High-content synaptic phenotyping in human cellular models reveals a role for BET proteins in synapse assembly
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
Berryer et al. report on an automated and quantitative platform to study the number of synaptic inputs formed in networks of human excitatory neurons and astrocytes in vitro. The authors tested the utility of the platform by screening a large collection of small molecules and identified several modulators of synapse density, which were validated in follow-up experiments. The automated platform substantially extends what is currently available, particularly with respect to the automation of the initial analysis steps. The positive hits identified here, the inhibitors of bromodomain and extraterminal (BET) family of gene expression regulators, are important, and will likely contribute to the understanding of the mechanisms of human synapse assembly.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #2 and Reviewer #3 agreed to share their name with the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
- Neuroscience (eLife)
Abstract
Resolving fundamental molecular and functional processes underlying human synaptic development is crucial for understanding normal brain function as well as dysfunction in disease. Based upon increasing evidence of species-divergent features of brain cell types, coupled with emerging studies of complex human disease genetics, we developed the first automated and quantitative high-content synaptic phenotyping platform using human neurons and astrocytes. To establish the robustness of our platform, we screened the effects of 376 small molecules on presynaptic density, neurite outgrowth, and cell viability, validating six small molecules that specifically enhanced human presynaptic density in vitro. Astrocytes were essential for mediating the effects of all six small molecules, underscoring the relevance of non-cell-autonomous factors in synapse assembly and their importance in synaptic screening applications. Bromodomain and extraterminal (BET) inhibitors emerged as the most prominent hit class and global transcriptional analyses using multiple BET inhibitors confirmed upregulation of synaptic gene expression. Through these analyses, we demonstrate the robustness of our automated screening platform for identifying potent synaptic modulators, which can be further leveraged for scaled analyses of human synaptic mechanisms and drug discovery efforts.
Article activity feed
-
-
Author Response
Reviewer #2 (Public Review):
In this manuscript, Berryer et al describe a fully automated, scalable approach to quantify the number of synaptic inputs formed onto human iPSC-derived neurons (hNs) in 2D culture. They validate the sensitivity of their approach by synapsin1 knock-down and test almost 400 small molecules for their effect on synapses, and the role of astrocytes. They identify BET inhibitors as strong modifiers of synapse numbers in hNs and performed follow-up experiments to confirm the finding, characterize the effect further and demonstrate the critical role of astrocytes.
Every step of the protocol is automated to achieve high reproducibility and homogeneity throughout the experiments. This automated approach has great potential for scaling up drug screening, genetic perturbations, and disease modeling …
Author Response
Reviewer #2 (Public Review):
In this manuscript, Berryer et al describe a fully automated, scalable approach to quantify the number of synaptic inputs formed onto human iPSC-derived neurons (hNs) in 2D culture. They validate the sensitivity of their approach by synapsin1 knock-down and test almost 400 small molecules for their effect on synapses, and the role of astrocytes. They identify BET inhibitors as strong modifiers of synapse numbers in hNs and performed follow-up experiments to confirm the finding, characterize the effect further and demonstrate the critical role of astrocytes.
Every step of the protocol is automated to achieve high reproducibility and homogeneity throughout the experiments. This automated approach has great potential for scaling up drug screening, genetic perturbations, and disease modeling experiments related to synapses.
The authors successfully identified, in two independent hNs lines, three small-molecule inhibitors of transcription modifiers of the BET family as the strongest positive modifiers of synaptic inputs. The initial study performed with immunofluorescence was then validated by Western blot analysis and mRNA-seq analysis, which showed an increase in the expression of trans-synaptic signaling genes.
While accessing the molecular mechanisms of BET inhibitors, the authors observed that the increased synaptic inputs occurred only in cocultures of astrocytes and neurons, and not in hNs monoculture. Finally, the authors report that the presence of astrocytes alone is a major driving force to promote synaptic inputs.
Overall, the experiments are well conducted, and the conclusions are supported by the data. The new approach reaches beyond the current state of the field, especially in the first steps of automation and the identified modulators (BET inhibitors) are interesting and novel, and the subsequent validation is convincing.
On the other hand, the manuscript does not yet define the exact resolution and power of the new methods, and does not convincingly show that the observed synapsin-puncta are synapses and that the data of the validation experiments can be improved.
MAJOR POINTS:
- Although the manuscript contains a lot of quantitative data on variance, the current manuscript stops short of an exact definition of the resolution of the assay and its statistical power. With the real (measured) variance of the assay, the power to detect certain effects can be computed. To be relevant for other applications than the current (e.g. genetic perturbations and disease modelling), it is relevant to define this for smaller effects too: can this assay detect a 25% effect with reasonable numbers of observations? Such assessments can also provide important recommendations on when it makes sense to add more repeated measures of the same specimens (wells, ROIs) and when more independent inductions are required (and how much this adds to overall power). The manuscript would also benefit from a short discussion on how to optimize future study designs (repeated measures, independent inductions, number of subjects).
As mentioned above, we have now calculated Cohen’s d for: (1) the primary screen overall as well as for compound included in the primary screen, (2) validation experiments performed in neuron monocultures and (3) validation experiments performed in neuron + astrocyte co-cultures, and these data have been added to Figure 5, Figure 5-figure supplement 1 and Supplementary File 2. For the validation experiments, we have also added a discussion of study design, given the observed effect sizes. These analyses are discussed in depth on pages 19-20 of the Results section and page 26 of the Discussion section in the PDF. In brief, we obtained a Cohen’s d of -0.18 for the primary screen where individual small molecules increased as well as decreased synaptic density. Also from the primary screen, we obtained a Cohen’s d of 2.914 for JQ1 and 3.710 for I-BET151, indicating large effects for the BET inhibitors. We also noted large effects for BET inhibitors in the co-culture validation experiments, where we could have scaled down on the number of fields and wells analyzed. While we were reasonably powered to detect changes in the monoculture validation experiments, here, effect sizes were much smaller and required the 50+ wells that we analyzed in order to achieve 95% power. Example from Figure 5 below shows well level data for the co-culture and monoculture validation experiments -
- It is widely recognized that synapses formed in networks of NGN2-induced excitatory neurons only, may not model synapses in the real human brain very well (yet), especially not at DIV21. First, the authors can be more open/precise about this, e.g., in line 156 the authors indicate they use hNs at DIV21 because they are "electrophysiologically active" based on three references. However, (a) these references indicate that hNs cultures start to mature from DIV21 onwards but are not really mature yet, and (b) being "electrophysiologically active" seems not the most relevant criterion. Synaptic parameters like initial release probability, rise/decay time, and synchronicity are more relevant (none of which indicate synapses are mature at DIV21). Second, especially in the light of the claims the authors make regarding the effects of compounds on "synaptic connectivity" it seems essential to test, at least in a set of validation experiments, the distribution of postsynaptic markers. Synapsin-positive puncta may not be accompanied by a postsynaptic specialization and rather represent (mobile) vesicle clusters and/or release sites without postsynaptic partners. In addition, the authors claim synapsin1 is a pan-neuronal synapse marker. This is not yet validated for human neurons. A few control stainings with synaptic vesicle and active zone markers will secure this claim.
We thank the reviewer for this comment and have now updated the text to indicate and expand on the fact that we are looking at immature synapses at day 21 in vitro (e.g., please see pages 8 and 12 of the Results section in the PDF).
As mentioned above, we also tested conditions for four additional postsynaptic antibodies, drawing from those used in published studies of human cellular models (and species that would not cross-react with antibodies used for Synapsin1 and MAP2). Specifically, we tested antibodies against PSD-95, NLGN4, Homer1 and BAIAP2 at a range of concentrations in co-cultures generated from two independent cell lines. Of these antibodies, we only obtained quantifiable signal for PSD-95, while NLGN4, Homer1 and BAIAP2 appeared to be of poor quality in our culture systems (e.g., nonspecific signal, high signal in astrocytes, etc.). As shown below and in Figure 1-figure supplement 1, analysis of PSD-95 revealed that 43.1% of PSD-95 puncta on MAP2 also colocalized with synapsin1, and 28.8% of synapsin1 puncta on MAP2 also colocalized with PSD-95. Discussions of these data and limitations have been significantly elaborated upon on pages 10-11 of the Results section and pages 24 and 29 of the Discussion section in the PDF. For example, we discuss how the partial colocalization could be due both to the relative immaturity of the synapses discussed above (presynaptic assembly preceding postsynaptic assembly at this early stage of neuronal development) as well as the overall poorer quality of the PSD-95 signal in human cellular material (PSD-95 signal was of insufficient quality and consistency for screening applications and was generally quite difficult to resolve as compared to Synapsin1).
Additionally, we tested two additional presynaptic antibodies, including synaptophysin and SV2A. Of these antibodies, we obtained reasonable quality signal for synaptophysin, which we have quantified in Figure 1-figure supplement 1. While SV2A also gave some signal, it was of poorer quality and difficult to reliably quantify. We observed roughly half of the Synapsin1 signal on MAP2 colocalizing with synaptophysin, and vice versa. Lack of complete colocalization could be due to reports that synapsin1 expression precedes synaptophysin expression in the cortex (e.g., Pinto et al 2013), reports that synaptophysin is also expressed at extra synaptic sites (e.g., Micheva et al 2010), or the reduced quality of staining for synaptophysin that we obtained compared with synapsin1. These data are now elaborated upon on pages 10-11 of the Results section and page 24 of the Discussion section in the PDF.
We have also expanded our discussion of Synapsin1 as a presynaptic marker including additional references on the use of Synapsin1 to label cortical glutamatergic synapses in rodent (e.g., Micheva 2010) and the use of Synapsin1 on MAP2 as a pan-synaptic marker in human neurons (e.g., Chanda et al 2019, Pak et al 2015, Yi et al 2016; page 10). We have also included the use of Synapsin1 on MAP2 as a specific Limitation on page 29 where we discuss that reliance on this system in developing neurons may be capturing sites which do not then develop into fully functional synapses with postsynaptic partners.
- The analysis of the transcriptional effects of BET inhibitors is rather basic, especially given the rather strong claim: "BET inhibitors enhance synaptic gene expression programs". Which programs? Differentially expressed transcripts can at least be analysed further in terms of subcellular localization (pre/post) or synaptic functions, e.g. using SYNGO, also to address point 2 above.
We thank the reviewer for this comment and have now incorporated SynGO analysis into Figure 6 to examine the synaptic ontology terms. As shown below, Figure 6g now includes the top 5 significantly enriched terms and Figure 6h shows the gene counts by cellular component. Here, we focused on genes upregulated after both JQ1 and Birabresib treatment compared with a background list of expressed genes. The most enriched synaptic ontology terms related to the post-synaptic membrane, so we also validated protein level changes in two postsynaptic proteins (Homer1 and BAIAP2) by Western blot analysis in Figure 6. In addition to Figure 6, these data are now included in Supplementary File 5 and discussed on page 22 of the Results section.
-
Evaluation Summary:
Berryer et al. report on an automated and quantitative platform to study the number of synaptic inputs formed in networks of human excitatory neurons and astrocytes in vitro. The authors tested the utility of the platform by screening a large collection of small molecules and identified several modulators of synapse density, which were validated in follow-up experiments. The automated platform substantially extends what is currently available, particularly with respect to the automation of the initial analysis steps. The positive hits identified here, the inhibitors of bromodomain and extraterminal (BET) family of gene expression regulators, are important, and will likely contribute to the understanding of the mechanisms of human synapse assembly.
(This preprint has been reviewed by eLife. We include the public …
Evaluation Summary:
Berryer et al. report on an automated and quantitative platform to study the number of synaptic inputs formed in networks of human excitatory neurons and astrocytes in vitro. The authors tested the utility of the platform by screening a large collection of small molecules and identified several modulators of synapse density, which were validated in follow-up experiments. The automated platform substantially extends what is currently available, particularly with respect to the automation of the initial analysis steps. The positive hits identified here, the inhibitors of bromodomain and extraterminal (BET) family of gene expression regulators, are important, and will likely contribute to the understanding of the mechanisms of human synapse assembly.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #2 and Reviewer #3 agreed to share their name with the authors.)
-
Reviewer #1 (Public Review):
A high-throughput synaptic phenotyping platform targeting human synapses is highly valuable. The validity of the present system is supported by a small molecule inhibitor screen that has identified targets, including the BET family proteins, whose role in brain function has been previously demonstrated. The authors have gone one step further to analyze the gene expression programs impacted by the BET Inhibitors, and the observations that synaptic genes encoding proteins such as neurexin-3 and homer 1 are altered is reassuring. In addition, demonstrating that the presence of astrocytes crucially impacts the density of presynaptic marker protein is of relevance for the design of similar platforms. The general utility of the present platform in identifying synaptic changes, however, needs to be further …
Reviewer #1 (Public Review):
A high-throughput synaptic phenotyping platform targeting human synapses is highly valuable. The validity of the present system is supported by a small molecule inhibitor screen that has identified targets, including the BET family proteins, whose role in brain function has been previously demonstrated. The authors have gone one step further to analyze the gene expression programs impacted by the BET Inhibitors, and the observations that synaptic genes encoding proteins such as neurexin-3 and homer 1 are altered is reassuring. In addition, demonstrating that the presence of astrocytes crucially impacts the density of presynaptic marker protein is of relevance for the design of similar platforms. The general utility of the present platform in identifying synaptic changes, however, needs to be further substantiated by additional synaptic markers.
-
Reviewer #2 (Public Review):
In this manuscript, Berryer et al describe a fully automated, scalable approach to quantify the number of synaptic inputs formed onto human iPSC-derived neurons (hNs) in 2D culture. They validate the sensitivity of their approach by synapsin1 knock-down and test almost 400 small molecules for their effect on synapses, and the role of astrocytes. They identify BET inhibitors as strong modifiers of synapse numbers in hNs and performed follow-up experiments to confirm the finding, characterize the effect further and demonstrate the critical role of astrocytes.
Every step of the protocol is automated to achieve high reproducibility and homogeneity throughout the experiments. This automated approach has great potential for scaling up drug screening, genetic perturbations, and disease modeling experiments related …
Reviewer #2 (Public Review):
In this manuscript, Berryer et al describe a fully automated, scalable approach to quantify the number of synaptic inputs formed onto human iPSC-derived neurons (hNs) in 2D culture. They validate the sensitivity of their approach by synapsin1 knock-down and test almost 400 small molecules for their effect on synapses, and the role of astrocytes. They identify BET inhibitors as strong modifiers of synapse numbers in hNs and performed follow-up experiments to confirm the finding, characterize the effect further and demonstrate the critical role of astrocytes.
Every step of the protocol is automated to achieve high reproducibility and homogeneity throughout the experiments. This automated approach has great potential for scaling up drug screening, genetic perturbations, and disease modeling experiments related to synapses.
The authors successfully identified, in two independent hNs lines, three small-molecule inhibitors of transcription modifiers of the BET family as the strongest positive modifiers of synaptic inputs. The initial study performed with immunofluorescence was then validated by Western blot analysis and mRNA-seq analysis, which showed an increase in the expression of trans-synaptic signaling genes.
While accessing the molecular mechanisms of BET inhibitors, the authors observed that the increased synaptic inputs occurred only in cocultures of astrocytes and neurons, and not in hNs monoculture. Finally, the authors report that the presence of astrocytes alone is a major driving force to promote synaptic inputs.Overall, the experiments are well conducted, and the conclusions are supported by the data. The new approach reaches beyond the current state of the field, especially in the first steps of automation and the identified modulators (BET inhibitors) are interesting and novel, and the subsequent validation is convincing.
On the other hand, the manuscript does not yet define the exact resolution and power of the new methods, and does not convincingly show that the observed synapsin-puncta are synapses and that the data of the validation experiments can be improved.
Major points:
1. Although the manuscript contains a lot of quantitative data on variance, the current manuscript stops short of an exact definition of the resolution of the assay and its statistical power. With the real (measured) variance of the assay, the power to detect certain effects can be computed. To be relevant for other applications than the current (e.g. genetic perturbations and disease modelling), it is relevant to define this for smaller effects too: can this assay detect a 25% effect with reasonable numbers of observations? Such assessments can also provide important recommendations on when it makes sense to add more repeated measures of the same specimens (wells, ROIs) and when more independent inductions are required (and how much this adds to overall power). The manuscript would also benefit from a short discussion on how to optimize future study designs (repeated measures, independent inductions, number of subjects).
2. It is widely recognized that synapses formed in networks of NGN2-induced excitatory neurons only, may not model synapses in the real human brain very well (yet), especially not at DIV21. First, the authors can be more open/precise about this, e.g., in line 156 the authors indicate they use hNs at DIV21 because they are "electrophysiologically active" based on three references. However, (a) these references indicate that hNs cultures start to mature from DIV21 onwards but are not really mature yet, and (b) being "electrophysiologically active" seems not the most relevant criterion. Synaptic parameters like initial release probability, rise/decay time, and synchronicity are more relevant (none of which indicate synapses are mature at DIV21). Second, especially in the light of the claims the authors make regarding the effects of compounds on "synaptic connectivity" it seems essential to test, at least in a set of validation experiments, the distribution of postsynaptic markers. Synapsin-positive puncta may not be accompanied by a postsynaptic specialization and rather represent (mobile) vesicle clusters and/or release sites without postsynaptic partners. In addition, the authors claim synapsin1 is a pan-neuronal synapse marker. This is not yet validated for human neurons. A few control stainings with synaptic vesicle and active zone markers will secure this claim.
3. The analysis of the transcriptional effects of BET inhibitors is rather basic, especially given the rather strong claim: "BET inhibitors enhance synaptic gene expression programs". Which programs? Differentially expressed transcripts can at least be analysed further in terms of subcellular localization (pre/post) or synaptic functions, e.g. using SYNGO, also to address point 2 above.
-
Reviewer #3 (Public Review):
In this paper Berryer et al. developed an efficient automated and quantitative high-content synaptic phenotyping platform to be used for human neurons and astrocytes derived from iPSCs. With this quantitative platform, the authors screened the effects of 376 small molecules on presynaptic density, neurite outgrowth, and cell viability. Interestingly six small molecules were identified that specifically enhanced human presynaptic density in vitro and the presence of astrocytes in culture was essential for mediating the effects of the six molecules. Among these molecules, the bromodomain and extraterminal (BET) inhibitors were the most effective in increasing the presynaptic clusters and in upregulating synaptic gene expression programs. Thus this paper provides strong evidence for the possibility to use a …
Reviewer #3 (Public Review):
In this paper Berryer et al. developed an efficient automated and quantitative high-content synaptic phenotyping platform to be used for human neurons and astrocytes derived from iPSCs. With this quantitative platform, the authors screened the effects of 376 small molecules on presynaptic density, neurite outgrowth, and cell viability. Interestingly six small molecules were identified that specifically enhanced human presynaptic density in vitro and the presence of astrocytes in culture was essential for mediating the effects of the six molecules. Among these molecules, the bromodomain and extraterminal (BET) inhibitors were the most effective in increasing the presynaptic clusters and in upregulating synaptic gene expression programs. Thus this paper provides strong evidence for the possibility to use a reproducible and automated screening platform for the identification of synaptic modulators in human neurons.
-
