PET imaging-guided chemogenetic silencing reveals a critical role of primate rostromedial caudate in reward evaluation

The rostromedial caudate (rmCD) of primates is thought to contribute to reward value processing, but a causal relationship has not been established. Here we use an inhibitory DREADD (Designer Receptor Exclusively Activated by Designer Drug) to repeatedly and non-invasively inactivate rmCD of macaque monkeys. We inject an adeno-associated viral vector expressing the inhibitory DREADD, hM4Di, into the rmCD bilaterally. To visualize DREADD expression in vivo, we develop a non-invasive imaging method using positron emission tomography (PET). PET imaging provides information critical for successful chemogenetic silencing during experiments, in this case the location and level of hM4Di expression, and the relationship between agonist dose and hM4Di receptor occupancy. Here we demonstrate that inactivating bilateral rmCD through activation of hM4Di produces a significant and reproducible loss of sensitivity to reward value in monkeys. Thus, the rmCD is involved in making normal judgments about the value of reward.

technique. However, the Abstract and Introduction led me to assume initially that the main point of the manuscript is about the causal roles of rmCD. There was already a previous study demonstrating a causal relationship between caudate activity and reward-modulated behavior, using unilateral injection of dopamine antagonist ( Nakamura and Hikosaka 2006 J Neurosci). I was also not sure why a simple muscimol injection experiment, just like what the authors described later in the manuscript, would not be sufficient. The authors addressed the latter question in Discussion. But these concerns unnecessarily detract from the main achievement of the study. I would thus suggest a re-structuring of the manuscript to highlight the motivation for developing the new techniques early on.
-Pg. 6, near the bottom, "For all three monkeys, the reaction times were not affected by CNO administration (two-way ANOVA, main effect of treatment, F1, 134 = 0.12, p = 0.72). The total reward earned was not affected by CNO (ANOVA, main effect of treatment, F1, 32 = 3.1, p = 0.09)." I am puzzled by these results. Since error trials are defined as too-fast or too-slow responses, presumably increased error rates must reflect changes in the reaction time distribution? For the second finding, given the higher error rates, the only way total reward earned did not change is if the monkeys performed more trials. Could silencing rmCD increase motivational drive?
-Fig 2d can be improved for clarity. For example, as is, it appears that error rates did not change with reward size for the vehicle, 0-20 min condition. Given that there are only 5 timing conditions, it would be better in my opinion to just plot 5 panels of error rate vs. reward size curves (similar to Fig 2b) for CNO/vehicle pairs, side-by-side for different timing conditions, instead of the surface plot.
-Pg. 8 line 6 from bottom, "Unilateral inactivation of the rmCD did not change the overall error rates (main effect of treatment, p = 0.065; Supplementary Fig. 4d)". This statement is not well supported, given that the data came from only one monkey and there is a clear trend. -Pg 6. "The CNO treatment increased the overall error rates in these monkeys (main effect of treatment, p < 0.01, Fig. 3c, d)." Were there occupancy data from these two monkeys? If so, they should be included.
-Pg. 7, line 5-6. "Explain" implies mechanistic knowledge. Perhaps it is more appropriate to state that the equation "describes" the observed effect.

Reviewer #3 (Remarks to the Author):
The manuscript examines the use of PET-imaging guided chemogenetics to make perturbations of caudate nucleus neurons in the monkey. This is a sensible approach that appears to work well, and I support publication of the paper. One additional piece of information would be useful, which is examination of potential tachyphylaxis is both the behavioral perturbation and receptor availability with successive doses of CNO.

Reviewer #4 (Remarks to the Author):
In the present manuscript, Nagai and collaborators did investigate the impact of chemogenetic silencing of striatal neurons (by DREADD approach) on reward evaluation in non-human primates. After in vitro characterization of their viral construct (hM4Di, the modified muscarinic receptor), the authors did bilaterally inject it within the rostro-medial part of the primate caudate nucleus and did control longitudinally its expression level by PET imaging using the radioligand 11C-Clozapine. They have then looked for the dose of clozapine-N-oxide (CNO) to use, as clozapine-N-oxide binds to and activates hM4Di (and therefore to suppress neuronal activity of neurons expressing hM4Di). Once the right conditions were found, the authors did test the impact of such inactivation on reward value evaluation by using a reward-size task, which they have used extensively in the past. They show that inactivating the rostro-medial part of the caudate nucleus by systemic administration of the drug inducer leads to a specific increase of error rates (without any change in reaction time or total reward earned; but these data are not shown). The task's performance returned to baseline levels the day after, showing the reversibility of CNO treatment. In the last part of the manuscript, the authors did compare their DREADD results with bilateral injections of muscimol (a GABAa agonist) within the same region of the caudate nucleus. Muscimol also led to an increase of error rates and loss of sensitivity to reward value, as expected.
This study is well designed, written, novel and complete. The authors did provide several control experiments (from in vitro to in vivo), which strongly reinforce the impact of their results. Besides this methodological point of view, the study doesn't provide new advances on the function of the caudate nucleus but the study is still very important to the research community as it combines for the first time PET imaging with DREADDs approach in the monkey and demonstrates the critical role of the caudate in reward value processing.
I have a few remaining comments: The authors have been using three different monkey species (please specify male or female) (cynomolgus, rhesus and japanese) and different viral constructs (AAV2-CMV-hM4Di; AAV2-CMV-HA-hM4Di; Lenti-hSyn-hM4Di; Lenti-hSyn-CFP) for the experiments. I would suggest to add a table summarizing this information (monkey, viral construct, volume and titer, type of injection, type of experiment, etc). The relevant experiments are often performed on rhesus or japenese monkeys with AAV2-CMV-hM4Di, and the control experiments are performed on cynomolgus monkeys with lenti-hM4Di-CFP. The authors have also been using AAV-HA-hm4Di. Why? Lentivirus and adeno-associated virus are supposed to have different transfection properties, isn't it? Lentiviral injections were further performed within the putamen and not the caudate, the region of interest in the paper. Please explain why. Could you also explain why the muscimol experiments were not performed on the monkeys with AAV-CMV-hM4Di? The discussion is quite methodological.
Minor comments - Figure 1 (and methods): How is made the intravenous CNO injection? For PET imaging, the monkey is under anesthesia, but the monkey is awake during the task.

Reviewer #1 (Remarks to the Author):
This is a potentially interesting paper which demonstrates the potential utility of combining PET imaging of chemogenetic GPCRs with behavior in nonhuman primates. It is likely to be of interest to a wide variety of neuroscientists--in particular those who use primates--and potentially has translational value as well. The authors essentially show that hM4Di reliably apparently induces silencing of a similar valence and magnitude as muscimol and suggest advantages for a chemogenetic approach rather than muscimol injections. They also show that hM4Di can be apparently visualized via 11C-CLZ PET imaging which makes this potentially highly useful from both a basic science and logiistical perspective.
R: We appreciate the reviewer's favorable opinion.

The authors mention potential issues with CNO in primates and it would be useful to mention that other CNO analogues have been described in the literature and mention their potential advantages/disadvantages over CNO.
R: We thank the reviewer for suggesting that we mention CNO analogues. We have added the following discussion regarding potential issues with CNO metabolites and mentioning using CNO analogues. "[ 11 C]CLZ-PET also provides a way of measuring in vivo DREADD receptor occupancy by CNO. Based on the dose-occupancy curve (cf. Fig. 2d), we chose a dose of 3 mg kg -1 CNO, yielding approximately 60% receptor occupancy, to carry out our behavioral studies. This dose of CNO produced a repeatable behavioral modulation in reward sensitivity. At this dose there was no CLZ in cerebrospinal fluid (CSF). We did not see signs of other behavioral effects of CNO such as sedation or loss of motivation as were seen at a higher dose (10 mg kg -1 ) in our previous study 18 . This latter observation is relevant because there is some evidence that in humans and guinea pigs, at least, CNO can be metabolized to CLZ, an atypical antipsychotic drug with sedative effects 22, 23 . Because of the concern over side effects, other DREADD agonists such as compound 21 or Perlapine 24 should be investigated for their potential utility in primates, and perhaps in the long run in humans." (pg. 12, lines 6-17) SPECIFIC CONCERNS: 1. If the authors have validated chemogenetic neuronal silencing via slice ePhys than those data should be in the main text and a main figure. If the authors have not done this essential control they would need to do it prior to considering this paper for publication. R: Chemogenetic neuronal silencing was validated via an in vitro electrophysiological assay using cultured neurons expressing hM4Di after being infected using our AAV construct ( Supplementary Fig. 1). We believe that this substitutes for what the reviewer requested.
We refer to the result in the main text; "We verified that activation of the hM4Di receptor with CNO caused neuronal silencing in vitro ( Supplementary Fig. 1)." (pg 5, line 9). We appreciate the suggestion that the result be in the main figure. Because this replicates a finding from our previous study, with the difference being the viral vector (lentivirus vs. AAV), we believe that it is appropriate for inclusion as a Supplementary Figure. Although a limited number of studies report electrophysiological examination in monkey brain slice (e.g., Alexander et al., Neuroscience, 2006), it is notoriously difficult to prepare such material. Furthermore, this is not a standard preparation. It would be difficult on ethical grounds to defend sacrificing our rhesus monkeys to carry out such experiments.
We believe that we have presented a convincing chain of evidence that we achieved neuronal silencing in our behaving monkeys: 1) Chemogenetic neuronal silencing was validated via an in vitro electrophysiological assay using cultured neurons expressing hM4Di ( Supplementary Fig. 1). 2) Using the same AAV vector, hM4Di was expressed in bilateral rmCD in three monkeys (cf. Fig. 2a). 3) CNO dose of 3 mg/kg specifically blocked [ 11 C]CLZ binding to the striatal hM4Di and yielded about 60% occupancy without conversion to CLZ (Fig. 2 and Fig S3). 4) CNO administration specifically altered reward evaluation in 3 monkeys expressing hM4Di in bilateral rmCD ( Fig. 3 and 4). Given this chain of evidence, we feel strongly that recording from brain slices is not an essential control experiment for our study or future behavioral studies using DREADDs in macaque monkeys.

In this study, Nagai et al. pioneered a DREADD-based method to examine rostro-medial caudate's roles in reward value processing in monkeys. This is to my knowledge the first demonstration of the usefulness of DREADD techniques in studying the neural mechanisms of cognition in non-human primates. The application of PET imaging to measure the efficacy of DREADD expression in vivo demonstrates additional, significant advantages over traditional pharmacological/lesioning techniques. The techniques are novel, solid with rigorous controls and have the potential to open up new directions of research in the field. Below I describe some suggestions/comments that are intended to improve the clarity of the paper.
R: We appreciate the reviewer's favorable opinion. Major: -It seems to me that the manuscript's foremost contribution is the potentially powerful new techniques. The effects with rmCD serve as an indicator of the effectiveness of the new technique. However, the Abstract and Introduction led me to assume initially that the main point of the manuscript is about the causal roles of rmCD. There was already a previous study demonstrating a causal relationship between caudate activity and reward-modulated behavior, using unilateral injection of dopamine antagonist ( Nakamura and Hikosaka 2006 J Neurosci). I was also not sure why a simple muscimol injection experiment, just like what the authors described later in the manuscript, would not be sufficient. The authors addressed the latter question in Discussion. But these concerns unnecessarily detract from the main achievement of the study. I would thus suggest a re-structuring of the manuscript to highlight the motivation for developing the new techniques early on. R: We are sympathetic to the reviewer's point of view. However, a technique becomes most valuable when it can be applied to bring new insights, here about caudate function. The reviewer's comments overlook the level of current understanding, by which not all sites in the caudate are equal. We believe that there has been no previous work, certainly none that addresses the intimate connection between causality and neuronal activation, that shows that the rostromedial caudate is essential for normal reward evaluation. The Nakamura-Hikosaka study was largely centered more caudally and more laterally, and as was clear in their manuscript, they had trouble identifying a connection between neuronal activity and the likelihood of stimulation at the site to have an effect. Our result with the one monkey where DREADD activation failed makes it likely that a specific locus in the rostromedial caudate is where the inactivation is effective. Without PET, though, we would have had a difficult time reaching these conclusions.
-Pg. 6, near the bottom, "For all three monkeys, the reaction times were not affected by CNO administration (two-way ANOVA, main effect of treatment, F1, 134 = 0.12, p = 0.72). The total reward earned was not affected by CNO (ANOVA, main effect of treatment, F1, 32 = 3.1, p = 0.09)." I am puzzled by these results. Since error trials are defined as too-fast or too-slow responses, presumably increased error rates must reflect changes in the reaction time distribution? For the second finding, given the higher error rates, the only way total reward earned did not change is if the monkeys performed more trials. Could silencing rmCD increase motivational drive? R: Reaction times were measured for correct responses only (text has been added to the 'Results' section to make this clear; pg. 7, line 7). We might expect to see a slowing of reaction times if attention or motivation was influenced by rmCD silencing, but the consistency of the reaction time distribution in the face of increased errors is what leads us to conclude that rmCD silencing caused a specific loss of sensitivity to the relative reward value.
Monkeys generally reached satiety within the 100 min testing limit (text modified to include statement to this effect; pg. 7, line 8)thus the lack of effect of CNO on total reward earned indicates that the internal threshold for effort and motivational drive have not been affected.  (Fig. 3d).
-Pg. 8 line 6 from bottom, "Unilateral inactivation of the rmCD did not change the overall error rates (main effect of treatment, p = 0.065; Supplementary Fig. 4d)". This statement is not well supported, given that the data came from only one monkey and there is a clear trend.
R: We rewrote the sentence as follows. "Unilateral inactivation of the rmCD produced a trend toward increased overall error rates (main effect of treatment, p = 0.065; Supplementary Fig. 4d), but did not change the reward-size sensitivity (Fig. 5e, f, included in NC),... " (pg. 9, line 9).

Minor: -Are the results in supplementary Fig1 from caudate neurons?
R: The data were obtained from primary cultured neurons derived from whole brain. This is now clarified in the legend.  - Supplementary Fig. 3: the best-fit Hill coefficient should be given.
R: According to the reviewer's suggestion, we used a Hill function with variable coefficient for approximation. We provide the best-fit coefficient, n = 0.64 in Figure 2d, which we have moved from Supplementary Fig. 3 according to a comment by Reviewer #4.
-Pg 6. "The CNO treatment increased the overall error rates in these monkeys (main effect of treatment, p < 0.01, Fig. 3c, d)." Were there occupancy data from these two monkeys? If so, they should be included.
R: The occupancy data are available only for monkeys #171 and #157.
-Pg. 7, line 5-6. "Explain" implies mechanistic knowledge. Perhaps it is more appropriate to state that the equation "describes" the observed effect.
R: We have corrected this (pg. 7, last line). Thank you.

Reviewer #3 (Remarks to the Author):
The manuscript examines the use of PET-imaging guided chemogenetics to make perturbations of caudate nucleus neurons in the monkey. This is a sensible approach that appears to work well, and I support publication of the paper. R: We appreciate the reviewer's favorable opinion.
One additional piece of information would be useful, which is examination of potential tachyphylaxis is both the behavioral perturbation and receptor availability with successive doses of CNO.
R: We agree with the reviewer's comment on the importance of examination of potential tachyphylaxis. It would be important information in successive chemogenetic silencing, which is a very powerful tool to examine specific functions such as learning, in which longitudinal PET monitoring is also valuable. We admire and are very grateful for the reviewer's deep thoughts for trying to increase the impact of our study. This study focused on basic technical advantage of DREADD/CNO such as temporal and repetitive silencing. Successful silencing is somewhat beyond the current status of DREADD technology in monkeys. Therefore, we would like to examine potential tachyphylaxis in future study.

Reviewer #4 (Remarks to the Author):
In the present manuscript, Nagai and collaborators did investigate the impact of chemogenetic silencing of striatal neurons (by DREADD approach) on reward evaluation in non-human primates.
After in vitro characterization of their viral construct (hM4Di, the modified muscarinic receptor), the authors did bilaterally inject it within the rostro-medial part of the primate caudate nucleus and did control longitudinally its expression level by PET imaging using the radioligand 11C-Clozapine. They have then looked for the dose of clozapine-N-oxide (CNO) to use, as clozapine-N-oxide binds to and activates hM4Di (and therefore to suppress neuronal activity of neurons expressing hM4Di). Once the right conditions were found, the authors did test the impact of such inactivation on reward value evaluation by using a reward-size task, which they have used extensively in the past. They show that inactivating the rostro-medial part of the caudate nucleus by systemic administration of the drug inducer leads to a specific increase of error rates (without any change in reaction time or total reward earned; but these data are not shown). The task's performance returned to baseline levels the day after, showing the reversibility of CNO treatment. In the last part of the manuscript, the authors did compare their DREADD results with bilateral injections of muscimol (a GABAa agonist) within the same region of the caudate nucleus. Muscimol also led to an increase of error rates and loss of sensitivity to reward value, as expected.
This study is well designed, written, novel and complete. The authors did provide several control experiments (from in vitro to in vivo), which strongly reinforce the impact of their results. Besides this methodological point of view, the study doesn't provide new advances on the function of the caudate nucleus but the study is still very important to the research community as it combines for the first time PET imaging with DREADDs approach in the monkey and demonstrates the critical role of the caudate in reward value processing.
R: We appreciate the reviewer's favorable opinion.
I have a few remaining comments: