Intracranial self-stimulation (ICSS), where animals self-administer electrical stimulation to specific brain regions, has long been used to study reward pathways. Dopamine neurons in the ventral tegmental area (VTA) are crucial for ICSS, but the underlying cognitive mechanisms remain unclear. Phasic activity of VTA dopamine neurons, typically firing at 10-20 Hz, signals prediction errors during reward learning, traditionally viewed as representing the scalar value of unexpected rewards. However, recent research suggests these signals can support more general learning, such as cognitive map development, without necessarily assigning value to cues. This raises the question: if dopamine prediction errors aren't value signals, why do animals work for VTA dopamine stimulation? This study aimed to investigate the cognitive representation of dopamine stimulation during ICSS by examining its interaction with the Pavlovian-to-instrumental transfer (PIT) effect. PIT differentiates between specific and general reward representations: specific representations motivate actions associated with a specific cue-reward pairing, while general representations motivate actions associated with any reward. The study hypothesized that the cognitive representation underlying ICSS would differ depending on the frequency of dopamine neuron stimulation, testing whether a physiologically relevant frequency (20 Hz) versus a supraphysiological frequency (50 Hz) would be represented as a specific reward or a general value signal.

Extensive research demonstrates ICSS using VTA dopamine neuron stimulation, but few studies have explored its neural representation. The PIT effect, observed in both rodents and humans, provides a valuable tool for distinguishing specific and general reward representations. In PIT studies, subjects first learn associations between cues and different rewards, then learn to perform actions to obtain those rewards. Finally, in a test phase, cues are presented without reward delivery, revealing the nature of the reward representation. Sensory-specific reward representations lead to selective increases in responding to the action associated with the presented cue, while general value representations increase responding for both actions. Previous research has established the role of dopamine neurons in reward learning and prediction error signaling (Schultz et al., 1997; Roesch et al., 2007). Studies have also explored the role of dopamine in supporting more general learning phenomena, including the development of cognitive maps (Keiflin et al., 2019; Sharpe et al., 2020). However, the specific cognitive representation underlying ICSS remained unclear.

The study utilized a PIT paradigm in rats with optogenetic stimulation of VTA dopamine neurons. TH-Cre rats received bilateral infusions of a Cre-dependent adeno-associated virus (AAV) carrying channelrhodopsin-2 (ChR2) into the VTA. Optic fibers were implanted bilaterally for stimulation. Rats were assigned to one of two groups: a 20-Hz stimulation group or a 50-Hz stimulation group. In the Pavlovian phase, rats learned to associate auditory cues with either sucrose pellets or VTA dopamine stimulation (at either 20 or 50 Hz). In the instrumental phase, rats learned lever pressing for either reward. A progressive ratio schedule increased the effort required to obtain rewards. Finally, in the PIT test, cues were presented without reward, allowing assessment of cue-induced responding. To confirm that the 20-Hz stimulation was effective as a teaching signal, a separate blocking experiment was conducted. Rats learned associations between visual cues and rewards. Then, auditory cues were introduced in compound with the visual cues, and dopamine stimulation (20 Hz) was paired with reward delivery for one compound. A probe test assessed learning of the auditory cues. A devaluation test then assessed whether the association was mediated by a reward representation. A further experiment assessed how much rats would lever press for 50 Hz VTA stimulation (ICSS). Finally, electrophysiological recordings and fiber photometry were used to compare the firing rates and dopamine release in the nucleus accumbens (NAc) in response to 20-Hz and 50-Hz stimulation, comparing to food reward.

The 20-Hz stimulation group did not exhibit robust instrumental responding beyond continuous reinforcement, nor did it show specific PIT. The 50-Hz stimulation group, however, showed robust instrumental responding and specific PIT, comparable to food reward. In the blocking experiment, 20-Hz stimulation functioned as a teaching signal, driving sensory-specific learning of the stimulation-paired cue; however, it did not support robust lever pressing. The degree of lever pressing for 50-Hz stimulation (ICSS) was positively correlated with the unblocking effect, suggesting that high frequency stimulation produces a specific reward effect, distinct from a teaching signal. Electrophysiological recordings showed similar numbers of action potentials with both 20-Hz and 50-Hz stimulation, but 50-Hz stimulation evoked greater dopamine release in the NAc over a shorter time period, compared to 20-Hz stimulation or food reward. Food delivery, however, evoked a significantly more prolonged increase in dopamine release, greater than 20 Hz stimulation but not significantly different from 50 Hz stimulation.

The findings challenge the value hypothesis of dopamine signaling, suggesting that the physiological prediction error signal (20 Hz) does not inherently function as a reward. While 20 Hz is effective as a teaching signal, driving learning without acting as a reward, it lacks the reinforcing properties to convey value. In contrast, 50-Hz stimulation functions as a specific reward, motivating behavior through a sensory-specific representation. This highlights a dissociation between the role of dopamine in learning (teaching signal) and its role as a reward. The results offer a novel perspective on ICSS, emphasizing the frequency-dependent effects on dopamine release and cognitive representation. The difference in effect between 20 Hz and 50 Hz is attributed to the rate of action potentials, rather than their total number, impacting downstream dopamine release. This suggests that future research should focus on understanding the mechanisms that distinguish between dopamine’s roles in learning and reward.

This study demonstrates a frequency-dependent effect of dopamine neuron stimulation on cognitive representation and ICSS. 20-Hz stimulation, approximating a prediction error signal, acts as a teaching signal driving learning but not as a reward in itself. 50-Hz stimulation functions as a distinct sensory reward. Future research should investigate the neural mechanisms underlying this frequency-dependent difference and explore the potential implications for understanding reward processing and addiction.

The study primarily focused on rodents, limiting the generalizability to humans. The optogenetic stimulation, while effective, may not perfectly replicate natural dopamine release patterns. The sample size was moderate, and further research with larger samples could enhance statistical power and generalizability.