Inferring action-dependent outcome representations depends on anterior but not posterior medial orbitofrontal cortex
Introduction
The OFC has been argued to mediate abroad array of cognitive and behavioural functions in learning and decision-making, not all of which can be easily reconciled. This is due, in part, to the fact that the vast majority of studies examining the OFC have lacked specificity in detailing the particular subregion targeted (i.e. medial, ventral, or lateral OFC); something that is particularly true of studies into rodent OFC. One example is the well-established finding that OFC damage causes impairments in reversal learning (e.g. Izquierdo et al., 2004, Rolls et al., 1994, Schoenbaum et al., 2002, Schoenbaum et al., 2003). This effect was subsequently shown to be specific to lesions of the lateral portion of the OFC (lOFC), with medial OFC damage actually resulting in a facilitation rather than a deficit in reversal learning (Mar, Walker, Theobald, Eagle, & Robbins, 2011). Likewise, there have been several demonstrations of lOFC inactivation leaving instrumental outcome devaluation intact (Ostlund and Balleine, 2007, Parkes et al., 2017), whereas mOFC inactivation has been found to impair it (Bradfield et al., 2015). It is possible, therefore, that a number of seemingly inconsistent findings are in fact a result of functional heterogeneity across the OFC regions being manipulated. A recent review (Izquierdo, 2017) has added weight to this suggestion by detailing, with unprecedented specificity, the neuroanatomical placements described in studies of rodent OFC and how each subregion might be linked to its specific functions.
The impairment in outcome devaluation we observed as a result of mOFC inactivation was a part of a larger investigation into the function of the mOFC more generally (Bradfield et al., 2015). Specifically, we inactivated mOFC using both excitotoxic lesions and inhibitory (hM4Di) Designer Receptors Exclusively Activated by Designer Drugs (DREADDs) in an instrumental choice situation where food outcomes (pellets or sucrose) were either observable or unobservable. Inactivating the mOFC selectively impaired performance during tasks in which outcomes were unobservable, i.e., in which they had to be recalled from memory, including outcome devaluation and specific-Pavlovian-instrumental-transfer (specific PIT). In contrast, performance on tasks in which the outcomes were presented and so observable, i.e., reinforced devaluation, outcome-selective reinstatement, and instrumental contingency degradation tests, was intact. Together, these results suggest that the mOFC is critical for inferring the occurrence of outcomes when they are unobservable as opposed to when they need merely to be recognised in the environment.
Beyond localising our placements in the medial subregion of OFC, however, we did not explore any differences of function along its anterior-posterior gradient. This could be significant because there are several lines of evidence suggesting that a further anterior-posterior distinction might exist. First, although we did not intentionally target the anterior mOFC, the placements in our 2015 study did tend to omit its posterior regions in an attempt to avoid overlap with prelimbic cortex. By contrast, in a recent study Munster and Hauber (2017) used more posterior (and from our inspection of their lesion image (Fig. 2), more dorsal) mOFC lesion placements, and were unable to replicate the impairments we observed in outcome devaluation and specific PIT. Second, in her review spanning various rodent studies, Izquierdo (2017) suggested that the anterior and posterior regions of mOFC might be functionally distinct, proposing that unobservable outcome retrieval is restricted to anterior mOFC, whereas delay discounting involves the more posterior regions (Izquierdo, 2017; Fig. 3). Finally, a recent article has shown that the anterior and posterior sections of lateral OFC play functionally distinct roles in aspects of decision-making (Panayi & Killcross, 2018), highlighting the possibility of a similar distinction in medial OFC. Indeed, in other species and in humans especially, several studies have suggested the existence of functional distinctions between anterior and posterior OFC (e.g. Kringelbach and Rolls, 2004, Mansouri et al., 2017, Smith et al., 2010). One particularly interesting finding from a meta-analysis of human neuroimaging studies suggests that activity in anterior but not posterior OFC correlates with representations of more complex or abstract reinforcers (Kringelbach & Rolls, 2004). Putting the question of homologies aside for a moment, this function appears to align closely with our proposal that the mOFC is necessary to infer action outcomes from memory.
Based on these findings, therefore, it might be reasonable to expect that the anterior and posterior regions of rodent mOFC carry out functionally distinct roles within dissociable neural circuits. This was the hypothesis investigated in the current study. First, we explored whether there were any observable differences in the density of output pathways from anterior versus posterior mOFC by placing retrograde tracers into the basolateral amygdala (BLA), posterior dorsomedial striatum (pDMS), nucleus accumbens core (NAc core), the mediodorsal thalamus (MD), and ventral tegmental area (VTA) and then quantifying the number of retrogradely labelled neurons in each region of the mOFC. We chose these structures because they have all been reported to receive some degree of input from mOFC (Hoover & Vertes, 2011), and they have all been previously identified as critical for various aspects of goal-directed action (see Hart, Leung, & Balleine, 2014). We next compared the performance of rats with specific excitotoxic lesions of either anterior or posterior mOFC on instrumental tasks in which action outcomes are absent on test, including specific PIT and instrumental outcome devaluation, and a further task for which outcomes are present on test: outcome-selective reinstatement. We predicted that rats with anterior but not posterior mOFC lesions would display deficits in both specific PIT and outcome devaluation, as these tasks require rats to infer absent outcomes, which we hypothesise relies on the anterior mOFC. We further predicted that all rats would demonstrate intact performance on a test of outcome-selective reinstatement, as the outcomes are presented during this test, can be directly recognised and, therefore, do not need to be inferred.
Section snippets
Material and methods
Our first aim was to establish whether the output pathways of anterior vs. posterior mOFC to NAc core, pDMS, BLA, MD and VTA differed in density. Rats received unilateral injections of the retrograde tracer flurogold (FG) into the pDMS, NAc core, or VTA plus an injection of the retrograde tracer cholera toxin B (CTb) into the NAc core, BLA, MD or VTA. Ten days after surgery, rats were perfused and brains were processed for immunofluorescence identification of retrogradely labelled neurons in
Experiment 1. Comparison of afferents from anterior vs. posterior mOFC to pDMS, NAc core, BLA and MD using retrograde tracing
Of the 46 retrograde injections in 23 rats that were conducted, 16 injections were excluded from the analysis due to misplacement or spread of the tracer beyond the boundaries of the target structure. This left 28 injections in 20 rats for the subsequent analysis; 6 in the NAc core, 6 in the pDMS, 6 in the BLA, 4 in the MD and 6 in the VTA. The results of tracing from these injections are shown in Fig. 1. The top row (Fig. 1A–E) shows examples of retrograde labelling in the anterior mOFC
Discussion
Taken together, the current findings demonstrate that the anterior and posterior subregions of the rodent mOFC can be dissociated both with regard to the density of their projections to specific target regions and with regard to their functions. First, we used retrograde tracing to assess the density of projections from anterior and posterior mOFC to the BLA, pDMS, NAc core, MD and VTA and found that, whereas projections to DMS, MD and VTA were relatively similar across anterior and posterior
Conclusions
Together with the findings of our previous study (Bradfield et al., 2015) and those of Munster and Hauber (2017), the current findings suggest that it is the anterior portion of the mOFC that is critical for animals to infer action-dependent outcomes when they are unobservable, whereas the posterior mOFC subserves a distinct function, perhaps related to response effort. Overall, these findings add to a growing trend within the literature of producing more specificity and consistency with
Conflict of interest
The authors declare no conflicts of interest.
Acknowledgements
The research reported in the manuscript was supported by grants to BWB and LAB from the National Health and Medical Research Council (NHMRC);Project Grant 1087689 and Project Grant 1148244. BWB is supported by a Senior Principal Research Fellowship from the NHMRC of Australia, Research Fellowship 1079561.
References (34)
- et al.
Goal-directed instrumetnal action, contingency and incentive learning and their cortical substrates
Neuropharmacology
(1998) - et al.
Medial orbitofrontal cortex mediates outcome retrieval in partially observable task situations
Neuron
(2015) - et al.
The role of the prelimbic cortex in instrumental conditioning
Behavioural Brain Research
(2003) - et al.
Toward a neurobiology of obsessive-compulsive disorder
Neuron
(2000) - et al.
Dorsal and ventral streams: The distinct role of striatal subregions int eh acquisition and performance of goal-directed actions
Neurobiology of Learning and Memory
(2014) - et al.
Circuit-based corticostriatal homologies between rat and primate
Biological Psychiatry
(2016) - et al.
The functional neuroanatomy of the human orbitofrontal cortex: Evidence from neuroimaging and neuropsychology
Progress in Neurobiology
(2004) - et al.
Obsessive-compulsive disorder as a failure to integrate goal-directed and habitual action control
- et al.
The role of the nucleus accumbens in instrumental conditioning: Evidence of a functional dissociation between accumbens core and shell
Journal of Neuroscience
(2001) - et al.
Prefrontal cortex in the rat: Projections to subcortical autonomic, motor, and limbic centers
The Journal of Comparative Neurology
(2005)
Consolidation of goal-directed action depends on MAPK/ERK signaling in rodent prelimbic cortex
Journal of Neuroscience
Prefrontal cortico-striatal disconnection blocks the acquisition of goal-directed action
Journal of Neuroscience
Statistics for the social sciences
Projections of the medial orbital and ventral orbital cortex in the rat
Journal of Comparative Neurology
Functional heterogeneity within rat orbitofrontal cortex in reward learning and decision making
Journal of Neuroscience
Bilateral orbital prefrotnal cortex lesions in rhesus monkeys disrupt choices guided by both reward value and reward contingency
Journal of Neuroscience
The neural bases of obsessive-compulsive disorder in children and adults
Development and Pscyhopathology
Cited by (33)
Outcome-selective reinstatement is predominantly context-independent, and associated with c-Fos activation in the posterior dorsomedial striatum
2022, Neurobiology of Learning and MemoryCitation Excerpt :Finally, because little is known about the neural mechanisms underlying outcome-selective reinstatement, we aimed to identify its potential neural correlates by examining expression of the immediate early gene and activity marker c-Fos in various brain regions. One brain region that selective reinstatement has been demonstrated to depend upon is the posterior dorsomedial striatum (pDMS) (Yin et al., 2005) and one region it does not depend upon is the medial orbitofrontal cortex (mOFC) (Bradfield et al., 2015; Bradfield et al., 2018). Thus, we included both regions in our analysis with the expectation that c-Fos expression would reflect performance in the pDMS but not the mOFC.
Is the core function of orbitofrontal cortex to signal values or make predictions?
2021, Current Opinion in Behavioral SciencesInactivation of posterior but not anterior dorsomedial caudate-putamen impedes learning with self-administered nicotine stimulus in male rats
2021, Behavioural Brain ResearchCitation Excerpt :Using this stringent testing approach, we show that rats with lesions to the p-dmCPu had blunted nicotine-evoked goal-tracking in the form of dipper entry duration after the non-contingent nicotine infusion on Test 4 (Fig. 3D, black squares). These findings are consistent with the deficits seen in previous studies demonstrating that the inactivation of p-dmCPu impairs instrumental learning [43,46,55,56] and impairs the acquisition of learning with nicotine stimulus using the discriminated goal-tracking task [8]. Therefore, our findings further strengthen the understanding that p-dmCPu is central for a broad range of associative learning mechanisms that include learning involving pharmacological states and appetitive stimuli.
Controlling one's world: Identification of sub-regions of primate PFC underlying goal-directed behavior
2021, NeuronCitation Excerpt :Areas 14 (rostral vmPFC/mOFC) and 14-25 (caudal vmPFC) were not specifically involved in the response to changes in A-O contingencies. The complete lack of effects of inactivation is consistent with rodent studies, in which lesions of a putative homolog of these regions, the anterior mOFC, impaired the effects of outcome devaluation but not contingency degradation (Bradfield et al., 2015, 2018). Although marmosets receiving overactivation of area 14 did not differentiate between degraded and non-degraded sessions, the finding that baseline responding was also affected prevents any firm conclusions concerning contingency degradation.