Escrever um comentário

Review of “The encoding of interoceptive-based predictions by the paraventricular nucleus of the thalamus D2R+ neurons”

Briana Machen, Sierra N. Miller, Al Xin, Carine Lampert, Lauren Assaf, Julia Tucker, Sarah Herrell, Francisco Pereira, Gabriel Loewinger, Sofia Beas

This paper explores the role of Dopamine 2 Receptor (D2R+) expressing neurons in the paraventricular nucleus of the thalamus (PVT) in a conditioned approach task. The authors propose that PVT may be a candidate ‘integrator’ of interoceptive signals to give rise to internal states. As the authors mention, interoceptive integration has been less explored than the ‘sensing’ and ‘tracking’ aspects of interoceptive processing, making this question of high importance and ripe for investigation. Interestingly, the authors show that this population of neurons is active when hungry mice receive a caloric reward as when thirsty mice receive a water reward, providing evidence for PVT’s candidacy as an interoceptive integrator. Further, previous work from the senior author demonstrated that this population of neurons was important for the execution of goal-oriented actions using a linear maze task (Beas et al., 2024, PMID: 38458192), establishing a clear rationale for the experiments conducted in this manuscript. The strengths of this study include the use of variations of an established behavioral paradigm to explore the role of these D2R+ PVT neurons in motivated reward seeking. In accordance with their previous work, the authors demonstrate that this population of neurons are more active when mice are hungry (early in the session) compared to sated (late in the session) and that the neural responses are associated with response latency and reward value. Furthermore, they use chemogenetics to test if the inhibition of these neurons during task acquisition impairs performance on the linear maze task. This manuscript asks an important and timely question about how PVT D2R+ neurons integrate interoceptive state information to shape motivated behavior. The rationale is strong, and the within-subject design is a clear asset. However, key claims are currently undermined by analytical choices that do not fully control for trial number differences, potential motor confounds, and order effects in behavioral design. Addressing these concerns and clarifying several methodological details would substantially strengthen the scientific rigor of this work and more strongly support the authors’ central conclusions.  Below are a few suggestions and clarifying questions that would improve the scientific rigor of this work:

Major Concerns

1. The authors try to address varying trial numbers from different experimental conditions by using repeated measures ANOVA and post-hoc tests to test for any statistical differences in trial numbers between groups (ex: Fig. 1Q,V, Fig. 2N). However, no statistical difference does not imply equivalence between groups. Throughout the manuscript, equivalence is inferred from non-significant hypothesis testing. To support these claims, the authors can consider performing equivalence testing (ex: TOST procedure), randomly subsampling data to trial match, or bootstrapping could be used to trial match across groups that have different trial numbers to improve the rigor of this comparison and support the claims made throughout the manuscript.

2. The authors demonstrate lower responsiveness of these neurons in sated-SMS (Fig. 1D), hungry-H2O (Fig. 2B), sated-OM (Fig. 3B), training session 1 (Fig. 7B) conditions. For all of the effects, these responses scale with high approach latency. In other words, the slowest mice are the ones that show lower neural activity, and the fastest mice show the highest neural activity. Therefore, this raises the concern that these effects could be due simply to the motor response of the mouse during reward approach. Although the authors begin to address this by plotting the correlation between the approach AUC and the approach latency for one of these experiments (Fig. 1K), similar analyses for each experiment that shows a difference in response latencies between groups would strengthen the rigor of the analysis. To further strengthen the rigor of this analysis, the authors could consider instead plotting 1) the correlation between velocity and z-scored neural activity 2) a heatmap of reward approach responses sorted by velocity to exclude the possibility of a motor confound. 

3. As currently designed, the behavioral experiments do not rule out the possibility that the observed effects arise from the fixed order in which need states are experienced. For example, hungry-SMS is always conducted the day before sated-SMS, followed by hungry-NCal (Fig. 1C). To enhance experimental rigor and more confidently attribute the effects to physiological state rather than task order, the authors should consider counterbalancing the sequence in which mice experience these conditions for all behavioral experiments. At the same time, a notable strength of the current design (one that would be important to maintain) is the use of within-animal comparisons, which effectively controls for both surgical variability and individual differences across subjects. Similarly, the rigor of the chemogenetic manipulations (Fig. 8) could be strengthened by counterbalancing the order in which animals experience saline vs. DCZ/CNO.

Minor Concerns

The following points are offered as clarifications that may help improve the clarity and interpretability of the manuscript:

1. Abstract wording and chemogenetic interpretation: In the abstract, the authors state that chronic inhibition of PVT D2R+ neurons impairs long-term task performance. However, the inhibition used in these experiments (Fig. 8) appears to be chemogenetic and acute (minutes to hours) during early learning, rather than chronic. In addition, CNO and DCZ are treated as interchangeable agonists. Although the doses were adjusted to account for potency differences, these compounds have distinct pharmacokinetics, and using both may introduce mixed or confounding effects. The methods would be more rigorous if a single agonist were used consistently throughout.

2. Interpretation of response latency in G1–G5 groups: Early in the manuscript, the authors report no difference in average approach latency across trial groups (G1–G5) (Supp. Fig. 1D). Because response latency is later discussed as a proxy for motivational strength, these data show that hungry mice (G1) do not show faster approach times than less hungry mice (G5). Clarifying this point would help the reader understand how the authors are operationalizing “motivation” within the framework of their behavioral assay.

3. Reward habituation procedures: In the methods for the linear maze task, the authors note that mice were given 8 hours of SMS exposure in the home cage for two days prior to training to avoid neophobia. It is unclear whether similar habituation procedures were performed for the non-caloric or sucrose rewards. This information would help readers interpret potential differences in reward approach behavior.

4. Figure organization: In the current presentation, Supplementary Figure 1 is discussed before the corresponding main figure, which makes the results more difficult to follow. Reordering or restructuring this section could improve clarity.

5. Potential excitatory DREADD complement: The authors might consider an optional complementary experiment using excitatory DREADDs to test whether chronic excitation of PVT D2R+ neurons enhances performance in the linear maze. Such an experiment would nicely parallel and extend the findings associated with inhibition in Fig. 8.

Competing interests

The authors declare that they have no competing interests.

Use of Artificial Intelligence (AI)

The authors declare that they used generative AI to come up with new ideas for their review.