Neuromodulators of orbitofrontal cortex function: A review of endocannabinoid, neurotrophin, and cell adhesion systems

Abstract

The ability to incorporate changing information from the environment into our actions is essential for flexible behavior. The orbitofrontal cortex (OFC) is involved in integrating new information into on-going and future actions. As elsewhere in the brain, neuronal activity in the OFC is subject to fine-tuning by neurotransmitters like serotonin, dopamine, and norepinephrine. The purpose of this review is to highlight other neuromodulatory factors that have received less attention but are also potent regulators of OFC-dependent decision making. Specifically, we summarize the impacts of endocannabinoids, the neurotrophin Brain-derived Neurotrophic Factor, and cell adhesion systems in both medial and lateral compartments of the OFC. Generally speaking, perturbations to these systems disrupt flexible decision making, resulting in behaviors resembling poor learning, perseveration, and deferral to habitual actions. Understanding the molecular mechanisms supporting OFC function may drive the discovery of novel therapeutic approaches to alleviating these symptoms in various neuropsychiatric disorders.

1. Introduction

Individuals make myriad decisions every day, ranging from what to say, to where to go for dinner. Decision making often requires the individual to incorporate sensory information, past experiences, and emotional states to evaluate options, assess consequences, form a course of action, and ultimately execute an action. The ability to represent previously learned information and accurately integrate that information into future actions is indispensable for decision making. Numerous psychiatric disorders are defined by compromised decision making (Hélie et al., 2017). Therefore, it is imperative to study the neurobiological underpinnings that govern decision-making behavior to identify potential therapeutic approaches.

For decades, one region of the brain known to contribute to flexible choice behavior is the orbitofrontal cortex (OFC) (Butters and Pandya, 1969; Teitelbaum, 1964), a ventral subdivision of the frontal cortex located on the inferior surface of the frontal lobes, immediately above the orbits of the eyes. The OFC receives inputs from multiple sensory modalities (Carmichael and Price, 1995), which contribute to the ability to predict the expected outcome of an action, or cue, and to assign subjective value to potential actions (Banerjee et al., 2020; Rolls, 2004; Sadacca et al., 2018; Wallis, 2019). The OFC also plays a critical role in action strategy formation and execution by categorizing relevant task information such as value and outcome probability to guide subsequent decision making (Stalnaker et al., 2015; Zhou et al., 2021a). The OFC is thought to contribute to a so-called cognitive map of task space by encoding an abstract representation of stimulus-outcome (S-O) and response-outcome (R-O) contingencies (Qiu et al., 2024; Schuck et al., 2016; Wilson et al., 2014). The accurate representation of these associations allows for updating action strategies when familiar taskrelevant contingencies are violated (Niv, 2019), and evidence suggests that the OFC may be capable of storing long-term representations of task variables (Farovik et al., 2015; Li et al., 2022; Sequeira et al., 2024; Sias et al., 2021; Yount et al., 2025; Zhou et al., 2021b; Zimmermann et al., 2018).

The OFC is fairly conserved across species (Heilbronner et al., 2016; Ongür and Price, 2000; Price, 2007; Wallis, 2011), and recent technical advances in neuroscience including optogenetics, calcium imaging, and chemogenetics have allowed researchers to understand cell function with unprecedented detail. Moreover, combining these techniques with transgenic rodent models allows for the testing of exquisitely specific hypotheses, for instance to investigate the functions of activity-and/or molecularly-defined neuron populations. The purpose of this review is to highlight recent advances in our understanding of neuromodulatory factors that impact OFC-dependent decision-making behavior with a focus on rodent model systems. Common behavioral tasks are detailed in Box 1. We will specifically discuss endocannabinoids (eCBs), the neurotrophin Brain-derived neurotrophic factor (BDNF), and cell adhesion factors. These proteins have been the focus of multiple exciting investigations in recent years, which will be summarized. And although these signaling systems have largely been investigated independently, each regulates synaptic strength and ultimately influences the ability of OFC circuits to undergo and maintain experience-dependent plasticity. Our goal is to complement existing excellent reviews concerning the impacts of other, more commonly discussed factors, like serotonin (e.g., (Roberts, 2011)), dopamine (e.g., (Chau et al., 2018; Ott and Nieder, 2019; Volkow et al., 2019)), noradrenaline (e.g., (Cerpa et al., 2021)), and corticosteroids (e.g., (Sequeira and Gourley, 2021)).

Box 1.

Tasks commonly used to assess OFC-dependent decision making in rodents

a). Reversal Learning

Reversal learning tests for the capacity to update action strategies when subjects are presented with changing R-O or, more commonly, S-O contingencies. For instance, in classical assays, rats or mice learn to dig media with unique scents to retrieve food pellets, neglecting alternative scented media that do not predict reward. When the reward-predictive scents change, rodents must flexibly modify digging action strategies to continue retrieving pellets. In other versions, a behavioral apparatus may include levers, nosepoke ports, or a touch-sensitive screen with distinct displays, and rodents must orient their actions based on reinforcer availability associated with spatial location or again, reward-predictive cues. Then, the relationship is “reversed,” such that the previously incorrect response or stimulus becomes rewarded. These relationships can be deterministic, fully predictive, or probabilistic, partially predictive. Rodents that fail to modify their behavior are considered to have failed to reverse.

b). Impulsivity and risk assessment

1. Another area of OFC research has focused on the region’s contribution to impulsive-like behavior (Torregrossa et al., 2008). One aspect of impulsive-like behavior is impulsive choice, which can be indicated by the inability to wait for a large reward and instead pursue an immediate, small reward in the delay discounting procedure. Rodents undergo instrumental training in which responding on one nose port (or lever) results in a small reward, and a separate response results in a large reward. Once rodents display a preference for the large reinforcer, they undergo the delay phase when responding for the large reinforcer results in a progressively increasing delay before reward delivery. With increasing delay times, rodents will ultimately switch to prefer the small, immediate reward. Rodents displaying impulsive-like behavior will more readily acquire a preference for the small reward (i.e., shorter latency to switch strategies).

2. Impulsive-like behavior may also include an inability to inhibit motor responses, which can be assessed using choice serial reaction time tasks. The task typically takes place in an operant conditioning chamber equipped with multiple response ports. To initiate a trial, the subject must perform an action like nose-poking into a designated port. A light turns on in one of the ports for a variable duration. The subject must then quickly and accurately respond by nose-poking the illuminated port, with reaction time recorded. Correct responses within the response window result in a food pellet reward, while incorrect, premature, or omission responses result in no reward and are recorded. The number of premature responses can indicate impulsive-like behavior due to lack of inhibitory control.

3. Additional insights into OFC function have been made using related but distinct probabilistic discounting tasks. In these tasks, rodents are typically trained to choose between a small but certain reward that is delivered on every trial, and a larger reward that is delivered only with a certain probability. Across the session, the probability of obtaining the large reward is systematically decreased, forcing the subject to continually reassess the value of each option as the likelihood of receiving the larger reward diminishes. Performance on this task is often interpreted as a measure of risk-based behavior, a concept related to impulsivity. Rodents that persist in choosing the large reward even when its probability is low are considered more risk-seeking, as they overvalue the potential high payoff despite unfavorable odds.

c). Outcome Devaluation

Value-based choice refers to making decisions based on the perceived worth of anticipated outcomes and is often assessed via devaluation procedures. Generally, rodents are trained to either attend to cues that predict delivery of distinct food reinforcers or to respond for distinct food reinforcers, measuring either S-O associations or R-O associations, respectively. In a different setting, one of the foods is then “devalued” either by pairing it with a lithium chloride injection, which induces gastric malaise, or by providing subjects with ad libitum access to it, which induces satiety. Subsequently, rodents will stop seeking the reinforcer because it has been devalued, whereas the other type of reinforcer maintains its original value. Failure to do so indicates a failure to integrate outcome value into choice behavior.

d). Contingency Degradation

Contingency degradation most commonly determines whether rodents engage in instrumental behaviors based on the expectation that they will be reinforced. Rats or mice are typically trained to generate two distinct responses for food reinforcers, then one response fails to be reinforced reliably, shifted to a zero or null contingency. First developed by Hammond (Hammond, 1980), it has since been adapted and modified for various purposes, for example to enable experimenters to differentiate epochs of time in which rodents must encode and retrieve new memories as they track changing reward likelihood (Gourley et al., 2012a). If rodents fail to update their behavior, they are considered to have failed to integrate action-reward expectancy into action strategies.

1.1. The medial and lateral OFC are functionally distinct

The OFC can be broadly divided into medial and lateral compartments (mOFC and lOFC). The mOFC generally corresponds with the medial portion of Brodmann’s areas 11 (anteriorly) and 14, as well as portions of area 10. The lOFC corresponds with lateral area 11, as well as Brodmann’s 47/12 and 13. These compartments are considered functionally distinct (Izquierdo, 2017). For example, inactivation of excitatory neurons in the lOFC renders rodents unable to update action strategies following reinforcer devaluation (Gremel and Costa, 2013; Zeeb and Winstanley, 2013), while chemogenetic stimulation of neurons in the lOFC restores sensitivity to outcome value following insults that impair this process (Renteria et al., 2018). The Coutureau lab reported that chemogenetic inhibition of neurons in the lOFC fails to alter instrumental devaluation following stable reinforcement contingencies; however, when rodents have to track multiple task features, such as new contingencies, lOFC cell inactivation impairs this ability (Fresno et al., 2019; Parkes et al., 2018). This pattern is consistent with evidence that the lOFC is necessary for mice to flexibly update action strategies based on changing reinforcement contingencies in an adaptation of the instrumental contingency degradation task (Gourley et al., 2013; Li et al., 2022; Whyte et al., 2019; Zimmermann et al., 2018). These and other patterns support a model in which the lOFC is thought to incorporate task rules and specific associations between events and states to guide choice (Costa et al., 2023; Tan et al., 2025).

Excitatory neuronal activity in the mOFC is also necessary for rodents to alter action strategies following devaluation, but specifically when outcomes are not readily observable and must be inferred (Bradfield et al., 2015; Bradfield et al., 2018; Gourley et al., 2016; Woon et al., 2022). For instance, rats with lesions of the mOFC can avoid devalued outcomes when they are presented, but they cannot modify responding when the reward is not immediately observable (Bradfield et al., 2015). And in contrast with the lOFC, neurons in the mOFC do not appear to be necessary for action selection during contingency degradation (Bradfield et al., 2015), in theory because rodents can use strategies developed when reinforcers were delivered noncontingently (and were thus observable) to perform optimally. In probabilistic discounting, damage to the mOFC increases risk selection, particularly following successful trials (win/stay), suggestive that the mOFC tracks reward history, and loss of this ability leads to a deferral to immediate reward feedback (Stopper et al., 2014). Altogether, the mOFC is thought to track hidden or higher-order task features generated from prior learning and use these task features to guide decision making during uncertainty (Lopatina et al., 2017).

These models concerning functional specializations between the lOFC and mOFC can account for response patterns in even simple tasks. For instance, one study directly compared the impacts of cell body lesions in the lOFC vs. mOFC in an uncued spatial reversal task. Lesions of the lOFC delayed acquisition of the reversal – as mice were unable to incorporate new reward contingencies into choice, they disengaged from the task, not making any choices at all (Gourley et al., 2010). Meanwhile, lesions of the mOFC caused mice to persist in the old strategy. In this case, the reinforcer was largely unobservable because mice initially persist in the previously reinforced response. With time, intact mice redirected their behavior, but damage to the mOFC interfered with the ability to react efficiently to uncertainty. Relatedly, mice and rats with damage or cell inactivation within the mOFC respond excessively in progressive ratio tests, wherein acquiring each reinforcer requires increasing effort relative to the prior reinforcer, and as such, reinforcers become less and less observable and more uncertain with time (Gourley et al., 2010; Gourley et al., 2016; Münster and Hauber, 2018; Münster et al., 2020). In the absence of the healthy mOFC, rodents are unable to gate responding as the reinforcer becomes increasingly unobservable and respond in excess relative to the worth of the reinforcer. Meanwhile, rodents with damage to the lOFC tend to perform normally in related effort-based tasks (reviewed (Hauber, 2026)).

Altogether then, both lOFC and mOFC subregions are thought to contribute to maintaining cognitive maps of task space; however, they seem to do so via tracking separate task components (Bradfield and Hart, 2020). To summarize, the mOFC is thought to anticipate the outcome of a given course of action (the terminal state of the task), tracking hidden task features generated from prior learning, and using these task features to guide decision making during instances of outcome uncertainty (Bradfield et al., 2015; Bradfield et al., 2018; Dalton et al., 2016; Tan et al., 2025). In contrast, the lOFC is thought to track one’s current position in the task, incorporating current information with immediate prior event history, choice behavior, and outcome receipt (Costa et al., 2023; Hocker et al., 2010; Tan et al., 2025). Damage to these regions – for instance, due to prolonged alcohol or cocaine exposure – causes subjects to either abandon goal-directed response strategies or fail to adjust actions when task demands change. Repeatedly, these patterns have been associated with drug-induced changes in endocannabinoid, neurotrophin, and cell adhesion systems (e.g., (DePoy et al., 2016; DePoy et al., 2017; Li et al., 2023; Morisot et al., 2019; Renteria et al., 2018; Renteria et al., 2021)), which emphasizes the importance of understanding these systems and their impact on OFC function. Next, we will summarize current knowledge regarding endocannabinoid (Part 1), neurotrophin (Part 2), and cell adhesion (Part 3) control of OFC-dependent decision making. Each section begins by describing these systems, writ large. We then discuss investigations concerning the OFC. Investigations overwhelmingly focus on the lateral compartment, as it has been the historical focus of OFC research, but we highlight investigations concerning the mOFC when possible.

1.2. Part 1. Endocannabinoid (eCB) signaling

Activation of group 1 metabotropic glutamate receptors (mGluR1/5), muscarinic acetylcholine receptors, or increase in intracellular Ca⁺ induce synthesis of eCBs like anandamide and 2-Arachidonoylglycerol (2-AG) in the postsynaptic cleft (Araque et al., 2017). Once synthesized and released from postsynaptic neurons, eCBs diffuse “backward” across the synaptic cleft to exert their effects on receptors on presynaptic neurons, both inhibitory and excitatory in nature. The two primary cannabinoid receptors are CB1R and CB2R. CB1R is the most widely distributed and abundantly expressed within the central nervous system, while CB2Rs are primarily localized to microglia cells within the central nervous system. These cannabinoid receptors are predominantly Gi-coupled, and they inhibit neurotransmitter release, including glutamate and GABA. The termination of eCB signaling relies on reuptake mechanisms and subsequent enzymatic degradation by fatty acid amide hydrolase (FAAH) or monoacylglycerol lipase (MAGL) (Kendall and Yudowski, 2016).

eCB signaling generally suppresses excitatory neuron output by serving as an activity-dependent, retrograde negative feedback mechanism that limits glutamatergic transmission and stabilizes network activity. When excitatory pyramidal neurons become highly active, they synthesize and release eCBs, which diffuse backward across the synapse and activate CB1Rs on presynaptic glutamatergic terminals. CB1R activation then inhibits presynaptic calcium influx and adenylyl cyclase signaling, reducing glutamate release probability and producing short-term depression of excitation (Fig. 1). With repeated or patterned activity, this mechanism can engage eCB-mediated LTD, leading to persistent weakening of excitatory synapses. Through these actions, eCB signaling provides a homeostatic brake on cortical excitation, shapes dendritic integration and spike timing, and regulates experience-dependent plasticity, thereby preventing hyperexcitability while allowing flexible control of cortical information processing. This is complemented by similar negative regulation of GABA release from inhibitory interneurons and persistent weakening of GABAergic synapses with repeated inhibition (Younts and Castillo, 2014), plus considerable eCB modulation of glia, generally thought to support homeostatic synaptic plasticity and the development of certain neuronal networks (Caballero et al., 2016; Martinez Ramirez et al., 2023).

Fig. 1.

A view of endocannabinoid-mediated signaling with emphasis on excitatory OFC neurons. Endocannabinoid-mediated signaling influences OFC function by dampening neurotransmitter release from presynaptic boutons. Stimuli that disrupt this mechanism may contribute to aberrant decision making. Obesogenic diet increases synthesis of eCBs and induces LTD at GABAergic synapses in the OFC. Along the same vein, eCB release from striatal medium spiny neurons (MSNs) dampens glutamate release from OFC terminals in the striatum. Alcohol exposure results in unchecked eCB production from D1R-containing MSNs and thus LTD at OFC-dorsal striatal synapses. Zoom in: eCB released from post-synaptic OFC neurons transverse backwards across the synaptic cleft. Binding of cannabinoid receptors (CBR) on the presynaptic neuron decreases the release of GABA (not pictured) or glutamate. eCB signaling within presynaptic inputs to the OFC is halted by degradation of eCBs by fatty acid amide hydrolase (FAAH) or monoacylglycerol lipase (MAGL). See above and main text for references. Image created with assistance from Biorender.

Cannabis use alters these processes by hijacking the brain’s endogenous eCB signaling, leading to a stronger, longer-lasting, and less spatially precise suppression of excitatory cortical transmission. The primary psychoactive component of cannabis, Δ9-tetrahydrocannabinol (THC), activates CB1Rs on presynaptic terminals, mimicking eCBs but without the normal activity-dependent control. As a result, glutamate release from excitatory cortical neurons is broadly reduced, rather than being selectively and transiently suppressed only at highly active synapses. Acute THC exposure therefore exaggerates the normal inhibitory role of eCBs, dampening cortical excitability, impairing synaptic plasticity, and disrupting the precise timing of neuronal firing that supports learning and working memory. With repeated cannabis exposure, persistent activation of CB1Rs by THC can lead to long-term suppression of excitatory cortical activity. Over time, this can result in reduced intrinsic excitability of pyramidal neurons, impaired induction of LTP, and a bias toward LTD-like synaptic states. With repeated or heavy use, chronic CB1R activation can also lead to receptor desensitization and downregulation, weakening endogenous eCB signaling over time and potentially impairing normal forms of synaptic regulation and LTD.

Disruptions in the eCB system have been implicated in substance misuse (Bedse et al., 2019; Mohammadkhani and Borgland, 2022; Navarrete et al., 2022) and schizophrenia (Fakhoury, 2017), in which decision making and impulse control are affected (Hoptman, 2015; Kozak et al., 2019; Sterzer et al., 2019; Verdejo-Garcia et al., 2018). Compared to non-dependent cannabis users, dependent cannabis users exhibit decreased medial and lateral OFC volumes (Chye et al., 2017), suggesting that drug-induced changes to OFC neurobiology could contribute to changes in behavior. This perspective is supported by investigations in model systems, in which case, cannabinoid signaling can be manipulated genetically or pharmacologically in highly controlled settings (Table 1). For instance, activation of eCB signaling within the lOFC induces impulsive choice patterns, such that rats prefer small immediate rewards compared to larger delayed rewards (Fatahi et al., 2018; Khani et al., 2015). These patterns concord with earlier investigations demonstrating that systemic administration of THC or CB1R agonists induce perseverative behaviors resembling those following damage to the lOFC and prefrontal brain regions and suppress immediate-early gene levels in the lOFC (Boucher et al., 2009; Egerton et al., 2005; Hill et al., 2006; Sokolic et al., 2011). Additionally, overexpression of Cnr1, which encodes CB1R, in the medial prefrontal cortex causes reversal deficits characteristic of damage to the lOFC (Klugmann et al., 2011). This pattern could reflect the impacts of CB1R on prefrontal-to-OFC axon terminals, as the medial prefrontal cortex is a principal input to the OFC.

Table 1.

Investigations into the behavioral consequences of endocannabinoid-mediated signaling within the OFC. Endocannabinoid (eCB)-mediated signaling within the OFC impacts impulsive choice and flexible action, potentially via interaction with vanilloid and inhibitory neurotransmitter systems. Entries are clustered by task.

Citation	Species, sex	Region	Task	Manipulation	Major Finding	Implication
Khani et al., 2015	Rat (male)	lOFC	T-maze decision-making task (delay-based)	CB1R agonist in OFC	Rats prefer small, immediate reward over large, delayed reward	CB1R activation in the OFC induces impulsive choice
Fatahi et al., 2018	Rat (male)	lOFC	T-maze decision-making task (delay-based)	CB1R agonist in OFC	Rats prefer small, immediate reward over large, delayed reward	eCBs in the OFC induce impulsive choice via CB1R and TRPV1
				TRPV1 channel antagonists in OFC	Partially reverses CB1R-induced preference for small, immediate reward
Ucha et al., 2019	Rat (male)	mOFC, lOFC	DDT 2-CSRT	Classified high and low impulsive choice (DDT) and impulsive action (2-CRSTT) rats. Measured gene expression of Cnr1 and Gabra1	High impulsive choice: higher Cnr1 expression in mOFC	Cnr1 expression correlates to impulsive behavior
					High impulsive action: lower Gabra1 expression in lOFC
Gremel et al., 2016	Mouse (male & female)	lOFC	Devaluation	Deletion of CB1R from OFC neurons projecting to the DS	Prevents mice from forming habitual response strategies	CB1R-mediated dampening of OFC-DS circuit activity is necessary for habit
Renteria et al., 2021	Mouse (male & female)	lOFC	Devaluation	Chronic intermittent ethanol exposure	Upregulates eCB signaling in D1R-containing spiny neurons, dampens glutamate release from OFC inputs	Ethanol exposure increases eCB signaling at OFC-DMS synapses which disrupts flexible action
				CB1R antagonist in DMS	Restores flexible responding after ethanol exposure
Seabrook et al., 2023b	Mice (male)	lOFC	Devaluation	Obesogenic diet	Increased excitability of lOFC pyramidal neurons via decreased GABAergic signaling Elicits habit-like response strategies	GABA-mediated regulation of excitatory neurons lOFC is necessary for goal-directed behavior
Seabrook et al., 2023a	Mice (male)	lOFC	Devaluation	Obesogenic diet	Decreased sensitivity to reinforcer devaluation	Obesogenic diet elicits habit-like responding via changes to GABAergic tone in the lOFC
				DREADD inhibition of vGAT+ interneurons in lOFC	Replicates effect of obesogenic diet impairing behavioral flexibility
				Increasing inhibitory tone in lOFC via local drug infusion or optogenetics	Restores goal-directed behavior in mice with prior history of obesogenic diet
Lau et al., 2021	Rat (male)	lOFC	NA	Obesogenic diet + CB1R antagonist, miniature and evoked inhibitory postsynaptic currents recorded in lOFC pyramidal neurons	Hypertrophy of astrocytes, increased extra synaptic glutamate stimulates mGluR5, induces eCB signaling, suppresses GABAergic signaling	Obesogenic diet leads to eCB-mediated LTD of GABAergic transmission

Abbreviations: cannabinoid type 1 receptor (CB1R), delay discounting task (DDT), dopamine type 1 receptor (D1R), dorsal striatum (DS), dorsomedial striatum (DMS), endocannabinoid (eCB), gamma-aminobutyric acid (GABA), gene encoding GABA_A receptor alpha-1 subunit (Gabra1), lateral OFC (lOFC), long term depression (LTD), medial OFC (mOFC), metabotropic glutamate receptor subtype 5 (mGluR5), orbitofrontal cortex (OFC), response-outcome (R-O), two-choice serial reaction time task (2-CSRT, vanilloid receptor 1 (TRPV1).

While few investigations have focused on the mOFC, it notably receives inputs from the amygdala, prelimbic cortex, and other cortical neurons, which are enriched in CB1R, including in excitatory (projection) cell populations (Domenici et al., 2006; Fortin and Levine, 2007). Interestingly, OFC-targeted projections arising from the hippocampus overwhelmingly target the mOFC, largely sparing more lateral subregions (Yang et al., 2025). CB1R is present at both inhibitory and excitatory (albeit to a far lesser degree) hippocampal synapses, which could conceivably impact hippocampal-mOFC interactions (Kawamura et al., 2006; Monory et al., 2015; Ruehle et al., 2013). And finally, mice displaying greater impulsive choice in a delay discounting task also have higher expression of Cnr1 within the mOFC (Ucha et al., 2019).

Because CB1R is predominantly located on presynaptic axon terminals, gene expression patterns within the OFC likely influence both local network activity and neurotransmission within downstream targets, a major one being the dorsomedial striatum (Schilman et al., 2008). Indeed, LTD in the striatum is strongly impacted by eCB release (Gerdeman et al., 2002; Moreira et al., 2015), particularly at corticostriatal synapses, only nominally at thalamo-striatal synapses, and independent of postsynaptic striatal spiny neuron subtype (Wu et al., 2015). Presence of eCB signaling within lOFC-dorsal striatal circuits accordingly attenuates goal-directed behaviors and allows for the predominance of competing habitual action control (Gremel et al., 2016), such that depletion of CB1R on OFC terminals in the striatum or the striatal patch compartments that receive abundant OFC inputs disrupt habit formation (Gremel et al., 2016; Nadel et al., 2020). These investigations reveal a mechanism by which organisms might engage familiar behaviors that have been reliably reinforced in the past (habits), freeing attentional resources to attend to other stimuli. Addictive drugs can bias reward-seeking behaviors toward habit, which could, for some individuals, contribute to maladaptive drug seeking. Again, eCB may be a mechanism: eCB signaling at lOFC synapses with striatal medium spiny neurons containing dopamine-type 1 receptors (D1R) appear to underlie alcohol-induced deficits in outcome-guided decision making, eliciting habitual behavior (Renteria et al., 2021). Specifically, alcohol-mediated disruption of D1R activation results in unchecked eCB signaling at lOFC synapses and subsequent LTD.

Another source of input to the dorsomedial striatum is the mOFC, with projections terminating in the medial-most compartments lining the lateral ventricles (Schilman et al., 2008). Nevertheless, very little eCB research in rodent models has focused on the mOFC. Interestingly, though, CB1R stimulation disrupts working memory in non-human primates and alters mOFC-striatal connectivity (Kohut et al., 2022). Taking into account the work described in the prior paragraph, elevated or unchecked CB1R signaling at mOFC-striatum synapses would be expected to moderate the typical behavioral impact of these connections, potentially unleashing competing behaviors. The mOFC is thought to support goal-directed action control particularly in ambiguous circumstances, at least in part via outputs to the dorsomedial striatum, which appears to be particularly attuned to frequencies of nonrewarded actions in order to adjust choices accordingly (Jenni et al., 2022). Should this function be mitigated, for example through excess CB1R stimulation, habit-like response patterns would be expected, such that organisms engage in familiar routines when they encounter uncertainty.

In another series of experiments, prolonged exposure to obesogenic diet resulted in hypertrophied astrocytes within the lOFC and subsequent excess extra-synaptic glutamate (Lau et al., 2021). This excess glutamate stimulated group 1 metabotropic glutamate receptors on excitatory lOFC pyramidal neurons, which led to production of eCBs and LTD of GABAergic transmission within the lOFC (Lau et al., 2021). The resulting disinhibition of lOFC neurons by obesogenic diet induced habit-based food seeking, as measured using multiple behavioral assays including reinforcer devaluation and contingency degradation, and which was causally associated with insufficient GABAergic signaling and hyper-excitability amongst OFC neurons (Seabrook et al., 2023a; Seabrook et al., 2023b) (see also (Mukherjee et al., 2026)). Independent investigations concord with the notion that hyper-activation of neurons within the lOFC induces habit-like response biases: GABA_Aα1 receptor depletion in the lOFC (Swanson et al., 2015) and chemogenetic hyper-activation of excitatory neurons in the healthy lOFC (Hinton et al., 2019; Li et al., 2022) both induce habit-like responding for food reinforcers. While these investigations did not investigate eCB signaling, they highlight abundant evidence that control of goal-directed behavior by the lOFC concords with an inverted U-shaped curve, with “too much” or “too little” activity amongst excitatory neurons causing organisms to defer to competing habit-like behaviors.

Interestingly, THC and obesogenic diet cause dendritic spine loss on lateral (but not medial) OFC neurons (Kolb et al., 2018; Thompson et al., 2017). This could relate to desensitization of the CB1R, as Cnr1 knockout has the same effects in the dorsomedial prefrontal cortex (Lee et al., 2014), as well as behavior: The densities and learning-related plasticity of dendritic spines on excitatory lOFC neurons have been repeatedly coupled with the ability of organisms to maintain goal-directed behavior in the face of changing task requirements (Allen et al., 2022; Butkovich et al., 2025; Li et al., 2022; Li et al., 2023; Sharp et al., 2017; Whyte et al., 2019; Yount et al., 2025). Conversely, Cnr1 knockout mice are impaired in reversal-based tasks (Varvel and Lichtman, 2002), presumably driven in large part by loss of CB1R in the OFC, given its prominent role in reversal learning. Altogether, eCB systems appear important for adjusting behavioral strategies, with both “too much” and “too little” signaling imperiling action flexibility.

1.3. Part 2. Neurotrophin signaling

Neurotrophins are secreted growth factor proteins that act as signaling molecules that interact with extracellular receptors and initiate intracellular signaling cascades to influence neuronal growth, differentiation, and function. The most well-known neurotrophins are nerve growth factor (NGF), neurotrophin-3, neurotrophin-4/5, and brain-derived neurotrophic factor (BDNF). In the cortex, NGF facilitates the connection and function of cholinergic neurons during development (Martínez et al., 1985). In adulthood, NGF is primarily produced via GABAergic neurons and likely aids in the preservation and function of cholinergic projections throughout the cortex (Biane et al., 2014). Neurotrophin-3 and −4/5 are relatively understudied compared to earlier discovered neurotrophic factors, NGF and BDNF. Overall, they appear to support the survival and differentiation of neurons within both the central and peripheral nervous systems (Huang and Reichardt, 2001). This review will focus on BDNF.

BDNF is an integral signaling molecule in neurogenesis and neural development, such that constitutive Bdnf knockout mice suffer developmental defects and die shortly after birth. It is expressed in the vast majority of excitatory neurons, including in the OFC (Ehinger et al., 2023). BDNF exerts its actions predominantly by binding tropomyosin receptor kinase B (TrkB), a cell surface receptor tyrosine kinase. Upon binding, TrkB receptors dimerize and trigger activation of intracellular kinase domains. This activation leads to phosphorylation of specific tyrosine residues on the intracellular domain of TrkB. These phosphorylated residues serve as docking sites for various adaptor proteins. Once the adaptor proteins are recruited, they initiate multiple downstream signaling cascades including MAPK/ERK, PI3K/Akt, and PLCγ. These activated signaling cascades ultimately mediate expression of genes involved in neuronal survival, differentiation, synaptic plasticity, and neurotransmitter function. Systemic administration of TrkB agonists generally improve OFC-dependent reversal learning, while antagonists interfere with it, suggesting that TrkB-mediated signaling helps to optimize OFC function (Barfield and Gourley, 2017, 2018).

In lateral portions of the OFC, BDNF is necessary for flexibly modifying learned R-O associations when new learning allows organisms to successfully obtain reward (Gourley et al., 2013; Pitts et al., 2016; Zimmermann et al., 2017). Stimulation of TrkB can overcome deficits in outcome-based decision making after Bdnf knockdown in the lOFC (Zimmermann et al., 2017), or exposure to cocaine (Pitts et al., 2020), a stimulus that reliably degrades flexible action and lOFC neuron structure alike (DePoy and Gourley, 2015; DePoy et al., 2014). These behavioral impacts are likely executed by coordination of circuit-wide BDNF-mediated signaling through TrkB, occurring locally within the lOFC and in interconnected regions, such as the basolateral amygdala and dorsal striatum (Ehinger et al., 2023; Gourley et al., 2013; Li et al., 2023; Pitts et al., 2020; Pitts et al., 2018; Zimmermann et al., 2017).

Interestingly, cocaine upregulates Calcium-dependent activator protein for secretion 2 (CAPS2; also referred to as CADPS2), a dense core vesicle-associated protein that promotes the activity-dependent release of neuropeptides including BDNF, in the lOFC (Trinoskey-Rice et al., 2021). This effect may be a compensatory response to dendritic spine and synaptic marker loss in this region following cocaine exposure (Li et al., 2023). Another possibility is that CAPS2 up-regulation in the lOFC is one way that addictive drugs hijack brain biology to promote further drug seeking: Excessive alcohol intake triggers BDNF-related signaling pathways in the lOFC (Laguesse et al., 2017), which promotes habit-like alcohol seeking (Morisot et al., 2019). Relationships between BDNF and alcohol consumption are, however, complex, as other BDNF-containing cells populations in the lOFC appear to be protective against drug seeking (Gunasekaran et al., 2026). Parcellating specific circuits and circumstances in which BDNF-mediated signaling impacts drug-vs. non-drug-seeking may be useful.

BDNF within the mOFC is necessary for mice to sustain stable representations of outcome value, particularly when outcomes are not readily observable and must be inferred (Gourley et al., 2016). Bdnf-deficient mice thus cannot differentiate between high-and low-value reinforcers, nor can they habituate to progressive ratio schedules of reinforcement, which require mice to expend progressively more effort for a given reinforcer over the course of a session. Deficits cannot be attributed to alterations in extinction learning (Gourley et al., 2009b). As such, the effects of Bdnf silencing in the mOFC resemble the effects of lesions of the same region (Bradfield et al., 2015; Bradfield et al., 2018), wherein rats fail to update behavior following changes to reward value. Subsequent experiments confirmed that value processing by the mOFC requires local TrkB binding (Woon et al., 2022), with reductions in TrkB in the mOFC resulting in impaired value updating. Meanwhile, Li and colleagues revealed that BDNF in the mOFC also controls sociability in mice (Li et al., 2021), which opens up the possibility that it controls social decision making – for instance, assessing the potential value in approaching a novel conspecific vs. avoiding it, or even more complex behaviors like integrating prior social experiences into later choice behavior. Studies focusing on the contributions of how BDNF-mediated signaling influences OFC function in reward-related behaviors are summarized in Table 2.

Table 2.

Investigations into the behavioral impacts of BDNF-mediated signaling within the OFC. Neurotrophin-mediated signaling within the OFC impacts cognitive flexibility and decision making, with some evidence for interactions with cytoskeletal regulatory systems. Entries are clustered by task.

Citation	Species, sex	Region	Task	Manipulation	Major Finding	Implication
Gourley et al., 2013	Mouse (male)	lOFC	CD	Bdnf knockdown	Insensitive to changes in R-O contingency	BDNF in the OFC is necessary for flexible decision making
			Conditioned place preference		Resistance to the extinction of place preference for cocaine
Zimmermann et al., 2017	Mouse (male)	lOFC	CD	Bdnf knockdown	Insensitive to changes in R-O contingency	BDNF in the OFC coordinates flexible decision making via interaction with the BLA and cytoskeletal regulatory systems
				+ Rho-kinase inhibition	Restores sensitivity to changes in R-O contingencies after Bdnf knockdown
				Unilateral OFC Bdnf knockdown + contralateral unilateral BLA lesion	Insensitive to changes in R-O contingency
				+7,8-DHF	Restores sensitivity to changes in R-O contingencies after Bdnf knockdown
Pitts et al., 2018	Mouse (male)	lOFC	CD	Trkb.t1 overexpression	Insensitive to changes in R-O contingency	TrkB is necessary for flexible decision making
DePoy et al., 2019	Mouse (male)	lOFC	CD	Itgb1 knockdown in adolescence	Insensitive to changes in R-O contingency in adulthood	Cell adhesion systems interact with TrkB during early-life critical periods to optimize later decision-making capacity
				+7,8-DHF in early-life sensitive period	Restores sensitivity to changes in R-O contingencies after Itgb1 knockdown
Pitts et al., 2020	Mouse (male & female)	lOFC	CD	Adolescent cocaine exposure	Induces insensitivity to changes in R-O contingency	Stimulation of TrkB confers resilience to the impact of cocaine on decision-making behavior, countering cocaine-induced biases toward habit-like action
				+7,8-DHF	Corrects cocaine-induced decision-making biases
				Bdnf knockdown + MDMA paired with new action-outcome memory encoding	MDMA pairing overcame decision making deficits induced by Bdnf knockdown in lOFC
Li et al., 2022	Mouse (male & female)	lOFC	CD	Unilateral Bdnf knockdown + contralateral unilateral inhibition of neuron activity within the BLA or DMS during encoding new R-O associations	Insensitive to changes in R-O contingency	Encoding new R-O memories requires BDNF in the OFC and interactions with the BLA and DMS
Gunasekaran et al., 2026	Mouse (male & female)	lOFC	Ethanol self-administration CD	Bdnf overexpression in sparse lOFC-to-DLS neurons following chronic volitional alcohol intake	Decreased ethanol intake and seeking behaviors	BDNF in specific lOFC projections is protective against habit-like ethanol consumption and seeking
				TrkB agonist following chronic volitional alcohol intake	Reinstates goal-directed ethanol seeking
Gourley et al., 2016	Mouse (male)	mOFC	Devaluation	Bdnf knockdown	Decreased sensitivity to reinforcer devaluation, increased break point ratios	BDNF in mOFC is necessary for value-based choice, particularly when outcomes are partially observable and must be inferred
			Progressive ratio
Woon et al., 2022	Mouse (male & female)	mOFC	Devaluation	Ntrk2 knockdown and Trkbt.1 overexpression	Decreased sensitivity to reinforcer devaluation	TrkB in the mOFC is necessary for outcome-specific value memory updating to guide future action selection, coordinating with the vHC
				Trkb.t1 overexpression	Decreased the density of mushroom-shaped dendritic spines and lengthened the remaining spines Decreased sensitivity to reinforcer devaluation
				Trkb.t1 overexpression in mOFC + contralateral unilateral inhibition of vHC during encoding new R-O associations
Gourley et al., 2009a	Rat (male)	lOFC	Fear conditioning	Chronic corticosterone exposure	Impairs extinction of conditioned fear and decreased Bdnf mRNA	Stress hormone exposure impairs fear memory extinction and is associated with decreased Bdnf
Wang et al., 2018	Rat (male)	lOFC	AST	Chronic social defeat stress	Increased number of trials needed to reach criteria in reversal learning and extradimensional set shift stages of the AST and decreased BDNF levels	Lower BDNF in the OFC correlates with reduced cognitive flexibility

Abbreviations: attentional set-shifting task (AST), basolateral amygdala (BLA), brain-derived neurotrophic factor (BDNF), contingency degradation (CD) variant of the original task created by Hammond (1980), conditioned taste aversion (CTA), dorsal striatum (DS), dorsomedial striatum (DMS), dorsolateral striatum (DLS), forced swim test (FST), gene encoding integrin-β1 (Itgb1), lateral OFC (lOFC), medial OFC (mOFC), orbitofrontal cortex (OFC), response-outcome (R-O), gene encoding truncated TrkB (Trkb.t1), Tyrosine receptor kinase B (trkB), gene encoding trkB (Ntrk2), ventral hippocampus (vHC), trkB agonist 7,8-dihydroxyflavone (7,8-DHF), 3,4-methylenedioxymetham-phetamine (MDMA).

1.4. Part 3. Cell adhesion molecules

Cell adhesion molecules (CAMs) are cell surface proteins that help to coordinate cellular interactions with the extracellular matrix (ECM). There are four main groups of CAMs, including immunoglobulin-like adhesion molecules, integrins, cadherins, and selectins. In general, CAMs act as transmembrane linkers by integrating extracellular signals to drive changes in cell shape, adhesion, and movement by initiating intracellular signaling cascades. In particular, integrins are heterodimeric receptors composed of two associated transmembrane subunits called α and β. The α subunit predominantly determines the extracellular partner, while the β subunits determine intracellular signaling partner(s) (Kerrisk et al., 2014). Ligand binding and subsequent activation of the heterodimerized integrin results in focal adhesion – or cell adhesion – to the extracellular matrix, referring to a complex, three-dimensional network of molecules that surrounds neurons, glia, and blood vessels, filling the space between cells and providing structural support (Bradley and Koleske, 2009).

In general, CAMs and their signaling partners are crucial for dendritic spine and synapse stabilization (Cameron and McAllister, 2018; Duncan et al., 2021; Mohan et al., 2019; Sytnyk et al., 2017). As dendritic spines are the primary sites of excitatory inputs onto neurons, it is sensible to imagine that spine stability is necessary for optimal OFC-dependent behavior: Consistent with this notion, when we selectively bred mice that perform poorly in the contingency degradation task, their off-spring also perform poorly, and they suffer dendritic spine attrition in the lOFC (Butkovich et al., 2025). Excitatory inputs from the basolateral amygdala, ventral hippocampus, and thalamus appear to coordinate goal-seeking behavior (e.g., see (Barfield and Gourley, 2019; Leung et al., 2024; Li et al., 2022; Wassum, 2022; Zong et al., 2025)). And in the mOFC, inputs from the prelimbic cortex optimize risk assessment behavior (Jenni et al., 2025), while inputs from the amygdala mediate adaptive responses to cues based on the desirability of associated rewards (Lichtenberg et al., 2021). Finally, the OFC famously receives input from sensory modalities, and these inputs are likely necessary for its function in integrating sensory information into value processing, decision making, and choice behavior (Rudebeck and Rich, 2018). It may be no surprise, then, that empirical investigations have confirmed that integrins and their substrates support lOFC function. β1-integrins in particular have been a focus of multiple investigations because they are found on a majority of vertebrate cells, importantly including throughout excitatory postsynaptic densities within the frontal cortex (Mortillo et al., 2012; Schuster et al., 2001; Shapiro et al., 2017a). Neuronal depletion of β1-integrin within the lOFC early in life causes dendritic spine collapse and obstructs lOFC-dependent updating of reward-and fear-related behavior (DePoy et al., 2019; Whyte et al., 2021). Thus, integrin-mediated development of neurons in the lOFC is necessary for action flexibility.

Integrin-mediated adhesion results in phosphorylation and activation of multiple proteins including Arg (Abelson-related gene) kinase (Bradley et al., 2006; Lapetina et al., 2009). This is important because Abl-family-kinases (Abl and Arg) promote neurite outgrowth and branching (Moresco et al., 2005). Arg kinase phosphorylates p190RhGAP, which in turn inhibits Rho-kinase by stimulating inherent GTPase activity. Rho-kinase typically phosphorylates LIM-kinase, which phosphorylates the master cytoskeletal regulator cofilin, inactivating it. In sum, integrin-Arg kinase-mediated signaling acts as a brake on Rho-kinase, preventing structural plasticity, stabilizing dendrites, spines, and synapses (Gourley et al., 2011; Kerrisk and Koleske, 2013; Lin et al., 2013; Park and Goda, 2016; Sfakianos et al., 2007). Moreover, inhibition of Rho-kinase can promote neuron growth, for instance, elongating dendritic branches in certain circumstances (Couch et al., 2010).

As in Itgb1-deficient mice, in Arg⁻/⁻deficient mice, dendritic spines in the lOFC collapse during adolescence, resulting in cortical spine and synapse loss and poor capacity to modify learned instrumental behaviors (Gourley et al., 2009b; Gourley et al., 2012b). Experiments locally infusing Arg kinase inhibitors into specific brain regions revealed that behavioral deficits are attributable to Arg loss in the lOFC (DePoy et al., 2017; Gourley et al., 2012a). Further, Arg kinase activators can improve OFC-dependent reversal learning following stress hormone exposure (Shapiro et al., 2017b). Finally, a series of investigations combining viral-mediated gene silencing and pharmacological interventions revealed that β1-integrin control of reward-related behaviors is attributable to interactions with Arg and downstream effects on Rho-kinase in the lOFC (Whyte et al., 2021). Altogether, a β1-integrin-Arg-Rho-kinase signaling axis appears essential for dendritic spine stabilization within excitatory neurons in the brain (Warren et al., 2012) and, concurrently, for lOFC function.

Rho-kinase levels in the frontal cortex modestly decrease during adolescent development (Shapiro et al., 2017a), thought to create a permissive environment for the structural changes occurring on excitatory neurons during this time (i.e., the elimination of some dendritic spines and stabilization of others). Recent investigations into the impact of social isolation in adolescence revealed that isolation increases cortical Rho-kinase levels (Li et al., 2024), and accordingly, spines on excitatory neurons in the lOFC and medial prefrontal cortices apparently fail to be pruned as would otherwise be expected (Hinton et al., 2019; Li et al., 2024). Suppressing Rho-kinase signaling (to redirect developmental trajectories) ameliorated at least some structural and behavioral consequences of isolation, including poor ability to engage in goal-directed actions and atypical structures of excitatory neurons in the lOFC (Hinton et al., 2019; Li et al., 2024).

Multiple gene variants of CAMs have been linked to drug misuse in humans (Drgon et al., 2010; Liu et al., 2006). These investigations led to the discovery that ablation of neuronal β1-integrin in the lOFC induces sustained cocaine seeking in mice (Whyte et al., 2021). Meanwhile, stimulating integrin-mediated signaling can mitigate cocaine seeking once cocaine has been devalued (Swanson et al., 2017). It is possible that β1-integrin-mediated signaling is just one of several CAM mechanisms affecting drug seeking (DePoy and Gourley, 2015; Gourley et al., 2011).

Another CAM is Neuronal Cell Adhesion Molecule (NrCAM), which supports both excitatory and inhibitory synapses and has been associated with schizophrenia, leading to investigations of NrCAM knockout mice (Buhusi et al., 2024; Buhusi and Buhusi, 2026). These mice have several executive functioning deficits, particularly when encountering stressors (Buhusi et al., 2024; Buhusi and Buhusi, 2026). Interestingly, stressors and pharmacological interventions modulate immediate-early gene levels throughout several regions of the brain, but the lOFC is insensitive to these stimuli in NrCAM-deficient mice (Buhusi et al., 2024; Buhusi and Buhusi, 2026). It thus appears that this CAM (and likely others) is necessary for recruitment of lOFC systems in behaviorally relevant contexts.

2. Conclusions

The OFC is critical for constructing and updating representations that guide adaptive, goal-directed behavior. The studies reviewed here highlight eCB, neurotrophin, and cell adhesion systems as important modulators of OFC function that complement the actions of more commonly discussed neurotransmitter systems. eCB signaling regulates synaptic transmission and plasticity within OFC networks and connected corticostriatal circuits, influencing impulsivity and behavioral flexibility. BDNF-TrkB signaling in both lateral and medial OFC subregions supports the encoding and updating of outcome-related information, enabling flexible decision making. Meanwhile, cell adhesion molecules and their downstream signaling partners maintain the structural architecture of OFC neurons, supporting the stability of dendritic spines and synapses required for adaptive behavior.

Although these signaling systems have largely been investigated independently, a central theme of this literature is their considerable convergence. Each regulates synaptic strength and ultimately influences the ability of OFC circuits to undergo and maintain experience-dependent plasticity. Perturbations to these pathways produce similar behavioral outcomes and shifts toward habit-like response strategies. These shared consequences suggest that OFC-dependent cognition relies not on any single molecular pathway, but rather coordinated regulation of synaptic and structural plasticity across interconnected neural circuits. Future work could determine whether and how system interaction within specific OFC cell populations and projection-defined circuits support cognitive maps, value representations, and adaptive action selection. For instance, trkB transactivates integrin-β1, and one report suggests that these interactions within the lOFC supports flexible, goal-directed action (DePoy et al., 2019). There is also considerable evidence that BDNF-mediated signaling leads to the release of eCBs and on the flip side, that CB1R activation induces BDNF release, leading to neuroprotective consequences (Martinez Ramirez et al., 2023). Better understanding points of convergence amongst these systems may reveal common molecular vulnerabilities that contribute to neuropsychiatric disorders characterized by impaired decision making, compulsivity, and maladaptive habits.

Data availability

No data was used for the research described in the article.