A medullary centre for lapping in mice

It has long been known that orofacial movements for feeding can be triggered, coordinated, and often rhythmically organized at the level of the brainstem, without input from higher centers. We uncover two nuclei that can organize the movements for ingesting fluids in mice. These neuronal groups, IRtPhox2b and Peri5Atoh1, are marked by expression of the pan-autonomic homeobox gene Phox2b and are located, respectively, in the intermediate reticular formation of the medulla and around the motor nucleus of the trigeminal nerve. They are premotor to all jaw-opening and tongue muscles. Stimulation of either, in awake animals, opens the jaw, while IRtPhox2b alone also protracts the tongue. Moreover, stationary stimulation of IRtPhox2b entrains a rhythmic alternation of tongue protraction and retraction, synchronized with jaw opening and closing, that mimics lapping. Finally, fiber photometric recordings show that IRtPhox2b is active during volitional lapping. Our study identifies one of the subcortical nuclei underpinning a stereotyped feeding behavior.

The anatomical and genetic dissection of the of orofacial premotor circuits is extremely elegant and well carried out using an impressive number of single and double conditional mice together with viral approached. The results are convincing and the data support the claims about the details of the network structure. I am very enthusiastic about the story but have some problems the text and the strong statements about the rhythmogenic potential of the network.
A. 1) The introduction. I find that the introduction does not properly introduce what the authors are aiming at. The general intro to the complexity of the brainstem and the need for molecular markers etc is appropriate but there is no real Intro to the problem of the orofacial CPG -what is known and excepted form it and why Phoxb2 is connected to this function. I suggest to include such a description to make the study more appealing to a broader audience.
2) The text is generally compact. I am looking for more text to explain the methods used and some of the results and findings.  Transcriptional signature: It is obvious why Atoh1 need to be brought in -but not that the Atoh1/Phox2b population is representative for back labelled Peri5 neurons. I think it needs a better explanation. How do we know that it is not the Atoh1-/Phoxb2+ neurons that are the interesting ones? It is not clear to this reader why we need to know that IRTPhoxb2 and nTS are two structures related to lineage but with distinct molecular identities and why we need this information.
Reciprocal pattern: the paradigm for revealing the reciprocal connections between IRT and Peri5 could be better explained.
3) I think that it is too strong a statement that the functional exps. show that that IRTPhoxb2 must be the long-hypothesised licking rhythm generator. It is very difficult to show this since stimulation only produce parts of the licking behaviour and that in this set up it cannot be excluded that these cells act on other cells in IRT or elsewhere in the brainstem that generate the rhythm. In a strict sense this claim should have been supported by transsynaptic ChR2 labelling experiments form the muscles to show that pre-motor neurons indeed can trigger the rhythm. Such experiments are very difficult and not requested. But in the absence of such evidence the authors should use more careful wording also taking into account that it is discussed whether a rhythm-generating circuits are premotor or pre-premotor (see e.g in the respiratory system (Feldman and others) and in the mammalian spinal cord (Kiehn, Gosgnach, McCrea, Dougherty etc).
4) The calcium imaging is nice but does only show correlation and does no added further evidence to the role as rhythm generation. Perhaps these data could have been analyzed more extensively? 5) Line 222: why will the interconnectivity support a rhythm generating role 6) Discussion I suggest that the authors bring their findings into a general picture about the need for a licking CPG and what they have found before entering the discussion about pre-motor and coordination of movement and rhythm generation. The latter issue would need more of a background for the reader to understand the point and to carefully consider the options -also weith reference to other systems (see above).
The last part of the discussion contain mainly speculations about the functional role of the transcriptional landscape and seems like an unnecessary appendix.
Reviewer #2: Remarks to the Author: This is a well-structured work that defines two populations of neurons (IRtPhox2b and Peri5Atho1) implicated in the control of bona fide consummatory oro-facial movements. Of the two the most interesting appears to be the IRtPhox2b population, hence its slightly more detailed functional characterization. The highlight of the work is without doubt the onto/genetic dissection of these neuronal populations and the anatomical characterization of their input output with viral techniques. Their functional characterization is less thorough but appropriate (see caveats below) to support the key claims of the paper. From a conceptual ground, the work doesn't hugely alter our understanding of the function of the reticular formation in the control of orofacial movements. Also, in terms of the circuitry involved in orfacial control (both jaw and tongue), other works already pointed at the IRt, NST and supratrigeminal region (the Peri5) (perhaps in greater details) (see for example Takatoh et al., elife 2021). However, the genetic dissection of the responsible neuronal populations is, potentially (see caveats below), a significant step forward in our understanding of this circuit, whose relevance might become more evident as future works will begin to assess the descending volitional control on these populations, which is made possible by this study.
Major points: -As far as I understand the authors used a single continuous light pulse of varying duration (100-1000ms). These are highly non-physiological conditions. What would happen if the authors stimulated with frequencies likely closer the actual firing rate of these neurons (e.g. 5-10Hz for 50-200ms)?
-Concerning the relevance of having identified a specific IRt population responsible for jaw/tongue movements, this point would be made clearer by comparing the results of the optogenetic stimulation and of the fibre photometry between the Phox2b population and the general IRt Vglut2 (excitatory) population. Is there any difference between the two? -The introduction and discussion seem to lack of a fair description of the state of the art of the field with respect to circuitry and function of the brainstem control of orofacial movements (e.g. what is already known and how exactly does this work move the field forward) Reviewer #3: Remarks to the Author: The study is predicated on the assumption that the early developmental fate of lower brainstem neurons and the transcription factors implicated at this stage is a major determinant of their mature physiological role. This approach, usually combined with the clever exploitation of other population-defining markers (vesicular transporters, receptors etc.) has been especially successful in the hands of the present investigators in defining a group of brainstem neurons with a specialized role in central respiratory chemoreception (retrotrapezoid nucleus) and by others in defining functional subgroups of serotonergic or respiratory-rhythm neurons. Here the authors focus on two Phox2b-dependent neuronal clusters that they had identified in prior studies. The first cluster is a ring of neurons that surround the trigeminal motor nucleus, a region already suspected to harbor trigeminal premotor neurons. The peritrigeminal ring is also atoh-1 dependent which distinguishes it from the motor nucleus itself and allowed the authors to manipulate the interneuronal ring selectively with intersectional genetic approaches. Thus authors were therefore able to determine the connectivity and physiological function of this ring of neurons. The results are extremely convincing. The second focus of this study is a group of Phox2b-derived neurons located in the IRt (intermediate reticular formation). These Phox2b-derived neurons could be selectively accessed and transduced based on their stereotaxic location in Phox2b-Cre mice. The IRt is an extraordinarily complex portion of the medullary reticular formation formerly believed to be primarily implicated in autonomic regulations. As shown here this region also plays a key role in the control of orofacial movements and may contain rhythm generator. This is an important and technically impressive study describing very novel findings regarding the genesis of orofacial movements implicated in drinking.

Reviewer #(Remarks to the Author):
This study uses anatomical, developmental and functional approaches to identify brainstem circuits involved in the control of rhythmic orofacial movement. With elegant genetic circuit tracing the authors identify two brainstem nuclei -in the intermediate reticular formation (IRt) and around the trigeminal nucleus -that are premotor to most of the muscle controlling jaw and tongue muscles. They hypothesize that this anatomical substrate might be a brainstem central pattern generator for feeding. When stimulated these two sites cause either jaw opening or jaw opening and lapping reproducing some of the movements needed for feeding. The neurons in IRT show correlated activity during natural lapping. The main conclusion of the study is that is has identified a licking CPG in the brainstem. The anatomical and genetic dissection of the of orofacial premotor circuits is extremely elegant and well carried out using an impressive number of single and double conditional mice together with viral approaches. The results are convincing and the data support the claims about the details of the network structure. I am very enthusiastic about the story but have some problems with the text and the strong statements about the rhythmogenic potential of the network.

1) The introduction. I find that the introduction does not properly introduce what the authors are aiming at. The general intro to the complexity of the brainstem and the need for molecular markers etc is appropriate but there is no real Intro to the problem of the orofacial CPG -what is known and expected from it and why
Phoxb2 is connected to this function. I suggest to include such a description to make the study more appealing to a broader audience.
We did not aim at finding an orofacial CPG, we rather stumbled on a likely CPG while exploring the function of unknown, but genetically defined, interneurons in the hindbrain. The introduction currently reflects this. To ameliorate the introduction along the lines suggested by the referee, we have now added two phrases in bold below, a reference to Takatoh et al 2021 (at the suggestion of referee #2) and we end the introduction, as per a more classical format, with a few sentences which sum up our findings (main changes in bold): Over decades, the reticular formation has slowly emerged from "localizatory nihilism" 2 , and regions defined by stereotaxy [e.g. 3 ], or cell groups defined by their projections [e.g. 4 ] have been implicated in a variety of roles: premotor neurons to orofacial or respiratory muscles 5,6 , and -underpinning the sophisticated residual behaviors observed in decerebrate animals 7 -rhythm and pattern generators for chewing, whisking, breathing and sighing 3,8,9,10,11,5 . Licking is another rhythmic behavior for which a hindbrain rhythm generator is predicted 12 although the evidence is mostly extrapolated from chewing, the two behaviors possibly sharing some substrate 9 . However, the parsing of the reticular formation […] […] thus "visceral" indeed. To this broadened picture of the visceral nervous system, in charge of vital functions and maintenance of the interior milieu, we now add two groups of Phox2b interneurons, located in the reticular formation of the hindbrain, that are premotor to orofacial muscles and can command licking or lapping, a rhythmic feeding behavior essential for the intake of liquids in many terrestrial vertebrates.
2) The text is generally compact. I am looking for more text to explain the methods used and some of the results and findings. For example, there is limited help to the reader to follow and understand the complicated crosses and the anatomy. The authors used two double conditional mice, one in Fig. S1A and one in Fig. 1A but without a clear explanation of the first (and why it is needed). It Is not described probably how cells become tdT, receive only GFP terminals or are just Chat positive MNs in Fig. 1B To remedy this problem, we have now extensively modified the technical explanations as follows, making things more explicit and splitting long sentences into shorter ones: The vast majority of these neurons are glutamatergic, thus express the glutamate vesicular transporter Vglut2 as shown by expression of the Cre and Flpo-dependent reporter RC::Fela in a Phox2b::Flpo;Vglut2::Cre background (Fig. S1A). We used this neurotransmitter phenotype to implement an intersectional strategy that excludes the potentially confounding widespread projections of other Phox2b + neurons, in the locus coeruleus 21 , which are noradrenergic. We designed a novel intersectional allele (Rosa FRTtomato-loxSypGFP or Rosa FTLG ) (Fig. 1A) which expresses one of two fluorophores, exclusively: action of flippase (FLPo) will trigger cytoplasmic expression of tdTomato (tdT), while additional action of Cre recombinase, will extinguish tdT in the cell soma and switch on instead a fusion of synaptophysin with GFP (Syp-GFP) transported to pre-synaptic sites 22 . When FLPo was driven by the Phox2b promoter, and Cre by the Vglut2 promoter, i.e. in Phox2b::Flpo;vGlut2::Cre;Rosa FTLG pups at P4, tdT was expressed, as expected, in the soma of the singly recombined motoneurons (which are Phox2b + , but not glutamatergic), but lost from the doubly recombined interneurons (which are Phox2b + and glutamatergic) (Fig. S1B). The latter, in turn, had switched on Syp-GFP + in their synaptic boutons, which covered remarkably discrete structures of the hindbrain (Fig. S1B, Fig. 1B), among which motor nuclei (whose function will be discussed later) featured prominently:

It will be useful to have the MN pools attached a function so the reader can understand what they are used for.
At this point in the text we would rather keep the attention on the simple notion that many Phox2b interneurons are premotor. Moreover, at this stage of the narrative, some motor nuclei are irrelevant to the rest of the study (MoA or Mo6, targeted by Phox2b premotor neurons, but not located in IRt or peri5). The role of the relevant nuclei is expounded later, as an introduction to the functional experiments, and repeating them would be awkward. To address the concern of the referee we have now added "(whose function will be discussed later)", as cited above.

Fig. 1 c is essential and need to be described in more detail.
We have now reformulated the description as follows: We injected a G-defective rabies virus variant encoding the fluorophore m-Cherry 24 together with a helper virus encoding G and the fluorophore YFP (HSV-YFP-G) in the posterior belly of the digastric muscle (Fig. 1C) (a jaw-abductor), known to be innervated by Acc7 25,26 . Predictably, the only seed neurons (i.e. that co-express the rabies virus encoded mCherry and the helper virus encoded YFP) were found in Acc7 (right panel in Fig. 1C). Premotor neurons, presynaptic to the seed motoneurons, (i.e. that express only the rabies virus encoded mCherry ) and which, in addition, were Phox2b+, were found at two sites only:

Fig S2 should be in the main text so we get the combined picture for the tracing and rationale connected with it (some sort of n number should be included in the text).
We respectfully decline to execute this change. Figure S2 is huge, cannot be combined with Figure 2, which would have to be reconfigured. Moreover, we do not provide the same detail in figure S2 than in Fig2, forcing us to an acrobatic formatting exercise, disproportionate to the benefit. Finally, Figure S2 contains parts which are merely contextual (i.e. the unrelated premotor landscape of laryngeal and masseter).

Transcriptional signature: It is obvious why Atoh1 need to be brought in -but not that the Atoh1/Phox2b population is representative for back labelled Peri5 neurons. I think it needs a better explanation. How do we know that it is not the Atoh1-/Phoxb2+ neurons that are the interesting ones?
The fact that neurons back-labeled from the posterior digastric INCLUDE Atoh1/Phox2b neurons in the peri5 region is shown in Fig. S2. However, we do not claim that they are representative of anything else other than themselves. It is almost certain that other neurons in the same region, which are Phox2b but not Atoh1, have different roles. In this paper, we are just limited, for technical reasons, to optogenetically manipulate the Atoh1/Phox2b ones.

It is not clear to this reader why we need to know that IRTPhoxb2 and nTS are two structures related to lineage but with distinct molecular identities and why we need this information.
We agree that most physiologists will not need this information. However, it is an integral part of our genetic characterization of a new neuronal population, and this information might turn out helpful to guide further genetic dissections of these neurons (for example using single cell transcriptomics) or to target or interpret future optogenetic or chemogenetic manipulations.

Reciprocal pattern: the paradigm for revealing the reciprocal connections between IRT and Peri5 could be better explained.
We have now reformulated this point as follows: In addition, anterograde tracing from IRt Phox2b in a Phox2b::Cre background and from Peri5 Atoh1 in a Phox2b::flpo;Atoh1::Cre background revealed, respectively, massive projections of IRt Phox2b to the peri5 region (Fig. 3H) and of Peri5 Atoh1 to the IRt region. (Fig. S3C). We could not assess the precise cellular target of the former, but those of the latter included IRt Phox2b (Fig. S3D, inset), suggesting reciprocal connections of the two nuclei.
3) I think that it is too strong a statement that the functional exps. show that that IRTPhoxb2 must be the long-hypothesised licking rhythm generator.
The original text says "the CPG […] or part thereof", but see below for further softening of our conclusions, along the lines of the referee's request.

It is very difficult to show this since stimulation only produce parts of the licking behavior…
We respectfully do not understand this remark. We think that the entire licking behavior is triggered.

…and that in this set up it cannot be excluded that these cells act on other cells in IRt or elsewhere in the brainstem that generate the rhythm. In a strict sense this claim should have been supported by transsynaptic ChR2
labelling experiments form the muscles to show that pre-motor neurons indeed can trigger the rhythm. Such experiments are very difficult and not requested. But in the absence of such evidence the authors should use more careful wording also taking into account that it is discussed whether a rhythm-generating circuits are premotor or pre-premotor (see e.g in the respiratory system (Feldman and others) and in the mammalian spinal cord (Kiehn, Gosgnach, McCrea, Dougherty etc).
Indeed, the experiment mentioned by the referee is exceedingly difficult, if not impossible, because the rabies virus would kill cells before they express enough ChR2. Along the line suggested by the referee, we have now added the following caveats on the diagnosis of a 4 CPG, together with four additional references to the authors mentioned by the referee. (Note that a discussion of the optogenetic experiments has been transported from the results to the discussion section, making the latter more complete and coherent).
In addition, one of them, IRt Phox2b , translates a tonic stimulation into a rhythmic behavior. The most parsimonious interpretation of IRt Phox2b is that its neurons are bifunctional: premotor through their collaterized inputs on motor nuclei, and rhythm generators, corresponding to the hypothetical licking CPG 11 or at least an element thereof, in the precise region where many lickrhythmic neurons were previously recorded ( 8,7 for reviews). It is of note that another nearby Phox2b + nucleus, the RTN, has intrinsic rhythmic properties, in that case related to breathing, in the neonate 55,14 . At this stage, though, we cannot exclude that IRt Phox2b contains two subtypes of neurons, one premotor and the other pre-premotor, and that it is the latter which, upon photostimulation, triggers rhythmic repetition; in other words, that IRt Phox2b encompasses a two (or more)-level architecture, akin to models proposed for other motor behaviors 56,57,58,59 . This possibility is made less likely by the apparent genetic homogeneity of IRt Phox2b , whose neurons all co-express the transcriptional signature Phox2b/Cited1. Finally, the possibility that the rhythm would be generated by neurons elsewhere in the brainstem (recruited by IRt Phox2b and feeding back on it) is constrained by the limited output of IRt Phox2b : to motor nuclei and the peri5 region.
In addition to rhythmic tongue protrusion and jaw opening, the entrainment of a full licking cycle requires the delayed activation of antagonistic muscles, as in several "burst generator" models of the locomotor CPG (e.g. 58 ). One substrate for such rhythmic alternation might comprise the reciprocal projections of IRt Phox2b and Peri5 Atoh1 (Fig. 3H, Fig. S3C,D), the former targeting tongue protractors and the latter tongue retractors.

4) The calcium imaging is nice but does only show correlation and does no added further evidence to the role as rhythm generation. Perhaps these data could have been analyzed more extensively?
Fiber photometry (the current gold standard for recording calcium activity in deep brain regions) is used here to show correlation between activity of IRt Phox2b and licking bouts. The limited temporal resolution of calcium indicators make this approach inherently inadequate to capture a ~7Hz rhythm, and no amount of additional analysis will resolve this.

5) Line 222: why will the interconnectivity support a rhythm generating role
This sentence was an allusion to the recurrent synaptic interconnections providing positive feedback that are proposed to play a role, for example in the preBötC. This is now made more explicit as follows: "suggesting local interconnectivity of IRt neurons, possibly related to rhythmogenesis, through recurrent synaptic connections, as hypothesized for other rhythm generating structures (Del Negro et al 2006)" 6) Discussion I suggest that the authors bring their findings into a general picture about the need for a licking CPG and what they have found before entering the discussion about pre-motor and coordination of movement and rhythm generation. The latter issue would need more of a background for the reader to understand the point and to carefully consider the options -also with reference to other systems (see above).
We hope that the above-mentioned modifications to the introduction, restructuring and lengthening of the discussion on the likelihood that IRt Phox2b is a CPG, and how it would mechanistically compare to other ones, plus added bibliographic references to other CPGs take care of this point.
The last part of the discussion contains mainly speculations about the functional role of the transcriptional landscape and seems like an unnecessary appendix.
The last part treats the subject of evolution of physiological functions and of neuron types, and is admittedly of no import to the physiologist. However, we believe that part of the originality of our paper is to sit at the border of several disciplines (physiology, development and evolution) and we would like to leave it that way.

Reviewer #2 (Remarks to the Author):
This is a well-structured work that defines two populations of neurons (IRtPhox2b and Peri5Atho1) implicated in the control of bona fide consummatory oro-facial movements. Of the two the most interesting appears to be the IRtPhox2b population, hence its slightly more detailed functional characterization. The highlight of the work is without doubt the onto/genetic dissection of these neuronal populations and the anatomical characterization of their input output with viral techniques. Their functional characterization is less thorough but appropriate (see caveats below) to support the key claims of the paper. From a conceptual ground, the work doesn't hugely alter our understanding of the function of the reticular formation in the control of orofacial movements. Also, in terms of the circuitry involved in orofacial control (both jaw and tongue), other works already pointed at the IRt, NST and supratrigeminal region (the Peri5) (perhaps in greater details) (see for example Takatoh et al., elife 2021). However, the genetic dissection of the responsible neuronal populations is, potentially (see caveats below), a significant step forward in our understanding of this circuit, whose relevance might become more evident as future works will begin to assess the descending volitional control on these populations, which is made possible by this study. Major points: -As far as I understand the authors used a single continuous light pulse of varying duration (100-1000ms). These are highly non-physiological conditions. What would happen if the authors stimulated with frequencies likely closer the actual firing rate of these neurons (e.g. 5-10Hz for 50-200ms)?
In our view, what the light pulse should ideally emulate is the firing pattern of input structures to the IRt. This pattern is unknown and, for all we know, could be a tonic drive (for example from the cortex, as already mentioned in the text), in which case constant illumination might be a decent approximation, after all. This, said, as requested by the referee, we have now added 100ms stimulations delivered at 4, 6 and 7Hz and we show that we analogically entrain the lapping movements. This is commented in the main text as: […] while IRt Phox2b but not Peri5 Atoh1 can protract the tongue, in line with the targeting of hypoglossal motoneurons for tongue protractors by the former and tongue retractors by the latter (Fig. 3D,E,J,K). Delivering the stimulus at 4, 5 or 7Hz led to an analogical repetition of the movement (Fig. S4A) showing a lack of refractory period in that frequency range. Lengthening the light pulse […] And illustrated as a new panel A in Fig. S4: