Dopamine neurons develop phasic increases in population-level activity to dopamine-predictive cues

Cues paired with natural reward evoke phasic activity in dopamine neurons, and dopamine release in striatal projection targets ^13–16. Given that we found optogenetic stimulation of dopamine neurons induced conditioned behavior to discrete paired cues, we asked if dopamine neurons might acquire phasic neural responses to these paired cues, using fiber photometry ¹⁷. For simultaneous optogenetic stimulation and activity measurement in the same neurons, we co-transfected dopamine neurons with ChrimsonR, a red-shifted excitatory opsin, and the fluorescent calcium indicator GCaMP6f (Fig. 2a and b). This strategy led to a ~90% overlap of GCaMP6 and ChrimsonR expression in TH+ neurons below optic fiber placements in the midbrain (Supplementary Fig. 5a–c). Photoactivation of ChrimsonR (590-nm laser) led to rapid, stable increases in GCaMP6f fluorescence that tightly tracked the length of optogenetic stimulation (Supplementary Fig. 5d). To test the behavioral specificity of light activation of ChrimsonR, we confirmed that 590-nm activation of ChrimsonR-expressing dopamine neurons supported robust intracranial self-stimulation behavior (Supplementary Fig. 5e and f), which rapidly extinguished when the 590-nm laser was switched to a 473-nm laser (Supplementary Fig. 5f; session 3). We also recorded photometry signals as rats consumed 10μl of a 10% sucrose solution. This produced a multi-second calcium fluorescence increase that was similar to that evoked by a 5-sec laser train in the same animals (Supplementary Fig. 5g), indicating that, at least at the population level, the optogenetic conditioning procedure taps into innate reward mechanisms. We note, however, that optogenetic stimulation produces artificial neural activation patterns that do not fully mimic natural dopamine neuron activity.

Open in a separate window

Figure 2

Dopamine neurons develop phasic activity in response to cues that predict their activation

(a) Schematic of fiber photometry system. Fiber photometry fluorescence measurements and optogenetic stimulation in the same dopamine neurons was achieved by co-transfecting TH+ neurons with DIO-GCaMP6f and DIO-ChrimsonR containing AAVs. (b) ChrimsonR and GCaMP6f co-expression in the same TH+ neurons in midbrain (n=13). (c) Fiber photometry measurements were made during optogenetic Pavlovian conditioning where neutral cues were paired with orange laser for activation of dopamine neurons. (d) Cues paired with optogenetic activation of dopamine neurons with ChrimsonR (n=8) develop conditioned stimulus properties to evoke locomotion, relative to unpaired (UP, n=5) controls (2-way RM ANOVA session by group interaction, F_(1,12)=52.53, p<.0001; post hoc comparison between groups). (e) Phasic activity in dopamine neurons in response to dopamine-neuron-activation-paired cues developed across Pavlovian training, shown as ΔF/F of the normalized photometry signal. Shaded area represents analysis window. (f) Summary of mean normalized ΔF/F response during the 1^st 1 s of cue presentations on the first and last session (2-way RM ANOVA session by group interaction, F_(1,668)=48.30, p<.0001). (g) Scatterplot of the relationship between conditioned response latency on individual trials and change in fluorescence measured in the first 1 s after cue presentation, compared to the 1 s period before cue onset. A significant negative relationship emerged later in training, (h) where larger changes in fluorescence during the 1^st 1-s of the cue occurred on trials where rats initiated conditioned locomotion faster (n= R²= 0.14, p=0.012). Data expressed as mean ± SEM. ***p< 0.0001.

To assess cue-evoked neural dynamics, we monitored dopamine neuron population fluorescence during optogenetic Pavlovian conditioning (Fig. 2c). As with the ChR2 experiments (Fig. 1), cues paired with ChrimsonR-mediated optogenetic activation of dopamine neurons came to reliably evoke conditioned behavior (Fig. 2d), relative to unpaired controls. In these cue-laser paired rats, we observed an increase in fluorescence at cue onset (preceding laser onset) that grew in magnitude across training (Fig. 2e & f), while no such cue-evoked signals emerged for unpaired control rats (Fig. 2e & f). A further trial-by-trial analysis revealed that, across training, on trials where a CR occurred, cue-evoked dopamine neuron activity became predictive of the latency of locomotion onset; larger magnitude cue-evoked fluorescence was associated with faster conditioned response initiation (Fig. 2g & h). These results show that dopamine neurons develop phasic activity to CSs they have directly established via their associated activation, in the absence of the constellation of sensory inputs that typically accompany seeking and consumption of natural rewards. Further, the magnitude of phasic cue-evoked population-level dopamine neuronal activity encodes the vigor of conditioned behavior.

Cue-evoked dopamine neuron activity reflects expectation of dopamine neuron activity

On the first and last day of optogenetic conditioning, we included probe trials (Fig. 3a, 25% of the total) in which cues were delivered but laser was omitted. On these trials, dopamine neuron activity in paired rats decreased at the time laser would have been delivered (Fig. 3b). This omission-related decrease in fluorescence developed across conditioning, and was not seen in unpaired controls (Fig. 3b–e). These results parallel electrophysiological recordings of dopamine neurons demonstrating a pause in their firing during the omission of expected food or water ¹⁶, which is thought to be mediated by recruitment of local GABAergic neuron activity ¹⁸. Our data suggest that natural reward exposure is not necessary to engage such midbrain plasticity mechanisms in Pavlovian learning ¹⁹.

Open in a separate window

Figure 3

Rapid emergence and extinction of dopamine expectation signals in dopamine neurons

(a) On 25% of trials in the first and last training session, laser was omitted from the paired groups. (b) Photometry signal from probe trials in the paired groups, with unpaired control shown for comparison. On session 12 for the paired group, following the cue-evoked spike in activity, a corresponding dip in fluorescence occurs at the time when laser stimulation would have been delivered. (c) Summary of mean normalized ΔF/F response during the omission period (session by group interaction, F_(1,108)=11.943, p=.0008). (d and e) Trial-by-trial heatmaps for a paired rat during Day 1 (d) and 12 (e) of conditioning. Cue, laser, and laser-omission related responses are evident on Day 12. (f) An extinction session occurred after training, where all cues were presented without laser stimulation. (g) The cue and omission-related fluorescence changes extinguishes rapidly, compared to the final day of paired conditioning. (h) Summary of mean normalized ΔF/F response during the 1^st 1 s of cue presentations during extinction and the probe trials in session 1 and 12 (effect of session, F_(2,459)=10.03, p=.0016). (i) Summary of mean normalized ΔF/F during the laser omission period across sessions (effect of session, F_(2,179)=46.276, p<.0001). (j) Behavior evoked by the cue rapidly extinguished, compared to the final session of training (effect of session, F_(2,18)=23.37, p<.0001). Data expressed as mean ± SEM. ***p< 0.0001.

We next conducted an extinction session, during which these rats received 25 cue presentations, but no laser stimulation was delivered (Fig 3f). This resulted in rapid reduction, to levels observed in session one, of the cue-evoked fluorescence spike (Fig. 3g,h) as well as the omission-related dip in fluorescence (Fig. 3i). Extinction of the neural signal tracked behavioral extinction, such that after one extinction session, behavior evoked by the cue diminished to a level comparable to the first training session (Fig. 3j). Thus, rapid adjustments in the population-level activity of dopamine neurons track extinction of stimulation-evoked learning. Together, the above findings demonstrate neural encoding of the predictive relationship between these cues and direct dopamine neuron activation.

VTA and SNc dopamine neurons confer distinct motivational properties to cues

Reward-associated CSs direct actions not only by serving as reward predictors that come to evoke neural activity, but also by acquiring reward-like incentive value. ‘Incentive value’ here is defined as that property of cues that lends them motivational power to attract attention and become desirable in the absence of reward, an important process that may contribute to compulsive seeking in addiction ¹. The acquisition of incentive value does not necessarily accompany the acquisition of predictive value ¹⁵, and so we next asked if VTA and SNc dopamine-associated CSs acquired incentive value. To do this, we examined the detailed structure of behavior during Pavlovian conditioning in ChR2-conditioned groups. In response to cue presentations, cre+ paired VTA rats (Fig. 4a) showed cue-directed approach behavior, moving to come into proximity (< 1 in) with the cue light while it was illuminated (Fig 4b, c; Supplementary Fig. 6; Video 1). This “attraction” conditioned response is a standard behavioral index of the attribution of incentive motivational value to a CS ^15,20. Critically, cre+ paired SNc rats did not develop approach behavior (Fig. 4d–f). VTA approach probability did not relate to subjects’ proximity to the cue before cue onset (Supplementary Fig. 7), and was not observed in unpaired or cre− controls (Fig 4b, c; Supplementary Videos 2–5). These results suggest that VTA, but not SNc, dopamine neurons confer incentive value to CSs, and this process does not require typical reward-elicited neuronal processes other than dopamine neuron activation.

Open in a separate window

Figure 4

VTA, but not SNc dopamine neurons instantiate Pavlovian cue attraction

(a) VTA dopamine-paired cues support cue approach/interaction. (b) Approach and interaction with the visual cue associated with optogenetic stimulation developed for VTA cre+ paired rats, but not control groups (session X group interaction, F_(6,57)=2.304, p<0.05; post hoc comparisons with unpaired and cre− groups). (c) VTA cre+ paired rats made significantly more total cue approaches across training, compared to controls (main effect of group, F_(6,57)=8.394, p<0.001; post hoc comparisons with unpaired and cre− groups). (d) SNc dopamine-paired cues do no support approach. (e) In contrast to the VTA group, cue approach did not develop for cre+ paired SNc rats, relative to controls (no session X group interaction, F_(6,48)=0.637, p=0.7). (f) SNc groups made almost zero total approaches across training. Data expressed as mean ± SEM. **p< 0.01.

While SNc dopamine paired rats did not develop approach behavior, we observed vigorous locomotion in these rats during cue presentations, but which was not directed at the cue. To quantify this conditioned movement, and compare it to the other groups, we analyzed behavioral videos using motion tracking to interpolate rats’ positions in the experimental chamber in the periods before, during, and after cue presentations on the final day of conditioning (Fig. 5a). We used this position information to determine the frame-by frame velocity and the distance of the rats’ heads from the cue. This tracking revealed that, at cue onset, cre+ paired SNc rats initiated rapid movement, reaching peak velocity within ~1 sec (Fig. 5b). The timing of movement onset and mean velocity in the first 2-sec of the cue was significantly greater for SNc rats than for cre+ VTA paired rats. In this analysis, unpaired and cre− controls did not show cue or laser evoked movement (Fig. 5a–d), consistent with our earlier results. During the last 5-sec of cue presentations, corresponding to the laser stimulation period, SNc and VTA paired rats exhibited stable and similar movement speeds, and velocity quickly returned to low levels after cue offset (Fig. 5a–d). Together with our photometry data, the movement analysis suggests that Pavlovian conditioned dopamine neuron signals, in addition to spontaneous dopamine neuron signals ^10,11, are involved in the generation of vigorous movements, and more so for SNc, compared to VTA, dopamine neurons.

Open in a separate window

Figure 5

SNc-dopamine neuron-paired cues evoke vigorous conditioned movement

(a) Rat position before, during, and after cue presentations was quantified with automated motion tracking software (b) This resulted in velocity (cm/s) traces and position information for each experimental animal. (c) Before cue onset, all rats were at rest, exhibiting low levels of movement (Pre-cue velocity, no effect of group, F_(3,21)=1.165, p=.347). Cue onset elicited rapid, vigorous movement for SNc cue paired rats (n=8), relative to VTA paired (n=8) and unpaired (UP, n=13) and cre− (n=11) controls (1^st 2-sec velocity, effect of group, F_(3,21)=32.39, p<.0001; ***post hoc comparison vs VTA and UN, p<.0001; #post hoc comparison vs UN, p=.01). SNc and VTA paired rats exhibited similar sustained velocity during the rest of the cue and laser period, while UN and cre− controls remained immobile (Last 5-s cue, effect of group, F_(3,21)=16.45, p<.0001; post hoc comparison to UP). (d) Heat maps depicting velocity on individual trials for a representative rat from each group. UP and cre− rats exhibit almost no movement. (e) Cue-directed movement (approach) and rotational movement on the final session were compared using the automated behavior tracking data. (f) VTA paired were biased toward cue-directed movement, reaching a significantly smaller minimum distance to the cue, compared to SNc and control groups, who did not differ (effect of group, F_(3,21)=18.06, p<.0001, post hoc comparison between groups). (g) SNc paired rats showed a bias towards rotational movement, reaching a faster angular velocity compated to VTA rats (effect of group, F_(3,21)=17.02, p<.0001, post hoc comparison between groups). (h) Directly comparing the likelihood of each type of movement, only cue-evoked rotational movement developed for SNc cre+ paired rats, which was expressed exclusively on nearly every trial by the end of training (interaction of conditioned response (CR) type x session, F_(3,21)=30.88, p<.0001; post hoc comparison between CR types). (i) VTA paired rats showed cue-directed and rotational, which became intermixed across Pavlovian training (interaction of CR type x session, F_(3,21)=4.341, p=0.016). (j) To quantify rats’ cue-directed/rotational bias, a CR Score was calculated, consisting of (X + Y)/2, where Response Bias, X, = (# of turns – # of approaches)/(# of turns + # of approaches), and Probability Difference, Y = (p[rotation] – p[approach]). VTA rats transitioned from an initial cue-directed bias to a mixed cue-directed/rotational score, while SNc rats showed an early and stable rotational bias (interaction of group x session, F_(3,42)=3.933, p=0.015); post hoc comparison between groups; unpaired 2-tailed t test on 4-day mean, t₁₄=7.287, p<0.0001). Data expressed as mean ± SEM. ***p< 0.0001.

During extended conditioned movements, rats turned in circles within the chamber, directed contralateral to the simulation hemisphere. We quantified this movement as “rotations” and compared their occurrence to cue-directed movement (Fig. 5e, Supplemental Fig. 8). In agreement with the experimenter-scored behavior data (Fig. 4), motion tracking analysis revealed a cue-directed movement bias for VTA paired rats, who, on average, came closer to the cue during its presentation, compared to SNc rats and controls (Fig. 5f, Supplemental Fig. 8). SNc paired rats, in contrast, showed a bias towards rotational movement, reaching a faster angular velocity, relative to VTA rats (Fig. 5g). While SNc rats showed an exclusive rotational movement phenotype throughout training (Fig. 5h & j), VTA paired rats also developed rotational movement as training progressed, resulting in a mixed cue-directed vs. rotational movement behavioral phenotype (Fig. 5i & j). The transition of VTA rats from purely linear, cue-directed movement to rotational movement could reflect the progressive recruitment of ascending serial midbrain-striatal circuits across extended training ²¹, culminating in cue-related dorsal striatal dopamine release and behavioral control ^22,23, especially if more lateral VTA dopamine neurons, which may contribute more directly to movement ¹⁰, are engaged. Together these results show that VTA and SNc dopamine neurons contribute to conditioned cue attraction and conditioned movement invigoration in distinct ways, and on different timescales throughout the progression of Pavlovian learning.

In addition to being attractive, cues with incentive value can also become desirable, in that they reinforce actions that lead to their procurement. This process is critical for durable reward-seeking behaviors when reward is not immediately available. Building on the results shown above (Fig. 4), we next asked if VTA and SNc dopamine optogenetically-conditioned CSs could subsequently serve as conditioned or “secondary” reinforcers, to support performance of a new action in the absence of optogenetic stimulation (Fig. 6a). Previously-conditioned Cre+ paired VTA (Fig. 6c), but not SNc (Fig. 6d) rats readily pressed a lever to receive conditioned cue presentations in absence of laser activation, indicating that the instantiation of conditioned reinforcement, a canonical test of incentive value properties of cues, is specific to VTA dopamine neurons. Furthermore, this shows that, while SNc-paired cues can generate vigorous movement (Fig. 5) the content of the signal conditioned through SNc dopamine neurons in Pavlovian learning is fundamentally distinct from VTA dopamine neurons.

Open in a separate window

Figure 6

VTA and SNc dopamine neurons differentially create conditioned, but not primary, reinforcement

(a) Conditioned reinforcement test, where lever presses produced the cue previously paired with dopamine neuron stimulation, but no laser. (c) VTA cre+ paired rats made instrumental responses for cue presentations in the absence of laser, relative to controls (2-way RM ANOVA, main effect of group, F_(2,19)=27.18, p<0.0001; post hoc comparisons with unpaired and cre− groups). (d) SNc cre+ paired rats did not respond for cue presentations, relative to controls (2-way RM ANOVA, no effect of group, F_(2,16)=1.407, p=0.274). (b) Primary reinforcement test, where nose poke responses produced optogenetic stimulation of dopamine neurons. (e) VTA (n=16) and SNc (n=13) cre+ rats made a similar number of instrumental responses for dopamine neuron activation (2-way RM ANOVA, no effect of group, F_(1,27)=0.227, p=0.638). Data expressed as mean ± SEM. ***p< 0.0001.

Finally, we assessed the primary reinforcing value of dopamine neuron activation in an intracranial self-stimulation paradigm ⁹, where nose pokes resulted in a brief laser train delivery, with no associated cues (Fig. 6b). Unlike the anatomical dissociation in conditioned reinforcement, VTA and SNc dopamine neuron stimulation produced similar levels of primary reinforcement (Fig. 6e). Taken together, our results show that brief, phasic activity of VTA dopamine neurons is sufficient to apply incentive value to previously neutral environmental cues to promote attraction and create conditioned reinforcement. SNc dopamine neuron activity, alternatively, imbues cues with conditioned stimulus properties that promote movement invigoration more generally. Direct reinforcement of an instrumental action, in contrast to these divergent Pavlovian cue conditioning functions, is perhaps a common currency across VTA and SNc dopamine neurons ^9,24.

Dopamine neurons have heterogeneous motivational functions

Our results confirm a longstanding, fundamental assumption in reward neuroscience – that activity in dopamine neurons can create Pavlovian conditioned stimuli that elicit conditioned behaviors. While our activation can be seen as essentially mimicking the phasic activity proposed to act as a reward prediction error, we show that dopamine neurons do not merely update Pavlovian associations between cues and external rewards ⁴, they directly confer value and can do so in the absence of normal sensory inputs and the other corresponding brain processes that typically accompany natural reward exposure and consumption. We extend previous studies ⁷ by showing that discrete, transient cues become conditioned stimuli via association with relatively brief bursts (~1–5 sec) of dopamine neuron activity. Importantly, our results define the default behavioral responses conditioned by cue-paired phasic dopamine signals in relation to different dopamine neuron populations.

We demonstrate that conditioned stimulus instantiation is a function generally present in the major dopamine neuron output systems in the ventral midbrain, the VTA and SNc (Fig. 1). We found that these conditioned stimuli came to evoke population-level activity in dopamine neurons themselves (Fig. 2), in line with what has been previously demonstrated with single unit recordings during natural (i.e., food) cue conditioning ^13,16,30. This cue-evoked dopamine neuron activity evinced a relationship with behavior as a consequence of Pavlovian conditioning, and fluctuations in this signal tracked a learned expectation of dopamine neuron activity (Fig. 3). Thus, a basic function of dopamine neurons in Pavlovian conditioning is to accumulate and signal predictive information about the probability of future dopamine neuron activity, a phenomenon that fundamentally does not require elaboration of further brain processes normally triggered by an external food reward beyond the elicitation of dopamine activity itself. Once learned, the phasic, cue-evoked dopamine neuron signals functionally represent the motivational value of cues that predict future dopamine neuron activity, which manifests as the vigor or intensity of conditioned behavior ^5,31,32, the details of which depend on the dopamine projection engaged. Given the absence of an external reward, our studies do not reveal how dopamine neuron activation may contribute to associations among specific sensory/identity properties of external rewards and the cues that predict them. Dopamine can contribute to both a general and an outcome-specific form of learning in natural reward conditioning ^33,34, suggesting that the sensory qualities and concomitant brain dynamics of the unconditioned stimulus that produces dopamine neuron activation are likely critical for determining the content of the association learned. Our experiments probed the function of dopamine systems absent outcome-specific information, revealing fundamental roles of dopamine neurons that may well be recruited in situations requiring more complex associations.

A primary finding here is that dopamine neurons in the VTA and SNc exhibited divergent conditioned motivational functions. VTA, but not SNc dopamine neurons conferred a signal that made those cues attractive and reinforcing on their own (Figs. 4 and ​and6),6), two properties demonstrating the incentive value of reward cues, and most likely the explanation for the behavioral patterns observed here. These results build on a large body of research implicating dopamine signaling in cue attraction and conditioned reinforcement ^{15,20,26,27,35,36}, by showing that some dopamine neurons create these properties during Pavlovian conditioning, in the absence of reward receipt or consumption. We found that SNc dopamine neurons, alternatively, conferred a more general movement invigoration signal (Fig. 5); cues paired with their activation evoked vigorous movement not directed at the cue, and they failed to serve as conditioned reinforcers. Thus, distinct components of conditioned reward are represented and controlled by different dopamine output systems. While nigrostriatal dopamine neurons do not appear to instantiate incentive value per se in Pavlovian conditioned stimuli, they do confer their own important motivational properties, however, evidenced here by vigorous cue-evoked movement, perhaps akin to a learned “motor motivation” ³⁷. In the absence of a VTA-mediated incentive component to orient and guide animals to a specific target, direct engagement of SNc-dopamine-mediated learning might manifest as a general increase in locomotion. Notably, dorsal striatal dopamine transmission is not required for the expression of approach to Pavlovian conditioned cues ³⁸, but the dorsal striatum is necessary for the ability of Pavlovian conditioned cues to invigorate ongoing instrumental actions ³⁹, which could be an expression of the conditioned movement invigoration reported here. Thus, in natural learning situations involving external rewards, motivation to engage in specific movements/actions, signaled by SNc dopamine, would be incorporated with motivation to achieve specific rewarding outcomes, signaled via VTA dopamine. Future work will be needed to explore the motivational content conferred by SNc dopamine neurons, how dorsomedial and dorsolateral projecting dopamine neurons may differ ^40–42 and, critically, how VTA and SNc dopamine circuits interact across learning to fine-tune reward seeking.

Here we show that at least some types of movements reflect a conditioned state resulting from an association between dopamine neuron activity and the presentation of external sensory cues: un-cued dopamine neuron activation did not generate locomotion. This provides context for important recent work assessing the role of dopamine neuron activity during self-initiated or spontaneous movements ^10–12. Our findings suggest that dopamine-mediated movements that are not self-generated are gated by the presence of salient sensory stimuli. Notably, movement patterning is dependent on sensory input ⁴³ and conditioning with visual cues can improve movement deficits in Parkinson’s patients ⁴⁴. Thus, external signals are critical for normal expression of movement, and our results suggest that dopamine neurons contribute to this process perhaps by assigning motivational value to cues, allowing them to draw attention and invigorate, shaping locomotion.

Viral Vectors

For optogenetic conditioning experiments, Cre-dependent expression of channelrhodopsin was achieved via injection of AAV5-Ef1α-DIO-ChR2-eYFP (titer 1.5–4e¹² particles/mL, University of North Carolina) into the VTA or SNc. For projection-specific experiments, AAV2/5-Ef1α-DIO-hChR2(H134R)-eYFP-WPRE-hGH (1.5–4e¹² particles/mL, University of Pennsylvania), which exhibits retrograde transport ⁵¹, was injected into the NAc core or dorsal striatum. For combined optogenetic stimulation and photometry experiments, a mixture of AAVDJ-Ef1α-DIO-GCaMP6f (titer 1.0–3.9e¹², Stanford University) and AAV9-hSyn-FLEX-ChrimsonR-tdTomato (1.5–4e¹² particles/mL, University of Pennsylvania) was injected into the VTA.

Surgical Procedures

Viral infusions and optic fiber implants were carried out as previously described ⁵². Rats were anesthetized with 5% isoflurane and placed in a stereotaxic frame, after which anesthesia was maintained at 1–3%. Rats were administered saline, carprofen anesthetic, and cefazolin antibiotic intraperitoneally. The top of the skull was exposed and holes were made for viral infusion needles, optic fiber implants, and 5 skull screws. Viral injections were made using a microsyringe pump at a rate of 0.1μl/min. Injectors were left in place for 5 min, then raised 200 microns dorsal to the injection site, left in place for another 10 min, then removed slowly. Implants were secured to the skull with dental acrylic applied around skull screws and the base of the ferrule(s) containing the optic fiber. At the end of all surgeries, topical anesthetic and antibiotic ointment was applied to the surgical site, rats were removed to a heating pad and monitored until they were ambulatory. Rats were monitored daily for one week following surgery. Optogenetic manipulations commenced at least 4 weeks (6–8 weeks for photometry and projection-specific studies) after surgery.

Midbrain cell body targeting

AAV5-Ef1α-DIO-ChR2-eYFP was infused unilaterally (0.5 to 1 μl at each target site, for a total of 2–4 μl per rat) at the following coordinates from Bregma for targeting VTA cell bodies: posterior −6.2 and −5.4mm, lateral +0.7, ventral −8.4 and −7.4. For targeting SNc dopamine cell bodies: posterior −5.8 and −5.0, lateral +2.4, ventral −8.0 and −7.0. Custom-made optic fiber implants (300-micron glass diameter) were inserted unilaterally just above and between viral injection sites at the following coordinates. VTA: posterior −5.8, lateral +0.7, ventral −7.5. SNc: posterior −5.3, lateral +2.4, ventral −7.3.

Projection-specific ChR2 targeting

The retrogradely-traveling AAV2/5-Ef1α-DIO-hChR2(H134R)-eYFP-WPRE-hGH was infused unilaterally into the NAc core, shell, or dorsal striatum. Two injections of 0.5 μl each (1μl total per rat) were given along the anterior-posterior axis at these coordinates from Bregma. NAc core: anterior +2.2 and +1.6, lateral +1.6, ventral −7.0. NAc shell: anterior +1.8 and +1.2, lateral +0.75, ventral −7.5. Dorsal striatum: anterior +1.8 and +1.0, lateral +2.6, ventral −4.2. Optic fiber implants were inserted above the ipsilateral VTA (for NAc injections) or SNc (for dorsal striatal injections) at the coordinates listed above.

Photometry

A mixture of AAVDJ-Ef1α-DIO-GCaMP6f and AAV9-hSyn-FLEX-ChrimsonR-tdTomato (0.5–1 μl of each, for a total volume of 1–2 μl per rat) was injected into the VTA (posterior −5.8, lateral +0.7, ventral −8.0) or SNc (posterior −5.3, lateral +2.4, ventral −7.4). Low-auto-fluorescence optic fibers (400 micron, Doric) were inserted just dorsal to the injection site at the same coordinates as above.

Optogenetic Stimulation

ChR2 studies utilized 473-nm lasers and ChrimsonR studies utilized 590-nm lasers (OptoEngine), adjusted to read ~10–20mW from the end of the patch cable at constant illumination. Light output during individual 5-ms light pulses during experiments was estimated to be ~2 mW/mm² at the tip of the intracranial fiber. Light power was measured before and after every behavioral session to ensure that all equipment was functioning properly. For all optogenetic studies, optic tethers connecting rats to the rotary joint were sheathed in a lightweight armored jacket to prevent cable breakage and block visible light transmission.

Video Scoring

Behavior during Pavlovian conditioning sessions was video recorded (Media Recorder 4.0, Noldus) using cameras positioned a standardized distance behind each chamber. Videos from sessions 1, 4, 8 and 12 were scored offline by observers who were blind to the identity and anatomical target group of the rats. Each cue (7-sec, 25 per session) and laser (1 or 5-sec, 25 per session) event was scored for the occurrence and onset latency of the following behaviors. Locomotion: Defined as the rat moving all four feet in a forward direction (i.e., not simply lifting feet in place). Cue Approach: Defined as the rat’s nose coming within 1 in of the cue light (trials in which the rat’s nose was in front of the light when it was presented were not counted in the approach measure). Approach often involved the rat moving from another area of the chamber to come in physical contact with the cue light while it was illuminated. Rearing: Defined as the rat lifting its head and front feet off the chamber floor, either onto the side of the chamber, or into the air. Rotation: Defined as the rat making a complete 360-degree turn in one direction.

Automated Motion Tracking and Analysis

We supplemented experimenter scored video analysis with automated behavior tracking, to provide a detailed quantitative assessment of cue-evoked movement patterns. Behavioral videos from the final session of Pavlovian conditioning were analyzed using Noldus Ethovision XT software to automatically track the position of the rats’ heads. The frame by frame location of the head within the video was transformed into a position coordinate within the experimental chamber. These coordinates were used to determine velocity (cm/s) and distance from the cue (cm) during pre-cue, cue, laser, and post-cue periods (Figs. 5 & Supplementary Fig. 8).

Ex vivo electrophysiology

5–6 weeks following virus injection (described above), rats were deeply anesthetized with isoflurane, decapitated, and brains were removed. 200 μM horizontal slices of the midbrain were cut in ice cold aCSF, then maintained at 33°C for current clamp recording as in previous studies ⁵³. ChR2-expressing neurons were identified with epiflourescence on the recording scope (AxioExaminer A1, also equipped with infrared and Dodt optics, Zeiss). ChR2 was activated by transmitting 470-nm light generated by an LED (XR-E XLamp LED; Cree) coupled to a 200 μm fiber optic pointed at the recorded cell and powered by an LED driver (Mightex Systems) and triggered by a Master 8. Cells were filled with biocytin during the recording, and when the recording was complete, the slice was fixed in 4% formaldehyde for 4 hr. Immunocytochemistry was completed as in previous studies ⁵³.

Fiber Photometry

Fiber photometry allows for real time excitation and detection of bulk fluorescence from genetically encoded calcium indicators, through the same optic fiber, in a freely moving animal. We first assessed dopamine neuron activity, via GCaMP6f fluorescence, in response to sucrose consumption, in order to determine the duration of activity during a reward exposure event, which we mimicked with optogenetic conditioning parameters. Rats underwent Pavlovian training wherein an auditory cue was presented on a 45-sec variable time schedule. VTA dopamine neurons in TH-cre+ rats (n=5) were transfected with GCaMP6f and implanted with optic fibers for photometry. Rats first received magazine training during which sucrose was periodically delivered into a reward port. We then conducted photometry recordings during sessions where sucrose was delivered to the port on a 45-sec VT schedule. We observed a rapid increase in fluorescence as animals consumed sucrose, lasting several seconds. These data show that natural reward consumption produces multi-second activation of dopamine neurons, at a comparable magnitude and duration, as measured by calcium fluorescence, to the 5-sec laser stimulation train we employed in optogenetic conditioning studies (Fig. S5).

To assess dopamine neuron activity during optogenetic Pavlovian conditioning, we co-transfected dopamine neurons were with GCaMP6f and ChrimsonR, a red-shifted excitatory opsin ⁵⁴. This approach allowed for simultaneous measurement of activity-dependent fluorescence, excited by low power blue light, and optogenetic activation using orange light, in the same neurons ⁵⁵. The photometry system was constructed similar to previous studies ⁴⁰. A fluorescence mini-cube (Doric Lenses) transmitted light streams from a 465-nm LED sinusoidally modulated at 211 Hz that passed through a GFP excitation filter, and a 405-nm LED modulated at 531 Hz that passed through a 405-nm bandpass filter. LED power was set at ~100 microwatts. The mini-cube also transmitted light from a 590-nm laser, for optogenetic activation of ChrimsonR through the same low-autofluorescence fiber cable (400nm, 0.48 NA), which was connected to the optic fiber implant on the rat. GCaMP6f fluorescence from neurons below the fiber tip in the brain was transmitted via this same cable back to the mini-cube, where it was passed through a GFP emission filter, amplified, and focused onto a high sensitivity photoreceiver (Newport, Model 2151). Demodulation of the brightness produced by the 465-nm excitation, which stimulates calcium-dependent GCaMP6f fluorescence, versus isosbestic 405-nm excitation, which stimulates GCaMP6f in a calcium-independent manner, allowed for correction for bleaching and movement artifacts. A real-time signal processor (RP2.1, Tucker-Davis Technologies) running OpenEx software modulated the output of each LED and recorded photometry signals, which were sampled from the photodetector at 6.1 kHz. The signals generated by the two LEDs were demodulated and decimated to 382 Hz for recording to disk. For analysis, both signals were then downsampled to 40 Hz, and a least-squares linear fit was applied to the 405-nm signal, to align it to the 465-nm signal. This fitted 405-nm signal was used to normalize the 465-nm signal, where ΔF/F = (465-nm signal – fitted 405-nm signal)/(fitted 405-nm signal). Task events (e.g., cue and laser presentations), were time stamped in the photometry data file via a signal from the Med-PC behavioral program, and behavior was video recorded as described above.

Photometry rats (Cue Paired group, n=8) underwent opto-Pavlovian conditioning, similar to that described above, but the intertrial interval for these experiments was halved to 100-sec VT, for a ~40-min session length. This was done to shorten the overall length of photometry measurement periods to minimize photobleaching of GCaMP-expressing cells. Photometry measurements were made on training sessions 1, 4, 8, and 12, during which both LED channels were modulated continuously, as described above. On these 4 sessions, 20% of trials (5/25), pseudo-randomly presented, were “probes”, where cues were presented without accompanying optogenetic stimulation. Unpaired rats (n=5) received 12 sessions of cue and laser presentations (25 each) separated by a 70-sec VT.

Following conditioning, a subset of paired animals (n=5) received one session of extinction training, during which cues were presented as before, but laser was omitted, while photometry measurements were made.

For baseline characterization of ChrimsonR-activated GCaMP6f signals, rats (n=5) were tethered to the photometry apparatus, and continuous photometry measurements were made during a series of 60 unsignalled 590-nm laser presentations (30 trials of 1-sec, 20 Hz stimulation trains, 30 trials of 5-sec, 20 Hz trains, counterbalanced), delivered on a 30-sec VT schedule.

ChrimsonR ICSS

Th-cre+ rats (n=7) were given the opportunity to respond for 590-nm laser pulses (1 s, 20 Hz), in 2 1-hr sessions, similar to above, to validate ChrimsonR support of dopamine-mediated primary reinforcement. On a third session, the laser was switched from orange to blue (473-nm), to verify that ChrimsonR activation necessary to support behavior is specific to red-shifted light.

Statistics, Data Collection, and Analysis

Rats were randomly assigned to conditioning groups (paired, unpaired) following surgery. Behavioral data from optogenetic conditioning experiments was automatically recorded with Med-PC software (Med Associates) and analyzed using Prism 6.0. Video of conditioning sessions was recorded using Noldus Media Recorder 4.0, and automated behavior tracking data was generated using Noldus Ethovision XT and analyzed in MATLAB. For manual video scoring, experimenters were blind to the anatomical and conditioning group identify. Experimenters were otherwise not blinded. Non-parametric tests were used when data distributions were non-normal. Two-way repeated measures ANOVA was used to analyze changes in behavior among the groups across training. Bonferroni-corrected post hoc comparisons were made to compare groups on individual sessions. No statistical tests were used to predetermine sample sizes, but our sample sizes were similar to previously published studies. Rats were included in optogenetic behavioral analyses if optic fiber tips were no more than ~500 microns dorsal to the target region (VTA or SNc). Photometry data were collected with TDT OpenEx and Synapse software and analyzed using MATLAB. To assess the change in fluorescence across training days we fit a linear mixed-effect model for ΔF/F during each period of interest (0–1 s post-cue and laser omission period), with fixed effects for day and random effects for subject. To assess the relationship between the magnitude of cue-evoked fluorescence and CR latency, we fit a linear mixed-effect model for latency with fixed effects for cue-evoked fluorescence magnitude and random effects for subject. All comparisons were two tailed. Data in figures are expressed as the mean ± SEM. Statistical significance was set at p<0.05. Please see the Life Sciences Reporting Summary for additional information.

We thank R. Keiflin and all members of the Janak laboratory for discussion and comments on the manuscript; A. Haimbaugh, D. Acs, H. Pribut, K. Lineback, N. Pettas, B. Persaud, and L. Kinny for assistance with histology and behavioral video scoring; P. Fong for conducting surgical procedures for ex vivo physiology studies; K. Deisseroth (Stanford) for the ChR2 construct; E. Boyden (MIT) for the ChrimsonR construct; and the Janelia Research Campus GENIE Project and Stanford Gene Vector and Virus Core for the GCaMP6f construct. This work was supported by National Institutes of Health grants DA036996 (BTS), DA042895 (BTS), AA022290 (JMR), AA025384 (JMR), DA030529 (EBM), and DA035943 (PHJ), as well as grants from the Brain and Behavior Research Foundation (BTS and JMR).