Deciphering an AgRP-serotoninergic neural circuit in distinct control of energy metabolism from feeding

Contrasting to the established role of the hypothalamic agouti-related protein (AgRP) neurons in feeding regulation, the neural circuit and signaling mechanisms by which they control energy expenditure remains unclear. Here, we report that energy expenditure is regulated by a subgroup of AgRP neurons that send non-collateral projections to neurons within the dorsal lateral part of dorsal raphe nucleus (dlDRN) expressing the melanocortin 4 receptor (MC4R), which in turn innervate nearby serotonergic (5-HT) neurons. Genetic manipulations reveal a bi-directional control of energy expenditure by this circuit without affecting food intake. Fiber photometry and electrophysiological results indicate that the thermo-sensing MC4RdlDRN neurons integrate pre-synaptic AgRP signaling, thereby modulating the post-synaptic serotonergic pathway. Specifically, the MC4RdlDRN signaling elicits profound, bi-directional, regulation of body weight mainly through sympathetic outflow that reprograms mitochondrial bioenergetics within brown and beige fat while feeding remains intact. Together, we suggest that this AgRP neural circuit plays a unique role in persistent control of energy expenditure and body weight, hinting next-generation therapeutic approaches for obesity and metabolic disorders.

clear effect of alpha-MSH on DRN 5-HT neurons. This discrepancy should be addressed/clarified. Do the authors have data directly showing an effect of alpha-MSH on DRN MC4R neurons (or data from those same neurons that are inhibited by AgRP neuron terminal stimulation?)? If not, why do they believe they see an effect on DRN 5-HT neurons? These potential mechanisms also deserve more attention in the discussion, where the MC4R KO/restoration studies (in the setting of AgRP presence/ablation) could be addressed. I assume that the authors are suggesting that loss of AgRP neurons here leads to unopposed action of alpha-MSH in the DRN, upon loss of AgRP (which would nicely explain the MC4R KO phenotype)?
Minor comments: Fig. 2: Panels s-w are very hard to interpret (as noted above). No data directly display excitatory effects of alpha-MSH on DRN MC4R neurons. However, there appears to be a direct effect of alpha-MSH on DRN 5-HT neurons. How do the authors reconcile this? Also, does AgRP inhibit DRN 5-HT neurons (as it directly inhibits DRN MC4R neurons)? It also appears that there is an error or extra panel in Fig. 2v that should be either explained in the figure legend or corrected in the figure (i.e. the first two pairwise data sets are labeled the same, but have different results).

Reviewer #1 (Remarks to the Author):
This study reports a circuit from hypothalamic AGRP neurons to MC4r-expressing glutamatergic cells in the dlDRN which themselves activate a subset of mDRN serotonergic neurons. This is an extremely comprehensive study using a variety of cutting edge genetic and neuroscience techniques. I think the major claims of the manuscript are supported by the experiments and the remaining questions are primarily clarifications that need to be addressed. However, despite my admiration for most of the experiments, the text of the manuscript is extremely hard to follow in places. It probably took 3-times longer than it should have for me to understand the experiments that were reported here. This is in addition to a large number of minor errors that I will leave to the copy editors. The impact of this work would be increased if the purpose of experiments and techniques used were better explained in the text.
Major concerns: 1. Fig. 2e-h. Is it only the ZsGreen cells being patched here-this is unclear in the text? Is there any evidence of fast synaptic transmission between AGRP neurons and dlDRN neurons? The characterization of the AGRP neuropeptide on the dlDRN neurons is surprisingly superficial given all of the other details (only one example cell showing a weak transient effect with no control). This needs to be expanded. NPY could also be contributing to the inhibition and this should be checked.
Thanks for your comments. We recorded the ZsGreen neurons in Fig. 2e-h. We clarified this in our updated manuscript. Per your suggestion, we further examined the effects of AgRP on the dlDRN neural circuit. Our new in vitro recording data indicated that AgRP robustly suppressed the neural activities of 5-HT neurons within the dmDRN ( Supplementary Fig. 14). To reveal the potential effect of NPY on MC4R neurons we performed in vitro patch-clamp recording. Our data showed that NPY elicit weaker inhibition upon the activities of MC4R neurons as comparing to the effects of AgRP ( Supplementary Fig. 12), suggesting that AgRP is the major functional signaling pathway within the AgRP-DRN thermogenesis circuit.
2. Figure 4 is poorly explained and potentially poorly performed, but it was unclear. How were Chr2-expressing neurons defined based on electrical response to blue light (for example see UChida and Cohen Nature 2012 as well as Lima and Zador PLOS1 2009 for a good approach. In Fig 4f there appears to be only a general increase in firing but no phase locked responding to the 20Hz photostimulus. The recordings were in Mc4R-cre mice, so how were 5-HT neurons identified? I am skeptical about all of the claims regarding cell type specificity in this section. If these experiments were not performed with rigorous criteria for establishing cell type identity, then the claims in this section will need to be restated to be more representative of the nature of the experiment. We appreciate these valuable comments. We added more details about how the ChR2 neurons were picked and validated following the previous studies 1 . First, we optogenetically evoked the action potentials under high-frequency photostimulation ( Supplementary Fig. 18a, b). The evoked spikes precisely followed the laser pulse within 1-ms latency. Second, we compared the evoked waveform with the spontaneous spikes ( Supplementary Fig. 18c). Only the units showing evoked waveforms matching the spontaneous spikes were identified as the ChR2 neurons.
3. Line 238-252 and associated context is just confusing. For example: "In all postsynaptic dmDRN neurons of MC4RdlDRN neurons the 5-HTdmDRN←dlDRN neurons accounted for 27.6%." I don't know what this sentence means and, also, why the backwards arrow here? Does it mean dlDRN->dmDRN(5HT)?
Thanks for your comments. To rule out any confusion, we updated text by adopting "5-HT dmDRN neurons receiving projections from MC4R dlDRN neurons".
4. There are many more instances where clarity could be improved and the authors should make an effort to do a better job here.
Thanks for your comments. We thoroughly revised the manuscript as to improve the clarity and readability of our manuscript.
Minor concerns: 5. Title: Why deciphering? This word implies something about language or codes. In several previously published papers, we adopt this word to suggest the identification of a novel neural circuit. 6. In the Introduction on lines 62 there are statements indicating uncertainty ("indicates" on line 62) whether there is a relationship of homeostatic systems between hypothalamus and hindbrain. However, the citations show that this relationship has been established. I think on line 65 this could also be corrected by replacing "this link" with "an additional link" Thanks for the comments. We updated the sentences in the revised manuscript.

Line 100: Should be Supp Fig 4a
We corrected the error in our revised manuscript. 8. Lines 182 and 188: the levels are not normalized. They are larger than the controls even if these differences are not statistically different, although it is not clear whether this statistical test was even performed, so I am not even sure if the statements are statistically true.
Thanks for your comments. We updated the details about the method of photometry. We performed the calcium imaging and analyzed data according the protocols from the paper of Cui, G. et al. (Nat Protoc. 2014, (9) 1213-1228). The data showed in the original manuscript are normalized within animal. It is the comparison between baseline and manipulation within individual mouse. We added more details in the Methods to clarify the issues. 9. In addition, page and figure numbers would be appreciated in the future. We added the page and figure numbers in the revised manuscript.

Reviewer #2 (Remarks to the Author):
This manuscript by Han et al. proposes a novel, multi-synaptic mechanism through which hypothalamic AgRP neurons regulate energy expenditure. The claim is that this is mediated by a direct arcuate AgRP neuron projection to melanocortin 4-receptor-positive neurons in the dorsal raphe nucleus (DRN), which in turn activate DRN serotonin neurons, upon loss of local AgRP signaling. The findings are quite novel and the work presents some exciting data on mechanisms through which the DRN may regulate energy expenditure through thermogenesis. Furthermore, these findings are of significant interest to the fields of neuroscience and metabolism. The data on local genetic manipulations in Figure 5 of MC4R in the DRN are particularly exciting and believable, and the manuscript would be fitting for Nature Communications. However, a number of key issues, such as claims of specificity, proper controls, presentation of the data, and clarity of the writing would need to be addressed prior to acceptance of the manuscript. Due to the SARS-CoV2 pandemic, it is understandable that it might be exceptionally difficult, if not impossible, to perform additional experiments at this time. There are thus no new added/suggested experiments. It is recommended that the authors do address the major and minor comments (below), to put this manuscript into a form suitable for publication. Thanks for your thoughtful consideration. We appreciate these positive comments.
Major comments: 1. Claims of specificitythe authors suggest that AgRP projections to the dlDRN (and not the vlPAG, for example) are responsible for mediating the proposed effects. However, this is not clearly shown. The authors do demonstrate some degree of specificity to the DRN, but it is highly improbable that these manipulations selectively targeted the dlDRN but not the vlPAG (frankly, a semantic difference, depending on which mouse atlas is used). The authors also do not offer coordinates for the vlPAG in the manuscript, and A/P axes are not shown for manipulations to distinguish between differential effects on these two loci (some slices are more rostral, and some more caudal). Altering these claims does not at all diminish the novelty of this manuscript, and would greatly improve it (without requiring new experiments). The same goes for the claim concerning dmDRN serotonin neurons. Why are the authors claiming that these neurons are selectively mediating the downstream effect, when it is clear that there are ventral DRN neurons that are also downstream of DRN MC4R neurons?
Thanks for your comments. We located each of the brain nuclei based upon the Atlas of the Mouse Brain (Franklin and Paxinos). The coordinate of vlPAG and other mentioned brain regions were added in the revised Methods. To reveal the anatomical and functional distinction of the AgRP ARC →dlDRN projections from the nearby vlPAG projections, we performed both anterograde and retrograde tracing studies as summarized below:

1) We ablated a subpopulation of AgRP neurons by microinjection of DT into the dlDRN of
Agrp Cre/DTR ::Ai14 mice. Ablation of AgRP ARC→dlDRN neurons abolished the axonal fibers in the dlDRN, but the integrity of AgRP fibers in the vlPAG were unchanged ( Supplementary Fig. 8) 2) We injected the Cre-dependent trans-synaptic AAV-DIO-WGA-ZsGreen into the ARC of Agrp Cre mice. The tracing data showed that majority of the ZsGreen-positive neurons are located within the dlDRN (Supplementary Fig. 10)

3) We labeled the AgRP ARC→dlDRN neurons by injecting HSV-hEF1α-LSL-hM3Dq-mCherry
into the dlDRN of Agrp Cre ::Npy GFP mice. The retrograde tracing data showed that the projections from AgRP neurons to the dlDRN showed none collateral projections to other downstream regions ( Supplementary Fig. 3a-j) 4) We performed optogenetic experiment to examine whether the AgRP→vlPAG circuits mediate metabolic phenotypes. Photostimulation within the vlPAG axonal terminals in the Agrp Cre ::Ai32 mice showed no significant change in core body temperature or iBAT temperature ( Supplementary Fig. 2d-e).
Together, our results suggest that the AgRP ARC →dlDRN projections are anatomically and functionally distinct from the AgRP ARC →vlPAG projections.
2. Proper controlsthe authors frequently add extraneous data, which are not properly controlled (and are, in many cases, not relevant to the claims of the paper anyhow). For example, in Figs. 3 and 4, the authors perform calcium and tetrode recordings, respectively, which suggest effects on the baseline activity of MC4R neurons (and, questionably, 5-HT neurons). They then co-administer CNO to demonstrate inhibition of MC4R neurons in various states (e.g. AgRP ablation, cold, thermoneutrality, etc.). However, these results add no new information or validation to the manuscript (and they are, in fact, not controlled properly in a 2x2 designi.e. with saline/vehicle treatment). It is recommended that the authors compare across conditions, while omitting the CNO treatment component. This will open up more space for clarifying schematics or more detailed figure legends. Many proper controls are also missing from Fig. 7. This figure is also least important to the paper's central thesis.
We used in vivo photometry and in vivo tetrode recording in order to reveal how the MC4R dlDRN neurons respond to ambient temperature and activities of AgRP neurons. To maintain the significance while eliminating extra redundancy, we updated manuscript by keeping the in vivo tetrode recording data with additional control groups ( Supplementary Fig. 19). We also updated the Fig.6 with new origination.  Fig. 6. For example, the Veh/CNO studies in Fig. 6b,d- We added more control data and reorganized the data presentation in the new Fig 5a-g, m-n to improve the clarity and significance. These new data clearly showed that the DRN Pet1 neurons play a role in control of thermogenesis. added in the main or supplemental text. Thanks for your comments. We corrected all these errors in the revised manuscript.
5. Potential mechanisms through which AgRP ablation regulates DRN functionthe authors show effects of AgRP, but critically do not show any data on alpha-MSH bath application, on MC4R neurons (at least, I could not find any data on this in the manuscript, except possibly by inference in Fig. 2h, comparing baselines of control versus 300nM alpha-MSH). However, they also show a clear effect of alpha-MSH on DRN 5-HT neurons. This discrepancy should be addressed/clarified. Do the authors have data directly showing an effect of alpha-MSH on DRN MC4R neurons (or data from those same neurons that are inhibited by AgRP neuron terminal stimulation?)? If not, why do they believe they see an effect on DRN 5-HT neurons? These potential mechanisms also deserve more attention in the discussion, where the MC4R KO/restoration studies (in the setting of AgRP presence/ablation) could be addressed. I assume that the authors are suggesting that loss of AgRP neurons here leads to unopposed action of alpha-MSH in the DRN, upon loss of AgRP (which would nicely explain the MC4R KO phenotype)?
We appreciated these valuable comments. Our data showed that treatment of α-MSH significantly potentiated the spontaneous firing activity of ~87% of those post-synaptic neurons (baseline level: 1.8 Hz vs 2.4 Hz) indicating MC4R dlDRN neurons could be regulated by α-MSH ( Fig. 2f-i). We further perform patch-clamp recording to demonstrate the role of α-MSH on MC4R dlDRN neurons. The results showed that α-MSH could significantly enhance the firing of MC4R dlDRN neurons from 1.9 Hz to 4.4 Hz (Supplementary Fig. 13).
To establish whether α-MSH enhanced the firing of 5-HT dmDRN neurons directly or indirectly, a cocktail of TTX, CNQX, AP5 and bicuculline was applied to block presynaptic inputs during the recording. The results showed that the 5-HT dmDRN neurons could not respond to α-MSH after blocking the glutamatergic inputs of 5-HT dmDRN neurons (Fig. 2s, t, v, w). These results demonstrate that the α-MSH enhanced the firing of 5-HT dmDRN neurons through indirectly action on the upstream MC4R-expressing neurons. Meanwhile the 5-HT dmDRN neurons received monosynsptic glutamategic inputs from MC4R dlDRN neurons ( Fig. 2n-r). These results suggest that α-MSH enhance neural activities of MC4R dlDRN neurons, which in turn excite 5-HT dmDRN neurons.
Minor comments: As discussed in our previous responses the data collectively showed that the α-MSH regulate the firing of 5-HT dmDRN neurons indirectly through the MC4R dlDRN neurons. To further show the effects of AgRP on the DRN we added additional in vitro data showing that AgRP also robustly suppressed the neural activities of 5-HT neurons in the DRN (Supplementary Fig. 14). Blocking the presynaptic inputs of 5-HT dmDRN neurons AgRP could not inhibit the activities of 5-HT dmDRN neurons, indicating that AgRP inhibit neural activities of 5-HT dmDRN neurons through suppression of MC4R dlDRN neurons.
In Fig. 2v we totally recorded 28 5-HT dmDRN neurons. Firstly, we applied TTX and CNQX to block the AMPA/kainate receptors. There were 18 5-HT dmDRN neurons (the first pairwise data) could not be activated by α-MSH which means α-MSH could excite the 18 neurons by AMPA/kainate receptors. To test the left 10 5-HT dmDRN neurons, the AP5 was added during ephys recording. The results showed after blocking the AMPA/kainate and NMDA receptors, α-MSH failed to excite the 10 neurons (the second pairwise data). Therefore, α-MSH excited the 10 neurons by NMDA receptors. We have explained this part in the figure legend. Fig. 3: In panels g-h,k-l, the claims should be that calcium transients are more "dynamic," "variable," or "fluctuating," not necessarily at increased baseline activity (because there's no comparison of different treatments, within groups across time). In Fig. 3m, why do the authors believe they see a decrease in iBAT temp? Under cold challenge, shouldn't iBAT temperature increase? (This does not happen in vehicle animals, as it should.) Thanks for your comments. We have updated the description with more accurate words per reviewer's suggestion.
We measured the iBAT temperature under cold challenge following the protocol as described previously 2,3 . Briefly, the iBAT temperature was monitored using biocompatible temperature transponders (dimension: 2 mm × 14 mm; model: IPTT-300 transponders; BioMedic Data Systems) implanted subcutaneously in the region between the scapulae. The mice were maintained under room temperature (23 °C) and then exposed to acute cold (4 °C). Consisting with the findings in literatures, we observed that the iBAT temperature decreased (~1.0 °C) in the mice treated with vehicle within 4 hours of cold exposure. According to the literatures, the 5-HT neurons display diverse in vivo spiking behaviors 4-7 . One subgroup of 5-HT neurons could be identified by electrophysiological characteristics including firing rate, firing rhythmicity, and spike distribution, which displayed slow-firing clock-like pattern [8][9][10] . In our studies the 5-HT neurons showed the low frequency (1.65 Hz) with a highly regular pattern that was revealed by the narrow interspike interval (Fig. 3j, m). Moreover, 5-HT neurons displayed a pacemaker pattern where autocorrelation histograms typically exhibited two or three regular peaks. All of characteristics indicated that these recorded cells are 5-HT neurons. However, classical electrophysiological identification criteria may misidentify a subpopulation of non-5-HT neurons 8,10 . In our studies there were 40% putative 5-HT neurons responding to the switch from room temperature to thermoneutral condition (Fig. 3m, o), suggesting that false-positive 5-HT neurons may be inadvertently included by the electrophysiological recording method. A combination of optogenetic and in vivo electrophysiological recording approaches may minimize this problem and identify a cohort of 5-HT neurons 11,12 . We added these discussions to the manuscript accordingly.