In relentless pursuit of the white whale: a role for the ventral midline thalamus in behavioral flexibility and adaption?

Abstract


Introduction
When associated with terms like cognitive or behavioral, the word flexibility defines one of the main executive functions (besides e.g., working memory, planning, inhibitory control, self-J o u r n a l P r e -p r o o f monitoring), and as such contributes to the control and adaption of goal-directed behaviors.
In humans, the concept of cognitive flexibility defines a mental ability to transition efficiently between different ideas, thoughts, or concepts in order to conform to a change of context or constrain during an ongoing situation or task.A typical neuropsychological task to assess cognitive flexibility in humans is the Wisconsin Card Sorting Test described by Grant and Berg in 1948.Cards (2 sets of 64) are distinguishable according to three object characteristics (4 colors, 4 numbers, and 4 shapes).The evaluator chooses one of these characteristics as a sorting rule (e.g., yellow color regardless of shape and number) but does not inform the tested subject about this rule.The tested subject must find it out based on the evaluator's feedback (correct, not correct).Once the consecutive choices of the tested subject show that the evaluator's selection rule has been discovered and is now used consistently, the evaluator shifts in petto to another sorting criterion (e.g., number '3' regardless of color and shape), which, again, the tested subject has to find out.Retarded shift to this new sorting rule, or retarded 'set-shifting', may account for perseveration, and thus lack of flexibility.
In the neuroscience literature, the concept of cognitive flexibility is often used interchangeably with its nearly twin, behavioral flexibility.The latter defines, although preferentially in animal models, an ability to adapt a behavioral response to a modification of environmental contingencies.Most tasks assessing behavioral flexibility in animal models consist of objectifying a perseverative behavior after a change of instruction or demand (e.g., Floresco et al., 2009).For instance, in a spatial context, after having learnt to swim to the location of an escape platform hidden in a water pool, the platform is displaced by the experimenter to a new location, which the tested animal has to search for, find and of course learn.This shift requires abandoning a search behavior based on memory of the previous location of the target.Typically, it is called reversal.In case of behavioral inflexibility, a shift from the former to the new location is significantly retarded (e.g., de Bruin et al., 1994).
One of the brain structures whose involvement in behavioral flexibility is crucial is the medial prefrontal cortex (mPFC).Apparently, this holds true across a variety of species, from primates (e.g., Moore et al., 2009) -starting with humans (e.g., Demakis, 2003) -to rodents (e.g., Bissonette and Powell, 2012;Floresco et al., 2009;Moore et al., 2009).In rodents, the mPFC mirrors the functional and neuroanatomical characteristics of the primate dorsolateral PFC, highlighting its significance across species (Seamans et al., 2008).Usually (but see below), the J o u r n a l P r e -p r o o f rodent mPFC is anatomically divided into two main regions: the dorsal mPFC, comprising the dorsal prelimbic (PL) and anterior cingulate cortex (ACC), and the ventral mPFC, including the ventral PL and infralimbic cortex (IL), each with distinct connectivity and functions (Howland et al., 2022;van Heukelum et al., 2020).After permanent or reversible lesions of the medial prefrontal cortex (mPFC) in rodents, shifting between strategies or rules (i.e., set-shifting) in a behavioral task is impaired (Birrell and Brown, 2000;Dalley et al., 2004;Floresco et al., 2008;Klune et al., 2021).In monkeys, similar results are obtained when the dorsolateral PFC is altered (Klein-Flügge et al., 2022;Moore et al., 2009).Regarding reversal learning in rodents, impairments are observed following damage to the orbitofrontal cortex, whereas the same damage does not seem to affect set-shifting (McAlonan and Brown, 2003).This could point to a certain degree of regional specialization within the prefrontal cortex as to the latter's contribution to different aspects of behavioral flexibility (e.g., Dalton et al., 2014;Hanganu-Opatz et al., 2023).
Regions of the mPFC exhibit dense interconnections with cortical association areas, limbic structures, thalamic nuclei, and various midbrain and brainstem nuclei, facilitating complex behaviors shaped by prolonged environmental interaction (Lakshminarasimhan et al.,, 2024;Klüne et al., 2021;Uddin, 2021;Friedman and Robbins, 2022).The interactions among mPFC regions are thought to mediate various aspects of flexible behavior, particularly through the mPFC-thalamus axis.Indeed, several thalamic nuclei are bidirectionally connected with the mPFC, as illustrated in Figure 1 (Re), perhaps also the rhomboid nucleus (Rh), in behavioral flexibility.This region of the ventral midline thalamus is bidirectionally connected with the PFC, but also with region CA1 of the hippocampus (HIP; e.g., Cassel et al., 2021), and with the amygdala, among other structures.Interestingly, some experiments indicate that the HIP, in addition to the PFC, may also contribute to flexibility (e.g., Avigan et al., 2020;Cernotova et al., 2021;Lehr et al., 2023;Li et al., 2024;Malá et al., 2015).

J o u r n a l P r e -p r o o f
Several reviews on the functions of the ReRh nuclei have pointed to a role for them in a variety of cognitive tasks in which success depends upon information exchange between the PFC and the HIP (e.g., Cassel et al., 2021;Dolleman-van der Weel et al., 2019;Ferraris et al., 2021;Pereira de Vasconcelos and Cassel, 2015;Vertes et al., 2022).Behavioral flexibility has received relatively little attention in these reviews.Herein, we compile and discuss the various studies having tackled the question of a possible implication of the ventral midline thalamus, more specifically of its Re nucleus, in behavioral flexibility.To the best of our knowledge, this question has been addressed experimentally only in rodent models so far.Only recently has a map of the Re and Rh nuclei become available in humans (Reeders et al., 2023), what now offers an opportunity for investigating this region in human primates using functional imaging approaches (e.g., fMRI).In the following sections, we will start with anatomical considerations about the ventral midline thalamus, then summarize findings from behavioral studies having tackled behavioral flexibility by focusing on perseveration, alternation, fear extinction, reversal learning, and set-shifting tasks, before concluding.

Overall connectivity pattern of the ventral midline thalamus
The thalamus is a diencephalic structure consisting of two ovoid masses made of gray matter and located on each side of the third ventricle.It is divided into not less than 50 nuclei, each with specific functions, and which are grouped in ensembles named according to their location along the medio-lateral, dorso-ventral and antero-posterior orientations of the brain.The main ensembles are formed by the anterior, medial, lateral, intralaminar, midline nuclei (+ the reticular nucleus, the pulvinar, and the lateral and medial geniculate nuclei).As indicated by its name, the midline thalamus is located medially, where it forms a thin strip of cells going from the dorsal to the most ventral limits of the structure.It encompasses the paraventricular, parataenial, intermediodorsal, reuniens and rhomboid nuclei.These nuclei are generally distinguished according to whether their position is dorsal or ventral: the paraventricular, parataenial and intermediodorsal form the dorsal midline thalamus, when the reuniens and rhomboid nuclei form the ventral midline thalamus.The nucleus reuniens is the largest of all J o u r n a l P r e -p r o o f midline thalamic nuclei.The rhomboid nucleus, which lays just dorsally to the reuniens nucleus, is relatively small.Another difference between both nuclei is that the reuniens receives inputs only from limbic structures whereas the rhomboid nucleus receives inputs from both limbic and non-limbic (sensori-motor) ones (Vertes et al., 2015).For a dense connection between the Re and the HIP, the connection of the Rh with the HIP is sparser.
Furthermore, whereas the Re contains numerous calretinin and calbindin-D28k neurons with a similar distribution for both proteins, there is less of such proteins in the Rh (Vertes et al., 2015).The Re, not the Rh, also contains a small population of dopaminergic neurons projecting contralaterally and to the hypothalamus (Ogundele et al., 2017).Both nuclei, however, share similarities between their respective connection patterns, although differences exist, and not only in afferent and efferent innervation densities to and from the same brain structure, but also in terms of structures with which they are interconnected.
The connectivity of the ventral midline thalamus has been considered over the recent past in several review articles (e.g., Cassel et al., 2021;Dolleman-van der Weel et al., 2019;Pereira de Vasconcelos and Cassel, 2015;Vertes et al., 2022), a reason why we do not reconsider it in detail herein.Briefly, the neurons of the Re nucleus receive more or less dense inputs from over thirty brain regions, including the hippocampal formation (exclusively the CA1 region and the subicular region), other deep or superficial telencephalic structures (including several regions of the cortical mantle), diencephalic structures (including a few other thalamic nuclei), the hypothalamus, and the brainstem.With a significant number of these afferent sources, the connections of the Re nucleus are reciprocal, without there necessarily being a balance between the respective densities of afferent and efferent fibers (e.g., Cassel et al., 2021).The connectivity of the Rh nucleus is lesser known.It receives afferent fibers from a dozen of brain structures, essentially from cortical and brainstem regions, and sends efferent fibers to about 30 brain structures, including the hippocampal formation, several cortical regions, other thalamic nuclei, as well as the nucleus accumbens and the caudate-putamen (e.g., Cassel et al., 2021).We mention en passant the connectivity of the Rh nucleus, because this nucleus is both small and closely associated with the Re nucleus, so it seems to us delicate, with classical tools (e.g., lesions, pharmacological inactivation), to functionally manipulate the Re nucleus without affecting all or part of the functioning of the Rh nucleus, and vice versa.It is also worth noticing that, at least for the Re nucleus, there is a large overlap of the connectivity J o u r n a l P r e -p r o o f pattern between mice and rats, but differences have been described for a significant number of structures, including the amygdala and the lateral entorhinal cortex (Scheel et al., 2020).
Therefore, functions of the ventral thalamic midline identified in one model species may not be fully generalized to the other, but from a functional point of view, both species have never been compared directly in the same experiment.

Focus on the connectivity between the ventral midline thalamus and the prefrontal cortex
Given the topic of the current review, we will focus primarily on the connectivity pattern of the Re nucleus with the PFC.In rodents, there is no consensus about what regions precisely the PFC is made of, most probably because of a still influent, although dated, definition of it.
This definition is by Rose and Woolsey (1948): the PFC is the cortical area to which the mediodorsal (MD) thalamus projects.In a more recent critical synthesis, Carlén (2017) considers that, in mice (see Figure 4 of her article and video S5), the PFC encompasses the anterior cingulate area, the prelimbic area, the infralimbic area, the orbital area, the secondary motor area and the agranular insular area; the latter two areas are not illustrated in Figure 1 (see caption for their connections).However, delimitations and even denominations of these regions may differ among authors having provided an atlas and/or neuroanatomical descriptions of this cortical region in mice, rats and other species (Carlén, 2017), so that interspecific overlap is all but a rule.This problem has also been the central concern of the critical review by Laubach et al. (2018).It is noteworthy, however, that for the cingulate cortex a homologous nomenclature has been proposed, which does not only permit aligned results between studies in rodents as compared to other species, but also provides a better structural and functional organization of this part of the PFC (van Heukelum et al., 2020).
Based on Carlén's subdivisions of the prefrontal cortex, the nucleus Re receives a dense innervation originating in the prelimbic, infralimbic and orbital areas, as illustrated in Figure 2A.Afferents from the anterior cingulate and the agranular insular areas are less dense.The weakest innervation originates in the secondary motor area.Regarding efferent connections, the Re nucleus has dense projections to the prelimbic, infralimbic and orbital areas.It also projects, although less densely, to the anterior cingulate and the agranular insular area.
J o u r n a l P r e -p r o o f Finally, it has sparse projections to the secondary motor area.An interesting picture in this connectivity schema is that the areas to which the Re projects most densely are also those from which it receives its densest innervation.This reciprocity in connection density also holds true for regions of the prefrontal cortex less densely interconnected with the Re nucleus.
Regarding the topic of the current review, all prefrontal areas that are densely connected with the Re nucleus have been, in a way or another, associated with behavioral flexibility (for a few review articles, see: Barker et al., 2014;Boorman et al., 2021;Dalley et al., 2004;Hangana-Opatz et al., 2023;Klein-Flügge et al., 2022;Klüne et al., 2021;Nett and LaLumiere, 2021;Park and Moghaddam, 2017;Parnaudeau et al., 2018;Rich et al., 2018).As the HIP and amygdala have also been linked to certain behaviors that can be considered reflective of (or at least related to) flexibility (e.g., the extinction of a conditioned fear response), it may be useful to add a few words about the connections of the Re nucleus with these two structures.Regarding the HIP, it is quite simple: the Re nucleus projects exclusively to the CA1 region (Figure 2B), predominantly in the ventral part, less so in the dorsal one.
Although properly speaking this is not the hippocampus, it can be useful to remind that there are dense projections of the Re nucleus to the subiculum, less to the pre-and postsubiculum.
As for the afferents to the Re nucleus, they originate from the same regions.Morever, in a study examining Re nucleus afferents/efferents, a strong convergence of fibers from the mPFC on Re nucleus cells projecting to the HIP has been shown (Vertes et al., 2007), providing anatomic support to an mPFC modulation of Re neurons that project to the hippocampus.
Finally, concerning the amygdala (Figure 2C), although sparse, there are reciprocal connections between the Re nucleus and all amygdala nuclei (basolateral, basomedial, lateral, medial, and posterior, perhaps a bit more with the basolateral than with the others).

Implications of thalamic nuclei other than ReRh in behavioral flexibility
J o u r n a l P r e -p r o o f In humans, functional magnetic resonance imaging (fMRI) studies indicate that the mediodorsal (MD) thalamus provides feedback to the dorsolateral PFC during behavioral strategy switching (Hummos et al., 2022).Similarly, in rodents, projection neurons from the PFC play a role in set-shifting by encoding feedback information to downstream targets such as the MD and ventral medial striatum (VMS) (Spellman et al., 2021).Furthermore, animal studies suggest an interconnected network involving the MD, PFC, and nucleus accumbens core (NAc), or even more generally striatum, that may collaborate to facilitate certain forms of behavioral flexibility (Bradfield and Balleine, 2017;Block et al., 2007;Gmaz and van der Meer, 2022;Liu et al., 2020;Rikhye et al., 2018;Wang et al., 2019).This phenomenon may be attributed to the extensive and bidirectional connectivity observed between the MD and most of the mPFC regions, as demonstrated in studies involving both humans and animals (Alcaraz et al., 2016;Groenewegen, 1988;Mengxing et al., 2023;Oyama et al., 2022).
Furthermore, using an NMDA-mediated, fiber-sparing lesion in rats, Ouhaz et al. (2022) reported on extra-dimensional (not intra-dimensional) set-shifting impairments after damage to the medial thalamus.Although this region is strongly connected with the PFC, the study design did not allow concluding whether a disrupted cortico-thalamic or thalamo-cortical connection was implicated in the induced dysfunction.Using optogenetic tools, Marton et al. (2018) demonstrated that projections from PFC to dorsomedial thalamus were important for behavioral flexibility in a decision-making task guided by auditory or visual cues.Rikhye et al. (2018) found that projections from the medial thalamus to the PFC regulated flexibility by acting on cortical interneurons; in this study, mice had to switch between sets of learned cues guiding their attention towards visual or auditory targets.Parnaudeau et al. (2015) showed that DREADD-induced hypofunction of the MD thalamus was accompanied by behavioural inflexibility in an instrumental discrimination task.Some evidence for an implication of intralaminar thalamic nuclei in behaviroal flexibility has also been provided (Brown et al., 2010;Kato et al., 2018).While the PFC demonstrates extensive connectivity with various thalamic nuclei (Oyama et al., 2022;Yang et al., 2022), the anterior thalamic nuclei (ATN), known for their importance in cognition and spatial memory, were also investigated for their potential role in behavioral flexibility.Lesions to the ATN produced a temporary impairment in strategy-shift experiments, but the extent of this deficit did not conclusively demonstrate that ATN are involved in modulating flexibility behavior (Kinnavane et al., 2019).Finally, J o u r n a l P r e -p r o o f Nakayama et al. (2018) provided experimental arguments implicating projections from the PFC to the midline thalamus (besides cortico-striatal ones) in a behavioral flexibility task assessing a probabilistic reversal in an operant chamber.Interestingly, in this study, the thalamic regions of the cortical projections predominantly encompassed the mediodorsal, parafascicular, but also the ventromedial thalamic nuclei, i.e., the region including the Re and Rh nuclei.

Early observations
Possibly one of the first papers to point on an implication of the Re nucleus in aspects of behavioral adaptions implicating flexibility is the report by Dolleman-van der Weel et al.
( 2009).These authors compared in rats the behavioral effects of bilateral neurotoxic lesions (using ibotenate) of the hippocampus, the MD thalamus or the Re nucleus.To this end, they used a water maze task assessing spatial learning and memory.Rats were trained over 3 days (5 trials/day) for a maximum of 2 min/trial, resting on the platform for 30 s when they had reached it.They were tested in a probe trial 24 hours after the last training, for which they were left in the pool for 60 s, the escape platform having been removed.All rats were subsequently retested, on the same day, using a visible platform to evaluate possible biases such as sensorimotor or motivational alterations, or difficulty to shift to a different problem solving strategy.As expectable, Dolleman-van der Weel et al. found impaired learning in rats subjected to hippocampal lesions.In rats with mediodorsal thalamic lesions, there was almost no alteration, whereas rats with Re nucleus lesions and sham-operated ones performed close to each other, and indistinctly from a statistical viewpoint.Interestingly, overall differences in swim strategies were noticed: rats with MD lesions swam more frequently along the pool edge (a behavior called thigmotaxis) and showed less direct swim paths.Rats with hippocampal lesions also displayed a lower number of direct swim paths and their swim patterns appeared disorganized.Finally, rats with Re lesions swam like the sham-operated controls did.In the probe test, only rats with hippocampal lesions showed an impairment that was significant when compared to their sham-operated controls.Most interestingly regarding the topic of this review, the rats with Re lesions swam directly to the former platform location, but then, Regarding the Re nucleus, its main lesion effect was reflected in a modification of the strategy used by the rats in the probe trial.

The Re nucleus and perseveration
Here, we consider only experiments in which perseverative behaviors have been evaluated without manipulations of explicit environmental contingencies informing the animal that the rule has changed.As in the Wisconsin card-sorting test for humans, rats have to find out by themselves that something has become different in the situation that they have been facing so far.Perseverative behaviors are behaviors that continue to be repeated in a particular situation in which contingency changes have rendered this behavior inadequate to achieve the goal it enabled to reach so far (as shown by Captain Ahab in Herman Melville's novel Moby Dick, hence the title of the current review).Perseveration corresponds to a form of behavioral inflexibility.Two classical tasks measuring perseveration in rodent models are the spontaneous or reinforced alternation task (e.g., rats do not consistently alternate between the target arms) and an operant conditioning task (e.g., rats do not consistently alternate between target levers on which they have to alternatively press in order to get a reward).Lack of (or delayed) extinction after classical or operant conditioning may also reflect a perseverative tendency (see below).

Alternation tasks
Using a modified T-maze in which rats could return to the start location by their own after having chosen an arm where they had been rewarded (or not in case of an incorrect choice),

Fear conditioning and extinction of conditioned fear
Perseverative behaviors, and thus lack of flexibility, can also be regarded as persistence of behavioral responses that should be extinguished following removal of an unconditional stimulus associated with an explicit conditioned stimulus (e.g., a tone) or a configuration of multiple stimuli (i.e., a context).As such, it may be interesting to consider experiments that used a conditioning paradigm implicating cooperation between the mPFC and the HIP (or J o u r n a l P r e -p r o o f amygdala).Two behavioral paradigms have been used, some with an explicit conditioned stimulus (a tone at a given frequency), others with contextual fear conditioning.

Connections between Re nucleus and mPFC
Ramanathan et al. (2018a) used a tone.Their rats were exposed to tone(CS)-shock associations in a given context, and, with a 24-h delay, re-exposed to the CS in a different context to induce extinction (tone was presented, but no electrical shock was delivered).
Before this, the Re nucleus was infused with a saline solution (control) or with muscimol.
Muscimol disrupted extinction to the point that no extinction at all could be observed.
Ramanathan et al. (2018a) also tested extinction retrieval of the extinguished fear by placing their rats back into the extinction context.They found that muscimol infused before this retrieval test disrupted the retrieval of the extinguished fear (rats behaved as if extinction had not taken place).In intact rats, they also found that both encoding and retrieval of extinction resulted in increased c-Fos expression and single unit firing in the Re nucleus.Furthermore, using an inhibitory DREADD (hM4Di) approach, they established that silencing (with Clozapine N-oxide or CNO) the projections from the mPFC to the Re nucleus reduced extinction efficiency, and they obtained similar results when the terminals of these projection neurons were silenced directly within the Re nucleus.Silencing projections from the infralimbic or prelimbic PFC, or from both, yielded very similar results.
Ramanathan and Maren (2019) performed contextual fear conditioning in rats that had been equipped with infusion canulae targeting the Re nucleus.Among other findings, they observed that Re inactivation disrupted both encoding and extinction.On the one hand, rats conditioned right after muscimol infusion and subjected (after 24 h) to extinction following saline infusion did not extinguish.On the other hand, rats conditioned after saline infusion and subjected to extinction after muscimol infusion also failed to extinguish, indicating perseverating fear.The latter observation is compatible with a Re-mediated suppression of fear memory expression in the on-going extinction trial, and the former with a participation of Re in encoding or consolidation of unpleasant circumstances.These authors also reported that the Re nucleus displayed theta-range rhythms that correlated with freezing during the extinction phase.

J o u r n a l P r e -p r o o f
In a more recent study from the same group, Totty et al. ( 2023) recorded LFPs in the mPFC (prelimbic and infralimbic cortices) and in the HIP (in fact in region CA1) while rats extinguished/retrieved an auditory fear conditioning in a context different from the conditioning one.They found that coherence (at theta frequency; 8 Hz) between mPFC (infralimbic cortex) and HIP activity was increased during extinction retrieval.Interestingly, they also reported that Re nucleus inactivation with muscimol reduced c-Fos expression in both the mPFC and HIP during fear relapse, a treatment that also reduced theta-range spectral coherence between the mPFC and HIP.Finally, optogenetic inhibition of the Re impaired extinction-memory retrieval.

Connections between Re nucleus and HIP
Ratigan et al. ( 2023) used a fear conditioning approach in mice placed into a virtual reality context, in which they could be delivered electrical shocks on the tail.Over three consecutive post-conditioning days, the mice froze significantly more in the shocked as compared to the neutral (no shock delivered) context, with freezing location evenly distributed over the entire virtual context; it did not depend on the mice location in the context.The duration of the average freezing epochs was longer in the freezing than in the neutral context.Ratigan et al.
(2023) subsequently used an inhibitory DREADD approach in the projections from the Re nucleus to region CA1 of the hippocampus.The DREADD was activated with deschloroclozapine (DCZ) during the conditioning session; DCZ is more potent than CNO and has a direct action of the DREADD.When tested subsequently, mice treated with DCZ during conditioning froze more in the shocked context, but an increase of freezing behavior was also observed in the neutral context, although with freezing levels lower than in the shock context.
In an additional experiment, the acute effects of DCZ were assessed on the first retrieval day.
Under this condition, the freezing response under DCZ was increased in the shock as compared projections originates in the mPFC, presumably in the infralimbic cortex given its implication in fear suppression (e.g., Giustino and Maren, 2015).They also mentioned that this system most probably operated within a large coordination dynamic implicating amygdala-dependent mechanisms that support the fear response and its related physiological and behavioral modifications.This is compatible with data showing that muscimol-induced inactivation of the dorsal hippocampus also results in a reduced rate of fear extinction (Corcoran et al., 2005), although in this case an auditory, not a contextual, conditioning was used.

Conclusions about extinction of conditioned fear
If one considers all aforementioned data at a glance, it appears that the Re is likely at the core of a system encompassing the mPFC, the HIP, and the amygdala; the connectivity of the Re nucleus is largely compatible with this idea (Figure 2).This system regulates fear responses that the animal expresses in reaction to environmental stimuli which it has previously learned to be afraid of, whether these stimuli are explicit (e.g., a tone) or not (e.g., a context).More generally, this system seems to adjust the expression of this emotion to the characteristics of the environment and to changes thereof.While it has been shown that the Re nucleus plays a role in different phases of associative fear learning (i.e., consolidation, retrieval, reconsolidation, and perhaps even encoding; e.g., Jhuang et al., 2023;Lin et al., 2020 ;Quet et al., 2020 ;Ramanathan et al., 2018b ;Sierra et al., 2017;Troyner et al., 2018;Troyner and Bertiglio, 2021 ;Vasudevan et al., 2022 ;Wu and Chang, 2022), the articles considered in this section clearly indicate that it also contributes to the extinction of a conditioned fear response.
Indeed, consistent data (see Figure 3 for an attempt at a summary transversal to the different studies mentioned above) show that inactivation of the mPFC or of its prelimbic and/or infralimbic components, of the HIP, or of the Re nucleus have a marked impact on fear extinction.Furthermore, projections from the mPFC to the Re nucleus, or terminations of J o u r n a l P r e -p r o o f these projections in the Re nucleus, or even projections from the Re nucleus to CA1, as well as projections from the Re nucleus to the amygdala, all have an impact on the extinction of conditioned fear, and thus behavioral flexibility.Altogether, these observations are highly coherent: the mPFC exercises top-down regulation over the retrieval of memory components associated with fear.Specifically, it influences the hippocampus (HIP) and basolateral amygdala (BLA), key brain regions involved in memory and fear processing, respectively.This regulatory mechanism involves the suppression of fear memory retrieval, with a notable contribution of the Re nucleus; see Plas et al. (2024) for details.It is even possible that elements upstream of this system could be involved, since the mPFC-to-Re pathway is involved in conveying information/details essential for understanding the relevance of sensory stimuli in associative learning contexts (e.g., Xu and Südhof, 2013;Yu et al., 2022).For each test trial, rats were given access to only 4 out of the 8 arms of the apparatus, and only when they had visited these 4 baited arms were they allowed to explore the four remaining ones, either without a delay, or after a 10 or 30 min interruption.For the 10-and 30-min delays, there was no difference between rats with lesions and their sham-operated counterparts.However, when there was no delay, rats with lesions committed more than twice as much perseverative errors (i.e., returns into the first four accessible arms that had been visited before the four last ones became accessible) as their sham-operated counterparts.While the task typically taxes spatial working memory, at least in the no delay condition, as mentioned by the authors, this difference cannot be easily explained by a working memory failure, which would suppose that more errors be committed also in the 4 lastly accessible arms.Furthermore, it cannot be due to an alteration of spatial computation capabilities as the same rats performed like controls under the 10-and 30-min delay conditions.Hence, for the authors, this difference is attributable to perseverative behavior, as J o u r n a l P r e -p r o o f it resembled consequences on working memory of prefrontal lesions.Because the increasing delays were introduced sequentially (0s, then 10 min, then 30 min), however, one cannot exclude the possibility that, at the start of the test, the rats were disturbed by the opening of the doors leading to the remaining four baited arms, a potential distractor, and that they then gradually became accustomed to this manipulation.If so, the Re nucleus could be involved in controlling interference phenomena, a hypothesis consistent with studies reporting alterations of working memory in situations with high proactive interference potential (e.g., Layfield et al., 2015;Hallock et al., 2016;Viena et al., 2018) vs. a situation in which such interference is lower (Boch et al., 2022).The perseveration in the radial maze, however, did not reflect a generalized lack of flexibility as the rats performed like controls in the first reversal of a visual discrimination in a touch screen task, and even better than controls on the second reversal.

The Re nucleus and reversal learning
Reversal learning refers to a situation in which subjects have to learn a new rule following the learning of a previous one that, following modification of contingencies, is no longer efficient (in getting a reward) and must therefore be 'unlearned'.For instance, they first learn to establish a link between a particular stimulus and a particular outcome (e.g., get a reward) vs.
another stimulus and an opposite outcome (e.g., no reward or a punishment).Once training has become efficient, the previously rewarded stimulus is associated with the opposite outcome and the previously non-rewarded stimulus now becomes rewarded.Lack of flexibility is reflected in delayed acquisition of the reversal, which, from another perspective, can also be regarded as perseveration.Linley et al. (2016) used an attentional set-shifting/reversal task previously used by the Brown group (e.g., Birrell and Brown, 2000) to assess the effects of electrolytic lesions of the Re nucleus.The lesions also encroached onto the Rh nucleus.The task was in a rectangular enclosure divided across the entire width of one end in two compartments of equal surface.
Each compartment was equipped with a ramekin (i.e., a small circular dish).The ramekins contained discriminative odors (used as olfactory cues) and discriminative digging mediums (used as tactile cues), and the food reward (fruit loops) was buried in the bottom of each J o u r n a l P r e -p r o o f ramekin.Rats were trained to associate cues with the location of the hidden food reward.In a first simple discrimination task design, they had to locate the food by discriminating two odors (O1, O2) using the same digging medium in both places.In a compound version of the task, they had to locate the food based on the same odor as before (O1) regardless of the digging medium in which the reward was buried (two medium were used, allowing to define four possible odor-medium pairs: O1-M1, O2-M1, O1-M2, O2-M2).Then intradimensional shift was tested by introducing four novel odor-medium pairs (O3-M3, O4-M3, O3-M4, O4-M4), rats having to choose the odor (O3) regardless of the medium.Finally, extradimensional shift was tested by introducing four unknown odor-medium pairs (O5-M5, O6-M5, O5-M6, O6-M6).For each of these conditions (compound, intradimensional shift and extradimensional shift), a test of reversal was made in order to evaluate perseveration.In the simple discrimination task, be it olfactory or tactile, rats with the lesion showed no deficit.In the reversal version of the compound task, however, rats with the lesion required more trials to complete the task (the results of intra-and extradimensional shifting are reported below, in section 2.4).
Klintsova and her group have developed a model of postnatal alcohol intoxication in rats, which results in the destruction of neurons in the Re nucleus -in a sufficient number to reduce the volume of this structure -without affecting those of the Rh nucleus (Gursky et al., 2019;Gursky and Klintsova, 2022).Other structures like the HIP and the PFC may also have been affected by early postnatal exposure to ethanol (Miki et al., 2000(Miki et al., , 2003;;Skorput et al., 2015).
Using this postnatal intoxication protocol, Gursky et al. (2021) examined, in adulthood, some cognitive abilities, including executive functions such as working memory, reversal learning, and set-shifting (Gursky et al., 2021).In a plus-maze, the authors trained rats to retrieve a food reward based on either a response strategy (always making the same turn when leaving the starting arm) or a place strategy (always going to the same target arm regardless of which starting arm was used).Once the trained strategy was well integrated (for instance a response strategy), the instruction was modified in two possible ways.Either the opposite arm was now baited (allowing for measurement of reversal), or the alternative strategy had to be engaged (what corresponds to a rule shift : use 'place' instead of 'response') because this strategy now led to a 100% rewarding rate (allowing for evaluation of set-shifting; see below).In the reversal tasks (reversal was tested thrice), rats exposed to ethanol required more trials to reach J o u r n a l P r e -p r o o f criterion, but the difference to their controls was not significant, suggesting weak alteration of reversal.However, this could be the consequence of damage too weak to affect Re nucleusdetermined functions significantly.Furthermore, ethanol-induced damage was most probably not restricted to the sole Re nucleus neurons.

The Re nucleus and set-shifting
Typically, set-shifting refers to the ability to shift attention in order to respond to a modification of environmental circumstances by adopting a new strategy, either ex nihilo, or previously learned but not efficient/useful so far.
In the aforementioned task (see above, section 2.3) based on discriminations of olfactory and tactile cues in combination, Linley et al. (2016) also assessed intradimensional (change in one dimension of the task not the other to which attention could be conserved; e.g. the same odor in a novel yet unknown medium) and extradimensional set-shifting (shifting to new dimensions of the task, i.e., a new set of odors and a new set of mediums).In the intradimensional set-shifting task, rats with the Re nucleus lesion required more trials than their controls, but were not impaired in the extradimensional shift task, although the mean number of trials required was also higher, but not significantly.
In the aforementioned study by Gursky et al. (2021), rats trained in a simple navigation task (see above, section 2.3 for detail) had to shift from a response strategy to a place strategy in a plus maze to get a food reward.Rats subjected to postnatal ethanol exposure, and thus showing damage to neurons of the Re nucleus, required more trials to become efficient after a rule shift in this navigation task, indicating a clear impairment in set-shifting.Cholvin et al. (2013) have assessed set-shifting in the context of a water-escape, spatial navigation task (in the double-H maze) and, in parallel, specifically evaluated spatial memory performance (in a classical water maze) in rats that were subjected to reversible inactivation of the ReRh nuclei.Each rat performed both tasks in a sequential order (water maze test was given first).The effects of this inactivation were systematically compared to those of inactivation of either the mPFC or the dorsal HIP.In all experimental groups and whatever the maze, inactivation was performed with muscimol (two doses) half an hour before a probe trial was performed.Although beyond of the current review's topic, results in the water maze can J o u r n a l P r e -p r o o f be briefly reminded as they shed light on the interpretation of the data obtained in the setshifting test.As expectable, dorsal HIP inactivation disrupted memory retrieval in the probe trial at the highest of both doses (i.e., 0.70 nmol/side).Neither dose affected retrieval when muscimol was infused into the mPFC.Interestingly, when infused into the ReRh nuclei, muscimol did not obliterate retrieval, but rats spent less time in the target quadrant than the controls, an observation reminiscent to a previous report by Dolleman-van der Weel et al.
In the other maze (i.e., the double-H maze), some remarks about the task may be necessary before discussing the results.Cholvin et al. (2013) used a training protocol that randomly alternated between two release points from which two swim paths, always identical for each release point when considered direct, allow the animal to escape from the water on a platform.In one case, the rat had to make two successive left turns (one at each choice point), while in the other, the rat had to make a right turn followed by a left turn (see Figure 4).We know from previous studies that such a training protocol allows a rat to learn both routes while quickly constructing a cognitive map in parallel, which can be used would it become necessary and, of course, would the rat show behavioral flexibility (Cassel et al., 2012).For the probe trial, we removed the platform and changed the release point so that the route consisting of two successive left turns could no longer be used from the start.The other route, however, which consisted in successive right and left turns, was still possible, but it now led to the wrong target arm (the N arm), what more than 80% rats usually do in first instance (Cholvin et al., 2013).Rats were then allowed to correct their initial choice.If so, when flexible, what they usually do until the end of the probe trial is spending more time in the correct arm (where the platform was located during training) than in the others, and in any case above chance level.It can therefore be considered that they have shifted from an egocentric response strategy to an allocentric place strategy, what can be observed as evidence for setshifting, and thus for behavioral flexibility (Figure 4).What we found in this task is that rats subjected to inactivation of the mPFC or the dorsal HIP failed to show the behavior accounting for efficient set-shifting between the two learned strategies.As the inactivation of the mPFC did not affect spatial recall performance in the water maze, the shifting deficit in the double-H maze cannot be explained by a spatial memory impairment.Since inactivation of the dorsal HIP altered recall performance in the water maze, the deficit observed in the set-shifting task could possibly be due to the rats' inability to engage their cognitive map.And what about the ReRh nuclei?At the high dose, inactivation of these nuclei produced the same effect as after mPFC (HIP) inactivation, most probably because the shift signal from the mPFC to the hippocampus could not be relayed in the ReRh nuclei, or because the ReRh nuclei could not activate the mPFC to trigger strategy shifting.
This experiment clearly showed that in the absence of normally operating ReRh nuclei, setshifting (in this case substituting a place strategy for a response one) does not seem possible in a spatial context.
Providing further support to our initial findings, a similar conclusion was made by Quet et al.
(2020) following an experiment which used a protocol identical to the one used by Cholvin et al. (2013), although with a DREADD approach for the acute inactivation of ReRh nuclei neurons.Before the probe trial assessing set-shifting, the authors administered CNO (intraperitoneal) to trained rats of which the ReRh nuclei were infected with hM4Di-mCherrybearing virus.The infection enabled to express the inhibitory DREADD in excitatory neurons of these thalamic nuclei.As was the case with muscimol in the Chovin et al. ( 2013) study, CNO disrupted the possibility to shift from an egocentric to an allocentric approach of the task following a contingency modification.This observation, which was fully confirmed using the same DREADD in a more recent and independent experiment (Boch et al., 2022), also echoes a result obtained by Gursky et al. (2021) in a neonatal alcohol intoxication model based on less specific neuronal alterations of the Re nucleus and a different behavioral task (see above, section 2.3).

General conclusions
In this review, we have considered several tasks assessing behavioral flexibility in mice and rats.All of them were shown to be sensitive to manipulations of the ventral midline thalamus, J o u r n a l P r e -p r o o f and perhaps more particularly of the Re nucleus.Controlling or modulating behavioral flexibility, however, is not the sole contribution of the Re nucleus.Indeed, as documented in several recent and older review articles, permanent damage to, or reversible inactivation of this thalamic region is known to affect many facets of cognition.These facets encompass spatial working memory, systems-level consolidation of spatial and contextual fear memory, reconsolidation of fear memory, inhibition of inherited defense reactions, memory for sequence of events, influence of internal states on behavioral responses to threats, and a couple of others (e.g., Cassel et al., 2013;2021;Dolleman-van der Weel et al., 2019;Ferraris et al., 2021;Saalmann, 2014;Vertes et al., 2022).Moreover, the ventral midline thalamus is not the sole thalamic region to contribute to behavioral flexibility.Other nuclei of the thalamus, also connected with regions of the prefrontal cortex, have been implicated in this type of executive function.The mediodorsal thalamic nucleus is one of them, as largely documented in a recent review article by Wolff and Halassa (2024; see also Hummos et al., 2022).Furthermore, the intralaminar nuclei, which project to the mPFC but not to the HIP, have also a role in shifting behavior (Brown et al., 2010;Kato et al., 2018), and the reticular thalamic nucleus was shown to contribute to fear extinction (Lee et al., 2019).These are just a few examples.Therefore, the Re nucleus, perhaps together with its neighbor the Rh nucleus, has not a unique cognitive implication, and clearly, it is not the only thalamic region involved in behavioral flexibility.
The Re and Rh nuclei are separate entities, as shown by their anatomical organization and respective connectivity pattern, although this connectivity is partly overlapping, at least in terms of efferent projections.However, in many studies, it is difficult if not impossible to have acted on one nucleus (e.g., the Re) without encroaching structurally or functionally onto the other, given their size and their proximity to each other.In the current article, we have considered studies having evaluated effects of ventral midline thalamus alterations on reversal, extinction, set-shifting and perseveration, and most of them have attributed the observed effects to manipulations of the Re nucleus, what is perhaps a bit reductionist in several of them.It is difficult a posteriori to discuss about which studies have or have not achieved a high degree of specificity in manipulating the Re nucleus with minimal or no effects on the Rh nucleus.For lesion studies, post-hoc verifications can be very precise and exclusions for lack of specificity easy to apply.For classical reversible inactivation studies (i.e., those using J o u r n a l P r e -p r o o f sodium channel blockers or a GABA receptor agonist), a high precision level is not only difficult to verify but also most probably impossible to achieve.Finally, regarding the approaches relying on the use of viral vectors to induce DREADD expression, the diffusion radius around the injection site will be the determinant factor: if too limited, the infection might be insufficient to produce effects; if too large, it will become unspecific.A way to circumvent this drawback would consist in combining an anterograde viral vector delivering a floxed gene (e.g., of a caspase or a DREADD) to neurons of a source region and a retrograde viral vector delivering a cre recombinase to axon terminals in a target region.By this way, only neurons interconnecting the source and target structures will be affected.However, this much more specific approach may have other drawbacks related to possible neurotoxicity of the cre recombinase-bearing construct (e.g., Panzer et al., 2024).
Moreover, there is nothing to assert that the alterations induced in the Re nucleus stemmed from the functional manipulation of the Re nucleus as a whole.Indeed, it may be that within the very midst of this nucleus, specific groups of neurons rather than the entire nucleus as a whole, were involved in only one of the deficits reported and not in the others.Indeed, within the Re nucleus, there could exist an organization corresponding to a functional topology in which a group of neurons that would be involved in reversal learning would not participate in set-shifting, and vice versa.Compatible with this view is the fact that the Re nucleus encompasses different subpopulations of neurons (Viena et al., 2021) that are clustered in specific zones.This heterogeneity provides a challenge that future experimental approaches will have to face by aiming to elucidate as finely as possible how different parts of the 'Re<->mPFC' system are working in relation with different facets of flexible behavior, and do so not just from the sole overall perspective of executive functions.
From a phenomenological point of view, behavioral flexibility, although defining an adaptive capacity in response to changing circumstances in the environment, nonetheless remains a complex phenomenon, especially in terms of supporting processes.Indeed, this flexibility may be affected by alterations of functions such as planning, working memory, inhibition, and attention, as well as by learning and memory capacities, emotions and the ability to cope with stress (e.g., Girotti et al., 2018), each of them separately or even in combination.If it is not particularly difficult to phenomenologically objectivate a lack of flexibility in a variety of behavioral tasks, it is more challenging to attribute a possible deficiency to an alteration of J o u r n a l P r e -p r o o f one or more of the aforementioned functions.Indeed, the same phenomenological manifestations interpreted as behavioral inflexibility could just as easily be supported by an inability to overcome stress in a given situation or, for instance, by dysfunctions of working memory or an attention disorder.Thus, depending on the task, errors, repetitions, lack of shifting or delayed reversal may be interpreted as reflecting perseverative behavior when in fact corresponding to a genuine disruption of memory processes, and variables accounting for bonafide perseverative behavior may be interpreted as the consequence of memory dysfunctions when in fact perseveration has actually led to task errors.All these functions appear intermingled to some degree.Therefore, future research would probably need to integrate more targeted complementary approaches into studies investigating the involvement of the ventral midline thalamus in behavioral flexibility.As a matter of fact, these approaches might, in addition to more classical tasks taxing behavioral flexibility, also focus on attention, working memory, inhibition abilities, etc, the attempt being to understand more precisely the underlying factors of any potential deficit affecting what is usually attributed to a flexibility dysfunction.The interest of doing so is sketched in the article by Cholvin et al. (2013).These authors paralleled spatial navigation capacities in the water maze and setshifting in the double-H maze, both tasks relying on comparable motivation leading to a water escape response.Whereas inactivation of the mPFC and the Re nuclei had little effect on spatial navigation performance in the probe trial, inactivation of the HIP unambiguously impaired it.In the set-shifting task, rats were impaired to comparable levels whatever the inactivated region.It could therefore be reasoned that the apparent set-shifting impairment reflected disruption of different processes depending on the region that had been inactivated.
The proposal of the authors was that HIP inactivation disrupted spatial navigation, mPFC disruption altered set-shifting, and Re inactivation either hindered information exchange between mPFC and HIP that was crucial to efficient spatial navigation, or prevented an activation of the mPFC for efficient set-shifting.However, set-shifting may have been genuinely altered only after mPFC inactivation.
Finally, it is also possible that the Re nucleus is not necessarily implicated in behavioral flexibility underlying reversal responses, suspension of perseveration and set-shifting.In fact, it could be part of a more general adaptive function by which it would contribute to the integration of new environmental contingencies when a given behavior is going on and J o u r n a l P r e -p r o o f contingencies change.Indeed, the Re nucleus could be involved in a more general regulation that would be implemented whenever a behavior needs to integrate changes occurring in the environment, which the subject has to take into consideration in order to revise the behavioral pattern engaged, and adjust it as best as possible to these changes.Understanding the nature of these modifications with regard to the overall context might then involve a circuit encompassing the hippocampus, evaluating and assimilating the emotional dimensions thereof would require the amygdala, and perhaps more generally the limbic system, and  et al., 2010).Indeed, all these disorders encompass aspects of behavioral inflexibility.To some respect, models of post-traumatic stress disorder, Alzheimer's and Parkinson's disease, and perhaps major depression might also be worth a look to the Re nucleus.

Declaration of interest
The authors have no conflict of interest to declare J o u r n a l P r e -p r o o f  2018); Ray and Price (1993); Vertes (2002;2004); Vertes et al., 2006;Xiao et al. (2009).Caution: this figure is an attempt at synthesis; at no point does it aspire to be a faithful reflection of the anatomical reality.Indeed, the illustrated densities correspond to rough estimations, as they are necessarily relative densities when compared with each other, and as the highest density between a given ensemble of thalamic nuclei and a cortical region (and vice versa) does not necessarily compare with the highest density between another ensemble of thalamic nuclei and another cortical region.
Furthermore, a high density of efferents from region 'A' to region 'B' does not necessarily compare with the exactly same high density of afferents to region 'A' from region 'B', among other reasons because afferents and efferents have been assessed in different studies, based on different tracing methods, by different laboratories.
J o u r n a l P r e -p r o o f  of the ReRh nuclei (Cholvin et al., 2013) or their CNO-induced inactivation after infection of their excitatory neurons with an hM4Di-bearing viral vector (Quet et al., 2020;Boch et al., 2024).Top: two trajectories trained to reach the escape platform located in the NE arm, one consisting in two successive left turns (left panel), the other consisting in a right turn next followed by a left turn (right).When the rat was released form the S arm, the N arm was closed, and when it was released from the N arm, the S arm was closed.Rats were trained on a total of 16 trials (2 trials for each trajectory/day) and, after a delay of 24 hours from the last training trial, they were tested in a probe trial for which i) the escape platform was removed, J o u r n a l P r e -p r o o f . ****************** Insert Figure 1 about here ****************** A growing number of reports open the possibility for an implication of the reuniens nucleus

*
*************** gave up searching at the right place, what clearly contrasted with the insistence displayed by the sham-operated control rats.According to the authors, this difference can be regarded as a behavior accounting for a change in the ability of 'shifting strategy' (dixit), possibly J o u r n a l P r e -p r o o f because of the disruption of Re to mPFC connections.In the cued test, only rats with MD lesions tended towards some impairment as they persisted to swim to the location of the hidden platform (what accounts for a clear reversal impairment, and thus reduced behavioral flexibility, an observation all but surprising given the strong connectivity between the MD and PFC).
Viena et al. (2018) examined the effects of reversible Re inactivation on alternation behavior.Delays of 30, 60 or 120 s were interposed between a sampling and an alternation trial.Two doses of the GABA A receptor agonist muscimol were used, 62.5 ng and 125 ng, and delivered to the Re in a volume of 0.5 µL PBS.When a rat made an incorrect choice on the alternation run (i.e., returned to the arm selected in the sampling trial), it was given up to 10 additional trials to correct this error.The later variable (number of successive uncorrected trials) J o u r n a l P r e -p r o o f accounted for perseveration.There was a 30 min gap between the end of muscimol infusion and the start of behavioral testing.At both doses and for all delays, the muscimol infusion disrupted the number of correct trials, what may be regarded as reflecting disruption of working memory.Most interestingly given the topic of this review, inactivation of the Re also resulted in a dramatic increase of the number of errors accounting for perseverative behavior, and thus for impaired behavioral flexibility.Using a maze similar to the one used byViena et al. (2018), Stout et al. (2022)  examined in rats the implication of the Re nucleus in vicarious trial and error (VTE).VTE refers to a situation in which animals behave as if exploring the possibilities provided at a choice point before properly engaging into the behavior.VTE, which emerges when flexible decision-making is needed, translates deliberation about the correct choice leading to the reward.From a behavioral point of view, at the choice point, it is accompanied by head-sweeping, reduced displacement velocity, and/or even short pausing, which can all be used as markers of VTE as long as they are noticed at the choice point (i.e., in the T-maze, this point is where the three arms cross).In rats previously trained to alternate the arm choices, Stout et al. used muscimol to reversibly inactivate the Re nucleus.The authors found that, over all choices (correct and incorrect trials collapsed), the inactivation did not affect the frequency of VTEs, although reducing the percentage of correct trials (i.e., trials leading to the reward; from almost 70% to about 40%).When, however, only incorrect trials were considered, muscimol increased the frequency of VTE substantially.Interestingly, Stout et al. (2022)  also analyzed theta activity synchrony between the mPFC and the hippocampus (HIP) and found that muscimol reduced coherence during VTE on incorrect choices, what suggests disruption of activity coordination in the prefrontal-thalamo-hippocampal circuit when the Re nucleus is inactivated.Overall, the reported results indicate that Re inactivation produces an increase of deliberation in repeated choice errors trials, hence in trials corresponding to a perseverative behavior.
to the neutral context, or in comparison with shocked controls bearing no DREADDs that were treated with DCZ in the target context.Interestingly, on day 2, mice given DCZ on the previous day still showed increased freezing in the shocked context (not in the neutral one), what can be regarded as evidence for delayed extinction.When freezing was increased, this was due to a lengthening of the freezing epochs.As the increased freezing response appeared in both the shocked and neutral contexts, it seems that mice showed reduced discrimination capabilities after inhibition of the Re-to-CA1 projections.None of these modifications was due to J o u r n a l P r e -p r o o f unspecific DCZ effects as mice did not respond to DCZ in darkness, which prevented perception of contextual cues.An additional step consisted in visualizing Re-to-CA1 activity by 2-photon calcium imaging in the HIP.Re axons ending in the HIP showed an activity that, postshock, correlated with freezing epochs; it was higher during freezing than during running.Furthermore, when freezing responses decayed as a result of extinction, this firing decayed in parallel with extinction of the fear response.The relationship between post-shock Re-to-CA1 activity and freezing responses was confirmed using a computational encoding model.In this article,Ratigan et al. (2023) rose the possibility that the input driving the activity of Re-to-CA1 Silva et al. (2021) focused on the connections of the Re nucleus with the basolateral amygdala and, using contextual fear conditioning, they examined their role in fear extinction.Both recent and remote fear could be extinguished by repeated non-shock exposure to the conditioned context.Specifically for extinction of remote fear, not for that of recent fear, Silva et al.(2021)  found that Re neurons projecting to the basolateral amygdala were activated, as was also the case for neurons from the infralimbic cortex projecting to the Re nucleus.Their next step consisted in inhibiting the Re nucleus during extinction (using a DREADD strategy based on expression of hM4Di in Re neurons, and CNO given systemically or infused into the amygdala).They found that under CNO influence, the freezing level remained higher than in the controls.Next, they expressed the hM3Di DREADD in excitatory Re neurons and tested extinction under CNO treatment.Extinction was faster.Using in vivo fiber photometry recordings, they could observe that during the extinction trials an activity increase in the Re was time-locked to behavior: it appeared upon freezing cessation.Finally, to demonstrate a causal relationship between extinction and this particular activation of Re neurons, Silva et al.(2021) used an optogenetic approach to investigate the effects of both gain and loss of J o u r n a l P r e -p r o o f function.When Re neurons were stimulated, freezing bouts were shorter and extinction was faster.Inversely, when inhibited, the freezing bouts were longer and freezing cessation was delayed.Additional experiments enabled to demonstrate a direct participation of the Re neurons that project to the basolateral amygdala in the extinction of remote fear.Here, it might be worth reminding a series of experiments byKitamura et al. (2017).These authors also used contextual fear conditioning and followed the systemic consolidation of the memory.At a recent post-conditioning time point (2 days), they found that the re-exposure to the context activated a hippocampal engram which produced the fear response by triggering the emotional engram in the amygdala.At a remote time point (12 days posttraining), however, the re-exposure to the context activated an engram that had 'migrated' to the PFC, from where the emotional engram in the amygdala was now triggered to produce fear.Although this possibility was not addressed byKitamura et al. (2017), one cannot exclude that, as in the study bySilva et al. (2021), the PFC-induced activation of the engram in the amygdala was relayed in the Re nucleus in the study byKitamura et al. (2017).

*
Prasad et al. (2017) have used a particular protocol in a radial maze to evaluate spatial memory (among other functions) in rats subjected to fiber-sparing, NMDA-induced lesions of the Re nucleus.

Figure captions Figure 1 :
Figure captions

Figure 2 :
Figure 2: Synthetic figure illustrating the interconnectivity pattern between the different regions of the prefrontal cortex (A; according to the delimitation proposed by Carlén, 2017), the hippocampus (B), the amygdala (C) and the reuniens nucleus (Re).Distinction is made between afferent projections (Input from, on the left) and efferent projections (Output to, on the right).Innervation density goes from highest (thick arrows) to sparsest (thin arrow), with in between densities (arrow of intermediate thickness).The arrow density code is strengthened by the fill pattern of the rectangles corresponding to each cortical area (from deep color, densest, to faded color, the least dense).This coloration nicely illustrates the