The medial PFC (specifically, the PL) is associated with action-outcome associative learning (Balleine & Dickinson, 1998
; Killcross & Coutureau, 2003
), while lateral orbitofrontal cortex lesions retard stimulus-response learning in reversal tasks (e.g.
, Schoenbaum et al., 2002
). Situated at the junction of the ventrolateral orbital and PL cortices in both primates and rodents, the mOFC may be expected to influence behavioral responding in tasks that require updating stimulus-response or
action-outcome associations, but little is known about this structure in these contexts. Here, mice were initially trained to respond on a nose poke aperture in the northeastern corner of an operant conditioning chamber, then were required to shift responding to an aperture in the northwestern corner or vice versa
for reinforcement. A variable ratio schedule of reinforcement was used, with no discrete cues signaling reinforcement delivery. Thus the task required mice to update action-outcome—as opposed to stimulus-response—associative relationships in order to obtain food pellets, and subsequent sensitivity to outcome devaluation confirmed mice responded based on action-outcome response contingencies (below).
In agreement with a classic report on the effects of mOFC lesions in monkeys (Iverson & Mishkin, 1970
), mOFC lesions increased “perseverative” responding—responding on the previously-reinforced aperture despite non-reinforcement [main effect of lesion F(1,21)=5.5, p=0.03; lesion × session F(6,126)=2.1, p=0.06] ()—while acquisition of the newly reinforced response was unaffected [main effect of lesion F<1; lesion × session F(6,126)=1.7, p=0.1] (). Mice with PL lesions also acquired the newly reinforced response [main effect and interaction Fs<1] (), but in this case, perseverative responding was unaffected [effect of lesion F(1,7)=1.2, p=0.3; session × lesion F<1] (), indicating distinct roles for these adjacent medial PFC structures.
Distinct response patterns in an instrumental reversal task in mice with discrete medial prefrontal lesions
mOFC and PL lesions in this and all other experiments were largely separated on both dorso-ventral and rostro-caudal planes, with the typical mOFC lesion at the rostral-most tip of the frontal pole and larger mOFC lesions spreading laterally to include the ventral OFC (). mOFC lesions were rostral enough to avoid the infralimbic cortex, although some GFAP staining in the rostral PL was noted. Approximately 50% of mOFC lesion mice had some degree of GFAP staining along the needle track in at least 1 hemisphere. PL lesions were caudal to mOFC lesions and encompassed the PL and anterior cingulate cortex, with spread to the infralimbic cortex in some mice (). Two mOFC and 6 PL lesions were unilateral, and 2 “mOFC” mice appeared to have lesions only in the infralimbic cortex; these animals were excluded.
To further characterize the role of the mOFC in behavioral inhibitory processes, we generated another group of lesion mice and conducted 5 test sessions in which mice were required to respond on a progressive ratio schedule of reinforcement for food. When break point ratios were analyzed, an interaction between lesion and session was detected [F(8,160)=2.6, p=0.01], and subsequent post-hoc tests indicated mice with mOFC lesions escalated responding, achieving break point ratios that differed from sham mice at a trend level of significance during session 2 (p=0.057), and that were significantly higher than sham levels during subsequent test sessions (ps≤0.03) ().
Medial prefrontal lesions regulate progressive ratio break point ratios
By contrast, mice with PL lesions differed from sham mice at a trend level of significance during the final session (p=0.07), suggestive of a declining response pattern (). To clarify this possibility, we calculated each animal's break point ratio as a percentage of its day 1 baseline. Both sham and mOFC lesion mice shifted responding upward to 135% and 159%, respectively, of day 1 baseline. By contrast, mice with PL lesions shifted downward, achieving break point ratios that were, on average, 72% of baseline across several sessions [main effect of lesion F(2,40)=4.6, p=0.02; post-hoc ps<0.05] (). Representative GFAP staining in lesion mice from these experiments is shown (); as indicated in , mOFC and PL lesions were distinguishable by rostro-caudal position within the mPFC.
To confirm that the effects of PL lesions were not simply attributable to insensitivity to the previously learned action-outcome association, we devalued the food outcome with 30-min prefeeding with the reinforcer pellets used in the task. All mice consumed equivalent amounts of food during this prefeeding period (relative to shams, ps≥0.2; not shown). Subsequently, sham, mOFC, and PL mice showed the expected attenuation of instrumental responding with no difference between groups [F(2,32)=1.4, p=0.3] (). Mice also extinguished responding at equivalent rates when reinforcement was withheld across several sessions [effect of lesion and lesion × session Fs<1] (). These findings are in agreement with the argument that, under normal circumstances, the PL invigorates reinforced instrumental responding by maintaining sensitivity to the motivational value of the outcome (Corbit & Balleine, 2003
), but lesions do not impact upon action-outcome associative relationships acquired prior to lesion placement (Ostlund & Balleine, 2005
Sensitivity to outcome devaluation, non-reinforcement, and progressive ratio training prior to lesion placement
Implicit in this interpretation is the idea that if lesions were placed after the progressive ratio response requirements had been learned, responding would not be affected. To address this possibility, we trained another group of mice to acquire reinforcement, then further trained these animals to respond on a progressive ratio schedule of reinforcement with 5 test sessions before mOFC or PL lesion placement. We then tested the same animals on a progressive ratio schedule after recovery (5 sessions). Before lesion placement, responding did not differ by group designation as determined by break point ratio [main effect of group and group × session Fs<1]. After lesions were placed, break points remained unchanged [main effect of group and group × session Fs<1] (). Our findings thus suggest that neither the mOFC nor PL is required for progressive ratio performance once response parameters have been learned. These data also provide the first evidence for mOFC involvement in the acquisition, and not expression, of an instrumental response schedule in rodents.
vHC lesions disinhibit instrumental responding
In addition to its well-established role in spatial learning and memory, the hippocampus regulates motivational sensitivity to food and drug reward. Moreover, stimulation of the ventral sector uniquely excites neurons within the mOFC (Ishikawa & Nakamura, 2003
), suggesting this region may provide the mOFC with information regarding the motivational salience of an appetitive outcome and thereby contribute to its regulation of goal-directed behavior. In this case, lesions of the vHC might be expected to result in similar response patterns relative to lesions of the mOFC. Indeed, vHC lesion mice successfully shifted responding to a newly reinforced aperture in an instrumental reversal task [main effect of lesion F<1; lesion × session interaction F(6,90)=1.6, p=0.15] (), but showed an impairment in response inhibition on the previously reinforced aperture, specifically during the initial test sessions [lesion × session interaction F(6,90)=3, p=0.01; sessions 1–3 post-hoc ps<0.05] (). Moreover, as with mOFC lesions, vHC lesions increased progressive ratio break points [main effect of lesion F(1,10)=6.5, p=0.03] (). Histological analyses indicated vHC lesions were largely limited to the ventral 50% of the caudal hippocampus, though some larger lesions spread dorsally, resulting in GFAP staining in the intermediate hippocampus. Lesions tended to be biased towards the rostral extent (e.g., Bregma −2.7) of the vHC or the caudal extent (e.g., Bregma −3.7) (), but this distinction did not appear to affect behavioral responding in our tasks.
vHC lesions mimic mOFC lesions
lOFC and dHC lesions have distinctive effects in an instrumental reversal task
A premise of this manuscript is that the mouse mOFC regulates action-outcome response flexibility in a manner that is unique relative to related prefrontal structures. In a final series of experiments, we dissociated the mOFC from lOFC by placing lesions in the lOFC and testing mice in the reversal and progressive ratio tasks. Mice with dHC lesions were also generated, as the dHC has no projections to the OFC (Cenquizca & Swanson, 2007
), and recent studies highlight its functional and genetic dissociation from the vHC (Dong et al., 2009
; cf., Fanselow & Dong, 2010
), so lesions of this region might also be expected to produce distinctive effects in these two tasks. Saline-infused mice did not differ and were combined for representative purposes.
As predicted, lOFC and dHC lesions produced distinctive response patterns in the instrumental reversal task that were dissimilar to mOFC lesion response profiles. First, both lOFC and dHC lesions delayed
the acquisition of the reversal [session × lesion interaction F(12,132)=3.9, p<0.001], though in distinct ways: Mice with lOFC lesions responded less than sham mice during session 3 (p=0.02), but not later (). By contrast, mice with dHC lesions appeared to reverse during early sessions, but were unable to achieve optimal responding, as indicated by fewer responses during the final test sessions (sessions 6–7, ps<0.006), perhaps because optimal responding depended on the spatial location of the aperture within the operant conditioning chamber (Mahut, 1971
; Whishaw & Tomie, 1997
). Somewhat surprisingly, lOFC lesions facilitated the extinction of responding on the previously reinforced aperture [session × lesion interaction F(12,132)=2.6, p=0.004, session 1 p<0.001], thus the medial and lateral OFC compartments were dissociable on both instrumental reversal response and response suppression measures. dHC lesions had no effect on response inhibition (all ps≥0.09) ().
lOFC and dHC lesions produce distinct response patterns
Unlike mOFC lesions, lOFC lesions had no significant effects on break point ratios when mice were required to respond on a progressive ratio schedule of reinforcement, and mice with dHC lesions achieved higher ratios [main effect of lesion F(2,30)=8.5, p<0.001; post-hoc p=0.004] (), as has been previously reported in rats (Schmelzeis & Mittleman, 1996
). Histological analyses indicated lOFC infusions resulted in prominent GFAP staining in the lOFC and lateral ventral OFC that spared medial prefrontal structures in all mice (). Twenty-eight percent of lOFC infusions resulted in particularly large lesions that spread laterally to affect the dorsolateral orbital cortex (“DLO” in Paxinos & Franklin, 2003
). dHC lesions were restricted to the rostral dHC, and most encompassed all major subregions (). In several mice, NMDA spread ventrally such that GFAP staining was detected in the intermediate hippocampus, but the ventral hippocampus was spared. In fact, a subset of animals in both the dHC and vHC groups had prominent GFAP staining within the intermediate hippocampus. Thus, disparate behavioral response patterns in these groups are presumed to be due to cell death within the non-overlapping dorsal and ventral regions, respectively. Two dHC lesions were unexpectedly non-detectable, and 2 were unilateral; these animals were excluded.