Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
[Preprint]. 2023 Aug 21:2023.08.17.553745.
doi: 10.1101/2023.08.17.553745.

Dorsal raphe neurons signal integrated value during multi-attribute decision-making

Yang-Yang Feng  1   2 , Ethan S Bromberg-Martin  1 , Ilya E Monosov  1   2   3   4   5
Affiliations

Dorsal raphe neurons signal integrated value during multi-attribute decision-making

Yang-Yang Feng et al. bioRxiv. .

Update in

Abstract

The dorsal raphe nucleus (DRN) is implicated in psychiatric disorders that feature impaired sensitivity to reward amount, impulsivity when facing reward delays, and risk-seeking when grappling with reward uncertainty. However, whether and how DRN neurons signal reward amount, reward delay, and reward uncertainty during multi-attribute value-based decision-making, where subjects consider all these attributes to make a choice, is unclear. We recorded DRN neurons as monkeys chose between offers whose attributes, namely expected reward amount, reward delay, and reward uncertainty, varied independently. Many DRN neurons signaled offer attributes. Remarkably, these neurons commonly integrated offer attributes in a manner that reflected monkeys' overall preferences for amount, delay, and uncertainty. After decision-making, in response to post-decision feedback, these same neurons signaled signed reward prediction errors, suggesting a broader role in tracking value across task epochs and behavioral contexts. Our data illustrate how DRN participates in integrated value computations, guiding theories of DRN in decision-making and psychiatric disease.

PubMed Disclaimer

Figures

Figure 1.
Figure 1.. Monkeys exhibit preferences for expected reward, delay, and uncertainty during multi-attribute decision-making.
(A) Two offers (Offer 1 and Offer 2) are presented sequentially and then the monkey has 5 seconds to free view and make a choice by fixating one offer for 0.5 seconds. After a delay, the outcome is revealed and the reward is delivered. (B) Left: Vertical bars represent the probability distribution over rewards, indicating the expected reward amount (E[R]) and reward uncertainty (Unc[R]). Right: Horizontal bar acts as a clock, representing reward delay (T[R]). To allow monkeys to prepare to consume the upcoming reward, when the reward delay has elapsed, the stimulus changes to reveal the outcome, and then the reward is delivered 1.1 s later (Methods). The total time between choice and end of the trial was always 8s. (C) Fitted weights describing each monkey’s subjective preferences for E[R], T[R], and Unc[R]. Error bars denote s.e.m. Both monkeys placed significant weights on each reward attribute (** p<0.01, *** p<0.001).
Figure 2.
Figure 2.. Recording single neurons in primate dorsal raphe nucleus.
(A) MRI taken with electrode in recording location, verifying its position in DRN. Shown are coronal views slightly tilted to align with the electrode track. The electrode is visible as a black ‘shadow’ on the MRI. (B) Nissl staining showing electrode track and marking lesions (Red arrows), delineating approximate range of recordings. The trochlear nucleus is visible just ventral to the recording track. (C) Anatomical reconstruction of recorded neurons, based on previously published methods. Empty circles represent recorded neurons that were not task-responsive. Filled circles are task-responsive neurons. Top: Neurons from +3 to +4.5 interaural are shown on a plane corresponding to approximately +3.5 interaural. Bottom: Neurons from +2 to +2.5 interaural are shown on a plane corresponding to approximately +2.5 interaural. (D) SERT staining of tissue slices cut along the transverse axis of the brainstem, approximately corresponding to reconstructions shown in (C). Aq, cerebral aqueduct; MLF, medial longitudinal fasciculus; PAG, periaqueductal gray.
Figure 3.
Figure 3.. DRN neurons signal expected reward, delay, and uncertainty during multi-attribute decision-making.
(A) Rasters and SDFs showing example neurons’ responses to Offer 1 on trials where it had Large vs Small E[R] (left, red), Short vs Long T[R] (center, blue), or Uncertain vs Certain Unc[R] (right, green). (B) Mean cross-validated normalized activity for task-responsive DRN neurons on trials where Offer 1 E[R] (left, red), T[R] (center, blue), or Unc[R] (right, green) was preferred or non-preferred by the neuron. (C) Difference in the activity shown in B. Error bars denote s.e.m. Average discrimination during the offer response analysis window (150–575 ms after Offer 1 presentation, black bar) was significant for E[R], T[R], and Unc[R] signals (two-tailed signed-rank test, *** p<0.001). (D) Proportion of task-responsive DRN neurons with significant fitted weights for each attribute, fitting normalized activity in response to Offer 1 (one-tailed binomial test, ** p<0.01, *** p<0.001). Error bars indicate 68% confidence interval (Clopper–Pearson method).
Figure 4.
Figure 4.. DRN neurons positively signal the value of expected reward, delay, and uncertainty.
Histograms showing E[R] (left, red), T[R] (center, blue), Unc[R] (right, green) value indices (see Methods) of task-responsive neurons. Signals are biased in the direction of monkey preferences (dashed lines show mean; median significantly different from 0 by signed-rank test). The proportions of neurons with significant signaling of each attribute in the direction of monkey preferences are significant, while the proportions signaling the opposite are not (two-tailed binomial test). One outlier with an extreme E[R] value index was omitted from the left plot for presentation purposes (E[R] value index = 1.0, sig.). Colors indicate that neither (black), the x-coordinate (red), the y-coordinate (blue), or both (magenta) are significant (p < 0.05).
Figure 5.
Figure 5.. DRN expected reward signals integrate the value of delay or uncertainty.
(A) Scatter plots showing E[R] vs T[R] (left) or E[R] vs Unc[R] (right) value signals across task-responsive neurons. E[R] signals are strongly positively correlated with T[R] and Unc[R] (Pearson and Spearman correlations, two-tailed permutation tests). Gray lines show Type 2 regression. One outlier with an extreme E[R] value index was omitted from the scatters for presentation purposes (E[R] value index = 1.0, sig.; T[R] value index = 0.14, n.s.; Unc[R] value index = 0.30, sig.). Colors indicate that neither (black), the x-coordinate (red), the y-coordinate (blue), or both (magenta) are significant (p < 0.05). (B) Top: Average normalized activity for E[R]-signaling neurons on trials where Offer 1 T[R] (left) or Unc[R] (right) was valued in the direction of the neuron’s E[R] value signal (Higher) or the opposite (Lower) (see Methods). Bottom: Difference in the activity shown in B. Error bars denote s.e.m. Average discrimination during the offer response analysis window (150–575 ms after Offer 1 presentation, black bar) was significant (two-tailed signed-rank test, ** p<0.001, *** p<0.001).
Figure 6.
Figure 6.. DRN expected reward signals integrate the value of information to reduce uncertainty
The task also measured the value of information to reduce uncertainty about upcoming rewards. Trials could either be informative (Info, orange), for which reward uncertainty would be resolved soon after choice, or non-informative (Noinfo, purple), for which reward uncertainty would not be resolved until immediately before reward delivery (see Methods). One outlier with an extreme E[R] value index was omitted from the scatter for presentation purposes (E[R] value index = 1.0, sig.; InfoxUnc[R] value index = 0.10, n.s.). (B) Bar plot showing fitted InfoxUnc[R] interaction weight (see Methods). Error bars denote s.e.m. Monkeys strongly scaled the value of information with uncertainty. (C) Same as Figure 5A, but for E[R] vs InfoxUnc[R] value signals. E[R] and InfoxUnc[R] are strongly positively correlated (Pearson and Spearman correlations, two-tailed permutation tests). Gray line shows Type 2 regression. Colors indicate that neither (black), the x-coordinate (red), the y-coordinate (blue), or both (magenta) are significant (p < 0.05). (D) Same as Figure 5B but showing differences in normalized activity between Info and Noinfo, splitting trials based on Offer 1 Unc[R].
Figure 7.
Figure 7.. DRN neurons that signal expected reward during decision-making signal reward prediction errors after decision-making.
(A) Rasters and SDFs showing an example neuron’s responses to Info Cue (left, orange) or Noinfo Reveal (right, purple) on trials where it elicited an RPE that was positive (RPE +, solid, dark), 0 (RPE 0, solid, light) or negative (RPE −, dashed, dark). (B) Same as A but showing average normalized firing across E[R]-signaling neurons. Info and Noinfo RPE indices (see Methods) were significant (two-tailed signed-rank test, *** p<0.001). Activity was analyzed in the analysis windows shown in the black bars (250 to 750 ms post-cue, Info; 250 to 1000 ms post-reveal, Noinfo). (C) Histogram showing RPE indices (see Methods) of task-responsive neurons. Signals are predominantly positive (dashed lines show mean; median significantly different from 0 by signed-rank test). The proportion of neurons with significant positive RPE indices is large and strongly significant, while the opposite is small and weakly significant (two-tailed binomial tests). (D) Scatter plot showing RPE indices vs E[R] value signals. RPE indices and E[R] value signals were strongly positively correlated (Pearson and Spearman correlations, two-tailed permutation tests). Gray line shows Type 2 regression. One outlier with an extreme E[R] value index was omitted from the scatter for presentation purposes (E[R] value index = 1.0, sig.; RPE index = 0.84, sig.) Colors indicate that neither (black), the x-coordinate (red), the y-coordinate (blue), or both (magenta) are significant (p < 0.05).
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. Ishimura K., Takeuchi Y., Fujiwara K., Tominaga M., Yoshioka H., and Sawada T. (1988).Quantitative analysis of the distribution of serotonin-immunoreactive cell bodies in the mouse brain. Neuroscience Letters 91, 265–270. 10.1016/0304-3940(88)90691-X. - DOI - PubMed
    1. Daw N.D., Kakade S., and Dayan P. (2002). Opponent interactions between serotonin and dopamine.Neural Networks 15, 603–616. 10.1016/S0893-6080(02)00052-7. - DOI - PubMed
    1. Dayan P., and Huys Q.J.M. (2009). Serotonin in Affective Control. Annual Review of Neuroscience 32,95–126. 10.1146/annurev.neuro.051508.135607. - DOI - PubMed
    1. Cools R., Nakamura K., and Daw N.D. (2011). Serotonin and Dopamine: Unifying Affective, Activational, and Decision Functions. Neuropsychopharmacology 36, 98–113. 10.1038/npp.2010.121. - DOI - PMC - PubMed
    1. Rogers R.D. (2011). The Roles of Dopamine and Serotonin in Decision Making: Evidence from Pharmacological Experiments in Humans. Neuropsychopharmacology 36, 114–132.10.1038/npp.2010.165. - DOI - PMC - PubMed

Publication types

-