines will not only shed light on mechanisms of fear extinction but also continue to bridge the gap between basic experimental studies of extinction and clinical exposure therapy for PTSD.

2.3. Protection from extinction.

Keane et al. (Keane et al., 1985) noted an apparent paradox that repeated recollections or reminders of trauma maintain PTSD symptoms, rather than reduce symptoms as would be expected considering trauma reminders generate affective responses without an accompanying aversive event. Put another way, why don’t those with PTSD naturally extinguish over time? They proposed that incomplete exposure to the trauma memory, either through only partial reactivation or incomplete knowledge for the details of the trauma in the first place, preclude full extinction. Another explanation derives from Mowrer’s two-factor theory (Mowrer, 1939) applied to the reinforcing power of avoiding people, places, situations to help prevent intrusive thoughts and unwanted emotions (Williams and Moulds, 2007), thereby maintaining symptoms.

From a learning theory perspective, stimuli or actions that accompany extinction training can get in the way of extinction learning (Figure 2). For example, the ability to avoid a shock by pressing a button when a CS is presented diminishes autonomic arousal; but arousal returns when there is no longer a button to press, even when there are no more shocks (Lovibond et al., 2009). This effect is known as “protection from extinction” (Lovibond et al., 2000; Rescorla, 2003), and has an analogue to safety behaviors that some clinicians consider contraindicated for treatment of anxiety disorders (Blakey and Abramowitz, 2016; Salkovskis, 1991). In this way, external stimuli, thoughts, or actions (e.g., keeping ones back to the wall) might preserve beliefs that the world is still dangerous, and that only through engaging in safety behaviors are negative consequences averted. Whether individuals with PTSD are more or less prone to these types of “protection from extinction” effects in laboratory experiments is unknown. Notably, the inability to inhibit conditioned responses in the presence of a concurrent safety cue (Laing and Harrison, 2021) is a characteristic phenotype that may distinguish PTSD from other psychiatric conditions like depression (Jovanovic et al., 2010). As such, it may be more difficult for individuals with PTSD to form what ultimately amounts to a maladaptive safety association in the first place.

Figure 2.

Examples for how seemingly small changes in the presentation of conditioned and unconditioned stimuli can affect the ability to learn fear and safety. These subtle nuances are detailed in the associative learning literature, but basic research evaluating these effects in clinical populations, particularly PTSD, is lacking.

2.4. Higher-order conditioning.

In real world learning situations stimuli often form an indirect association with an event via an association with another stimulus (Figure 1). For example, someone attacked by a dog might later avoid parks where they are likely to encounter dogs, even if they were originally attacked by a dog inside of a house. In this case, prior knowledge that dogs are associated with parks shapes avoidance generalization to unique situations where the feared stimuli are anticipated. Transferring learned experiences and generalizing knowledge through indirect associations is a hallmark of human cognition. In the case of highly emotional experiences, the higher-order transfer of learning provides a number of possible routes by which inconsequential stimuli acquire negative emotional value (e.g., parks are now a feared context despite lacking a direct link with a fearful event). As trauma-focused therapy hones in on major aspects of the trauma, these indirect associations could maintain long-term fear associations.

There are several recognized ways beyond stimulus generalization in which cues form indirect associations with a meaningful event (Gewirtz and Davis, 2000; Rizley and Rescorla, 1972). For instance, in second-order conditioning a CS (CS1) is first directly paired with the US, and is then later paired with another CS (CS2). The CS2 takes on associative value by virtue of the CS1-CS2 association, and is therefore capable of eliciting a conditioned response without ever coming into direct contact with the US.

Sensory preconditioning likewise involves an indirect transfer of learning, only in this instance CS1 and CS2 are paired together before conditioning in the absence of overt reinforcement. When CS1 is then paired with the US, the pre-established CS1-CS2 link allows the CS2 to indirectly acquire associative value and later elicit a conditioned response. Sensory preconditioning is a form of latent learning, because the initial link between CS1 and CS2 is purely associative and does not demand an observable behavior (e.g., the association between a light and a tone). Much of what we learn occurs through latent processes, including linking disparate pieces of information to a central concept. Consider for example the number of associations built up over a lifetime with a single concept, such as “vehicle.” Following a horrific motor vehicle accident, a number of stimuli could act as indirectly conditioned stimuli and may be sufficient to trigger a fear-related defensive response. In this way, the number of pre-associated stimuli capable of later serving as a trauma-related reminder is truly immense. Experimental paradigms such as sensory preconditioning provide a tractable laboratory model to investigate how these latent associations could serve as hidden links to trigger PTSD symptoms following trauma (Dunsmoor et al., 2011).

Contemporary research on the neurobiological substrates of higher-order conditioning are somewhat limited, but a number of influential network models center on the role of the hippocampus (Gluck and Myers, 1993). The hippocampus is implicated in pattern completion processes (O’Reilly and McClelland, 1994), which higher-order conditioning presumably involves as a means to imbue associative value based on retrieving stored stimulus representations. Sensory preconditioning protocols in rodents (Wong et al., 2019) and humans (Wimmer and Shohamy, 2012) show that the hippocampus integrates memory to promote generalization. Thus, the hippocampus connects disparate elements separated by time into a common association, enabling the transfer of new experiences.

Despite theoretical understandings that higher-order conditioning likely serves as a route to maintain or exacerbate fear in PTSD (Foa et al., 1989; Keane et al., 1985), higher-order conditioning protocols in PTSD patients are lacking almost entirely. As such it is not clear whether fear associations transfer more readily to indirectly associated cues in PTSD patients, or whether the tendency to readily transfer fear is a predisposing risk factor for development of PTSD.

2.5. Delayed-onset PTSD and modeling the persistence of fear.

A major point of criticism of fear conditioning models of PTSD is that these models more appropriately detail elements of Acute Stress Disorder (Bryant, 2010) than the development of PTSD, per se. That is, conditioned fear learning develops rapidly and is observed immediately, often requiring only a few CS-US trials. Diagnostic criteria for PTSD, however, require symptoms to persist for at least a month. As noted above, heightened fear responding following trauma is normative and what differentiates PTSD is the lack of attenuation of fear responding over time, necessitating delayed measurements in a laboratory model to capture the point at which response patterns fail to attenuate. Additionally, in some cases, PTSD symptoms may not develop until several months or longer after the traumatic experience, known as delayed-onset PTSD. Such manifestations are commonly reported in cases of chronic exposure to traumatic events. For example, in combat veterans, an initial semi-successful resolution of the initial response to a traumatic event may collapse due to ongoing and repeated exposure to traumatic stressors (Andrews et al., 2009). The typical conditioning protocol is optimized for generating rapid learning, and under most cases this behavior either stabilizes or starts to diminish over weeks and months. In humans, the typical one- or two-day conditioning protocol is ill-equipped to model distinctions in normative versus dysregulated fear over the time-course required for PTSD diagnosis.

Conditioning research reveals situations where fear learning can increase over time. This process could be related to incubation of the fear memory (Eysenck, 1968), as well as a loss of memory for the specific details of the CS that leads to greater behavioral generalization over time (Riccio and Joynes, 2007). Protocols in rodents that deliver footshock over the course of several days (Pickens et al., 2009), use much longer CSs (Pickens et al., 2010), or present more shocks during training (Poulos et al., 2016) show that fear emerges more slowly, is sustained for longer periods of time, and is more likely to generalize. The major practical limitation to testing fear incubation in humans is the length of time needed for extensive training. However, incubation-like effects in rats were obtained in a single session of contextual fear conditioning when using a high number of shocks (Poulos et al., 2016), a form of training referred to as stress-enhanced fear learning (Rau et al., 2005).

Emerging neurobiological evidence in rodents is revealing crucial information on a reorganization of fear memory over time that may have significant implications for neuroscience-informed treatment for PTSD. Specifically, auditory fear memory retrieval originally supported by the basolateral amygdala at recent time intervals (several hours) shifts to the paraventricular nucleus of the thalamus over longer retention intervals of several days to a month (Do-Monte et al., 2015). Projections from the prelimbic region of the mPFC, involved in the retrieval of fear memory, likewise shift from the basolateral amygdala at recent intervals, to the paraventricular nucleus of the thalamus at longer retention intervals (Do-Monte et al., 2015). This may bear on earlier findings in primates that the amygdala is critical for the initial acquisition, but not the retrieval and long-term expression of conditioned fear (Antoniadis et al., 2007). Further research in this area is warranted, including translation to human neuroimaging, but it might suggest that fear memories shift away from the basolateral amygdala over time and are reorganized in different neural structures with cortical-subcortical connections (Do Monte et al., 2016).

One straightforward modification for any conditioning protocol could involve testing the remote effects of fear and extinction training following a more prolonged duration than the typical 24-hour delay. This would simply require subjects to return for one additional visit after a month or so following fear conditioning to more closely approximate the diagnostic criteria of PTSD, which requires symptoms to persist for over one month. Altogether, there is currently extremely little research assessing how learned fear changes over time in humans, and whether fear memories in PTSD are more prone to stabilize or strengthen over an extended conditioning-to-retention interval.

2.6. Retrospective revaluation.

Another factor that may contribute to delayed-onset of PTSD is a reinterpretation of a past event given new insights or information. For example, individuals who survive a trauma may not immediately realize the severity of the event, but experience PTSD symptoms upon learning more about the event (Ehlers et al., 2002). Another instance of delayed onset can occur in military veterans returning from deployment to a combat zone, where situations back home are appraised to be unhelpful, maladaptive, or overwhelming.

Retrospective effects have been of particular interest for bringing cognitive processes into the fold of conditioned learning (Dickinson and Burke, 1996), as they reveal the influence of mental stimulus representations on learning. A classic example from the learning literature is known as US inflation (White and Davey, 1989), whereby exposure to an isolated and more intense version of the US increases the strength of the conditioned response to the CS it was originally paired with. Such effects could have valuable explanatory power for the etiology of PTSD. For instance, PTSD following a person’s first exposure to a traumatic event is considerably rare. By contrast, one of the strongest predictors of developing PTSD following trauma is the individual’s history of prior trauma. Learning models that describe how a new stressful event enhances fear associated with an accumulated prior history of trauma could be an important factor for a more integrated and comprehensive learning model of PTSD.

The opposite of US inflation, known as US deflation (or devaluation), is perhaps a useful and alternative construct to extinction for understanding aspects of successful psychotherapy. In post-conditioning deflation/devaluation the conditioned response to the CS is reduced through weakening the aversive US in isolation, rather than extinguishing the CS in the absence of the US (Colwill and Rescorla, 1985). Basic research on post-conditioning inflation and deflation is firmly grounded in learning theory, but there is minimal contemporary neuroscience research on these effects in humans, and minimal or no efforts to examine these processes in psychiatric populations.

Other well-known retrospective effects involve competition between different cues battling for association with the US (Miller and Escobar, 2002). For example, when two cues (A and X) are paired together as a compound, the more salient cue (say A) overshadows the weaker cue (say X) and appears to prevent it from forming an association with the US (Pavlov, 1927). Presenting the X alone after compound conditioning will therefore not elicit a conditioned response. An interesting effect can occur, however, if A is presented alone and extinguished after compound conditioning (A−); X alone might then elicit a response (Lovibond et al., 2003). This effect is known as release from overshadowing, and is a remarkably clever protocol to show that animals represent numerous associations even if they don’t initially express that learning. In this case, subsequent learning that the more salient CS does not predict the US retroactively renders the weaker CS as the sole predictor. Put another way, if A doesn’t predict the US, then X must.

Release from overshadowing is a complex phenomenon, and much of the basic research is in the realm of non-affective causal learning experiments in humans (Lovibond et al., 2003). It holds unique potential for models of PTSD, and in particular may help explain unpredicted effects of treatment. For instance, exposure treatment could simultaneously reduce the associative strength of salient trauma reminders, while in turn releasing other less salient reminders from their latent associations to the trauma. This could potentially render previously hidden stimuli to become powerful triggers.

2.6. Alternatives to conditioning models of PTSD.

While conditioning models and protocols are widely popular, stress models have contributed considerable knowledge on the etiology of PTSD. An advantage to stress models is they better mimic life-threatening aspects of trauma than the electrical shock unconditioned stimulus used in Pavlovian conditioning, which lacks ethological validity (Deslauriers et al., 2018). Specifically, animal protocols of forced swim, predator-stress, physical restraint, and social defeat produce enduring physiological effects and better reflect high-intensity events relevant to real-world human trauma. Notably, inescapable/unpredictable shock protocols toe the line between a stress protocol and a conditioning protocol; there are elements of contextual fear conditioning, though the parameters (e.g., shock intensity and frequency) are modified to induce either associative or non-associative (sensitization) fear (Siegmund and Wotjak, 2007).

Broadly speaking, animal stress models show changes in glucocorticoid receptivity, HPA axis functioning, and functional and structural changes in regions with a dense concentration of corticosteroid receptors, such as the hippocampus (Roozendaal et al., 2009). Fear conditioning can also be construed as a stress-inducing experience that produces effects on glucocorticoid systems and functional changes in neurocircuitry underlying emotional learning, memory, and regulation (Rodrigues et al., 2009).But a single incident of Pavlovian conditioning is not known to induce lasting structural abnormalities observed in certain stress protocols. And as discussed earlier, fear acquisition is a normative (i.e., non-pathological) process critical to learning and reacting to potential danger. In the context of human fear conditioning, the effects of stress is mostly considered for its neuromodulatory role on learning and memory processes critical to the consolidation and retrieval of fear and extinction (Raio and Phelps, 2015).

While stress-models offer certain advantages over fear conditioning models in producing neurobiological processes associated with human PTSD (e.g., altered fear circuitry structure), there are several disadvantages of stress models of PTSD. A practical disadvantage is the inability to ethically translate the conditions of high-intensity and long duration stress to human research. While there is considerable research on effects of social and physiological stress induction in humans, protocols that model life-threatening trauma and the resulting outcomes cannot be ethically reproduced in human research. Another disadvantage to stress models is they lack a clear learning and memory component. While an intense stressor provides a valuable model for the development of PTSD by producing a surge of stress hormones (among other outcome), the conditioning framework is better suited for understanding what information is extracted and later remembered from the experience itself. In addition, many of the stress models are not comparable to the index trauma required for a DSM5 diagnosis of PTSD but instead are characterized by chronic unpredictable stress. These models have been utilized in animal models of depression and PTSD because they result in alterations in HPA axis activity. However such models which include, for example, electric foot shock, reverse day-night cycle, restraint stress, unfamiliar home environment, etc. are more closely linked to the clinical picture of complex PTSD which is a very different phenotype, often associated with early life trauma as well as recent life events.

In sum, we do not to argue for benefits of one type of animal model of PTSD over another. (For a more comprehensive review detailing the pros and cons of different animal models of PTSD the reader is directed to (Bienvenu et al., 2021; Richter-Levin et al., 2019). Given the ubiquity of conditioning models of PTSD, used both in animal and human studies, we focused our critical evaluation toward basic research on fear conditioning. It is in the conditioning model where there is arguably more room for novel methodological conceptualizations to foster improved translation of animal models to human subjects research.

2.8. Summary

Animal models have been incredibly useful to provide needed mechanistic specificity to the study of PTSD through precise neurobiological experimentation. While chronic stress models demonstrate success in reproducing important phenotypes consistent with PTSD, they lack a strong learning and memory component that are central to PTSD models focused on explaining important aspects of human cognition. The fear conditioning model, while overly simplistic in assumptions of a single incident producing persistent phenotypes consistent with PTSD, has proved to be a dominant theoretical and empirical model for investigating PTSD related phenotypes across species. Conditioning models provide a clear explanation for learning and memory phenomena, and associative learning models are supported by a tremendous amount of animal and human basic laboratory research. A rich and nuanced associative learning literature provides a number of avenues to augment laboratory models for PTSD. This includes modeling the generalization of fear following initial learning, understanding extinction generalization to better model the goals of clinical treatment, higher-order learning to account for complex spread of fear, and the role of retrospective revaluation to model delayed onset PTSD. The effect of retention intervals comparing recent versus remote fear and extinction retrieval may prove to be a critically underappreciated factor in basic laboratory research, as emerging neurobiological evidence indicates a temporal shift in neural organization of fear memory over time. Fortunately, established paradigms for investigating the multiple ways animals form and extinguish fear associations can be straightforwardly adapted directly to the study of PTSD both in laboratory animals and in humans. Further research on these topics, and integration with existing models, will increase the explanatory power of conditioning-based models on basic mechanisms of PTSD and help bridge the gap between laboratory research and innovative clinical treatments.

3. Beyond Models of Fear Conditioning and Extinction for Understanding PTSD

In addition to the expansion of fear conditioning and extinction research discussed above, there are other domains of basic research on human cognition that can inform our understanding of PTSD and aid the translation of basic discovery to clinical practice. We highlight a few of these domains below, including episodic memory, emotion regulation, and social cognition. While it is certainly the case that these processes can to some degree be targeted with animal models, the subtlety and nuance of these processes, and how they manifest in PTSD, may be better suited to basic laboratory research with humans.

3.1. Episodic Memory.

Memory disturbances are a predominant feature in many influential theories of PTSD (Brewin, 2011; Elzinga and Bremner, 2002). The emotional nature of trauma lend these experiences to prioritized consolidation via stress-enhanced memory modulation (de Quervain et al., 2017). As a result, there is a bias to recall trauma-related details and elements of the event form the content of intrusive memories. Highly emotional memories can occur spontaneously (e.g., flashbacks) and are associated with a sense of reliving the experience. Involuntary memory retrieval can be triggered by a host of idiosyncratic environmental or internal cues, speaking to the widespread generalization of associative value captured by conditioning-based models. The frequency and intensity of involuntary emotional memory retrieval is a characteristic feature of PTSD that distinguishes it from other affective disorders (Brewin and Holmes, 2003). At the same time, detailed memories surrounding the traumatic event are sometimes impoverished or disorganized in time (Foa et al., 1995; Harvey and Bryant, 1999), and individuals can be challenged to retrieve specific autobiographical details (Buckley et al., 2000; Van der Kolk and Fisler, 1995).

Influential mnemonic theories of PTSD propose certain aspects of a trauma memory are not accessible in the same way as other types of declarative memory, giving rise to the sense of disorganization, decontexualization, and fragmentation (Brewin et al., 1996; Elzinga and Bremner, 2002). These characteristic memory disturbances may arise from difficulties incorporating the intensity of the emotional experience with prior knowledge structures that constitute a sense of self. Therefore, attempting to reconcile the memory of trauma into a larger schematic memory structure composing a sense of self remains a goal of certain forms of PTSD treatment (Brewin, 1989; Brewin et al., 1996).

Some core mnemonic symptoms of PTSD are well described by animal models of learning and memory. In particular, arousal triggers upregulation in hippocampal-amygdala circuits selectively biasing encoding and consolidation of memories encoded around the time of an emotional event (de Quervain et al., 2017; McGaugh, 2015). Other aspects of memory disturbances in PTSD are less adequately captured by animal learning models. More specifically, an animal learning framework does not straightforwardly address aspects of declarative memory on the whole, including PTSD related symptoms of temporally disordered memory or spontaneous and intrusive memories. In these cases, a cognitive framework has proved useful to extend empirically tractable theories of memory disturbances in PTSD (Brewin, 2001; Ehlers and Clark, 2000; McNally, 2006). Focusing on aspects of declarative memory allows for integrating and leveraging research on the cognitive neuroscience of memory to understand memory disturbances in PTSD.

3.2. Enhancement of emotional episodic memory.

Early information processing models of PTSD built on a conditioning framework by incorporating the concept of associative-semantic networks that are activated by the emotional event (Foa and Kozak, 1986; Lang, 1977). These links include related information, propositional knowledge, beliefs, meaning, and response elements. Through these diffuse triggers, the emotional memory is easily retrieved, leading to characteristic fear-related symptoms of PTSD such as hyperarousal, hypervigilance, and avoidance.

Animal models have been exceptionally applicable for understanding how emotional experiences persist in memory and strongly resist forgetting. In short, stress hormone activation shortly before, during, or after a learning event modulates consolidation processes via connections between the basolateral amygdala and hippocampus (McGaugh, 2015). This neuromodulation helps ensure that learning that involves the hippocampus, such as object recognition, spatial learning, and context conditioning, are secured in long-term memory. Widespread projections from the amygdala subnuclei to other brain regions provides a route for arousal to bias other types of learning supported by those regions (Packard et al., 1994).

This arousal-mediated framework has been effectively translated to research on human emotional memory, such that emotional material is easier to retrieve with a higher sense of vividness and confidence than mundane everyday experiences. Notably, the bias toward emotional memory tends to be most evident over time, due to a combination of selective preservation of emotional information and normal forgetting of neutral information (Kleinsmith and Kaplan, 1963; Sharot and Phelps, 2004). Emotional memory processes in humans likewise involve connectivity between the hippocampus and amygdala (Murty et al., 2010). Hybrid designs merging fear conditioning with episodic encoding in humans show selective biases in episodic memory for neutral CS+ items from a semantic category associated with shock (Dunsmoor and Kroes, 2019; Dunsmoor et al., 2015a). Similar to emotional memory enhancements, these conditioning-induced declarative memory enhancements correlate with amygdala activity at encoding (de Voogd et al., 2016b) as well amygdala-cortical connectivity during post-encoding rest (de Voogd et al., 2016a).

However, selective retention of emotional memory is an adaptive process and is clearly not itself pathological (Cowan et al., 2021). As such, emotional memory paradigms in animals and humans provide important evidence for the psychological and neurobiological processes that preserve these types of memories. However, whether these paradigms can be used to distinguish normative mnemonic process from memory disturbances seen in PTSD and other mental health disorders is less clear. The clinical literature suggests trauma-memories are persistent, intrusive, and easily retrieved. However, while subjects frequently recall more emotional items in laboratory experiments, evidence that PTSD is associated with consistently stronger enhancements of negative emotional memory is inconsistent (Durand et al., 2019). For example, recent studies using hybrid fear conditioning/episodic memory designs show better 24-hour recognition memory for the CS+ versus CS− items, but this selective emotional memory enhancement is not different between healthy and PTSD groups (Hennings et al., 2021b; Morey et al., 2020).

There is evidence that certain types of everyday memories are impaired in PTSD, indicating a general episodic memory deficit (Brewin, 2011); though whether these deficits provide a possible mechanistic account of PTSD etiology or occur as a consequence of trauma is unclear from the literature, which in general is dominated by cross-sectional rather than longitudinal studies. Indeed, to the degree that episodic memory is mediated by hippocampal-dependent processes, it is relevant to note that while decreased hippocampal volume is commonly observed in PTSD, it may actually be a pre-morbid risk factor for development of PTSD (Gilbertson et al., 2002). It is worth noting, however, this remains a very controversial area, as the precise etiology of reduced hippocampal volume remains obscure, and the underlying neurobiological mechanisms involved are presently unclear (e.g. reduced dendritic sprouting, synaptogenesis, reduction in glia or neuronal number).

Neuroimaging of emotional episodic memory in PTSD has also yielded mixed evidence for abnormal hippocampal-amygdala processing during encoding of emotional material (Durand et al., 2019; Hayes et al., 2012b). Using a dimensional approach in a sample of African American women at risk for PTSD, Stevens et al. (Stevens et al., 2018) found that re-experiencing symptoms, but not negative affect, correlated with encoding-related activity in the hippocampus and amygdala for items that were successfully recalled. Interestingly, these effects were not selective to negative emotional material. This finding is consistent with prior evidence of stronger hippocampal-amygdala engagement for neutral emotional material in PTSD versus controls (Hayes et al., 2011). Behaviorally, however, these findings are somewhat at odds with earlier research showing biases to recall trauma-related over neutral material in PTSD (e.g., Zeitlin and McNally, 1991). An additional complication is documentation of individual differences (e.g., variability in genes conferring risk for PTSD (Fani et al., 2013) modulating hippocampal responses to threat), underscoring the complexity of identifying neurocircuitry processes implicated in emotional memory as mechanisms of PTSD.

One explanation is that emotional memory enhancement is a normative process that simply fails to distinguish healthy and clinical populations. This explanation also applies to fear conditioning studies, which frequently shows equivalent fear acquisition to a CS+ between healthy and clinical groups, as described earlier. What is perhaps distinctive is an overreliance on emotional memory circuitry to encode the neutral information. This may indicate a failure to utilize discrete memory encoding processes for emotional and innocuous material when subjects are viewing intermixed evocative and neutral images. Again, this explanation would align with findings from conditioning research, wherein PTSD groups exhibit generalization to the CS− (Duits et al., 2015) and fail to inhibit fear to safety cues (Jovanovic et al., 2012). Thus, from a translational research perspective, emotional episodic memory paradigms may provide more utility for revealing abnormal processing of neutral stimuli presented in the temporal context of emotional material, rather than selective memory for emotional material per se.

3.3. Memory suppression.

The ability to stop memory retrieval processes to thwart an emotional response is characteristic of successful emotion regulation. People often develop strategies to reduce intrusions of unpleasant memories triggered spontaneously or by environmental cues. Akin to top-down inhibitory emotion regulation, described below, it is considered in many respects a uniquely human process that relies on the right dorsolateral prefrontal cortex (dlPFC) downregulating activity in the hippocampus (Anderson et al., 2004; Benoit and Anderson, 2012) and the amygdala (Benoit et al., 2016). As such, it is challenging to develop an animal model of memory suppression, because intrusive memory is problematic to model outside humans and the cortical regions implicated are less developed. Importantly, this ability to suppress negative emotional memories is impaired in PTSD (Brewin, 2011; Ehlers et al., 2004).

In a recent study of survivors of a 2015 terrorist attack in Paris, those who developed PTSD were impaired at regulating re-experiencing of neutral intrusive memories and failed to engage a memory control network centered on the right dlPFC (Mary et al., 2020). They proposed that inability to engage a normative memory control system, even for the types of neutral memoranda used in their task, may present a preexisting vulnerability to the development of PTSD; though the cross-sectional nature of the experimental design limits a strong conclusion of whether a general deficit in memory control is a predisposing risk factor for PTSD or a consequence of trauma.

Anderson & Floresco (2021) noted the resemblance between retrieval suppression and fear extinction in regards to reducing the affective salience of negative memories. Notably, there is considerable overlap in fMRI studies of extinction, memory suppression, and explicit emotion regulation (Fullana et al., 2018a; Picó-Pérez et al., 2019). They proposed an integrated framework for re-conceptualizing extinction as mnemonic inhibitory control. This could provide a rich avenue for novel research efforts in PTSD combining knowledge across these domains.

There have been two recent efforts at combining explicit memory control and fear extinction in healthy adults. Wang et al. (2021) found that instructions to consciously suppress a conditioned stimulus during extinction reduced the return of fear, and the effect was stronger in individuals who scored higher on thought-control ability. In contrast, Hennings et al. (2021a) found that thought suppression during extinction reduced extinction generalization to conceptually related cues presented at a 24-hour renewal test. In this case, the instruction to suppress thoughts during extinction was interpreted as a conditioned inhibitor (Laing and Harrison, 2021) that protected the CS from extinction, interfering with the ability to learn that the CS would be safe in a context devoid of instructions to suppress your thoughts. Methodological differences between these designs could explain the conflicting results, and speak to the need for more research in this area. It will be particularly important to examine memory control of conditioned fear directly in PTSD patient groups using neuroimaging, both to test the hypothesis that memory suppression during extinction utilizes an inhibitory network that involves the dorsolateral and ventromedial prefrontal cortex, and to link this work to recent findings of impaired memory suppression networks in PTSD (Mary et al., 2020).

3.4. Memory (dis)organization.

PTSD has been associated with fragmented memories of trauma that result in disorganized, short, and simplistic narratives that may contain important gaps (Brewin, 2011). One explanation involves the effect of intense stress disrupting encoding and consolidation processes in the hippocampus, a region involved in contextualizing details in long-term memory. Stress clearly affects the structure and function of the hippocampus (McEwen, 1999), and a recent meta-analysis demonstrates decreased hippocampus volumes among individuals with PTSD compared to traumatized and non-traumatized controls, as well as decreased hippocampal volumes among traumatized controls compared to non-traumatized controls (Bromis et al., 2018).

There has been increasing interest in understanding how the brain organizes events in memory, and the factors that link or separate events at different timescales (Clewett et al., 2019). Events presented sequentially are often bound together within a temporal context, such that memory retrieval facilitates retrieval of other information encoded in temporal proximity (Howard and Kahana, 2002). However, when sequential events are separated by an “event boundary” (Zacks, 2020) (such as a context shift or even a subtle change in the environment), retrieval for information encoded across the boundary is reduced (Clewett et al., 2019). These event boundaries also affect our ability to remember the temporal order of events across boundaries (DuBrow and Davachi, 2013; Heusser et al., 2016) and creates a subjective feeling that events across boundaries were further apart in time (Ezzyat and Davachi, 2014). There has been a surge of neurophysiological and neuroimaging research across laboratory animals and humans investigating the effects of event boundaries on integrating and differentiating events encoded over time. Altogether, event boundaries are important for chunking continuous experience into discrete segmented events, and play a crucial role in organizing our episodic memories.

Relevant to a discussion of memory organization in PTSD, emotional states can serve as an event boundary segmenting continuous encoding experiences. For example, Dunsmoor et al., (2018) proposed that the transition from fear conditioning to extinction acts as an event boundary segmenting the two experiences that may retroactively protect fear associations from being forgotten during extinction. Using a hybrid episodic/associative conditioning design, individuals associated a series of category exemplars with an aversive electrical shock (conditioning), and then encoded new category exemplars without shock (extinction). Participants then underwent a surprise recognition memory test 24-hours later for exemplars encoded the previous day. There was a sharp reduction in episodic memory for exemplars encoded during extinction, while exemplars encoded during conditioning were strongly remembered with high-confidence. Remarkably, the reduction in memory occurred for exemplars encoded immediately after the transition from conditioning to extinction, a period of time when subjects still expected the electric shock and evinced heightened autonomic arousal. Further, this segmentation of item memory between conditioning and extinction was only evident at a 24-hour memory test and when a short break of only a few seconds was injected between conditioning and extinction. There was no reduction in extinction-specific memory when the conditioning-to-extinction was continuous (i.e., without a perceptible contextual change), or when memory was tested immediately. The authors proposed a model whereby, over a period of consolidation, memories associated with fear are segmented and prioritized in relation to conceptually similar but conflicting memories of safety, in line with the idea that extinction is a weaker memory that can fail to override retrieval of fear associations over time (but see Totty et al., 2019).

3.5. Social cognition

There is increasing recognition that social functioning impairment is not just a core symptom of PTSD, but an impairment that underlies other symptoms. For example, network analysis shows impairment in relationships is among the top two types of functional impairment exerting influence on PTSD symptoms (Ross et al., 2018). One of the more common ways social cognition has been assessed in PTSD is through emotional facial recognition tasks. Individuals with PTSD show impaired emotional facial recognition and perspective taking (Stevens and Jovanovic, 2019), and these deficits appear to be distinct from impairments seen in anxiety disorders (Lavoie et al., 2014).

Another common laboratory measure of social cognition is the Trust Game. The classic Trust Game involves repeated monetary exchanges between participants and includes two roles: investor and investee. The investor determines an amount to invest in the investee, the investee receives tripled the investment amount and has to decide how much of the tripled investment to keep vs return to the investor. As such, the investor role operationalizes ‘trust’, and the investee role operationalizes ‘reciprocity.’ In one prior study, adult women with PTSD developed after exposure to interpersonal violence demonstrated decreased trust (i.e., lower investments) compared to controls in a manipulation of the Trust Game where a confederate investee provided relatively lower reciprocity (Cisler et al., 2015). Another recent study used a version of the Trust Game with varying confederates that displayed different levels of reciprocity and found lower trust among individuals with PTSD compared to controls towards the more cooperative confederates (Bell et al., 2019). While extant behavioral studies with objective behavioral measures of trust suggest decreased trust in PTSD, there have been no neuroimaging investigations to link these behavioral processes to neurocircuitry mechanisms.

Another literature demonstrates significant deficits in risk perceptions of social situations and increased risky social behavior among violence victims (Messman-Moore and Brown, 2006; Walsh et al., 2012). These deficits are notable as assaultive violence confers the greatest risk for PTSD among the various types of trauma (Kessler et al., 1995). One study among young women examined judgments of written descriptions of social situations with sexual victimization risk and found that greater histories of interpersonal violence were associated with higher thresholds for judging a situation as risky (Yeater et al., 2010). Another study found that the latency with which victimized young adult women decided to leave hypothetical risky social situations escalating towards rape significantly predicted subsequent revictimization (Messman-Moore and Brown, 2006). Data also suggests that college age women with histories of interpersonal violence demonstrate less response effectiveness and less response refusal in sexually risky situations (Yeater et al., 2011), less overall sexual assertiveness (Livingston et al., 2007), and greater sexual risk behaviors (e.g., number of sexual partners, frequency of unprotected sex, etc) (Messman-Moore et al., 2010). What is interesting about these data is that the common conceptualization of trauma exposure and PTSD would predict greater threat avoidance and threat hypervigilance; but these data suggest the opposite occurs uniquely when examined in a social-interpersonal setting, which clearly has implications for understanding heightened revictimization rates among individuals exposed to assaultive violence.

Motivated by this body of data on decreased risk perceptions in social situations and increased rates of revictimization, two prior studies have examined neurocircuitry encoding of social learning among adolescents exposed to physical or sexual assault with varying degrees of PTSD severity. In an initial pilot study of 30 adolescents (n=15 exposed to assault), participants completed a version of a three-arm bandit in which they observed financial investments towards female conspecifics that were either trustworthy (provided a return on investment) or untrustworthy (kept the investment) with varying probabilities across the task (Lenow et al., 2014). There was a dose-dependent relationship between assault and performance on the task, such that greater assault, independent of PTSD severity, was associated with worse identification of the empirically most trustworthy conspecific on the task. Additionally, there was a dose-dependent impact of assault, independent of PTSD, on neural encoding of negative social prediction errors (i.e., when trustworthy conspecific took the investment) on the anterior insula and dorsal anterior cingulate cortex, key regions of the salience network. A subsequent study of 60 adolescents girls (n=30 exposed to assault) performed a similar social learning task but also completed a standard facial emotion processing task (Cisler et al., 2019). A dose-dependent association between assault severity and decreased encoding of negative social prediction errors in the salience network was replicated, while concurrently observing heightened salience network as a function of assault on the facial emotion processing task. These data highlight unique cognitive deficits in social / interpersonal learning contexts vs standard threat detection tasks among individuals exposed to early life trauma. While forms of interpersonal / social behavior can certainly be modeled in animal studies, investigation of nuanced trust and social cognition relevant to understanding risk for, and maintenance of, PTSD is better suited to human laboratory studies. Given the importance of social support and risk for revictimization in PTSD, further study of these processes is required.

Other recent data directly integrates social support and fear learning, with implications for PTSD. A potent example comes from work demonstrating that social support reminders enhance fear extinction and lead to lasting reduction in response to fear reminders (Hornstein et al., 2016; Hornstein et al., 2018). When considering alongside the strong body of work showing that poor social support confers risk for PTSD after trauma exposure, and other work suggesting social environment is key to risk and resilience after trauma beginning in childhood (Hostinar et al., 2014; Stevens et al., 2016), it is clear social-emotional functioning is important in the development and maintenance of PTSD. However, better integration of social-emotional functioning into the study of cognitive processes that contribute to fear learning is necessary to understand this effect. For example, meta-analysis of studies distinguishing between social versus non-social fear stimuli has suggested that amygdala hyper reactivity to threat in PTSD appears to be limited to social stimuli (Hayes et al., 2012a). Although theories such as Hobfall’s conservation of resources and Cacippopo’s work on loneliness (Cacioppo and Cacioppo, 2014) provide compelling accounts for how PTSD symptoms may erode social relationships and persistent social isolation may increase attention bias toward threat, the transactional nature of associations between social-emotional functioning and fear learning have not been capitalized on in the laboratory.

3.6. The Human-Specific Dimensions of Regulating Emotions.

Although fear extinction is a learning process leading to the regulation of fear responses that is shared across numerous species, humans have the aptitude and capacity for numerous other methods for regulating emotional state. One major reason why the study of emotion regulation is important for the understanding and treatment of PTSD is that fear extinction procedures, even if optimized and perfectly delivered, will not eliminate all strong negative emotional responding. As such, the individual with PTSD additionally needs skills training and refinement in the cascade of cognitive processes required to effectively manage strong negative emotional responses in order to maintain goal-relevant behavior. These cognitive processes are unique and not shared with other species by virtue of the complexity and enlargement of the brain’s cortical mantle and complex gyrification patterns. Emotion regulation can be defined as the process by which one shapes or alters the trajectory of an emotional response over time (Gross, 2002), and this process has been conceptualized as reflecting at least two orthogonal dimensions: one denoting the level of conscious awareness of the emotion regulation goal (called an implicit/explicit or non-conscious/conscious dimension) and the other denoting the nature of the change process (from automatic to controlled) (Braunstein et al., 2017b).

With animal models of fear conditioning and extinction, the nature of the regulatory process induced is often limited to those that are implicit and automatic, as it is presumed that explicit or controlled efforts to reduce emotional state are not deployed. Whether or not rodents can deliberately engage in more explicit and controlled methods of emotion regulation, humans routinely do so in attempts to regulate emotional behavior, often with success. A framework for explicit and controlled emotion regulation posits that the individual form, at bare minimum, the following four mental symbols to support such a process: first) an abstract representation of the emotion regulation goal to be reached in the immediate future (e.g., feel more calm); second) an overall representation of one’s current affective state to serve as an internal marker for comparison to subsequent affective states; third) a goal-directed generative process (e.g., generating different ways of viewing a situation to change subjective meaning) or a context-dependent interpretive process (e.g., attempting to suppress an emotional response because it is not socially appropriate, or reminding oneself when at the zoo that the lion in front of one is behind a plexiglass wall, and therefore one is safe) that can serve to alter the trajectory of one’s emotional response; and fourth) the ability to compare the current affective state to either the prior affective state representation or the abstract desired affective state representation (Etkin et al., 2015). Variations on these components give rise to numerous and diverse forms of emotion regulation strategies in humans, ranging from cognitive reappraisal to emotional suppression to attentional distraction to explicit expectancy-based automatic processes (e.g., placebo effects) (Braunstein et al., 2017b).

The nature of complex forms of emotion regulation (e.g., reappraisal) render some aspects of this process impractical for developing a translatable animal model. This highlights the critical need for human subjects’ research in patients with PTSD to further elucidate potential novel strategies for enhancing diverse forms of emotion regulation. Consistent with animal models of PTSD, the typical manifestation of emotion regulation processes in PTSD is one of difficulty with adaptively altering the trajectory of fear, as well as numerous other emotions such as anger, guilt, shame, and disgust (Shepherd and Wild, 2014) Particular trauma-focused psychotherapy packages, such as cognitive processing therapy, extensively focus on cognitive reappraisal-based strategies to alter negative beliefs and develop more helpful and adaptive ways of thinking and interpreting life experiences. This type of explicit, controlled emotion regulation strategy is often considered to be one of the most helpful and adaptive forms of regulating emotions (Braunstein et al., 2017a) while other types of higher-order strategies, such as emotional suppression, have often been found to be associated with more severe forms of PTSD symptomatology (Khan et al., 2021). However, both approaches generally recruit extensive areas of the prefrontal cortex in humans (Morawetz et al., 2017) further distinguishing these processes from the types of implicit behavioral control in rodent fear learning models.

Brain imaging studies have begun to explore these explicit and controlled emotional regulatory processes in individuals with PTSD (MacNamara et al., 2016; Rabinak et al., 2014) typically yielding results for abnormal prefrontal responses when regulating emotional responses to affective stimuli. Recent research also suggests that the neural circuitry underlying explicit regulation of emotional state through cognitive reappraisal seems to demonstrate a particularly salient change following a course of prolonged exposure therapy, an empirically-supported PTSD psychotherapy that relies primarily on exposure-based interventions which ostensibly engage implicit/automatic emotion regulation circuitry via fear extinction to generalization stimuli (Fonzo et al., 2017). This highlights a somewhat surprising and potentially important point that implicit/automatic and explicit/controlled emotion regulation processes are not separate mechanisms for therapeutic effects in humans, an insight that cannot be provided by animal model-informed experimental designs alone. There is abundant opportunity in this area to expand on potential novel intervention approaches to enhancing various emotion regulation processes in PTSD, which are complex and varied in humans and thus offer many ripe targets for clinicians to intervene.

4. Diagnostic limitations, heterogeneity, and subtyping

An additional challenge to the translation of basic laboratory research, from either human or non-human animal models, to clinical treatments for PTSD is recognition of significant heterogeneity in the construct of ‘PTSD.’ Ignoring challenges in diagnostic nosologies used to define and classify mental health disorders (Insel et al., 2010; Regier et al., 2009), there is an inherent problem of heterogeneity in PTSD symptom presentation. To illustrate, there are 21 symptoms of PTSD across 4 clusters, resulting in 636,120 possible combinations of symptoms that could result in a PTSD diagnosis (Galatzer-Levy and Bryant, 2013). This would not be problematic if it could be assumed that a single set of mechanisms was responsible for producing any given combination of symptoms; however, there is growing recognition that heterogeneity in PTSD symptoms actually reflects distinct psychopathological mechanisms. One of the earlier proposals for a distinct subtype of PTSD was for a dissociative subtype, characterized symptomatically by the presence of depersonalization and dissociation and mechanistically by increased medial prefrontal cortical modulation of emotional state (Lanius et al., 2010; Nicholson et al., 2020; Wolf et al., 2012). Similarly, the construct of ‘complex PTSD’ has been proposed as a distinct subtype of PTSD distinguished by the presence of more chronic early life trauma and increased dysfunction across a range of domains (Herman, 2012; Resick et al., 2012), and there is some evidence complex PTSD responds better to psychological treatments that include emphasis on emotion regulation (Bohus et al., 2020; Cloitre et al., 2010).

While initial conceptualizations of mental health disorder subtypes, such as these examples in PTSD, were largely based on or initiated by observations of clinical symptoms, subtyping of mental health disorders has become increasingly popular in the past decade and has used increasingly sophisticated statistical methodology focused on mechanistic observations, rather than symptoms, to define subtypes (Drysdale et al., 2017; Miranda et al., 2021; Woo et al., 2017). Some more recent examples in PTSD have used voxelwise patterns of brain activity during a cognitive task as input features to a clustering algorithm (e.g., kmeans clustering). Based on these high dimensional patterns of activity, the algorithm attempts to separate the data into distinct clusters, with each cluster representing a unique ‘profile’ of brain activity pattern. In one dataset of 114 adolescent girls with varying histories of trauma exposure and PTSD severity who completed a facial emotion processing task (Ahrenholtz et al., 2021; Sellnow et al., 2020), the clustering algorithm identified 3 distinct profiles of activity differentiated by patterns of engagement in the medial PFC, anterior insula, and hippocampus. Moreover, these distinct profiles of brain activity were uniquely associated with histories of trauma exposure and symptom severity, which is notable because the clustering algorithm was blind to any symptomatic information and identified the clusters solely through brain activity profiles. In another example among 76 women with PTSD related to assaultive violence completed a fear conditioning and extinction task, a clustering algorithm identified three unique profiles of brain activity patterns during conditioning and extinction, characterized by differences in engagement of salience, default mode, and visual networks. These profiles of brain activity patterns were associated with unique clinical profiles, including differences in neurocognitive abilities, childhood sexual abuse, and PTSD symptom severity, as well as unique differences in psychophysiological and cognitive responses during the conditioning and extinction task. This is particularly noteworthy given that the sample was ostensibly homogeneous with respect to biological sex, index trauma, and diagnosis. Finally, in another example among 146 recently traumatized adults who completed a battery of cognitive tasks during MRI (Stevens et al., 2021), a clustering algorithm identified three profiles of brain activity patterns across the tasks. These profiles of brain activity patterns were differentially related to subsequent trajectories of PTSD symptoms across a six month period, with one profile characterized by greater insula and dorsal ACC activity during a threat task demonstrating greater subsequent PTSD symptoms.

The body of literature related to mechanistic subtyping of PTSD is still small; nonetheless, this growing area of research suggests distinct subtypes of PTSD characterized by distinct psychopathological mechanisms. The larger concept that PTSD is non-unitary, and cannot be well captured by a single set of dysfunctional processes, has significant implications for translating basic laboratory research to a clinical setting. For example, increased knowledge about fear generalization would not be informative for a subtype of PTSD characterized more by dissociation and social isolation than by increased fear generalization. This problem raises the important considerations for laboratory research that 1) different sets of potential mechanisms, characterizing different subtypes of PTSD, are equally studied in laboratory research (e.g., researchers should not just focus on fear learning as though this process is equally relevant to all individuals with PTSD), and 2) knowledge gleaned on a particular mechanism in basic laboratory research is translated to the appropriate PTSD subtype and not to all of PTSD as though it’s a unitary diagnostic entity. Clearly, embracing nuance and complexity, rather than simplistic reductionism, will be necessary for informative bidirectional translation between laboratory and clinical settings.

5. Linking the molecular pathology of PTSD across species

There is an emerging focus on molecular-genetic approaches to better understand the etiology, diagnosis, and treatment response in PTSD (Koenen et al., 2008; Nievergelt et al., 2019; Stein et al., 2002). This includes large scale genome-wide association study (GWAS) consortia devoted to identifying genetic variants associated with increased risk for PTSD (Duncan et al., 2018; Logue et al., 2015). An important goal of this line of research is to determine whether gene expression can be used as a biomarker that differentiates PTSD pathophysiology from other psychiatric conditions. Fundamentally: what differentiates those who develop PTSD from those who are resilient following trauma? While an in-depth review of transcriptomic organization in PTSD is beyond the scope of this review (see Girgenti and Duman, 2018), transcriptome based RNA sequencing of human blood serum and is beginning to reveal differential expression in PTSD in genes associated with glucocorticoid receptor activity, immune responses, amygdala function, and fear regulation (Breen et al., 2015; Kuan et al., 2017; Kuan et al., 2021; Wuchty et al., 2021). There is emerging evidence using RNA-Seq that these patterns of gene expression may be associated with PTSD dimensions characterized by severity in different symptom clusters (Kuan et al., 2021; Waszczuk et al., 2020), aiding important work on PTSD subtyping.

Experimental research in rodents using modified fear conditioning models have replicated the molecular pathology of PTSD obtained from human serum and postmortem brain tissue. For instance, Sillivan et al. (2020) used a stress-enhanced fear learning paradigm and found expression of genes associated with heightened PTSD-related fear expression in the basolateral amygdala in mice vulnerable to heightened fear expression. Targeted overexpression of microRNAs (mir-135b-5p) in the basolateral amygdala promoted long-term fear memory retrieval and renewal in mice who were otherwise stress-resilient, while inhibiting mir-135b-5p in stress-susceptible mice reduced fear memory retrieval. Importantly, mir-135b-5p was also identified in human blood serum and postmortem brain tissue of military veterans with PTSD, providing important cross-species evidence of molecular pathology.

Notably, while the fear conditioning model may be capable of replicating aspects of molecular pathology of human PTSD (Sillivan et al., 2020), much more research is needed to extend this molecular-genetic approach to understand phenotypes of persistent fear memory and fear renewal in humans with PTSD. One area that lacks behind animal research involves human postmortem multi-omic studies that focus on brain regions that have functional engagement with fear-related processes that are presumably abnormal in PTSD, such as the amygdala, hippocampus, and medial prefrontal cortex. Notably, postmortem studies inherently preclude a direct link to laboratory-based phenotypes in the same individuals (e.g., extinction-retention deficits). As such, a more direct comparison of fear conditioning and stress paradigms across humans and laboratory animals are needed to compare homologous genetic transcripts obtained from these tasks. To date, transcriptome alterations in humans with PTSD is largely based on samples collected from individuals with and without PTSD (and other comorbidities) outside the context of laboratory research protocols (Lori et al., 2021; Wuchty et al., 2021). Thus, an important question going forward is whether the dynamic behavioral and neural abnormalities associated with PTSD in human research (such as extinction-retention deficits, fear overgeneralization, social deficits, and other areas detailed above) can be linked to genetic profiles associated with PTSD. Another avenue for future research is to detail whether genetic profiles associated with vulnerability to PTSD differentiate responses to conditioned fear and stress in individuals prior to developing PTSD. In this regard, cross-translational research optimized for recapitulating dimensional aspects of PTSD symptom clusters (as well as memory, social and emotion regulation deficits detailed above) are of critical importance for linking PTSD models at the molecular, cellular, and systems level.

6. Conclusions

While current gold-standard treatments for PTSD have solid empirical support for efficacy, there is growing consensus that the absolute effectiveness of these treatments is lacking (Bradley et al., 2005; Hoskins et al., 2015; Lee et al., 2016; Levi et al., 2021; Steenkamp et al., 2015). At the same time, our knowledge of the neurobiology and genetics associated with PTSD has profoundly advanced in the past several years. These advances are strongly driven by innovations in technical tools available for animal models, allowing fine grain inferences on the molecular, cellular, and systems neuroscience involved in processing fearful and stressful events. Unfortunately, there remains a translational gap whereby groundbreaking discoveries from animal models of PTSD have yet to produce neuroscience-informed avenues for optimizing mental health treatment.

Here, we’ve outlined important domains for understanding the development, maintenance, and treatment of PTSD that expand beyond what is typical for simplified implicit fear learning models in animals. An integrated neuroscience-informed model of PTSD in humans requires basic research also accounting for higher-order learning systems, episodic memory, social-emotional processes, and emotion regulation processes affected by this disorder. Comparison of these various cognitive processes between healthy and PTSD samples will be important for further defining deficits that contribute to component symptoms of PTSD. Moreover, laboratory-based experimental manipulations will be necessary for identifying novel ways to modify deficits in these myriad cognitive processes and directly inform translation to clinical practice. Heterogeneity in PTSD poses a formidable challenge to translation of laboratory-based studies to clinical practice, but can be addressed with multidisciplinary research incorporating careful definition of PTSD subtypes into laboratory-based mechanistic studies. Building on the fine-grained circuit-level insights afforded by animal models, further laboratory-based mechanistic studies using human models along these lines will facilitate greater translation of growing research knowledge about PTSD and related processes into more effective treatments.