1. Animal models of trait anxiety

Anxiety-like behavior is not exclusive to humans. Many nonhuman animals also have biological mechanisms that enable them to anticipate and successfully cope with various threat-related stimuli in the environment. Except for phenomenological and highly subjective aspects that are inherent to reports of human anxiety, almost all other anxiety-related components show great similarity across mammalian species and have long been used to study anxiety in laboratory settings (for review, see Graeff, Reference Graeff, Koob, Le Moal and Thompson2010; Steimer, Reference Steimer2011). Since the seminal work of Hall (Reference Hall1934) that showed that rats with high and low levels of emotionality exhibit different patterns of exploration in an open field (i.e., more emotionality, less exploration), no other emotion has been more studied in animal models than anxiety and its related disorders.

Dozens of animal models of elicited anxiety have been validated, and many have been recognized as valuable or indispensable tools for studying defensive behaviors and searching for more effective and targeted treatments for specific anxiety disorders (Steimer, Reference Steimer2011). However, despite its obvious clinical important, trait anxiety is paradoxically omitted or rarely addressed in most of these studies. This is particularly concerning when attempting to model or simulate anxiety in anxiolytic screening studies and when investigating neurobiological mechanisms that underlie defensive behaviors and their possible associations with anxiety disorders. One reason omission of a trait perspective is problematic is that behavioral tests of anxiety in humans and animals always involve an interplay between a trait anxiety component (which reflects an individual’s vulnerability or susceptibility to anxiety) and the situation that elicits state anxiety at the time of testing. Thus, when extreme forms of these traits are ignored, it is difficult, if not impossible, to dissociate adaptive defensive reactions from eventual maladaptive defensive reactions that are supposedly associated with specific anxiety disorders. Moreover, trait anxiety is not directly observable. Instead, it is inferred as a tendency to anxiety that can only be phenotypically observed and assessed through a standardized anxiety-related measure (e.g., behavioral, physiological, and neural correlates) at the time of testing. Finally, interactions between trait and state anxiety have been found to influence both the direction and magnitude of a given treatment (Griebel et al., Reference Griebel, Belzung, Perrault and Sanger2000; Rao & Sadananda, Reference Rao and Sadananda2016).

One strategy to manipulate trait anxiety in animal models is bidirectional selective breeding for extremes in anxiety-related parameters (for review, see Steimer & Driscoll, Reference Steimer and Driscoll2003), such as high anxiety-related behavior (HAB) and low anxiety-related behavior (LAB) rats and mice (Carboni et al., Reference Carboni, Khoury, Beiderbeck, Neumann and Mathé2022; Landgraf & Wigger, Reference Landgraf and Wigger2003; Liebsch et al., Reference Liebsch, Montkowski, Holsboer and Landgraf1998), Roman high- and low-avoidance rats (Bignami, Reference Bignami1965; Giorgi et al., Reference Giorgi, Corda and Fernández-Teruel2019), Naples high- and low-excitability rats (Sadile et al., Reference Sadile, Cerbone, Lamberti and Cioffi1984; Pellicano & Sadile, Reference Pellicano and Sadile2006), the Syracuse (high- and low-avoidance) rat strains (Brush, Reference Brush2003; Brush et al., Reference Brush, Gendron and Isaacson1999), the Maudsley reactive and non-reactive strains (selected for emotional defecation; Broadhurst, Reference Broadhurst and Eysenck1960, Reference Broadhurst1975), the Tsukuba (high and low runway activity) rat strains (Blizard et al., Reference Blizard, Wada, Onuki, Kato, Mori and Makino2005; Fujii et al., Reference Fujii, Asada, Takata, Yamano and Imada1989; who also differ on defecation), Floripa H and L rat lines (Izídio, & Ramos, Reference Izídio and Ramos2007; Ramos et al., Reference Ramos, Correia, Izídio and Brüske2003), aggressive and non-aggressive mice (Benus et al., Reference Benus, Bohus, Koolhaas and van Oortmerssen1991; Miczek et al., Reference Miczek, Takahashi, Gobrogge, Hwa and de Almeida2015), and Carioca High-conditioned Freezing (CHF) and Carioca Low-conditioned Freezing (CLF) rats. The latter were developed by our group and associated laboratories, whose data and foundations as an animal model of GAD are discussed below.

2. CHF rat line as a model of GAD

Starting with a highly heterogeneous population of Wistar rats, our groups used a selective breeding protocol that has been in progress for the last two decades, to develop two new lines of rats. The lines differ in strong phenotypic divergence for high (CHF) and low (CLF) trait anxiety, respectively, with selection based on conditioned freezing scores in the well-known contextual fear conditioning paradigm (Bolles & Fanselow, Reference Bolles and Fanselow1982). The basic contextual fear conditioning protocol for our breeding separation involves two phases, trial and test sessions, that occur on two consecutive days. On the first day (conditioning trial), the rats are placed in a conditioning chamber. After 8 min (pre-shock period/baseline), they are exposed to three unavoidable mild electric footshocks. Twenty-four hours later, the rats are placed in the same conditioning chamber (context), but no shock is delivered. In this second exposure (test session), a trained observer records the occurrence of freezing behavior for 8 min according to a time-sampling schedule. Rats were scored every 2 s as either freezing or not freezing. Freezing was defined as a crouching, immobile posture with no movement other than that required for breathing. Contextual freezing behavior is then converted to a percentage as an anxiety-like measure. The breeding protocol, which we have uninterruptedly conducted in 42 successive generations since 2006, consists of the selective mating of male and female rats with their respective highest and lowest percentages of contextual freezing behavior. To better interpret differences between these two rat lines, a group of Wistar rats (CTL), composed of the offspring of randomized cross-breeding populations, is used as an additional control group in most of our studies. All animals were phenotyped at 2–3 months of age.

Over the past 16 years, more than 13 000 animals have already been phenotypically selected based on this protocol. Figure 1 presents the conditioned freezing behavior of our breeding lines across the 42 generations. As we reported in our first study (Castro-Gomes & Landeira-Fernandez, Reference Castro-Gomes and Landeira-Fernandez2008), CHF and CLF rats exhibited reliable differences in conditioned freezing after the first three generations of selection. Males from both lines consistently exhibit more conditioned freezing in response to contextual cues than females. We also noted that the shock parameters that we employed (i.e., three 1 mA, 1 s unsignaled electrical footshocks with an inter-shock interval of 20 s) were very high. After the fourth generation, the shock intensity was reduced until it reached 0·4 mA in the 12th generation. In the 13th and 14th generations, we increased the shock intensity to 0·5 mA and 0·6 mA, respectively. This shock intensity has remained until the present generation. Male and female CHF, CLF, and CTL animals systematically exhibit clear differences across the remained generations.

Figure 1.

Mean ± SEM percentage of the time spent freezing in male (top) and female (bottom) Carioca High Freezing (CHF) and Carioca Low Freezing (CLF) rats across 42 generations. Control animals (CTL) started in the fifth generation of the two breeding lines.

The CHF rat line has been identified as a valid and reliable animal model of GAD. We present a brief summary of how the main behavioral, pharmacological, physiological, and neurobiological findings from these animals constitute an animal model of GAD. Notably, no animal model of anxiety fully recapitulates all aspects of clinical anxiety in humans. Importantly, however, this is not the intention of such models. Typically, the validity of an animal model of anxiety or some other psychiatric condition is estimated by considering its degree of face validity, predictive validity, and construct validity (Treit, Reference Treit1985; Willner, Reference Willner1984), although there is often certain parsimony with regard to the need to fully meet these three criteria, depending on the purpose of the study.

Table 1 summarizes the major correspondence between findings from the CHF rat line and some of the main features of GAD in humans. One of the first behavioral findings from the CHF line was that these animals expressed their corresponding “anxious” characteristics not only in the contextual fear conditioning paradigm but also in the elevated plus maze (Cavaliere et al., Reference Cavaliere, Maisonnette, Krahe, Landeira-Fernandez and Cruz2020; Dias et al., Reference Dias, Bevilaqua, Silveira, Landeira-Fernandez and Gardino2009; Léon et al., Reference Léon, Castro-Gomes, Zárate-Guerrero, Corredor, Cruz, Brandão, Cardenas and Landeira-Fernandez2017; Salviano et al., Reference Salviano, Ferreira, Greidinger, Couto, Landeira-Fernandez and Cruz2014), one of the best-known animal models of anxiety (Cruz et al., Reference Cruz, Frei and Graeff1994; Handley & Mithani, Reference Handley and Mithani1984; Pellow et al., Reference Pellow, Chopin, File and Briley1985). The elevated plus maze is based on the naturally occurring approach-avoidance conflict in rodents that is related to their motivation to explore new environments and innate fear of heights and open spaces (Treit et al., Reference Treit, Menard and Royan1993). Importantly, this anxious behavioral profile that is observed in CHF rats in the elevated plus maze (i.e., decrease in open-arm exploration) was detected without significant changes in the total number of arm entries (open + closed arm entries) or absolute number of closed arm entries, thus indicating that the phenotyping protocol that is used for the selection of successive generations of these animals based on contextual fear conditioning does not produce significant general locomotor impairments, a key point in animal models of anxiety that require locomotor activity.

Table 1.

Correspondence between CHF rat line findings and some main features of generalized anxiety disorder in humans

A substantial body of evidence (Brandão et al., Reference Brandão, Zanoveli, Ruiz-Martinez, Oliveira and Landeira-Fernandez2008; Fanselow, Reference Fanselow2000; Luyten et al., Reference Luyten, Vansteenwegen, van Kuyck, Gabriëls and Nuttin2011; Phillips & LeDoux, Reference Phillips and LeDoux1992) indicates that conditioned freezing behavior in response to a context but not to an explicit cue (e.g., a tone that is previously associated with an aversive stimulus) share several behavioral, physiological, and neurobiological characteristics with GAD. Likewise, the elevated plus maze, at least in its usual form of a single 5-min session, also exhibits features of an animal model of GAD. An intricate relationship has been suggested between a type of behavior that is supposedly related to generalized anxiety, which would occur during the first 5 min of a single exposure to the test, with another type of anxiety (specific phobia) that results from a longer exposure (10 min) or second 5-min exposure to the test (File & Zangrossi, Reference File and Zangrossi1993). This view is consistent with findings that anxiolytics increase open-arm exploration during a single 5-min session in the elevated plus maze but lose this anxiolytic-like effect (i.e., “one trial tolerance”) when given in a single 10-min session or second 5-min session (File & Zangrossi, Reference File and Zangrossi1993). According to these authors, this mimics a condition that is supposedly related to specific phobia, for which anxiolytics are known to be ineffective (Bandelow, Michaelis, & Wedekind, 2017). Therefore, high trait anxiety in CHF rats exposed for 5 min to the elevated plus maze further corroborates the proposition that these animals are a model of GAD.

Other more recent behavioral findings also corroborate this view. For example, one of the main characteristics of GAD is its chronic course. Accordingly, Lages et al. (Reference Lages, Maisonnette, Rosseti and Landeira-Fernandez2021a) showed that CLF rats were unable to consolidate aversive memories, whereas CHF rats exhibited considerable percentages of freezing behavior even after multiple exposures to the context that was previously associated with the aversive stimulus, with the additional interesting feature of being susceptible to extinction. In another study that compared patterns of freezing behavior in CHF and CLF rats in response to the context or cue (a tone that was previously associated with an electric footshock), Macêdo-Souza et al. (Reference Macêdo-Souza, Maisonnette, Filgueiras, Landeira-Fernandez and Krahe2020) showed that CHF rats froze more than CTL rats and these more than CLF when exposed to the context, that is associated to generalized anxiety disorder (Luyten et al., Reference Luyten, Vansteenwegen, van Kuyck, Gabriëls and Nuttin2011) but not to the cue that is supposed linked to specific phobia (Garcia, Reference Garcia2017; Grillon et al., Reference Grillon, Baas, Pine, Lissek, Lawley, Ellis and Levine2006).

Potentially threatening situations are also known to activate the hypothalamic–pituitary–adrenal (HPA) axis (Hinds & Sanchez, Reference Hinds and Sanchez2022). Consequently, GAD patients have been found to have elevated cortisol levels (Lenze et al., Reference Lenze, Mantella, Shi, Goate, Nowotny, Butters, Andreescu, Thompson and Rollman2011). Again, CHF rats were equally selective for this parameter as reported by Mousovich-Neto et al. (Reference Mousovich-Neto, Lourenço, Landeira-Fernandez and Corrêa da Costa2015). These authors showed increased neuroendocrine responses (i.e., increased serum corticosterone) in CHF rats comparted to control animals.

Benzodiazepines are among the most commonly prescribed and effective medications for GAD (Gomez et al., Reference Gomez, Barthel and Hofmann2018). For this reason, many animal models of anxiety are pharmacologically validated based on this class of drugs. Thus, in one of our studies (Cavaliere et al., Reference Cavaliere, Maisonnette, Krahe, Landeira-Fernandez and Cruz2020), systemic injections of the benzodiazepine midazolam (0·25, 0·5, 0·75, and 1·0 mg/kg, i.p.) selectively increased open-arm exploration in CHF rats exposed for 5 min to the elevated plus maze test. Interestingly, however, this anxiolytic-like profile was only observed at the lowest dose tested (0·25 mg/kg) in CLF rats. This observation is consistent with previous findings that anxiolytic-like effects of benzodiazepines and mainly serotonergic anxiolytics appear to depend on the animals’ level of anxiety before testing (Blanchard et al., Reference Blanchard, Hynd, Minke, Minemoto and Blanchard2001). Accordingly, in another study, systemic (0·5 mg/kg, i.p.) and intra-infralimbic cortex (5 nmol/ml) injections of the preferential 5-HT_2A receptor antagonist ketanserin induced anxiolytic-like effects in CHF rats but anxiogenic-like effects in CLF rats in the elevated plus maze (León et al., Reference Léon, Castro-Gomes, Zárate-Guerrero, Corredor, Cruz, Brandão, Cardenas and Landeira-Fernandez2017).

Another interesting finding that is also consistent with characteristics of GAD in humans refers to alcohol consumption. Alcohol is known for its “anxiolytic” properties in humans and animals. So, since GAD has also been associated with alcohol abuse in clinical and nonclinical populations (Kushner et al., Reference Kushner, Sher and Beitman1990; Smith & Randall, Reference Smith and Randall2012), this association was also recently investigated in the CHF rat line (Bezerra-Krounis et al., Reference Bezerra-Karounis, Krahe, Maisonnette and Landeira-Fernandez2020). As expected, CHF animals consumed more alcohol than CLF and control animals, which opens the possibility of using this model to better understand the comorbidity between GAD and alcohol abuse.

It is important to mention that anxiety and depression are independent disorders, although, clinical studies have shown that there is a high level of comorbidity between them (Groen et al., Reference Groen, Ryan, Wigman, Riese, Penninx, Giltay, Wichers and Hartman2020), with a co-occurrence rate of 90% (Gorman (Reference Gorman1996). Results from our breeding line indicated that CHF animals from the fourth generation did not differ from control animals, as measured by the forced swimming test (Dias et al., Reference Dias, Bevilaqua, Silveira, Landeira-Fernandez and Gardino2009). However, more recent results, employing CHF animal from the 26th and 27th generation indicated a depressive like behavior when compared to control animals (Goulart et al., Reference Goulart, Rocha-Mendonça, Maisonnette, Pandolfo, Landeira-Fernandez and Campello-Costa2021). Further studies may explore this type of relationship and whether antidepressant drugs are capable of reversing the depressive and the anxiety like effects in these animals.

At the other extreme from animals with high trait levels, CLF animals exhibited a delayed response to haloperidol at lower doses, needing higher doses to reach similar levels of catatonia as control randomly bred animals. Moreover, methylphenidate increased freezing response and motor activity among CLF rats when compared to control animals (Lages et al., Reference Lages, Maisonnette, Rosseti, Galvão and Landeira-Fernandez2021b). Since haloperidol and methylphenidate are dopamine-related molecular targets, it is possible that the CLF line of rats might represent an animal model of hyperactivity and attention disorders. This hypothesis is currently under investigation in our laboratory.

Finally, other studies developed by our group but not discussed in the scope of this review (for details, see Dias et al., Reference Dias, Bevilaqua, da Luz, Fleming, de Carvalho, Cocks and Gardino2014; Lages et al., Reference Lages, Balthazar, Krahe and Landeira-Fernandez2023; Léon et al., Reference León, Brandão, Cardenas, Parra, Krahe, Cruz and Landeira-Fernandez2020) also indicate functional and structural changes in neural circuits underlying anxiety in the CHF rat line, which seems to indicate that the phenotyping process that strengthened this trait anxiety was also expressed in terms of a great capacity for neural plasticity.