Emergence of Functional Heart-Brain Circuits in a Vertebrate
Curation statements for this article:-
Curated by eLife
eLife Assessment
This important study establishes a zebrafish model to investigate the development of interoceptive circuits involved in the regulation of heart rate and their functions in vivo, in an intact vertebrate body. Using a combination of targeted functional imaging, manipulation, and mathematical modeling, the study presents compelling evidence for the staggered development of motor and sensory feedback elements of the interoceptive circuit.
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
Abstract
The early formation of sensorimotor circuits is essential for survival. While the development and function of exteroceptive circuits and their associated motor pathways are well characterized, far less is known about the circuits that convey viscerosensory inputs to the brain and transmit visceromotor commands from the central nervous system to internal organs. Technical limitations, such as the in utero development of viscerosensory and visceromotor circuits and the invasiveness of procedures required to access them, have hindered studies of their functional development in mammals. Using larval zebrafish—which are genetically accessible and optically transparent—we tracked, in vivo, how cardiosensory and cardiomotor neural circuits assemble and begin to function. We uncovered a staged program. First, a minimal efferent circuit suffices for heart-rate control: direct brain-to-heart vagal motor innervation is required, intracardiac neurons are not, and heart rate is governed exclusively by the motor vagus nerve. Within the hindbrain, we functionally localize a cholinergic vagal premotor locus that engages this early efferent control. Second, sympathetic innervation arrives and enhances the dynamics and amplitude of cardiac responses, as neurons in the most anterior sympathetic ganglia acquire the ability to drive cardiac acceleration. These neurons exhibit proportional, integral, and derivative–like relationships to heart rate, consistent with controller motifs that shape gain and dynamics. Third, vagal sensory neurons innervate the heart. Distinct subsets increase activity when heart rate falls or rises, and across spontaneous fluctuations, responses to aversive stimuli, and optogenetically evoked cardiac perturbations, their dynamics are captured by a single canonical temporal kernel with neuron-specific phase offsets, supporting a population code for heart rate. This temporally segregated maturation isolates three experimentally tractable regimes—unidirectional brain-to-heart communication, dual efferent control, and closed-loop control after sensory feedback engages—providing a framework for mechanistic dissection of organism-wide heart–brain circuits.
Article activity feed
-
eLife Assessment
This important study establishes a zebrafish model to investigate the development of interoceptive circuits involved in the regulation of heart rate and their functions in vivo, in an intact vertebrate body. Using a combination of targeted functional imaging, manipulation, and mathematical modeling, the study presents compelling evidence for the staggered development of motor and sensory feedback elements of the interoceptive circuit.
-
Reviewer #1 (Public review):
Summary:
The manuscript by Hernandez-Nunez et al. provides a comprehensive characterization of how heart-brain circuits develop in a vertebrate brain namely the zebrafish. The characterization is performed using a combination of modern and sophisticated imaging and neural manipulation techniques and achieves unprecedented clarity and detail to how the heart-brain communication develops early in life. The paper describes a three staged program, where first an efferent-circuit from the motor vagus to the heart develops, followed by sympathetic innervation and lastly sensory neurons innervate the heart.
Strengths:
The paper is very clearly and nicely written. The findings are novel and of high quality and relevance. The presentations are very clear and nicely interpreted. The analyses are well presented and …
Reviewer #1 (Public review):
Summary:
The manuscript by Hernandez-Nunez et al. provides a comprehensive characterization of how heart-brain circuits develop in a vertebrate brain namely the zebrafish. The characterization is performed using a combination of modern and sophisticated imaging and neural manipulation techniques and achieves unprecedented clarity and detail to how the heart-brain communication develops early in life. The paper describes a three staged program, where first an efferent-circuit from the motor vagus to the heart develops, followed by sympathetic innervation and lastly sensory neurons innervate the heart.
Strengths:
The paper is very clearly and nicely written. The findings are novel and of high quality and relevance. The presentations are very clear and nicely interpreted. The analyses are well presented and applied.
-
Reviewer #2 (Public review):
Hernandez-Nunez et al. investigate the development and function of neural circuits involved in the regulation of heart rate in larval zebrafish. Using conserved genetic markers, they identify neural pathways involved in the bidirectional control of heart rate and in providing sensory feedback, potentially enabling more precise tuning. The main observation is that the different elements of this circuit are laid down in a developmentally staggered manner.
At 4 days old, the heart rate is invariant to a range of sensory stimuli, and the vagal motor or sympathetic pathways could not be seen to innervate the heart. Progressively through development, the heart is first innervated by the vagal motor pathway, whose axons are cholinergic, before the formation of phox2bb+ intracardiac neurons (ICNs). At this stage, …
Reviewer #2 (Public review):
Hernandez-Nunez et al. investigate the development and function of neural circuits involved in the regulation of heart rate in larval zebrafish. Using conserved genetic markers, they identify neural pathways involved in the bidirectional control of heart rate and in providing sensory feedback, potentially enabling more precise tuning. The main observation is that the different elements of this circuit are laid down in a developmentally staggered manner.
At 4 days old, the heart rate is invariant to a range of sensory stimuli, and the vagal motor or sympathetic pathways could not be seen to innervate the heart. Progressively through development, the heart is first innervated by the vagal motor pathway, whose axons are cholinergic, before the formation of phox2bb+ intracardiac neurons (ICNs). At this stage, before the first ICNs are observed, activation of the vagal motor pathway by optogenetic activation of a localized population of cholinergic hindbrain neurons leads to bradycardia. After the vagal motor innervation begins, the sympathetic pathway innervates the heart, which could be visualized in the form of TH+ fibers from the anterior paravertebral ganglia (APG). The activity of the TH+ APG neurons was diverse and showed proportional, integral and derivative like relationships to the heart rate, suggesting a role in more precise tuning of the rate than what could be achieved through the vagal pathway alone. The sensory vagus innervation of the heart was identified to be the last stage to develop, however, neurons in the nodose ganglion exhibited diverse responses tuned to the heart rate well before the innervation reached the heart. The authors attribute this to the fact that other indirect sensory cues from the gills or vasculature could be used to sense heart rate prior to innervation.
This study identifies key components of the control loop required for the regulation of heart rate in zebrafish. The control mechanism appears to be independent of the cues that trigger heart rate changes, indicating that the circuit is indeed part of an interoceptive pathway for heart rate control. Evidence for the staggered development of the vagal-motor, sympathetic and sensory pathways is conclusive, and as the authors discuss, this phenomenon progressively allows for finer grained control of the heart rate. This could be achieved through proportional-integral-derivative like control properties emerging in a diverse set of neurons in the APG and sensory feedback of the state of the heart. In line with these findings, the baseline variability of heart rate prior to innervation at 4 days old appears to be comparatively lower than the later stages (Fig 1C, D, supp. Fig 1C-F) and increases over development. This has been quantified in supp. Fig 2.
Based on this observation and the time courses of the kernels identified by the GLMs, I would expect heart rate fluctuations of a finer time scale, ultimately limited by the time course of GCaMP6s, to be captured by the models in Figs 3, 5 and 7, in addition to the stimulus locked changes that are highlighted. While the models yield valuable insight in the form of the activation kernels and their potential roles, in one instance, it captures the potential contribution of either the motor vagus or the APG to the change in heart rate. This makes it challenging to identify where it falls short and the potential functions of pathways that have not been probed. Some other known modulatory influences on the cardiac rhythm are mentioned in the text, however exploring these pathways would be outside the scope of this study.
Lastly, the proposed anatomical connectivity of the heart-brain circuit is based on tracts observed in this study as well as those inferred from function and from previous studies. The authors have included a section on the limitations of the model, where they importantly clarify that the proposed mechanisms, although likely, do not establish direct synaptic connectivity, for example whether the brainstem neurons identified for their functional roles are the ones that give rise to efferent projections via the motor vagus nerve. The addition of clear labels in Fig 9 also helps the reader interpret relationships between model components more clearly.
To conclude, this study establishes the larval zebrafish as a tractable in vivo model to study heart-brain interoceptive and regulatory circuits through development and describes a detailed mechanistic framework to take forward.
-
Author response:
The following is the authors’ response to the original reviews.
Public Reviews:
Reviewer #1 (Public review):
Summary:
The manuscript by Hernandez-Nunez et al. provides a comprehensive characterization of how heart-brain circuits develop in a vertebrate brain, namely the zebrafish. The characterization is performed using a combination of modern and sophisticated imaging and neural manipulation techniques and achieves unprecedented clarity and detail in how the heart-brain communication develops early in life. The paper describes a three-stage program, where first an efferent-circuit from the motor vagus to the heart develops, followed by sympathetic innervation, and lastly sensory neurons innervate the heart.
Strengths:
The paper is very clearly and nicely written. The findings are novel and of high quality …
Author response:
The following is the authors’ response to the original reviews.
Public Reviews:
Reviewer #1 (Public review):
Summary:
The manuscript by Hernandez-Nunez et al. provides a comprehensive characterization of how heart-brain circuits develop in a vertebrate brain, namely the zebrafish. The characterization is performed using a combination of modern and sophisticated imaging and neural manipulation techniques and achieves unprecedented clarity and detail in how the heart-brain communication develops early in life. The paper describes a three-stage program, where first an efferent-circuit from the motor vagus to the heart develops, followed by sympathetic innervation, and lastly sensory neurons innervate the heart.
Strengths:
The paper is very clearly and nicely written. The findings are novel and of high quality and relevance. The presentations are very clear and nicely interpreted. The analyses are well presented and applied.
Weaknesses:
From the heart rate traces, heart rate variability seems to be prominent and changes across days post-fertilisation (dpf). That would be a useful dependent variable, considering that the variation captured by the models does not fully explain heart rate, both for sympathetic and parasympathetic efferents. Given the strong auto-rhythmicity of nodal tissue in neurogenic hearts, modulatory inputs could potentially predict heart rate variability with higher precision.
We thank Reviewer #1 for their generous assessment of the manuscript. We are grateful that the reviewer found the work clearly written, novel, technically strong, and relevant, and that they viewed the analyses and interpretation as clear and appropriate.
The reviewer’s main suggestion concerned heart-rate variability and the possibility that developmental changes in variability may provide an additional readout of autonomic circuit maturation. We thank the reviewer for this insightful comment. The developmental change in baseline cardiac variability was indeed apparent in the heartrate traces, and the reviewer’s observation prompted us to examine it directly. This analysis added a useful component to the revised manuscript by showing that the maturation of heart–brain circuits is accompanied not only by changes in stimulus-evoked cardiac control, but also by a marked developmental change in the temporal variability of the cardiac rhythm.
We therefore quantified variability in the sequence of interbeat intervals during prestimulus baseline periods. We measured short-timescale variability using the root mean square of successive differences between interbeat intervals, or RMSSD, and normalized this value to the mean interbeat interval to allow comparison across developmental stages with different baseline heart rates. Normalized RMSSD increased 2.66-fold between 4 and 5 dpf and remained elevated at 7 and 12 dpf (bootstrap 95% CI, 2.04–3.27; new Supp. Fig. 2 A, B). The coefficient of variation of interbeat intervals showed the same developmental pattern (new Supp. Fig. 2D).
We also tested whether this increase could be explained solely by developmental differences in mean heart rate. It could not. The best-supported mixed-effects model included an interaction between developmental stage and mean interbeat interval, indicating that the relationship between average cardiac timing and interval-to-interval variability itself changes across development (new Supp. Fig. 2C). We now describe these results in the last paragraph of the first result section and provide the full analysis and statistical procedures in the new Methods subsection “Analysis of developmental cardiac variability.”
We also clarified the relationship between this new analysis and the neural prediction models. Each value in the cardiac signal used in those models was derived from one interbeat interval and therefore already contained the beat-resolved variability present in the cardiac recordings. The models consequently test whether neural activity predicts fluctuations in that beat-resolved cardiac signal. However, the cardiac and neural measurements have different temporal bandwidths: the cardiac recordings resolve millisecond-scale differences between successive intervals, whereas calcium indicators integrate neural activity over substantially longer timescales. The models therefore recover the component of cardiac variation that evolves over timescales accessible to both measurements.
We are grateful to the reviewer for drawing our attention to this feature of the data. This new analysis helped clarify that cardiac innervation coincides with a substantial increase in heart rate variability. Thus, the resulting analysis strengthens the developmental physiology presented in the manuscript while also clarifying the temporal relationship between the cardiac recordings and the calcium-based neural models.
Reviewer #2 (Public review):
Hernandez-Nunez et al. investigate the development and function of neural circuits involved in the regulation of heart rate in larval zebrafish. Using conserved genetic markers, they identify neural pathways involved in the bidirectional control of heart rate and in providing sensory feedback, potentially enabling more precise tuning. The main observation is that the different elements of this circuit are laid down in a developmentally staggered manner.
At 4 days old, the heart rate is invariant to a range of sensory stimuli, and the vagal motor or sympathetic pathways could not be seen to innervate the heart. Progressively through development, the heart is first innervated by the vagal motor pathway, whose axons are cholinergic, before the formation of phox2bb+ intracardiac neurons (ICNs). At this stage, before the first ICNs are observed, activation of the vagal motor pathway by optogenetic activation of a localized population of cholinergic hindbrain neurons leads to bradycardia. After the vagal motor innervation begins, the sympathetic pathway innervates the heart, which could be visualized in the form of TH+ fibers from the anterior paravertebral ganglia (APG). The activity of the TH+ APG neurons was diverse and showed proportional, integral, and derivative-like relationships to the heart rate, suggesting a role in more precise tuning of the rate than what could be achieved through the vagal pathway alone. The sensory vagus innervation of the heart was identified to be the last stage to develop; however, neurons in the nodose ganglion exhibited diverse responses tuned to the heart rate well before the innervation reached the heart. The authors attribute this to the fact that other indirect sensory cues from the gills or vasculature could be used to sense heart rate prior to innervation.
This study identifies key components of the control loop required for the regulation of heart rate in zebrafish. The control mechanism appears to be independent of the cues that trigger heart rate changes, indicating that the circuit is indeed part of an interoceptive pathway for heart rate control. Evidence for the staggered development of the vagal-motor, sympathetic, and sensory pathways is conclusive, and as the authors discuss, this phenomenon progressively allows for finer-grained control of the heart rate. This could be achieved through proportional-integral-derivative-like control properties emerging in a diverse set of neurons in the APG and sensory feedback of the state of the heart. In line with these findings, the baseline variability of heart rate prior to innervation at 4 days old appears to be comparatively lower than the later stages (Figure 1C, D, Supplementary Figure 1C-F) and increases over development. Based on this observation and the time courses of the kernels identified by the GLMs, I would expect heart rate fluctuations of a finer time scale, ultimately limited by the time course of GCaMP6s, to be captured by the models in Figures 3, 5, and 7, in addition to the stimulus-locked changes that are highlighted. While the models yield valuable insight in the form of the activation kernels and their potential roles, in one instance, this captures the potential contribution of either the motor vagus or the APG to the change in heart rate. This makes it challenging to identify where it falls short and the potential functions of pathways that are yet to be discovered.
Lastly, the proposed anatomical connectivity of the heart-brain circuit is based on tracts observed in this study as well as those inferred from function and from previous studies.
(1) It is not clear from the images presented here whether the VSNs send feedback projections to the brainstem VPN.
(2) Do the brainstem neurons identified by their functional roles send efferent projections via the motor vagus nerve? This is unclear from the results presented and needs to be clarified in the text.
(3) Add appropriate clarifying annotations to Figure 9 and a section of text discussing the potential unknowns in the proposed circuit diagram.
We thank the reviewer for the thoughtful and constructive review. We greatly appreciate the careful synthesis of the developmental sequence described in the manuscript. The comments were particularly helpful in prompting us to clarify the temporal scope of the neural models, to quantify developmental changes in baseline cardiac variability, and to distinguish more explicitly between anatomical connections directly observed in this study, functional relationships supported by our experiments, and connections inferred from prior work. We have revised the manuscript and figures accordingly.
Developmental changes in cardiac variability and the temporal scope of the neural encoding or decoding models
We agree with the reviewer that baseline cardiac variability is lower at 4 dpf and increases at later developmental stages. To quantify this observation, we analyzed the sequence of all interbeat intervals during pre-stimulus baseline periods. We measured the root mean square of successive differences between cardiac intervals, or RMSSD, and normalized this value to the mean interbeat interval to permit comparison across stages with different baseline heart rates. Normalized RMSSD increased 2.66-fold between 4 and 5 dpf and remained elevated at 7 and 12 dpf (bootstrap 95% CI, 2.04– 3.27; new Supp. Fig. 2 A, B). The coefficient of variation of interbeat intervals showed the same developmental pattern (new Supp. Fig. 2 D). We also tested whether this increase could be explained solely by developmental differences in average heart rate. The best-supported model included an interaction between age and mean interbeat interval, indicating that the relationship between average cardiac timing and interval-to-interval variability itself changes across development (new Supp. Fig. 2 C).
We now describe this result in the revised Results section and provide the full analysis in the new Methods subsection, “Analysis of developmental cardiac variability.” We are grateful to the reviewer for drawing our attention to this feature of the data, as the resulting analysis provides an additional quantitative description of the developmental change in baseline cardiac dynamics.
We also agree that it is important to clarify what temporal features can be captured by the generalized linear models. Each value in the cardiac time series used in the models was calculated from the duration of an individual cardiac interval. The modeled cardiac signal therefore already contains the beat-resolved variation present in the recordings, including stimulus-locked and spontaneous fluctuations. The models are not limited to fitting only the mean stimulus response.
However, the cardiac and neural measurements have different temporal bandwidths. The cardiac recordings resolve millisecond-scale differences between successive intervals, whereas GCaMP6s integrates neural activity over substantially longer timescales. The models therefore recover the component of cardiac variation that occurs over timescales shared by the cardiac and calcium measurements. We have clarified this point in the revised Results and Methods.
We further revised the text to clarify the purpose and limits of the models. The models identify predictive relationships between the recorded neural populations and cardiac dynamics; they are not intended to provide an exhaustive causal decomposition of all sources of heart-rate variation. Unexplained variation may reflect the intrinsic autorhythmicity of pacemakers, endocrine or paracrine influences, metabolic state, or temporal information not recoverable from calcium imaging. We now discuss these possibilities explicitly.
(1) Do viscerosensory neurons project to the VPN?
We thank the reviewer for identifying this gap in the anatomical presentation. In the original manuscript, we showed the developing viscerosensory ganglia and their later peripheral innervation of the organs, but we did not show clearly that these neurons also extend central projections into the hindbrain.
We have now added a new supplementary figure showing the right and left viscerosensory ganglia at 2, 3, and 4 dpf. At each of these stages, labeled axons can be seen extending dorsally from the developing ganglia and entering the hindbrain, including at stages that precede detectable peripheral innervation of the organs (new Supp. Fig. 8). These observations demonstrate that the central projection of the sensory vagus is established early, while the ganglia are still forming and before completion of their peripheral organ innervation. We have clarified in the discussion that our data do not resolve whether these afferents form direct synaptic contacts with the specific functionally identified VPN described in this study.
(2) Do the functionally identified brainstem neurons project through the motor vagus nerve?
We agree that this distinction needed clarification. We have revised the text to distinguish the descending motor vagus pathway from the functionally defined cholinergic hindbrain locus identified in Fig. 3. Our data show that activity in a medial cholinergic hindbrain locus predicts cardiac inhibition and that targeted activation of this region elicits bradycardia, indicating that it engages descending vagal cardiac control. However, because the functionally identified neurons were not individually traced, this locus may include motor vagus neurons, premotor neurons upstream of the motor vagus, or both. We have clarified this point in the Results and Discussion.
(3) Clarification of Figure 9 and unresolved elements of the proposed circuit
We thank the reviewer for this suggestion. We have revised Figure 9 to make the evidence of each proposed interaction explicit. The updated legend now distinguishes functional relationships supported by an observed anatomical connection, functional relationships for which the anatomical connection was observed in fewer than 50% of animals, and functional relationships for which no anatomical connection was observed at the developmental stage examined. The revised figure therefore no longer presents all arrows as equivalent or as evidence of direct synaptic connectivity.
We have also added a paragraph to the Discussion clarifying the principal unresolved aspects of the model. Specifically, we now state that the functional relationships identified through encoding, decoding, and perturbation experiments do not by themselves establish direct synaptic connections; that the peripheral structures through which viscerosensory neurons detect heart-rate-related dynamics remain unknown; and that, although viscerosensory neurons project centrally into the hindbrain, their postsynaptic targets have not yet been identified.
Recommendations for the authors:
Reviewer #1 (Recommendations for the authors):
(1) Figure 3: Given that the paper is arguing for cells with an inhibitory coupling to heart rate to be major contributors to cardiac rate among cholinergic neurons in the early phase, it would be useful to report the outcomes of a GLM model with just the C1 cluster as input.
We thank the reviewer for this helpful suggestion. We performed the recommended analysis by retraining the heart-rate decoding model using only the calcium traces from neurons assigned to the C1 kernel class, while preserving the same model architecture and cross-validation procedure used for the original C1+C2 model. Because C1 and C2 represent opposing kernel polarities in our current notation, we also trained a complementary C2-only model. These analyses are now included in the new Supplementary Fig. 4.
At 5, 7, and 12 dpf, both cluster-restricted models retained substantial cross-validated predictive performance. We detected no significant differences in cross-validated R2 between the C1-only, C2-only, and combined C1+C2 models at any age examined. These results indicate that heart-rate-related predictive information is present in both kernel classes and that either class alone captures much of the signal recovered by the combined model.
(2) Figure 3J: A magnified version of the figure with larger time scales would help visualize the relationships more clearly.
We thank the reviewer for this helpful suggestion. We have revised Fig. 3J by magnifying the time axis and increasing the temporal scale of the displayed traces as much as possible.
(3) Figure 3: It would be useful to show the performance of the model at 4 dpf as a negative control for the argument that vagal drive is the key regulator of heart rate in the early window.
We agree with the reviewer that 4 dpf provides an important negative-control stage. However, we did not train the heart-rate decoding GLM at 4 dpf because the slower heart-rate dynamics that define the neural relationships at later ages are not yet detectable at this stage. The cardiac recordings nevertheless contain beat-to-beat interval variability, but much of this variation occurs on millisecond timescales that cannot be meaningfully related to GCaMP6s signals. Thus, low decoding performance at 4 dpf would conflate the absence of the slower cardiac dynamics accessible to the model with the temporal bandwidth of calcium imaging, rather than provide an interpretable negative control for vagal coupling.
(4) Figure 3: C1 and C2 neurons seem to be organised rostrocaudally at 5 dpf. Since C2 neurons have also been shown to encode signatures of cardiac events, along with C1 neurons, optogenetic activation of the rostral and caudal areas of hindbrain cholinergic neurons at 5 dpf would provide additional insights into downstream regulation.
We agree with the reviewer that spatially restricted activation of different rostrocaudal cholinergic hindbrain regions would be an informative experiment. We attempted this approach at 5 dpf; however, the C2 neurons are sparse and are not defined by an independent molecular marker, making it difficult to selectively activate this functionally defined class without also stimulating neighboring cholinergic neurons. In practice, broader activation of these regions frequently evoked variable motor contractions or struggling, which introduced motion artifacts that confounded simultaneous heart-rate and calcium measurements. We therefore do not think these experiments provide an interpretable test of the specific role of C2 neurons and have not included them as evidence.
(5) For both the vagal efferent decoding model and the sympathetic efferent decoding model, could noise terms be introduced based on the intrinsic variability of firing at 4 dpf, where no coupling is presumed and presumably governed by nodal tissue? From the model, firing of anterior sympathetic ganglia and heart rate is seen to be coupled at 5 dpf, despite the lack of direct anatomical innervation. The response to the stress test provides corroborative evidence.
We thank the reviewer for this astute suggestion. The proposed control would be informative only if we could measure the same vagal and sympathetic neuronal populations at 4 dpf, before they become functionally coupled to the heart, and use their activity to define the expected performance of an uncoupled model. We cannot do this reliably. In the later recordings, the neurons included in the decoding models are identified by their location and by reproducible relationships between their calcium activity and heart-rate dynamics. At 4 dpf, the corresponding neurons cannot yet be assigned to these functional classes because the slower heart-rate fluctuations used to define the kernels are not detectable. The remaining differences between individual cardiac intervals occur on millisecond timescales and therefore cannot be related meaningfully to GCaMP6s signals. A noise term derived from 4 dpf would consequently come from an arbitrary set of neurons, not from an identifiable uncoupled version of the populations modeled at later stages.
We agree that the relationship between anterior paravertebral ganglion activity and heart rate at 5 dpf is especially informative because sympathetic fibers have not yet reached the heart. We interpret this as coordinated recruitment of the APG and the cardiac response by a broader stress-related program. Our stress-test and optogenetic experiments support this interpretation by showing that APG neurons are already responsive at 5 dpf, whereas direct sympathetic control emerges later, together with TH-positive cardiac innervation and optogenetically evoked tachycardia. We have clarified this distinction in the revised manuscript.
(6) In the discussion and in the summary diagram, the autorhythmicity of the SA node should be highlighted. The possibility of other contributory variables, like endocrine and paracrine signals and metabolic cues, should be discussed. From the model, firing of anterior sympathetic ganglia and heart rate is seen to be coupled at 5 dpf, despite the lack of anatomical innervation. However, the model performance is comparably lower. This could be potentially due to the effects of stress hormones, given that the hypothalamic-pituitary-interrenal axis is functional at 5 dpf. The response to the stress test provides corroborative evidence.
We thank the reviewer for this important suggestion. We have revised the manuscript and the Fig. 9 legend to state explicitly that the larval heart is intrinsically autorhythmic and that the neural pathways described in this study modulate an ongoing rhythm rather than generate each heartbeat. We also now acknowledge that cardiac dynamics may be influenced by endocrine, paracrine, metabolic, and other non-neuronal physiological signals that are not represented in our neural recordings.
As the reviewer notes, this distinction is particularly relevant to the 5 dpf sympathetic results. At this stage, APG neurons respond to visual stress and their activity covaries with heart-rate dynamics, yet TH-positive cardiac innervation is not detectable and optogenetic activation of the APG does not evoke tachycardia. We therefore interpret that APG activity and the cardiac response may be recruited in parallel as components of a broader stress-related physiological state. As the reviewer notes, endocrine stress pathways, including the hypothalamic–pituitary–interrenal axis, are plausible contributors at this stage, although our experiments do not directly test their involvement. We have clarified the broader stress-related interpretation in the revised manuscript.
(7) In the discussion section, a brief mention of cardiac interoception and its relevance to arrhythmias, heart failure, and myocardial infarction would strengthen the clinical perspective.
We thank the reviewer for this helpful suggestion. We have added a brief clinical perspective to the Discussion linking the developmental assembly of cardiac interoceptive pathways to arrhythmias, heart failure, and myocardial infarction. This perspective is particularly important because current neurocardiac therapies predominantly target cardiac tissue or the efferent mechanisms that regulate heart rate, contractility, and vascular tone, whereas the sensory pathways that communicate cardiac state to the brain are rarely targeted directly. Defining how these afferent circuits are organized may therefore reveal an underexplored point of intervention for maladaptive autonomic responses following cardiac injury or dysfunction.
(8) Line 31: Though the findings and the mode of encoding are novel, the idea that vagal sensory neurons might encode heart rate information has been suggested. The following papers (or others) would be useful to cite.
- Nosaka, S., Murase, S., Murata, K. and Inui, K., 1995. 'Aortic baroreceptor' neurons in the nucleus tractus solitarius in rats: convergence of cardiovascular inputs as revealed by heartbeat-locked activity. Journal of the autonomic nervous system, 55(1-2), pp.69-80.
- Hayward, L.F. and Felder, R.B., 1995. Cardiac rhythmicity among NTS neurons and its relationship to sympathetic outflow in rabbits. American Journal of Physiology-Heart and Circulatory Physiology, 269(3), pp.H923-H933.
We thank the reviewer for pointing us to these relevant studies. We agree that prior work has shown heartbeat-locked activity and cardiovascular convergence in central brainstem neurons, including neurons in the nucleus tractus solitarius. We have added these citations to the Discussion. We have also revised the text to more clearly distinguish those observations from the encoding described here. Heartbeat-locked activity reflects neural modulation at the timescale of individual cardiac cycles, whereas our analysis identifies vagal sensory neurons whose calcium dynamics track slower changes in heart rate over seconds. Thus, the novelty of our finding is not that cardiac signals can influence vagal or brainstem circuits, but that heart-rate dynamics are represented in vagal sensory neurons by a stimulus-invariant, phase-shifted temporal kernel across spontaneous, sensory-evoked, and optogenetically induced cardiac fluctuations.
Reviewer #2 (Recommendations for the authors):
(1) The introduction will benefit from a section on the current understanding of the development of the motor and sensory innervations of the heart and their functions.
We thank the reviewer for this helpful suggestion. We have revised the Introduction to include a brief overview of the current understanding of cardiac autonomic innervation during development. In this new section, we summarize that the heart is intrinsically autorhythmic before neural control is established, that parasympathetic/vagal, sympathetic, and sensory innervation mature over development, and that most prior functional work has focused on adult or late-stage preparations rather than the initial emergence of these circuits in intact animals. This addition provides context for why larval zebrafish offer a useful system for resolving when motor and sensory heart–brain circuits first become functional.
(2) Figure 1 B: Why is the dilated ventricle brighter and the contracted ventricle darker in the reflected IR image?
We thank the reviewer for asking us to clarify this point. The contrast in Fig. 1B arises from near-infrared transmission through the cardiac chamber. When the ventricle is dilated, the density along the optical path is lower and/or spread over a larger volume, allowing more NIR light to reach the detector; the chamber therefore appears brighter. During contraction, the ventricular contents are compressed into a smaller volume, increasing optical density along the imaging path and reducing transmitted light, so the chamber appears darker. We have clarified this point in the figure legend.
(3) Include a scale for time in Supplementary Figures C-F.
We have added a scale for time.
(4) Figure 1 I: Add a panel showing the number of ICNs at each developmental stage. Insufficient information to infer the distributions from panel I.
We have added a new panel to Supplementary Fig. 1 showing the distribution of reliably countable intracardiac neurons at 12 dpf. Because individual somata within the intracardiac ganglion are often densely clustered and overlapping, we did not assign exact ICG neuron counts across all animals. Instead, the new panel reports the number of countable ICNs outside the ICG, including neurons associated with the SAP and AVP, for each individual fish. This provides the distribution requested by the reviewer while avoiding false precision in cases where individual neurons within the ganglion could not be reliably resolved.
(5) Figure 2F: a) Not referenced in text. b) Same comment as above. Include a panel showing the labeled ICN numbers.
We thank the reviewer for pointing this out. We have now referenced Fig. 2F explicitly in the Results. We have also revised the figure legend to specify the number of distinguishable labeled ICNs observed in each fish. Because adjacent labeled somata can be difficult to separate unambiguously in this dataset, we report these values descriptively in the legend rather than as a separate quantitative panel.
(6) Figure 3 K, O: The labeling of C1 and C2 is inverted with respect to panel G and text.
We thank the reviewer for catching this labeling error. We have corrected the C1/C2 labels in Fig. 3K and Fig. 3O so that they are now consistent with Fig. 3G and the corresponding text.
(7) P14, line 36: Figure 4R should be Figure 5R.
We have corrected this.
(8) Include quantification of the effect observed in Figure 5 P-R and Supplementary Figure 5.
We thank the reviewer for this suggestion. We have added quantification of the APG optogenetic activation experiments shown in Fig. 5O–Q and Supplementary Fig. 7. Because cardiac imaging was interrupted during optogenetic stimulation, we quantified the post-stimulation response rather than the stimulation period itself. For each fish, heart rate was averaged during a 12 s pre-stimulation baseline window and a 12 s post-stimulation response window after excluding the stimulation artifact. Responses were normalized to baseline and averaged across trials within fish. The new quantification shows the developmental emergence of APG-evoked tachycardia and confirms that propranolol reduces the response.
(9) Figure 5 P-R, Supplementary Figure 5: Annotation showing the stimulus window occludes the traces.
We thank the reviewer for pointing out this potential source of confusion. The stimulus window annotation does not occlude recorded data. During optogenetic activation in these experiments, cardiac imaging could not be acquired concurrently because the stimulation light interfered with data acquisition. The gap in the trace therefore reflects the absence of recorded data during the optogenetic stimulation period, rather than graphical occlusion by the stimulus annotation. We have revised the figure legend and annotation to clarify that no concurrent heart-rate measurements were acquired during the stimulation window.
(10) Figure 5 P-R, Supplementary Figure 5: Annotation showing the stimulus window occludes the traces.
We thank the reviewer for pointing out this potential source of confusion. The stimulus window annotation does not occlude recorded data. During optogenetic activation in these experiments, cardiac imaging could not be acquired concurrently because the stimulation light interfered with data acquisition. The gap in the trace therefore reflects the absence of recorded data during the optogenetic stimulation period, rather than graphical occlusion by the stimulus annotation. We have revised the figure legend and annotation to clarify that no concurrent heart-rate measurements were acquired during the stimulation window.
(11) Could the authors comment on the large negative dF/F signals in Figs 5 and 7? Are these neurons expected to be tonically active?
We thank the reviewer for raising this point. The negative ΔF/F values in Figs. 5 and 7 reflect decreases in fluorescence relative to the baseline period used for normalization. We do not interpret these signals as direct evidence that the neurons are tonically active, because calcium fluorescence reports activity relative to the chosen baseline and does not by itself establish absolute firing rate or tonic spiking. Rather, these negative responses indicate that, during the relevant behavioral or cardiac epoch, fluorescence in some neurons fell below their pre-stimulus baseline.
-
eLife Assessment
In this important manuscript, the authors establish a vertebrate model for studying the development of circuits that control heart rate. This contribution uses a combination of experimental techniques to provide compelling information for scientists looking to understand how heart rate regulation emerges during development.
-
Reviewer #1 (Public review):
Summary:
The manuscript by Hernandez-Nunez et al. provides a comprehensive characterization of how heart-brain circuits develop in a vertebrate brain, namely the zebrafish. The characterization is performed using a combination of modern and sophisticated imaging and neural manipulation techniques and achieves unprecedented clarity and detail in how the heart-brain communication develops early in life. The paper describes a three-stage program, where first an efferent-circuit from the motor vagus to the heart develops, followed by sympathetic innervation, and lastly sensory neurons innervate the heart.
Strengths:
The paper is very clearly and nicely written. The findings are novel and of high quality and relevance. The presentations are very clear and nicely interpreted. The analyses are well presented and …
Reviewer #1 (Public review):
Summary:
The manuscript by Hernandez-Nunez et al. provides a comprehensive characterization of how heart-brain circuits develop in a vertebrate brain, namely the zebrafish. The characterization is performed using a combination of modern and sophisticated imaging and neural manipulation techniques and achieves unprecedented clarity and detail in how the heart-brain communication develops early in life. The paper describes a three-stage program, where first an efferent-circuit from the motor vagus to the heart develops, followed by sympathetic innervation, and lastly sensory neurons innervate the heart.
Strengths:
The paper is very clearly and nicely written. The findings are novel and of high quality and relevance. The presentations are very clear and nicely interpreted. The analyses are well presented and applied.
Weaknesses:
From the heart rate traces, heart rate variability seems to be prominent and changes across days post-fertilization (dpf). That would be a useful dependent variable, considering that the variation captured by the models does not fully explain heart rate, both for sympathetic and parasympathetic efferents. Given the strong autorhythmicity of nodal tissue in neurogenic hearts, modulatory inputs could potentially predict heart rate variability with higher precision.
-
Reviewer #2 (Public review):
Hernandez-Nunez et al. investigate the development and function of neural circuits involved in the regulation of heart rate in larval zebrafish. Using conserved genetic markers, they identify neural pathways involved in the bidirectional control of heart rate and in providing sensory feedback, potentially enabling more precise tuning. The main observation is that the different elements of this circuit are laid down in a developmentally staggered manner.
At 4 days old, the heart rate is invariant to a range of sensory stimuli, and the vagal motor or sympathetic pathways could not be seen to innervate the heart. Progressively through development, the heart is first innervated by the vagal motor pathway, whose axons are cholinergic, before the formation of phox2bb+ intracardiac neurons (ICNs). At this stage, …
Reviewer #2 (Public review):
Hernandez-Nunez et al. investigate the development and function of neural circuits involved in the regulation of heart rate in larval zebrafish. Using conserved genetic markers, they identify neural pathways involved in the bidirectional control of heart rate and in providing sensory feedback, potentially enabling more precise tuning. The main observation is that the different elements of this circuit are laid down in a developmentally staggered manner.
At 4 days old, the heart rate is invariant to a range of sensory stimuli, and the vagal motor or sympathetic pathways could not be seen to innervate the heart. Progressively through development, the heart is first innervated by the vagal motor pathway, whose axons are cholinergic, before the formation of phox2bb+ intracardiac neurons (ICNs). At this stage, before the first ICNs are observed, activation of the vagal motor pathway by optogenetic activation of a localized population of cholinergic hindbrain neurons leads to bradycardia. After the vagal motor innervation begins, the sympathetic pathway innervates the heart, which could be visualized in the form of TH+ fibers from the anterior paravertebral ganglia (APG). The activity of the TH+ APG neurons was diverse and showed proportional, integral, and derivative-like relationships to the heart rate, suggesting a role in more precise tuning of the rate than what could be achieved through the vagal pathway alone. The sensory vagus innervation of the heart was identified to be the last stage to develop; however, neurons in the nodose ganglion exhibited diverse responses tuned to the heart rate well before the innervation reached the heart. The authors attribute this to the fact that other indirect sensory cues from the gills or vasculature could be used to sense heart rate prior to innervation.
This study identifies key components of the control loop required for the regulation of heart rate in zebrafish. The control mechanism appears to be independent of the cues that trigger heart rate changes, indicating that the circuit is indeed part of an interoceptive pathway for heart rate control. Evidence for the staggered development of the vagal-motor, sympathetic, and sensory pathways is conclusive, and as the authors discuss, this phenomenon progressively allows for finer-grained control of the heart rate. This could be achieved through proportional-integral-derivative-like control properties emerging in a diverse set of neurons in the APG and sensory feedback of the state of the heart. In line with these findings, the baseline variability of heart rate prior to innervation at 4 days old appears to be comparatively lower than the later stages (Figure 1C, D, Supplementary Figure 1C-F) and increases over development.
Based on this observation and the time courses of the kernels identified by the GLMs, I would expect heart rate fluctuations of a finer time scale, ultimately limited by the time course of GCaMP6s, to be captured by the models in Figures 3, 5, and 7, in addition to the stimulus-locked changes that are highlighted. While the models yield valuable insight in the form of the activation kernels and their potential roles, in one instance, this captures the potential contribution of either the motor vagus or the APG to the change in heart rate. This makes it challenging to identify where it falls short and the potential functions of pathways that are yet to be discovered.
Lastly, the proposed anatomical connectivity of the heart-brain circuit is based on tracts observed in this study as well as those inferred from function and from previous studies.
(1) It is not clear from the images presented here whether the VSNs send feedback projections to the brainstem VPN.
(2) Do the brainstem neurons identified by their functional roles send efferent projections via the motor vagus nerve? This is unclear from the results presented and needs to be clarified in the text.
(3) Add appropriate clarifying annotations to Figure 9 and a section of text discussing the potential unknowns in the proposed circuit diagram.
-
Author response:
We thank the reviewers for their thoughtful, constructive, and generous evaluations of our manuscript. We are encouraged by their overall assessment of the clarity, novelty, and significance of the work, and we appreciate the opportunity to further strengthen the manuscript.
Both reviewers highlight the central contribution of this study: a developmental, circuitlevel dissection of how heart–brain signaling emerges in a vertebrate. We are pleased that the evidence supporting the staggered assembly of vagal motor, sympathetic, and sensory pathways was found to be compelling, and that the computational and experimental framework was viewed as appropriate and informative.
Below, we briefly outline how we plan to address the main points raised in the reviews.
Heart rate variability and temporal structure
Both reviewers note …
Author response:
We thank the reviewers for their thoughtful, constructive, and generous evaluations of our manuscript. We are encouraged by their overall assessment of the clarity, novelty, and significance of the work, and we appreciate the opportunity to further strengthen the manuscript.
Both reviewers highlight the central contribution of this study: a developmental, circuitlevel dissection of how heart–brain signaling emerges in a vertebrate. We are pleased that the evidence supporting the staggered assembly of vagal motor, sympathetic, and sensory pathways was found to be compelling, and that the computational and experimental framework was viewed as appropriate and informative.
Below, we briefly outline how we plan to address the main points raised in the reviews.
Heart rate variability and temporal structure
Both reviewers note that heart rate variability (HRV) changes across development and suggest that HRV may provide additional insight into the function of autonomic circuits. We agree that HRV is an important physiological readout and that its developmental changes are consistent with the progressive emergence of autonomic control.
In the revised manuscript, we plan to (i) discuss heart rate variability more explicitly in the context of circuit maturation and (ii) clarify the temporal scales captured by our experiments and modeling framework. In particular, we will emphasize that our analyses focus on relationships between neural activity and heart-rate trajectories at timescales accessible given imaging rate and indicator kinetics, rather than beat-to-beat variability. We will also consider adding a supplementary analysis of the variability that can be reliably measured within these constraints, and, where appropriate, how neural activity predicts that measurable variation.
Scope and interpretation of the computational models
Reviewer #2 raises thoughtful points regarding what the generalized linear models can and cannot disambiguate, particularly when multiple efferent pathways may contribute to heart-rate dynamics. We will revise the text to more clearly distinguish between functional encoding relationships inferred from the models and anatomical connectivity that is directly demonstrated.
Our intent is to frame the kernels identified in the motor and sympathetic pathways as computational motifs that capture distinct dynamical contributions, rather than as exclusive or complete explanations of heart-rate control. We will clarify these limitations explicitly in the Results and Discussion.
Circuit diagram and anatomical interpretation
We appreciate the reviewer’s careful reading of the proposed circuit schematic. In the revised manuscript, we will revise the figure and accompanying text to clearly annotate which connections are directly observed, which are functionally inferred, and which remain hypothetical. We will also expand the Discussion to explicitly address open questions, including unresolved feedback pathways and the potential for additional nodes in the circuit.
We believe these revisions will improve clarity without altering the core conclusions of the study. We thank the reviewers again for their insightful feedback and look forward to submitting a revised version of the manuscript that addresses these points in detail.