High-resolution volumetric imaging constrains compartmental models to explore synaptic integration and temporal processing by cochlear nucleus globular bushy cells
Curation statements for this article:-
Curated by eLife
eLife assessment
This manuscript provides a structural analysis of bushy cells in the mouse cochlear nucleus. These neurons receive a large synaptic contact from the auditory nerve termed an endbulb that preserves the temporal information present in the auditory nerve and are key elements of binaural sound localization circuits. The analysis combines volume electron microscopy techniques with computational models to predict heterogeneous bushy cell responses. The analysis takes morphological analysis of bushy cells to a new level, and the modeling is well done.
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
- Neuroscience (eLife)
Abstract
Globular bushy cells (GBCs) of the cochlear nucleus play central roles in the temporal processing of sound. Despite investigation over many decades, fundamental questions remain about their dendrite structure, afferent innervation, and integration of synaptic inputs. Here, we use volume electron microscopy (EM) of the mouse cochlear nucleus to construct synaptic maps that precisely specify convergence ratios and synaptic weights for auditory nerve innervation and accurate surface areas of all postsynaptic compartments. Detailed biophysically based compartmental models can help develop hypotheses regarding how GBCs integrate inputs to yield their recorded responses to sound. We established a pipeline to export a precise reconstruction of auditory nerve axons and their endbulb terminals together with high-resolution dendrite, soma, and axon reconstructions into biophysically detailed compartmental models that could be activated by a standard cochlear transduction model. With these constraints, the models predict auditory nerve input profiles whereby all endbulbs onto a GBC are subthreshold (coincidence detection mode), or one or two inputs are suprathreshold (mixed mode). The models also predict the relative importance of dendrite geometry, soma size, and axon initial segment length in setting action potential threshold and generating heterogeneity in sound-evoked responses, and thereby propose mechanisms by which GBCs may homeostatically adjust their excitability. Volume EM also reveals new dendritic structures and dendrites that lack innervation. This framework defines a pathway from subcellular morphology to synaptic connectivity, and facilitates investigation into the roles of specific cellular features in sound encoding. We also clarify the need for new experimental measurements to provide missing cellular parameters, and predict responses to sound for further in vivo studies, thereby serving as a template for investigation of other neuron classes.
Article activity feed
-
Author Response
Reviewer #1 Public Review:
- “…The authors make reasonable assertions, but all of these need to be validated by electrophysiological studies before they can be treated as fact. Instead, they should be treated as predictions. For example, in the conclusions from the model section, that endbulb size does not strictly predict synaptic efficacy should be modified from an assertion to a prediction.”
The reviewer makes an important point. We realize that, despite describing the data as the output of a model, we needed to be clearer that the model output is in fact a set of predictions to be tested experimentally. In the reorganization of the results, we collect the model output explicitly in a section named “Model Predictions”, and list five classes of predictions that describe explorations of bushy cells. The fifth set of …
Author Response
Reviewer #1 Public Review:
- “…The authors make reasonable assertions, but all of these need to be validated by electrophysiological studies before they can be treated as fact. Instead, they should be treated as predictions. For example, in the conclusions from the model section, that endbulb size does not strictly predict synaptic efficacy should be modified from an assertion to a prediction.”
The reviewer makes an important point. We realize that, despite describing the data as the output of a model, we needed to be clearer that the model output is in fact a set of predictions to be tested experimentally. In the reorganization of the results, we collect the model output explicitly in a section named “Model Predictions”, and list five classes of predictions that describe explorations of bushy cells. The fifth set of predictions was previously a separate section but should now be better appreciated as conveying hypotheses since it is incorporated into this newly named section. Please note that the hypotheses are constrained to varying extents by the high-resolution structural data we present, such as the estimation of synaptic weights from the counts of synapses. The compartmental models for each bushy cell also are constrained by the structural data and published biophysical and electrophysiological properties of the cells. The pipeline to create the models is described in its own section now using that terminology: “A pipeline for translating high-resolution neuron segmentation into compartmental models consistent with in vitro and in vivo data.”, which we hope conveys the notion that the modeling framework is indeed a template that can be applied to future experimental data. We explicitly make this latter point in the new Discussion section “Toward a complete computational model for globular bushy cells: strengths and limitations”.
Reviewer #2 Public Review:
- …” While this is technically impressive (in regards to both the structure and modelling) there are significant weaknesses because this integration makes massive assumptions and lacks a means of validation; for example, by checking that the results of the structural modelling recapitulate the single-cell physiology of the neuron(s) under study. This would require the integration of in vivo recorded data, which would not be possible (unless combined with a third high throughput method such as calcium imaging) and is well beyond the present study.
We appreciate the support for our approach, and we now make explicit in the manuscript that the output of the models should be interpreted as predictions for eventual experimental testing. We also consider in the Discussion some experimental procedures that might be used to test the predictions. Ca2+ imaging is currently too slow a reporter for the rapid synaptic events and integration time constant for bushy cells, as the reviewer knows, and we think (and present in the Discussion, section 2) that focal optical stimulation simultaneous with recording from fast voltage sensors are potential avenues to achieve this goal.
- The authors need to be more open about the limitations of their observations and their interpretations and focus on the key conclusions that they can glean from this impressive data set.
As indicated in response to a similar comment from Reviewer 1, we have collected and discuss the primary limitations in a new section within the Discussion, entitled “Toward a complete computational model for globular bushy cells: strengths and limitations”.
- The manuscript would be considerably improved by re-writing to focus the science on the most important results and provide clear declarations of limitations in interpretation.
We have extensively re-organized and re-written the text to highlight the key structural observations (Figures 1-3, 7-8), the pipeline from structure to model (Figure 4) and interleave structural observations with the outputs of the model (Figures 5-6, 8). The latter are explicitly detailed in a new section called “Model Predictions”. These predictions are organized into five classes. We think that this new organization will improve communication of the key results, and further highlights the key discoveries from structural analysis and predicted functional mechanisms as explored in the compartmental models.
Reviewer #3 Public Review:
- The authors extract here from the longer introductory commentary a one-sentence summary of the strengths of the manuscript, and thereafter focus on the weaknesses, since this document emphasizes our response to those critiques. To quote reviewer #3: “The strengths of this paper are that the authors obtained unprecedented high-resolution 3-D images of the AN-bushy cell circuit, and they implemented a biophysical model to simulate the neural processing of AN inputs based on these structural data. … The biophysical modeling, although lacking comparison with in vivo physiological data due to the chosen species (mice), is also solid and well documented.”
We appreciate that the reviewer acknowledges the attention to detail that entered into the nanoscale imaging, cell reconstructions, building the modeling pipeline and constructing the compartmental models.
- Despite the high quality of the data, the paper is marred by the species they chose: there are very few published in vivo single-unit results from mouse bushy cells, so it is hard to evaluate how well the model predictions fit the real-world data, and how the structural findings address the “fundamental questions” in physiology. … No rationale (e.g. use of molecular tools or in vitro physiology) is given why the authors focus on the mouse. It seems that the analyses provided here could as well have done on a species with good low-frequency hearing, which may have provided a much more interesting case for understanding the spectacular temporal transformation performed by bushy cells.
We now report our reasons, in the first paragraph of the Results, for selecting the mouse. One reason for choosing mouse was that biophysical properties of bushy cells, which were important parameters to constrain the compartmental models, were collected from mice. These data are collected from dissociated cells and from brain slices, and these experiments continue to be more tractable in mice. The second reason is that mice are used in nanoscale and light microscopy connectomic studies because their neurons, cell groups and entire brain are smaller, so that a given volume of imaged brain will contain more cellular elements. These other connectomic studies provide a template for eventual comparisons among brain regions. Our overall goal is to image the entire cochlear nucleus, and the size of the mouse brain makes this goal tractable given current technology. Indeed, we are currently analyzing an image volume of the more rostral ventral cochlear nucleus that is about 5x larger than this image volume and collected with a much better signal to noise ratio. The third reason for choosing mouse was so that the current project could be augmented by genetic tools to further classify cochlear nucleus (CN) neurons and their extrinsic inputs, and potentially manipulate neural circuits in future studies. For example, the atoh7 (math5) and hhip gene products are markers for subsets of bushy cells, suggesting the presence of molecular subtypes of this cell class (Jing et al. 2023).
- If we look at data from other animals such as cats and gerbils, it is true that high-frequency (globular) bushy cells show envelope phase locking, but compared to ANs they are at best only moderately enhanced (gerbils: Frisina et al. 1990: Fig 7 and 10; cats: Joris and Yin 1998 Fig 4); the most prominent enhancement is actually to the temporal fine structures of low-frequency bushy cells (cells tuned to < 1 kHz), which mice lack. Furthermore, the temporal modulation transfer function (tMTF, i.e. the vector strengths vs modulation frequency plots in Fig 7O of the paper) of (globular) bushy cells are mostly low-pass filtered, with a cutoff frequency close to 1 kHz, and the highest vector strength rarely surpasses 0.9 (cats: Rhode 1994 Fig 9, 16, Rhode 2008 Fig 8G, Joris and Yin 1998 Fig 7; and there's one report from mice: Kopp-Scheinpflug et al 2003 Fig 8). Thus, the band-pass tMTFs tuned to 100-200 Hz with vector strengths > 0.9 or 0.95 in this paper (Fig 7O, Fig 8M) do not really match known physiology (in non-mouse species). Again, we know very little about in vivo physiology of mouse (globular) bushy cells and there is of course a possibility that responses in mice may be closer to the predictions of this paper.
We agree that there are (unfortunately) few studies in mouse that can be compared with our simulations. With regard to the tMTFs, we can make a couple of points. First, we note that the stimulus used for all the panels except P2 in Figure 6 (previous Figure 7) were at 15 dB SPL, which is the level where maximal envelope phase-locking occurs in the low-threshold ANF inputs. This choice was based on previous experimental work that examined the intensity dependence for SAM stimuli in the auditory nerve (Smith and Brachman, 1980; Joris and Yin, 1992; Cooper et al, 1993; Dreyer and Delgutte, 2006, Figure 2B, Figure 3). Second, Figure 6, Supplemental Figure 1 confirms the behavior of the auditory nerve model used for input to the bushy cells (Rudnicki and Hemmert (2017) implementation), replicating Zilany et al., 2009, Figure 13D. These results show that phase-locking decreases at higher intensities as expected from the experimental work. Relevant to this topic, the lone report of responses to SAM stimuli in mice (Kopp-Scheinpflug et al. 2003) used 100% SAM at CF at 80 dB SPL. At this high intensity, it is expected that the envelope phase locking at CF will be less than at lower intensities because of rate saturation in the high and medium spontaneous rate ANFs (Carney, JARO 2019; Joris and Yin, 1998). In guinea pig, envelope phase locking is greater in low-SR fibers at 80 dB SPL than in medium and high SR fibers, but it is still lower than at its peak at about 50 dB SPL (Cooper et al., 1993). All of these experimental observations therefore lead to the prediction that the SAM envelope locking in Kopp-Scheinpflug et al. (2003) should be lower than in our simulations.
In addition, Kopp-Scheinpflug et al. (2003) did not report which VCN cell populations cells were recorded. If the recorded cells were a heterogenous mixture of bushy and multipolar cells, then their data are not directly comparable to our model predictions. The stimulus intensity also needs to be considered for comparison with the work of Rhode (1994), whose lowest stimulus level is 30 dB SPL (Figure 9), and who also used a different stimulus, 200% SAM, and with the work of Frisina et al. (1990), who used 50 dB SPL. Interestingly, Figure 14D in Rhode (1994) shows a synchrony coefficient ranging from 0.5 to 0.9 at 30 dB SPL at 300 Hz modulation, which is similar to what we predict in Figure 6P2. We also remind the reviewer that our simulations did not include the effects of feed-back inhibition at CF (Caspary and Palombi, 1994; Campagnola and Manis, 2014; Xie and Manis, 2014, Keine et al. eLife 2016), which may affect phase synchrony in complex ways (Gai and Carney, 2008). One important feedback pathways arises from the tuberculoventral cells of the DCN (Wickesberg and Oertel, 1991; Campagnola and Manis, 2014), but the envelope synchrony behavior of those cells is not known.
Thus, we now emphasize in the revised manuscript (in the Discussion) considerations of stimulus intensity used across published studies, citing the works above, the relatively high vector strengths at low modulation frequency, and that these simulation results are currently predictive. The simulations are also limited in that we used only one configuration of ANF inputs (low-threshold, high SR). This ANF SR category was selected to be consistent with the suggestion by Liberman (1991) that the globular BCs receive input principally from the low-threshold high-SR fibers. Mixtures of input SR classes would be expected to change the envelope representation at higher intensities. Finally, the parameter space is quite large (intensity x frequency x [ANF distributions], x inhibition) and is better explored in a separate study once we are able to provide better or additional constraints to the modeling framework. Also, to put the selection of SAM stimuli in context, we indicate that mice can encode temporal fine structure although only as low at 1 kHz, but at similar VS to larger rodents such as guinea pig (Taberner and Liberman 2005; Palmer and Russell 1986).
Reviewer 4: Public comments
- The authors have collected an impressive array of physiological data and provided some beautiful 3D images of SBCs with dendrites. These are clearly strengths. The computational models for mechanisms of SBC responses, however, are made to fit what may be inadequate anatomical data. Instead of conclusions, perhaps they need to reword their discussions to refer to the anatomy as hypothetical substrates.
It is true that the SBEM image volumes have strengths and limitations. We now collect these considerations in the second section of the Discussion, “Toward a complete computational model for globular bushy cells: strengths and limitations”. One limitation of this volume is that we do not have sufficient resolution to categorize synaptic vesicles by shape and must infer their excitatory or inhibitory nature. Note that tracing inputs to a source neuron, such as tracing the endbulbs to parent auditory nerve fibers, solves this problem, but the smaller terminals remain problematic in this regard. The goal is to not only assign excitatory or inhibitory phenotype, but also a cell type of origin, so that actual spike patterns, evoked by sound, can be provided as inputs to the model. The compartmental model is detailed, and amenable to mapping this information from other experiments as it becomes available. Nanoscale imaging does provide detailed structural information in terms of surface areas, volumes and process diameters that is important in constraining the compartmental models, and that is not attainable by standard light microscopy approaches. These points are now made in the Results and in the Discussion, as mentioned earlier in this paragraph. And, as indicated in the responses to other reviewers, we highlight the model outputs as predictions to be tested experimentally.
-
-
eLife assessment
This manuscript provides a structural analysis of bushy cells in the mouse cochlear nucleus. These neurons receive a large synaptic contact from the auditory nerve termed an endbulb that preserves the temporal information present in the auditory nerve and are key elements of binaural sound localization circuits. The analysis combines volume electron microscopy techniques with computational models to predict heterogeneous bushy cell responses. The analysis takes morphological analysis of bushy cells to a new level, and the modeling is well done.
-
Reviewer #1 (Public Review):
This manuscript provides a structural analysis of bushy cells in the mouse cochlear nucleus. The analysis uses volume electron microscopy techniques to describe bushy cell-auditory nerve synapses and bushy cell dendrites. The analysis takes a morphological analysis of bushy cells to a new level, and the computational modeling is well done. The models are used to predict busy cell behavior, which leads to a major concern. The authors make reasonable assertions, but all of these need to be validated by electrophysiological studies before they can be treated as fact. Instead, they should be treated as predictions. For example, in the conclusions from the model section, that endbulb size does not strictly predict synaptic efficacy should be modified from an assertion to a prediction.
-
Reviewer #2 (Public Review):
This is an interesting manuscript in which the authors demonstrate the power of serial section reconstruction at the EM level of a volume within the anterior ventral cochlear nucleus (aVCN) containing bushy cells and their large afferent synapses - the endbulbs of Held. Integration of this information with compartmental modelling of the neuronal excitability is then used to make observations about the form and function of these neurons and their synaptic inputs. While this is technically impressive (in regards to both the structure and modelling) there are significant weaknesses because this integration makes massive assumptions and lacks a means of validation; for example, by checking that the results of the structural modelling recapitulate the single-cell physiology of the neuron(s) under study. This …
Reviewer #2 (Public Review):
This is an interesting manuscript in which the authors demonstrate the power of serial section reconstruction at the EM level of a volume within the anterior ventral cochlear nucleus (aVCN) containing bushy cells and their large afferent synapses - the endbulbs of Held. Integration of this information with compartmental modelling of the neuronal excitability is then used to make observations about the form and function of these neurons and their synaptic inputs. While this is technically impressive (in regards to both the structure and modelling) there are significant weaknesses because this integration makes massive assumptions and lacks a means of validation; for example, by checking that the results of the structural modelling recapitulate the single-cell physiology of the neuron(s) under study. This would require the integration of in vivo recorded data, which would not be possible (unless combined with a third high throughput method such as calcium imaging) and is well beyond the present study. The authors need to be more open about the limitations of their observations and their interpretations and focus on the key conclusions that they can glean from this impressive data set. The manuscript would be considerably improved by re-writing to focus the science on the most important results and provide clear declarations of limitations in interpretation.
-
Reviewer #3 (Public Review):
Bushy cells are one of the principal neurons in the cochlear nucleus that provide temporal information to higher auditory nuclei to compare sound signals from both ears. One prominent feature in the auditory processing of bushy cells is that they show enhanced temporal responses compared to the auditory nerve (AN) inputs, thus providing a better temporal representation of the acoustic signals. Another feature of the AN-bushy cell circuit is that AN fibers form large synapses termed "endbulbs" around the soma of bushy cells. Scientists have proposed that the temporal enhancement can be due to the coincidence detection of subthreshold convergent AN inputs, or a first-spike latency-based detection of convergent supra-threshold inputs. However, testing these hypotheses requires knowledge of the detailed …
Reviewer #3 (Public Review):
Bushy cells are one of the principal neurons in the cochlear nucleus that provide temporal information to higher auditory nuclei to compare sound signals from both ears. One prominent feature in the auditory processing of bushy cells is that they show enhanced temporal responses compared to the auditory nerve (AN) inputs, thus providing a better temporal representation of the acoustic signals. Another feature of the AN-bushy cell circuit is that AN fibers form large synapses termed "endbulbs" around the soma of bushy cells. Scientists have proposed that the temporal enhancement can be due to the coincidence detection of subthreshold convergent AN inputs, or a first-spike latency-based detection of convergent supra-threshold inputs. However, testing these hypotheses requires knowledge of the detailed anatomical arrangement of the AN inputs onto bushy cells. This paper provides a first look at the 3-D organizations of the pre- and postsynaptic structures of mouse bushy cells at a nanoscale resolution. Furthermore, the authors create a morphology-constrained biophysical model to examine how these structures may affect synaptic integration and auditory processing.
The main finding of the paper is that the authors found two input motifs in the AN-bushy cell circuit: one with all small, subthreshold endbulb inputs (all < 180 um2), and the other with 1-2 large, suprathreshold endbulb inputs (> 180 um2) plus other smaller endbulb inputs. Using modeling, the authors argue that the former group correlates with a physiological phenotype of "coincidence detection", and the latter correlates with a phenotype termed "mixed-mode detection". "Coincidence detection" cells require the coincident firing of many subthreshold presynaptic inputs to evoke a postsynaptic spike; "mixed-mode" cells can either have postsynaptic spikes evoked by the largest input(s) alone, or by the coincident firing of small (plus large) inputs. Interestingly, the authors found that even though the large inputs alone can trigger spikes in the "mixed mode" cells, smaller inputs can further enhance the temporal precision of the spikes. The structural data are of very high quality and clearly show the endbulb inputs comprise various sizes. Whether these inputs are really supra/sub-threshold remains to be explored physiologically, but nevertheless, the model provides a hypothesis for the functional roles of the endbulb of different sizes.
In addition to the finding of "two convergent motifs', the authors also report a first complete map of synaptic inputs to a single bushy cell, and structures that have not been observed before, such as synapses at axon-hillock and axon initial segment, dendritic "hub", "braided" dendrites, non-innervated dendrites, etc. These data, like the previous "two input motifs" observation, are also of very high quality and can be useful resources for the ultrastructural study of the bushy cells.
Strengths:
The strengths of this paper are that the authors obtained unprecedented high-resolution 3-D images of the AN-bushy cell circuit, and they implemented a biophysical model to simulate the neural processing of AN inputs based on these structural data. The 3-D reconstruction of the pre- (input organization) and post- (dendrites and axons) synaptic structures of bushy cells are of high quality, as exemplified by the high-resolution figures and animations. The biophysical modeling, although lacking comparison with in vivo physiological data due to the chosen species (mice), is also solid and well documented. The combination of high-resolution imaging and structure-based modeling, together with the detailed documentation, provides rich information for not only auditory scientists but non-auditory scientists who want to use similar techniques to explore neural circuits.Weaknesses:
Despite the high quality of the data, the paper is marred by the species they chose: there are very few published in vivo single-unit results from mouse bushy cells, so it is hard to evaluate how well the model predictions fit the real-world data, and how the structural findings address the "fundamental questions" in physiology. If we look at data from other animals such as cats and gerbils, it is true that high-frequency (globular) bushy cells show envelope phase locking, but compared to ANs they are at best only moderately enhanced (gerbils: Frisina et al. 1990: Fig 7 and 10; cats: Joris and Yin 1998 Fig 4); the most prominent enhancement is actually to the temporal fine structures of low-frequency bushy cells (cells tuned to < 1 kHz), which mice lack. Furthermore, the temporal modulation transfer function (tMTF, i.e. the vector strengths vs modulation frequency plots in Fig 7O of the paper) of (globular) bushy cells are mostly low-pass filtered, with a cutoff frequency close to 1 kHz, and the highest vector strength rarely surpasses 0.9 (cats: Rhode 1994 Fig 9, 16, Rhode 2008 Fig 8G, Joris and Yin 1998 Fig 7; and there's one report from mice: Kopp-Scheinpflug et al 2003 Fig 8). Thus, the band-pass tMTFs tuned to 100-200 Hz with vector strengths > 0.9 or 0.95 in this paper (Fig 7O, Fig 8M) do not really match known physiology (in non-mouse species). Again, we know very little about in vivo physiology of mouse (globular) bushy cells and there is of course a possibility that responses in mice may be closer to the predictions of this paper. No rationale (e.g. use of molecular tools or in vitro physiology) is given why the authors focus on the mouse. It seems that the analyses provided here could as well have done on a species with good low-frequency hearing, which may have provided a much more interesting case for understanding the spectacular temporal transformation performed by bushy cells. -
Reviewer #4 (Public Review):
The authors have collected an impressive array of physiological data and provided some beautiful 3D images of SBCs with dendrites. These are clearly strengths. The computational models for mechanisms of SBC responses, however, are made to fit what may be inadequate anatomical data. Instead of conclusions, perhaps they need to reword their discussions to refer to the anatomy as hypothetical substrates.
-
-
-