Synaptic polarity and sign-balance prediction using gene expression data in the Caenorhabditis elegans chemical synapse neuronal connectome network
Bank G. Fenyves, Gabor S. Szilagyi, Zsolt Vassy, Csaba Soti, Peter Csermely
RRESEARCH ARTICLE
Synaptic polarity and sign-balance predictionusing gene expression data in the
Caenorhabditis elegans chemical synapseneuronal connectome network
Ba´nk G. Fenyves ID , Ga´bor S. Szila´gyi ID , Zsolt Vassy ID , Csaba S ő ti ID ,Peter Csermely ID * Department of Molecular Biology, Semmelweis University, Budapest, Hungary, Department ofEmergency Medicine, Semmelweis University, Budapest, Hungary * [email protected] Abstract
Graph theoretical analyses of nervous systems usually omit the aspect of connection polar-ity, due to data insufficiency. The chemical synapse network of
Caenorhabditis elegans isa well-reconstructed directed network, but the signs of its connections are yet to be eluci-dated. Here, we present the gene expression-based sign prediction of the ionotropic chemi-cal synapse connectome of C . elegans (3,638 connections and 20,589 synapses total),incorporating available presynaptic neurotransmitter and postsynaptic receptor geneexpression data for three major neurotransmitter systems. We made predictions for morethan two-thirds of these chemical synapses and observed an excitatory-inhibitory (E:I) ratioclose to 4:1 which was found similar to that observed in many real-world networks. Our opensource tool (http://EleganSign.linkgroup.hu) is simple but efficient in predicting polarities byintegrating neuronal connectome and gene expression data. Author summary
The fundamental way neurons communicate is by activating or inhibiting each other viasynapses. The balance between the two is crucial for the optimal functioning of a nervoussystem. However, whole-brain synaptic polarity information is unavailable for any speciesand experimental validation is challenging. The roundworm
Caenorhabditis elegans pos-sesses a fully mapped connectome with an emerging gene expression profile of its 302neurons. Based on the consideration that the polarity of a synapse can be determined bythe neurotransmitter(s) expressed in the presynaptic neuron and the receptors expressedin the postsynaptic neuron, we conceptualized and created a tool that predicts synapticpolarities based on connectivity and gene expression information. Using currently avail-able datasets we propose for the first time that the ratio of excitatory and inhibitory synap-ses in a partial connectome of C . elegans is around 4 to 1 which is in line with the balanceobserved in many natural systems. Our method opens a way to include spatial and tempo-ral dynamics of synaptic polarity that would add a new dimension of plasticity in the PLOS COMPUTATIONAL BIOLOGY
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 1 / 19 a1111111111a1111111111a1111111111a1111111111a1111111111
OPEN ACCESS
Citation:
Fenyves BG, Szila´gyi GS, Vassy Z, S ő ti C,Csermely P (2020) Synaptic polarity and sign-balance prediction using gene expression data inthe Caenorhabditis elegans chemical synapseneuronal connectome network. PLoS Comput Biol16(12): e1007974. https://doi.org/10.1371/journal.pcbi.1007974
Editor:
Emma K. Towlson, CCNR, Northeastern,UNITED STATES
Received:
May 19, 2020
Accepted:
October 19, 2020
Published:
December 21, 2020
Copyright: © Data Availability Statement:
All relevant data arewithin the manuscript and its Supportinginformation files.
Funding:
This work was supported by theHungarian National Research, Development andInnovation Office [Hungarian Scientific ResearchFund, K131458 to P.C. and K116525 to C.S.](https://kormany.hu/emberi-eroforrasok-miniszteriuma), by the Higher EducationInstitutional Excellence Programme of the Ministry xcitatory:inhibitory balance. Our tool is freely available to be used on any networkaccompanied by any expression atlas.
Introduction C hemical synapses of a neuronal network are both directed and signed, since a neuron is ableto excite or inhibit another neuron. The nervous system of the nematode Caenorhabditis ele-gans has been fully mapped and reconstructed [1–3]. However, except for a few connectionsthere is no comprehensive chemical synapse polarity data available [1]. While the direction ofa synaptic connection can be inferred from its structure, the experimental determination ofits polarity requires delicate electrophysiological methods with limited system-level use (e.g.patch-clamping) or more recent calcium-imaging or optogenetic techniques. Instead, in silico approaches using reverse engineering and genetic algorithms have efficiently predicted synap-tic signs for different subnetworks of the C . elegans connectome [4–8].Many synaptic sign prediction models have relied on the widely accepted assumption thatthe polarity of a chemical synapse is solely determined by the type of neurotransmitter releasedby the presynaptic neuron [4,5]. Therefore, in C . elegans excitatory glutamatergic and choliner-gic, as well as inhibitory ɣ -aminobutyric acid (GABA)-ergic ionotropic chemical connectionshave been modeled. However, with this approach, approximately 6% of the connections turnedto be inhibitory [9,10]. A low proportion of inhibitory connections can result in an unbal-anced, over-excited network, as has been shown by previous publications [11–13]. Moreover,there is evidence of unconventional postsynaptic effects of neurotransmitters, such as choliner-gic inhibition [14,15] or glutamatergic inhibition [16–18], meaning that a neuron can simulta-neously excite and inhibit its postsynaptic partners with the same neurotransmitter due tovariable neurotransmitter receptor expression on the postsynaptic neuron membrane. Forexample, the cholinergic AIY interneuron can activate RIB neurons and inhibit AIZ neuronsin an acetylcholine-mediated fashion [15].We aimed to predict synaptic polarities in the C . elegans ionotropic chemical synapse con-nectome (297 neurons and 20,589 synapses) relying on presynaptic neurotransmitter andpostsynaptic receptor gene expression data for the three main neurotransmitters glutamate,acetylcholine and GABA. In this study of the C . elegans nervous system, we predicted thepolarity of more than 70% of ionotropic chemical synapses and predicted a sign-balance ofexcitatory:inhibitory connections close to 4:1 that has been observed as functionally stable inmany real-world circumstances. Presenting a new dataset, we show that the concept of geneexpression-based polarity prediction can efficiently be applied to demonstrate a balanced E:Iratio in a nervous system. Results
Creating a prediction tool of the synaptic polarities of the C . elegans connectome Our primary goal was to infer synaptic polarity from combining connectivity and gene expres-sion data. We created a simple, yet powerful algorithmic database (S1 Data) that takes as inputconnectome and gene expression data to predict synaptic polarity of ionotropic glutamatergic,cholinergic and GABA-ergic connections. We used the C . elegans WormWiring connectomedata primarily in the form of a weighted edge list representing 20,589 chemical synapses in3,638 connections. Our prediction tool is available here: http://EleganSign.linkgroup.hu.
PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 2 / 19 for Innovation and Technology in Hungary (https://nkfih.gov.hu) within the framework of theMolecular Biology thematic programme of theSemmelweis University [to P.C.] (https://semmelweis.hu), as well as by the ThematicExcellence Programme (Te´materu¨leti Kiva´lo´sa´giProgram, 2020-4.1.1.-TKP2020) of the Ministry forInnovation and Technology in Hungary (NemzetiKutata´si, Fejleszte´si e´s Innova´cio´s Alap), within theframework of the artificial intelligence, biomarkerthematic programme of the SemmelweisUniversity [to P.C]. B.G.F was supported by theHuman Capacities Grant Management Office inHungary(Emberi Eroforra´sok Miniszte´riuma) [NTP-NFTO¨-18-B-0179 and NTP-NFTO¨-19-B-0264], alsoby the Semmelweis University (EFOP-3.6.3-VEKOP-16-2017-00009). C.S. is a Merit Prizerecipient of the Semmelweis University (https://semmelweis.hu). The funders had no role in studydesign, data collection and analysis, decision topublish, or preparation of the manuscript.
Competing interests:
The authors have declaredthat no competing interests exist. pdate of the previous neurotransmitter expression tables
We updated the C . elegans neuronal neurotransmitter tables previously published [19–22]with recent evidence [10,23] (Methods). After this update, 256 neurons had a single neuro-transmitter expressed, while 12 neurons had double neurotransmitter expression (Fig 1A).There were 34 neurons which did not express any of the three neurotransmitters investigated. Extraction of gene expression data
Fig 1. Neurotransmitter and receptor expression patterns of C . elegans neurons. Expression data of the three major synaptic neurotransmitters andtheir receptors of C . elegans were collected from multiple datasets and were manually curated (see Methods). (A) Distribution of neurons according totheir neurotransmitter expression: glutamate (red), acetylcholine (green), GABA (blue) or none (grey). (B)
Number of receptor genes expressed byneurons, grouped by neuron modality. (C)
Distribution of neurons based on their neurotransmitter receptor gene expression (colors are the same as inpanel A ). (D) Distribution of neurons according to the number of neurotransmitters for which anion and/or cation channel receptor genes areexpressed. https://doi.org/10.1371/journal.pcbi.1007974.g001
PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 3 / 19 eceptor expression pattern to each neuron (see Methods). To do this we first sorted the previ-ously identified 62 ionotropic receptor genes into six functional classes based on their suggestedneurotransmitter ligand (glutamate, acetylcholine or GABA) and putative ion channel type (cat-ionic or anionic, i.e. excitatory or inhibitory), as shown in Table 1. We found evidence for post-synaptic neuronal expression of 42 out of the 62 receptor genes in the C . elegans nervous system(genes marked bold in Table 1; also S1 Data). We also found an increasing average number ofreceptor genes in sensory, inter- and motor neurons, respectively (Fig 1B).Next, for all the 302 neurons of the C . elegans connectome we determined which receptorclasses were expressed. 166 neurons had an overlapping expression of receptors for two orthree different neurotransmitters (Fig 1C). The distribution of neurons according to theirexpression of cationic and/or anionic glutamate, acetylcholine and/or GABA receptors sug-gested functional diversity due to the high number of neurons expressing both excitatory andinhibitory receptors (Fig 1D). Surprisingly, 85 neurons expressed both excitatory and inhibi-tory receptor genes for the same neurotransmitter (S1 Data). Forty out of 302 neurons showedno receptor expression, out of which 32 neurons were primarily sensory neurons (S1 Data).The average number of receptor genes expressed was 3.7 per neuron (S2 Table). Neurotransmitter and receptor gene expression-based polarity prediction
After assigning neurotransmitter and receptor expression patterns to each neuron, we pre-dicted synaptic polarities by looking for matches between the neurotransmitter expression ofthe presynaptic neuron and the receptor gene expression of the postsynaptic neuron (Fig 2A).This way, we labeled synapses as one of the following: excitatory, inhibitory, complex, or
Table 1. Neurotransmitter receptor genes. Glutamate Acetylcholine GABACation channel receptor gene glr-1glr-2glr-3glr-4glr-5glr-6glr-7glr-8nmr-1nmr-2 acr-1 acr-16acr-2 acr-17 acr-3 acr-18 acr-4 acr-19 acr-5 acr-20acr-6 acr-21acr-7 acr-23 acr-8 acr-25acr-9 deg-3 acr-10 des-2acr-11 eat-2 acr-12 lev-8 acr-13 unc-29acr-14 unc-38 acr-15 unc-63 exp-1lgc-35
Anion channel receptor gene glc-1glc-2 glc-3 glc-4 avr-14avr-15 acc-1acc-2 acc-3 lgc-47 lgc-48lgc-49 gab-1ggr-1ggr-2ggr-3lgc-36lgc-37lgc-38 unc-49
The
Caenorhabditis elegans genome contains 62 ionotropic postsynaptic receptor genes for glutamate, acetylcholine, and GABA. acc-4 and lgc-46 genes were excludedfrom our database due to suggested presynaptic expression (S1 Table). In this table genes are grouped according to their neurotransmitter ligand and whether formingcationic (+) or anionic ( − ) ion channels (based on [24] and other references listed in S1 Table). In C . elegans "unconventional signaling", namely, glutamate-mediatedinhibition, cholinergic inhibition and GABA-ergic excitation, is facilitated by 6, 6, and 2 receptor genes, respectively. In the gene expression database used in this work,expression in at least one neuron was found in the case of 42 genes (marked bold ), while for 20 genes no neuronal expression was found. https://doi.org/10.1371/journal.pcbi.1007974.t001 PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 4 / 19 npredicted (see Methods). “Excitatory” or “inhibitory” label were given when the neurotrans-mitter-matched postsynaptic receptor genes were only cation or anion channel related, respec-tively. A synapse was labeled as “complex” if data suggested both excitatory and inhibitoryfunction. With this approach, we predicted synaptic polarity for 73% of chemical synapses of
Fig 2. Prediction of synaptic polarities of the C . elegans ionotropic chemical synapse connectome. (A) Predictionmethod. Connectome and gene expression data were manually curated (see Methods). Polarities of chemical synapseswere predicted based on the neurotransmitter expression of presynaptic neurons and the matching receptor geneexpression of the postsynaptic neurons. (B)
Distribution of predicted and unpredicted synapses. We were able topredict polarity for 73% of chemical synapses (green). The polarities of the rest of synapses were unpredicted due tounknown neurotransmitter expression of the presynaptic neurons (dark grey) or non-matching receptor geneexpression of the postsynaptic neurons (light grey). https://doi.org/10.1371/journal.pcbi.1007974.g002
PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 5 / 19 he C . elegans connectome (Fig 2B). We could not predict polarity for the remaining synapsesdue to missing neurotransmitter data or mismatch in neurotransmitter/receptor expression(Fig 2B). We predicted that 9,070 of the synapses are excitatory and 2,580 are inhibitory, while3,413 synapses have complex function (Fig 3A and S1 Data). These findings suggest that theoverall ratio of excitatory and inhibitory synapses (E:I ratio) in the C . elegans ionotropic chem-ical synapse network is close to 4:1 (Fig 3A, NT+R method ). Alternative polarity prediction methods
To put our results in context, we applied two alternative prediction methods for comparison(
NT-only and
R-only ; see Methods). The
NT-only method yielded a much higher E:I ratio
Fig 3. Predicted synaptic polarities. (A)
Distributions of predicted polarities, using the method developed in thispaper (
NT+R ) and two alternative methods as comparison (
NT-only and
R-only ). Polarities were predicted byconsidering the neurotransmitter expression of the presynaptic neuron and/or the receptor gene expression of thepostsynaptic neuron (see Methods). (B)
Distributions of predicted synaptic polarities (using the
NT+R method )according to the presynaptic neurotransmitter. Colors are the same as in panel A . Unpredicted synapses are notshown. https://doi.org/10.1371/journal.pcbi.1007974.g003 PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 6 / 19
Fig 3A,
NT-only method ; S2 Data), which is in line with the dominance of purely glutama-tergic or cholinergic (traditionally excitatory) neurons over GABA-ergic (traditionally inhib-itory) neurons (Fig 1A). To explain the difference further, the
NT+R method predicted that30% of cholinergic and 5% of glutamatergic synapses were inhibitory (Fig 3B) which is a sig-nificant fraction of synapses otherwise predicted excitatory with the
NT-only method. A pair-wise comparison of polarities predicted with the
NT+R and
NT-only methods is presented inS2 Data. The other,
R-only method yielded a markedly smaller E:I ratio, however predictedan excessive number of complex synapses (Fig 3A,
R-only method , S3 Data). This is due tothe fact that many neurons express both cation and anion channel receptor genes (Fig 1D).
Feedback inhibition between neuron groups
Notably, in subsets of connections which connect neurons of different modalities of sensoryneurons, motor neurons, interneurons and polymodal neurons, the E:I ratios varied between1:10 (motor– > sensory) and 14:1 (inter– > motor). Importantly, we observed dominant inhi-bition in the motor– > sensory, motor– > inter, and inter– > sensory directions (Fig 4A),exhibiting inhibitory backward signaling as discussed in the literature previously [25,26], asopposed to a forward (sensory - > motor) excitatory excess. Besides, a significant presence ofinhibitory and complex connections was found in the locomotion circuit as well (Fig 4B andS2 Fig). Network representations of the signed C . elegans connectome Network representations of synaptic polarities in the C . elegans ionotropic chemical synapseconnectome using the EntOptLayout plugin of Cytoscape [27] are in Fig 5. Fig 5A shows thatthe modular structure of the C . elegans connectome visualized by the EntOptLayout methodnicely captures the anatomical locations of the anterior, ventral and lateral ganglions, as well asthe premotor interneurons of the worm. While the anterior and lateral ganglions show a largeglutamate expression, this is much less characteristic to the ventral ganglion (Fig 5A). Fig 5Bshows that the ventral ganglion has predominantly inhibitory connections, while connectionsin the other locations are predominantly excitatory if predicted by our NT + R method . Fig 5Cdemonstrates that the prediction of polarities by neurotransmitters only (
NT-only method )results in a large excitatory excess, mainly because the connections predicted as inhibitory orcomplex with the
NT+R method turn into excitatory. This difference can be observed amongsthead neurons and premotor interneurons, but less amongst motor neurons. Many connec-tions of the polymodal neurons are predicted as complex only with the
NT+R method . Theanatomical locations of neurons expressing various neurotransmitters (Fig 5D) correspondwell to the network representation shown in Fig 5A. Links between inhibitory function andanatomical structures have been shown in the human brain [28,29], but have not been demon-strated previously in the nematode.
Validation of our predictions
To validate our results, we contrasted our predictions to previously published in silico andexperimental work as well. Comparison to the computational findings of Rakowski & Karbow-ski [30] in the C . elegans locomotion circuit of 7 neuron groups and 652 synapses showed a70% consistency in predicted synaptic polarities (53% on the level of connections) (S3 Table),albeit using a completely different concept. When testing our predictions against experimentalevidence based on the literature we found that the majority of predicted polarities using ourmethod were consistent with earlier findings in C . elegans : only 1 out of 12 interneuronal con-nections (29 / 501 synapses) reviewed was predicted an opposing polarity to what has PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 7 / 19 ig 4. Excitatory:inhibitory balance of different neuron groups. (A)
Excitatory:inhibitory balance between neurongroups of different modalities. Nodes represent groups of neurons by modality. Edges are weighted according to theexcitatory:inhibitory (E:I) ratios (see numbers). Green and red colors represent excitatory (E:I >
1) and inhibitory (E:I <
1) excess in sign-balances, respectively. (B)
Network representation of the locomotion subnetwork. Edges representexcitatory (blue), inhibitory (red), or complex (black) chemical connections. Edges are weighted according to synapsenumber. The shape of vertices ( Δ , � , � ) represent the modality (sensory, inter, motor, respectively) of neurons. Separaterepresentations of the head circuit and the ventral nerve cord motor neurons are in S2 Fig. https://doi.org/10.1371/journal.pcbi.1007974.g004 PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 8 / 19 xperimentally been confirmed (S4 Table). This ratio is 6 /12 when validating the
NT-only pre-diction method, supporting the importance of receptor expression.To test the robustness of our
NT+R prediction method to predict E:I balance, we appliedthe same rules to predict polarities of connections regardless of synapse number data, andafter perturbations in the network like deletion of the 20 pharyngeal nervous system neuronsor deletion of potentially variable (i.e. single-synapse) connections (S4 Data). Furthermore, werepeated our analyses in other published connectomes [2,3] of different sizes as well (S5 Tableand S5 Data). In all five cases, the excitatory:inhibitory ratios were in the range of 3.1 to 4.1 (S1Fig). When predicting not using the yet preprint-published RNA-seq expression dataset [23],this range was 3.7–4.1 (S6 Data). All together, these findings suggest that the observed sign-bal-ance is a remarkably robust property of the C . elegans ionotropic chemical synapse network. Discussion
Nervous systems are not only directed but signed networks as well since neurons either acti-vate or inhibit other neurons [12]. The balance of excitatory and inhibitory connections (i.e.the sign-balance) is a fundamental feature of brain networks, clearly marked by the variety ofdisorders associated with its impairment [12,31,32]. However, direct evidence of single syn-apse polarity is rather sporadic even in simple species. In this work we predicted that the sign-balance in the C . elegans ligand-gated ionotropic chemical synapse network is approximately4:1 (excitatory-inhibitory, E:I). This is consistent with previous in vitro and in vivo studies ofnervous systems [33–36], and also with observations of different social networks [37,38], asshown in Table 2. However, this ratio can only be predicted if not only the neurotransmitterexpression of the presynaptic neuron but also the receptor gene expression of the postsynapticneuron is taken into consideration. Its significance is due to fourteen receptor genes that arepresumed to encode inhibitory glutamatergic/cholinergic or excitatory GABA-ergic postsyn-aptic ion channel receptors. This concept of unconventional signaling is not new, but hasalready been described in C . elegans [14,39,40] and other primitive species [41–43], and also in Fig 5. Network representations of the C . elegans chemical synapse network. (A-C) Network representations using the EntOpt layout plugin inCytoscape [27]. (A)
Color and shape of vertices represent neurotransmitter expression and modality of neurons, respectively (see inset for definitions). (B)
Edges represent excitatory (blue), inhibitory (red), or complex (black) chemical connections predicted by the NT+R method (see Methods),weighted according to synapse numbers. (C)
Colors of edges (see panel B ) represent the polarities of chemical synapses predicted by the NT-onlymethod. (D) Layout of vertices is representing the anatomic position of neurons. Node and edge colors are as in panels A and B , respectively. High-resolution representation is available in S1 File. https://doi.org/10.1371/journal.pcbi.1007974.g005 PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 9 / 19 ammals in the postnatal period [44,45]. This concept has already motivated the prediction ofconnection polarity instead of neuron polarity, yet on a subcircuit level [30]. Complementingrecent work that used gene expression data for structural and functional modelling of the ner-vous system [46–48], our prediction model is a novel attempt to predict polarities of all ligand-gated ionotropic chemical synapses of the C . elegans connectome.A surprisingly high proportion of synapses were predicted to have a complex, i.e. both excit-atory and inhibitory polarity. This is due to parallel expression of cationic and anionic receptorgenes–often for the same neurotransmitter–in half of the neurons. This suggests a highly com-plex functioning of neuronal connections that extends beyond the permanently exclusive con-cept of excitation-inhibition dichotomy. Since our work is mostly based on expression data ofsubunits instead of functional receptors, predictions made are derived from genetic permissi-bility rather than direct receptor complex presence. While ionotropic transmission in a singlesynapse is typically either excitatory or inhibitory, the predicted "complexity" can be resolvedmainly in two physiologically well-established ways. One is that postsynaptic receptors are nothomogenously distributed across the plasma membrane but their subcellular localization isregulated. This allows the receptors to act independently [50–55] and allows the same neuro-transmitter to excite and inhibit at distinct postsynaptic sites. For example, such mechanismshave been identified in the AIA [56–58] and AIB neuron groups [59].Another explanation of "complexity" is the dynamic change of gene expression in timewhich is observed all through the life-span of a worm e.g. during development, learning (syn-aptic plasticity), and aging [60–68]. Ultimately, changes in gene expression can lead to neuro-transmitter-switching and consequential up- and downregulation of receptors of opposingpolarity [69,70]. In its complexity, co-transmission by parallel expression of different neuro-transmitters and receptors is one of the mechanisms of plasticity [71].Currently, there is not enough data to address either the spatial or the temporal aspects ofreceptor expression regulation on the worm-scale. As expression-profiling methods will pro-vide whole-brain and dynamic proteomics data of subcellular resolution, complex synapsesmight be further resolved.There are several mechanisms of interneuronal communication acting in concert to transmitsignals while maintaining a responsive but balanced system. The balance of excitation and inhibi-tion—crucial for network stability—is reached via a number of mechanisms. Both synaptic andextrasynaptic, electrical and chemical, voltage-gated and ligand-gated, ion channel-mediated andG-protein coupled neurotransmission have diverse but intertwining roles in promoting and Table 2. Proportions of negative edges in signed networks.Network Proportion of negative edges Reference C . elegans chemical synapse network 20% this workRat hippocampus ( in vivo ) 5–30% [35] Rat excitatory neocortical neurons 20% [36]
Cerebral cortex ( in vivo ; GAD expression) 10–20% [17]
Cerebral cortex ( in vivo ; GABA neurons) 20–25% [17]
Optimal network for synchronized bursting activity ( in silico ) 10–20% [6]
Primary visual cortex (V1; in silico) [49]
Neuronal network (e x vivo ) and neuronal network model (in silico) [18]Wikipedia (social network) 21% [37]Epinions (social network) 15% [37]Slashdot (social network) 23% [37]University freshman network (social) 12–14% [38] https://doi.org/10.1371/journal.pcbi.1007974.t002
PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 10 / 19 odulating excitation and inhibition. As there are significant mechanistic and functional differ-ences (e.g. speed, modulatory role) between transmission modes, the neuronal connectome canbe comprehended as a multiplex network of partially independent layers with each layer repre-senting a distinct type of signaling [47,72,73]. Two dedicated cases were monoamine and meta-botropic transmission which were excluded from our workflow. Since monoamine transmissionvia serotonin, dopamine and tyramine typically occurs extrasynaptically (i.e. 94% in case of tyra-mine-responsive neurons), expression data of even ionotropic channels (e.g. mod-1 , lgc-55 )would have been difficult to apply on the hard-wired connectome used in our study [47,74,75].Additionally, in [47] the authors created a wireless (extrasynaptic) connectome of C . elegans based on matching monoamine/neuropeptide expression with receptor gene expression showinga network structurally different from the hard-wired connectome. Wireless networks possiblyexist for ionotropic receptors as well via mechanism of spillover transmission [10,76]. Likewise,metabotropic neurotransmission typically acting via G-protein coupled receptors plays a rathermodulatory than direct excitatory/inhibitory role in the nervous system by inducing broad, long-lasting, slow time-scale changes which is distinctive [71,73,77–80].Our paper covers ligand-gated ionotropic synaptic connections, which account for the fast-acting system of neurotrans-mitter-mediated synaptic signaling. Additional layers of neural signaling can be targets of polar-ity prediction in subsequent studies and potentially overlayed on this signed network.There are a number of limitations of our study which limit its generalizability at the currentstate: 1) although the most complete of any species, new connectivity data of the worm is stillemerging [3,60] as well as 2) gene expression data [23,81]; 3) although our assumption that cationand anion channels are consistently excitatory and inhibitory, respectively, is generally validbased on their ion selectivity, the direction of a channel-mediated ion current is ultimately deter-mined by a set of additional biophysical conditions e.g. ion gradients and the membrane potential[39,82,83] which were not considered in our work; 4) neurotransmission types other than ligand-gated ionotropic chemical signaling (e.g. G-protein coupled, monamine or neuropeptide) wereexcluded to avoid mixing different layers of neurotransmission (i.e. extrasynaptic, slow-scale,neuromodulatory transmission) [47,74,75];5) even in the case of ionotropic receptors, there arelikely a number of additional ligand-gated ion channels that are still uncharacterized [84].Although the strength of prediction of our work is generally acceptable ( > C . elegans using expressiondata, for the first time. We developed and applied a novel method that combines connectivitywith presynaptic and postsynaptic gene expression data and made its tool available for users atthe website http://EleganSign.linkgroup.hu/. Amongst ionotropic chemical synaptic connec-tions, balance of excitatory and inhibitory connections similar to other real-world networkscan be well approximated only if one considers both pre- and postsynaptic gene expression, aconcept that was lacking from previous work. Our method opens a way to include spatial andtemporal dynamics of synaptic polarity that would add a new dimension of plasticity in theexcitatory:inhibitory balance. When sufficient data is available, our polarity prediction methodcan be applied to any neuronal (and as a concept non-neuronal) network. Methods
Description of C . elegans connectome data Connectome reconstruction of the adult hermaphrodite worm published by WormWiring(http://wormwiring.org) consists of 3,638 chemical connections and 2,167 gap junctions,
PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 11 / 19 onnecting 300 neurons (the two canal-associated neurons, CANL and CANR remained iso-lated in this reconstruction, and therefore were omitted from the connectome-related analy-ses). Each of the connections has 4 attributes: the presynaptic neuron, the postsynapticneuron, the type of the connection (chemical or electrical), and the number of synapses. Thechemical connections subset consisting of 20,589 synapses connecting 297 neurons was usedin our work (the sensory neuron pair PLML and PLMR, and the pharyngeal neuron M5 is iso-lated in the chemical synapse network). Additionally, two other connectome reconstructions–both covering a smaller number of neurons and synapses–were used for validation (S5 Table).
Description of gene expression data and processing
Initially, neuronal binary gene expression data was obtained from a previous publication basedon Wormbase [21]. This was extended with data of neuronal neurotransmitter [10,19,20,61]and receptor [24] expression, and with expression data from the recently published CenGendatabase [23] after transformation to binary information (S1 Text, S6 Table, and S7 Data). Forreceptor expression scoring, only the genes coding postsynaptic ionotropic receptor subunitswere evaluated according to the six functional classes based on their suggested neurotransmit-ter ligand (glutamate, acetylcholine or GABA) and putative ion channel type. Cation andanion channel genes were categorized as excitatory and inhibitory, respectively. Expression ofone or more genes in a functional class was considered positive.
Prediction of synaptic polarities
Polarities were predicted for connections based on presynaptic neurotransmitter and postsynap-tic receptor expression data, using nested logical and conditional formulas. In case of themethod referenced as the
NT+R method throughout the paper, synapses were predicted as excit-atory or inhibitory if only cation channel or only anion channel receptor genes matched the pre-synaptic neurotransmitter, respectively; complex if both types of receptor genes matched; and unpredicted if no receptor gene matched. Alternative prediction methods were used accordingto the following rules. NT-only method : synapses were predicted excitatory or inhibitory if thepresynaptic neurotransmitter was acetylcholine and/or glutamate or GABA, respectively; com-plex if acetylcholine and/or glutamate and GABA; and unpredicted if the neurotransmitter wasnone of these. R-only method : synapses were predicted excitatory or inhibitory if the postsynapticreceptor genes expressed were only cation channel or anion channel coding, respectively; com-plex if both types of ion channel receptor genes were expressed; and unpredicted if no receptorgene was expressed. Exact formulas are available in Supplementary Data. Software and data
Data were processed and predictions were made using Microsoft Excel (ver. 16.32) and R(RStudio 1.1.456) using standard packages.
Supporting information
S1 Fig. Proportions of predicted synaptic polarities in alternative C . elegans neuronal net-works. Predictions were made based on the neurotransmitter and receptor gene expressionpatterns of the presynaptic and postsynaptic neurons, respectively (
NT+R method , see Meth-ods). Red, blue, and grey colors mark inhibitory, excitatory, and complex polarities, respec-tively. (A)
Excitatory-inhibitory balances in alternative networks of the WormWiringconnectome reconstruction. Bars from top to bottom: 1. synapse weighted network for com-parative purpose (same as in Fig 3A); 2. weak links (defined by synapse number of 1) deleted
PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 12 / 19 (B)
Predicted synaptic polarities for two connectome recon-structions other than Wormwiring, covering a variable number of neurons and synapses [2,3](S5 Table). In summary, excitatory:inhibitory sign-balance ratios were similar in all cases,ranging between 3.1–4.1. Source data are provided in S1, S4 and S5 Data.(TIF)
S2 Fig. Separate representation of the locomotion subnetwork of C . elegans . Figure is a splitnetwork representation of Fig 4B. Edges represent excitatory (blue), inhibitory (red), or com-plex (black) chemical connections. Edges are weighted according to synapse number. Theshape of vertices ( Δ , � , � ) represent the modality (sensory, inter, motor, respectively) of neu-rons. (A) Head circuit neurons. (B)
Ventral nerve cord motor neurons. Colors as in Fig 4B.(TIF)
S1 Data. Prediction of synaptic signs based on neurotransmitter and receptor expressiondata. (XLSX)
S2 Data. Prediction of synaptic signs based on neurotransmitter expression data. (XLSX)
S3 Data. Prediction of synaptic signs based on neurotransmitter receptor expression data. (XLSX)
S4 Data. Prediction of synaptic and edge signs based on neurotransmitter and receptorexpression data in different subnetworks. (XLSX)
S5 Data. Prediction of synaptic signs based on neurotransmitter and receptor expressiondata (alternative connectome reconstructions). (XLSX)
S6 Data. Summary of predictions as in S1, S4, and S5 Data after exclusion of the RNA-seqdataset. (XLSX)
S7 Data. Utilization and curation of data sources. (XLSX)
S1 File. High-resolution vector graphic version of Fig 5. (PDF)
S1 Table. Channel type (cation or anion) of ionotropic neurotransmitter receptor genes of C . elegans . The
Caenorhabditis elegans genome contains 62 ionotropic postsynaptic receptorgenes for glutamate, acetylcholine, and GABA. In this table genes are listed in alphabeticorder, and the type of channel ( + for cation channel,–for anion channel) is presented with rele-vant reference. The 42 genes that were expressed postsynaptically in at least one neuron in ourdatabase (marked bold ) have been validated for being cationic or anionic.(DOCX) S2 Table. Distribution of neurons according to the number of ionotropic neurotransmitterreceptor genes expressed.
Neuronal gene expression database was compiled from availabledatasets and manually curated (Methods). Genes encoding ionotropic receptors for glutamate,
PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 13 / 19 cetylcholine or GABA were grouped according to the type of ion channel (cation or anion),i.e. whether being excitatory or inhibitory. Bold numbers represent number of neuronsexpressing certain numbers of excitatory and/or inhibitory receptor genes. 63 neurons expressonly cation-channel receptor genes (green), while 48 neurons express only anion-channelreceptor genes (red). 151 neurons express a mixture of cation- and anion-channel receptorgenes (grey). Source data is in S1 Data.(DOCX)
S3 Table. Validation of results with a previous synaptic polarity prediction paper.
Pre-dicted polarities from our results (S1 Data) were compared to the polarities predicted byRakowski and Karbowski, 2017, for the locomotion circuit of the C . elegans connectome. Eachcell represents a connection between the named source and target neuron. Green colored cellmeans that the predicted polarity was the same in both cases. Orange colored cell means thatthe predicted polarity was different with the two methods. 456 of the 652 synapses (70%) werepredicted the same.(DOCX) S4 Table. Validation of predictions with previously published experimental results.
Pre-dicted polarities from our results (S1 Data) were individually compared to previously pub-lished experimental data. Each row represents a single connection. If validated (“Yes” incolumn “Validated?”), the
NT+R predicted polarity equals the reference polarity. Partialvalidation means that the predicted and/or reference polarity was complex or uncertain.(DOCX)
S5 Table. Comparison of three chemical synapse connectome reconstructions.
The threemost complete connectome reconstructions of C. elegans, namely the WormWiring (http://wormwiring.org), as well as published by Varshney et al., 2011, and Cook et al., 2019, havefundamental differences in their coverage of chemical connections and synapse numbers.(DOCX)
S6 Table. Manual curation and edits.
Bulk gene expression data was updated manuallyaccording to literature data. Green and red text shows receptor gene additions and deletions,respectively, to (from) specified neuron groups. Blue text shows neurotransmitter expressionadditions. acc-4 and lgc-46 genes were excluded to avoid false predictions because of literatureevidence supporting a presynaptic localization rather than postsynaptic. All neurons of a neu-ron group were updated unless specified otherwise.(DOCX)
S1 Text. Gene expression. (DOCX)
Acknowledgments
We thank members of the LINK network science group (http://linkgroup.hu) for their helpfulcomments.
Author Contributions
Conceptualization:
Ba´nk G. Fenyves, Peter Csermely.
Data curation:
Ba´nk G. Fenyves, Ga´bor S. Szila´gyi.
Formal analysis:
Ba´nk G. Fenyves, Ga´bor S. Szila´gyi.
PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 14 / 19 unding acquisition:
Ba´nk G. Fenyves, Csaba S ő ti, Peter Csermely. Investigation:
Ba´nk G. Fenyves, Ga´bor S. Szila´gyi.
Methodology:
Ba´nk G. Fenyves, Ga´bor S. Szila´gyi.
Project administration:
Peter Csermely.
Software:
Zsolt Vassy.
Supervision:
Csaba S ő ti, Peter Csermely. Visualization:
Ba´nk G. Fenyves.
Writing – original draft:
Ba´nk G. Fenyves.
Writing – review & editing:
Ba´nk G. Fenyves, Ga´bor S. Szila´gyi, Csaba S ő ti, Peter Csermely. References White JG, Southgate E, Thomson JN, Brenner S. The structure of the nervous system of the nematode
Caenorhabditis elegans . Philos Trans R Soc London B, Biol Sci. 1986; 314: 1–340. https://doi.org/10.1098/rstb.1986.0056 PMID: 22462104 Varshney LR, Chen BL, Paniagua E, Hall DH, Chklovskii DB. Structural properties of the
Caenorhabdi-tis elegans neuronal network. PLoS Comput Biol. 2011; 7: e1001066. https://doi.org/10.1371/journal.pcbi.1001066 PMID: 21304930 Cook SJ, Jarrell TA, Brittin CA, Wang Y, Bloniarz AE, Yakovlev MA, et al. Whole-animal connectomesof both
Caenorhabditis elegans sexes. Nature. 2019; 571: 63–71. https://doi.org/10.1038/s41586-019-1352-7 PMID: 31270481 Wicks SR, Roehrig CJ, Rankin CH. A dynamic network simulation of the nematode tap withdrawal cir-cuit: predictions concerning synaptic function using behavioral criteria. J Neurosci. 1996; 16: 4017–4031. https://doi.org/10.1523/JNEUROSCI.16-12-04017.1996 PMID: 8656295 Rakowski F, Srinivasan J, Sternberg PW, Karbowski J. Synaptic polarity of the interneuron circuit con-trolling C . elegans locomotion. Front Comput Neurosci. 2013; 7: 128. https://doi.org/10.3389/fncom.2013.00128 PMID: 24106473 Dong C-Y, Cho K-H. An optimally evolved connective ratio of neural networks that maximizes the occur-rence of synchronized bursting behavior. BMC Syst Biol. 2012; 6: 23. https://doi.org/10.1186/1752-0509-6-23 PMID: 22462685 Izquierdo EJ, Beer RD. Connecting a connectome to behavior: An ensemble of neuroanatomical mod-els of C . elegans klinotaxis. Graham LJ, editor. PLoS Comput Biol. 2013; 9: e1002890. https://doi.org/10.1371/journal.pcbi.1002890 PMID: 23408877 Cangelosi A, Parisi D. A Neural network model of
Caenorhabditis elegans : The circuit of touch sensitiv-ity. Neural Process Lett. 1997; 6: 91–98. https://doi.org/10.1023/A:1009615807222 McIntire SL, Jorgensen E, Kaplan J, Horvitz HR. The GABAergic nervous system of
Caenorhabditis ele-gans . Nature. 1993; 364: 337–341. https://doi.org/10.1038/364337a0 PMID: 8332191
Gendrel M, Atlas EG, Hobert O. A cellular and regulatory map of the GABAergic nervous system of C . elegans . Elife. 2016; e17686. https://doi.org/10.7554/eLife.17686 PMID: 27740909 Rubin R, Abbott LF, Sompolinsky H. Balanced excitation and inhibition are required for high-capacity,noise-robust neuronal selectivity. Proc Natl Acad Sci. 2017; 114: E9366–E9375. https://doi.org/10.1073/pnas.1705841114 PMID: 29042519
Bhatia A, Moza S, Bhalla US. Precise excitation-inhibition balance controls gain and timing in the hippo-campus. Elife. 2019; 8: e43415. https://doi.org/10.7554/eLife.43415 PMID: 31021319
Baker C, Ebsch C, Lampl I, Rosenbaum R. Correlated states in balanced neuronal networks. Phys RevE. 2019; 99: 52414. https://doi.org/10.1103/PhysRevE.99.052414 PMID: 31212573
Putrenko I, Zakikhani M, Dent JA. A family of acetylcholine-gated chloride channel subunits in
Caenor-habditis elegans . J Biol Chem. 2005; 280: 6392–6398. https://doi.org/10.1074/jbc.M412644200 PMID:15579462
Li Z, Liu J, Zheng M, Xu XZS. Encoding of both analog- and digital-like behavioral outputs by one C . ele-gans interneuron. Cell. 2014; 159: 751–765. https://doi.org/10.1016/j.cell.2014.09.056 PMID:25417153 PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 15 / 19 Chalasani SH, Chronis N, Tsunozaki M, Gray JM, Ramot D, Goodman MB, et al. Dissecting a circuit forolfactory behaviour in
Caenorhabditis elegans . Nature. 2007; 450: 63–70. https://doi.org/10.1038/nature06292 PMID: 17972877
Marom S, Shahaf G. Development, learning and memory in large random networks of cortical neurons:lessons beyond anatomy. Q Rev Biophys. 2002; 35: 63–87. https://doi.org/10.1017/s0033583501003742 PMID: 11997981
Pastore VP, Massobrio P, Godjoski A, Martinoia S. Identification of excitatory-inhibitory links and net-work topology in large-scale neuronal assemblies from multi-electrode recordings. PLOS Comput Biol.2018; 14: e1006381. https://doi.org/10.1371/journal.pcbi.1006381 PMID: 30148879
Loer CM, Rand JB. The evidence for classical neurotransmitters in
Caenorhabditis elegans . Altun ZF,Herndon LA, editors. WormAtlas. 2016. https://doi.org/10.3908/wormatlas.5.200
Pereira L, Kratsios P, Serrano-Saiz E, Sheftel H, Mayo AE, Hall DH, et al. A cellular and regulatory mapof the cholinergic nervous system of C . elegans . Elife. 2015; 4: e12432. https://doi.org/10.7554/eLife.12432 PMID: 26705699 Hobert O, Glenwinkel L, White J. Revisiting neuronal cell type classification in
Caenorhabditis elegans .Curr Biol. 2016; 26: R1197–R1203. https://doi.org/10.1016/j.cub.2016.10.027 PMID: 27875702
Serrano-Saiz E, Poole RJ, Felton T, Zhang F, De La Cruz ED, Hobert O. Modular control of glutamater-gic neuronal identity in C . elegans by distinct homeodomain proteins. Cell. 2013; 155: 659–673. https://doi.org/10.1016/j.cell.2013.09.052 PMID: 24243022 Taylor SR, Santpere G, Reilly M, Glenwinkel L, Poff A, McWhirter R, et al. Expression profiling of themature C . elegans nervous system by single-cell RNA-sequencing. bioRxiv. 2019; 737577. https://doi.org/10.1101/737577 Altun ZF. Neurotransmitter receptors in
Caenorhabditis elegans . WormAtlas. 2011. https://doi.org/10.3908/wormatlas.5.202
Martikainen MH, Kaneko KI, Hari R. Suppressed responses to self-triggered sounds in the human audi-tory cortex. Cereb Cortex. 2005; 15: 299–302. https://doi.org/10.1093/cercor/bhh131 PMID: 15238430
Dalenoort G, de Vries PH. The essential role of binding for cognition in living systems. In: Schaub H,Detje F, Bruggemann U, editors. Logic af artificial life: Abstracting and synthesizing the principles of liv-ing systems. Berlin: Aka GmbH; 2004. pp. 32–39.
A´ gg B, Csa´sza´r A, Szalay-Bek ő M, Veres DV., Mizsei R, Ferdinandy P, et al. The EntOptLayout Cytos-cape plug-in for the efficient visualization of major protein complexes in protein–protein interaction andsignalling networks. Bioinformatics. 2019; 35: 4490–4492. https://doi.org/10.1093/bioinformatics/btz257 PMID: 31004478
Buzsa´ki G, Kaila K, Raichle M. Inhibition and brain work. Neuron. 2007; 56: 771–783. https://doi.org/10.1016/j.neuron.2007.11.008 PMID: 18054855
Markram H, Toledo-Rodriguez M, Wang Y, Gupta A, Silberberg G, Wu C. Interneurons of the neocorti-cal inhibitory system. Nat Rev Neurosci. 2004; 5: 793–807. https://doi.org/10.1038/nrn1519 PMID:15378039
Rakowski F, Karbowski J. Optimal synaptic signaling connectome for locomotory behavior in
Caenor-habditis elegans : Design minimizing energy cost. PLOS Comput Biol. 2017; 13: e1005834. https://doi.org/10.1371/journal.pcbi.1005834 PMID: 29155814
Sohal VS, Rubenstein JLR. Excitation-inhibition balance as a framework for investigating mechanismsin neuropsychiatric disorders. Mol Psychiatry. 2019/05/14. 2019; 24: 1248–1257. https://doi.org/10.1038/s41380-019-0426-0 PMID: 31089192
Błaszczyk JW. Parkinson’s Disease and neurodegeneration: GABA-collapse hypothesis. Front Neu-rosci. 2016; 10: 269. https://doi.org/10.3389/fnins.2016.00269 PMID: 27375426
Liu G. Local structural balance and functional interaction of excitatory and inhibitory synapses in hippo-campal dendrites. Nat Neurosci. 2004; 7: 373–379. https://doi.org/10.1038/nn1206 PMID: 15004561
Markram H, Muller E, Ramaswamy S, Reimann MW, Abdellah M, Sanchez CA, et al. Reconstructionand simulation of neocortical microcircuitry. Cell. 2015; 163: 456–492. https://doi.org/10.1016/j.cell.2015.09.029 PMID: 26451489
Gulya´s AI, Megı´as M, Emri Z, Freund TF. Total number and ratio of excitatory and inhibitory synapsesconverging onto single interneurons of different types in the CA1 area of the rat hippocampus. J Neu-rosci. 1999; 19: 10082–10097. https://doi.org/10.1523/JNEUROSCI.19-22-10082.1999 PMID:10559416
Nakanishi K, Kukita F. Intracellular [Cl − ] modulates synchronous electrical activity in rat neocortical neu-rons in culture by way of GABAergic inputs. Brain Res. 2000; 863: 192–204. https://doi.org/10.1016/s0006-8993(00)02152-1 PMID: 10773207 PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 16 / 19 Leskovec J, Huttenlocher D, Kleinberg J. Signed networks in social media. Proceedings of the SIGCHIConference on Human Factors in Computing Systems. New York, NY, USA: ACM; 2010. pp. 1361–1370.
Kirkley A, Cantwell GT, Newman MEJ. Balance in signed networks. Phys Rev E. 2019; 99: 012320.https://doi.org/10.1103/PhysRevE.99.012320 PMID: 30780212
Beg AA, Jorgensen EM. EXP-1 is an excitatory GABA-gated cation channel. Nat Neurosci. 2003; 6:1145–1152. https://doi.org/10.1038/nn1136 PMID: 14555952
Martinez-Torres A, Miledi R. Expression of
Caenorhabditis elegans neurotransmitter receptors and ionchannels in Xenopus oocytes. Proc Natl Acad Sci. 2006; 103: 5120–5124. https://doi.org/10.1073/pnas.0600739103 PMID: 16549772
Kehoe J, McIntosh JM. Two distinct nicotinic receptors, one pharmacologically similar to the vertebrate α Aplysia neurons. J Neurosci. 1998; 18: 8198–8213.https://doi.org/10.1523/JNEUROSCI.18-20-08198.1998 PMID: 9763466
Cully DF, Paress PS, Liu KK, Schaeffer JM, Arena JP. Identification of a
Drosophila melanogaster gluta-mate-gated chloride channel sensitive to the antiparasitic agent avermectin. J Biol Chem. 1996; 271:20187–20191. https://doi.org/10.1074/jbc.271.33.20187 PMID: 8702744
Liu WW, Wilson RI. Glutamate is an inhibitory neurotransmitter in the
Drosophila olfactory system. ProcNatl Acad Sci U S A. 2013; 110: 10294–10299. https://doi.org/10.1073/pnas.1220560110 PMID:23729809
Kullmann PHM, Ene FA, Kandler K. Glycinergic and GABAergic calcium responses in the developinglateral superior olive. Eur J Neurosci. 2002; 15: 1093–1104. https://doi.org/10.1046/j.1460-9568.2002.01946.x PMID: 11982621
Wolstenholme AJ. Glutamate-gated chloride channels. J Biol Chem. 2012; 287: 40232–40238. https://doi.org/10.1074/jbc.R112.406280 PMID: 23038250
Baraba´si DL, Baraba´si A-L. A genetic model of the connectome. Neuron. 2020; 105: 435–445.e5.https://doi.org/10.1016/j.neuron.2019.10.031 PMID: 31806491
Bentley B, Branicky R, Barnes CL, Chew YL, Yemini E, Bullmore ET, et al. The multilayer connectomeof
Caenorhabditis elegans . Jbabdi S, editor. PLOS Comput Biol. 2016; 12: e1005283. https://doi.org/10.1371/journal.pcbi.1005283 PMID: 27984591
Sarma GP, Lee CW, Portegys T, Ghayoomie V, Jacobs T, Alicea B, et al. OpenWorm: Overview andrecent advances in integrative biological simulation of
Caenorhabditis elegans . Philos Trans R Soc BBiol Sci. 2018; 373. https://doi.org/10.1098/rstb.2017.0382 PMID: 30201845
Zhou D, Rangan AV, McLaughlin DW, Cai D. Spatiotemporal dynamics of neuronal populationresponse in the primary visual cortex. Proc Natl Acad Sci. 2013; 110: 9517–9522. https://doi.org/10.1073/pnas.1308167110 PMID: 23696666
Tao L, Porto D, Li Z, Fechner S, Lee SA, Goodman MB, et al. Parallel processing of two mechanosen-sory modalities by a single neuron in C . elegans . Dev Cell. 2019; 51: 543–658. https://doi.org/10.1016/j.devcel.2019.11.010 PMID: 31794713 Nusser Z. Subcellular distribution of neurotransmitter receptors and voltage-gated ion channels. In: Stu-art G, Spruston N, Hausser M, editors. Dendrites. Oxford University Press; 2012. pp. 154–188. https://doi.org/10.1093/acprof:oso/9780198566564.003.0007
Megı´as M, Emri Z, Freund TF, Gulya´s AI. Total number and distribution of inhibitory and excitatory syn-apses on hippocampal CA1 pyramidal cells. Neuroscience. 2001; 102: 527–540. https://doi.org/10.1016/s0306-4522(00)00496-6 PMID: 11226691
Zou W, Fu J, Zhang H, Du K, Huang W, Yu J, et al. Decoding the intensity of sensory input by two gluta-mate receptors in one C . elegans interneuron. Nat Commun. 2018; 9: 4311. https://doi.org/10.1038/s41467-018-06819-5 PMID: 30333484 Arey RN, Kaletsky R, Murphy CT. Nervous system-wide profiling of presynaptic mRNAs reveals regula-tors of associative memory. Sci Rep. 2019; 9: 20314. https://doi.org/10.1038/s41598-019-56908-8PMID: 31889133
Stetak A, Ho¨rndli F, Maricq AV, van den Heuvel S, Hajnal A. Neuron-specific regulation of associativelearning and memory by MAGI-1 in C . elegans . PLoS One. 2009; 4: e6019. https://doi.org/10.1371/journal.pone.0006019 PMID: 19551147 Choi S, Taylor KP, Chatzigeorgiou M, Hu Z, Schafer WR, Kaplan JM. Sensory neurons arouse C . ele-gans locomotion via both glutamate and neuropeptide release. PLOS Genet. 2015; 11: e1005359.https://doi.org/10.1371/journal.pgen.1005359 PMID: 26154367 Shinkai Y, Yamamoto Y, Fujiwara M, Tabata T, Murayama T, Hirotsu T, et al. Behavioral choicebetween conflicting alternatives is regulated by a receptor guanylyl cyclase, GCY-28, and a receptor
PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 17 / 19yrosine kinase, SCD-2, in AIA interneurons of
Caenorhabditis elegans . J Neurosci. 2011; 31: 3007–3015. https://doi.org/10.1523/JNEUROSCI.4691-10.2011 PMID: 21414922
Chalasani SH, Kato S, Albrecht DR, Nakagawa T, Abbott LF, Bargmann CI. Neuropeptide feedbackmodifies odor-evoked dynamics in
Caenorhabditis elegans olfactory neurons. Nat Neurosci. 2010; 13:615–621. https://doi.org/10.1038/nn.2526 PMID: 20364145
Kuramochi M, Doi M. An excitatory/inhibitory switch from asymmetric sensory neurons defines postsyn-aptic tuning for a rapid response to NaCl in
Caenorhabditis elegans . Front Mol Neurosci. 2019; 11: 484.https://doi.org/10.3389/fnmol.2018.00484 PMID: 30687001
Witvliet D, Mulcahy B, Mitchell JK, Meirovitch Y, Berger DK, Wu Y, et al. Connectomes across develop-ment reveal principles of brain maturation in C . elegans . bioRxiv. 2020. https://doi.org/10.1101/2020.04.30.066209 Serrano-Saiz E, Pereira L, Gendrel M, Aghayeva U, Battacharya A, Howell K, et al. A neurotransmitteratlas of the
Caenorhabditis elegans male nervous system reveals sexually dimorphic neurotransmitterusage. Genetics. 2017; 206: 1251–1269. https://doi.org/10.1534/genetics.117.202127 PMID:28684604
Kunitomo H, Sato H, Iwata R, Satoh Y, Ohno H, Yamada K, et al. Concentration memory-dependentsynaptic plasticity of a taste circuit regulates salt concentration chemotaxis in
Caenorhabditis elegans .Nat Commun. 2013; 4: 2210. https://doi.org/10.1038/ncomms3210 PMID: 23887678
Ho VM, Lee J-A, Martin KC. The cell biology of synaptic plasticity. Science. 2011; 334: 623–628. https://doi.org/10.1126/science.1209236 PMID: 22053042
Hadziselimovic N, Vukojevic V, Peter F, Milnik A, Fastenrath M, Fenyves BG, et al. Forgetting is regu-lated via Musashi-mediated translational control of the Arp2/3 complex. Cell. 2014; 156: 1153–1166.https://doi.org/10.1016/j.cell.2014.01.054 PMID: 24630719
Ingrosso A, Abbott LF. Training dynamically balanced excitatory-inhibitory networks. PLoS One. 2019;14: e0220547. https://doi.org/10.1371/journal.pone.0220547 PMID: 31393909
Freytag V, Probst S, Hadziselimovic N, Boglari C, Hauser Y, Peter F, et al. Genome-wide temporalexpression profiling in
Caenorhabditis elegans identifies a core gene set related to long-term memory. JNeurosci. 2017; 37: 6661–6672. https://doi.org/10.1523/JNEUROSCI.3298-16.2017 PMID: 28592692
Hangya B, Ranade SP, Lorenc M, Kepecs A. Central cholinergic neurons are rapidly recruited by rein-forcement feedback. Cell. 2015; 162: 1155–1168. https://doi.org/10.1016/j.cell.2015.07.057 PMID:26317475
Hoerndli FJ, Walser M, Fro¨hli Hoier E, de Quervain D, Papassotiropoulos A, Hajnal A. A conservedfunction of C . elegans CASY-1 calsyntenin in associative learning. PLoS One. 2009; 4: e4880. https://doi.org/10.1371/journal.pone.0004880 PMID: 19287492
Hammond-Weinberger DR, Wang Y, Glavis-Bloom A, Spitzer NC. Mechanism for neurotransmitter-receptor matching. Proc Natl Acad Sci. 2020; 117: 4368–4374. https://doi.org/10.1073/pnas.1916600117 PMID: 32041885
Spitzer NC. Neurotransmitter switching in the developing and adult brain. Annu Rev Neurosci. 2017; 40:1–19. https://doi.org/10.1146/annurev-neuro-072116-031204 PMID: 28301776
Tritsch NX, Granger AJ, Sabatini BL. Mechanisms and functions of GABA co-release. Nat Rev Neu-rosci. 2016; 17: 139–145. https://doi.org/10.1038/nrn.2015.21 PMID: 26865019
Bianconi G. Multilayer Networks: Structure and Function. Oxford: Oxford University Press; 2018.https://doi.org/10.1093/oso/9780198753919.001.0001
Pournaki A, Merfort L, Ruiz J, Kouvaris NE, Ho¨vel P, Hizanidis J. Synchronization patterns in modularneuronal n etworks: A case study of C . elegans . Front Appl Math Stat. 2019; 5: 52. https://doi.org/10.3389/fams.2019.00052 Chase D. Biogenic amine neurotransmitters in C . elegans . WormBook. 2007. https://doi.org/10.1895/wormbook.1.132.1 PMID: 18050501 Bentley B. Connectomics of extrasynaptic signalling: applications to the nervous system of Caenorhab-ditis elegans [Doctoral thesis]. University of Cambridge. 2017.
Jobson MA, Valdez CM, Gardner J, Garcia LR, Jorgensen EM, Beg AA. Spillover transmission is medi-ated by the excitatory GABA receptor LGC-35 in C . elegans . J Neurosci. 2015; 35: 2803–2816. https://doi.org/10.1523/JNEUROSCI.4557-14.2015 PMID: 25673867 Niswender CM, Conn PJ. Metabotropic glutamate receptors: physiology, pharmacology, and disease.Annu Rev Pharmacol Toxicol. 2010; 50: 295–322. https://doi.org/10.1146/annurev.pharmtox.011008.145533 PMID: 20055706
Crupi R, Impellizzeri D, Cuzzocrea S. Role of metabotropic glutamate receptors in neurological disor-ders. Front Mol Neurosci. 2019; 12: 20. https://doi.org/10.3389/fnmol.2019.00020 PMID: 30800054
PLOS COMPUTATIONAL BIOLOGY
Synaptic polarity balance in a neuronal networkPLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1007974 December 21, 2020 18 / 19 Koelle MR. Neurotransmitter signaling through heterotrimeric G proteins: insights from studies in C . ele-gans . WormBook. 2018. https://doi.org/10.1895/wormbook.1.75.2 PMID: 26937633 Reiner A, Levitz J. Glutamatergic signaling in the central nervous system: Ionotropic and metabotropicreceptors in concert. Neuron. 2018; 98: 1080–1098. https://doi.org/10.1016/j.neuron.2018.05.018PMID: 29953871
Yemini E, Lin A, Nejatbakhsh A, Varol E, Sun R, Mena GE, et al. NeuroPAL: A neuronal polychromaticatlas of landmarks for whole-brain imaging in C . elegans . bioRxiv. 2019; 676312. https://doi.org/10.1101/676312 Branicky R, Miyazaki H, Strange K, Schafer WR. The voltage-gated anion channels encoded by clh-3regulate egg laying in C . elegans by modulating motor neuron excitability. J Neurosci. 2014; 34: 764–775. https://doi.org/10.1523/JNEUROSCI.3112-13.2014 PMID: 24431435 Farrant M, Kaila K. The cellular, molecular and ionic basis of GABAA receptor signalling. Prog BrainRes. 2007; 160: 59–87. https://doi.org/10.1016/S0079-6123(06)60005-8 PMID: 17499109
Jones A, Sattelle D. The cys-loop ligand-gated ion channel gene superfamily of the nematode,
Caenor-habditis elegans . Invert Neurosci. 2008; 8: 41–47. https://doi.org/10.1007/s10158-008-0068-4 PMID:18288508
Milo R, Shen-Orr S, Itzkovitz S, Kashtan N, Chklovskii D, Alon U. Network motifs: simple building blocksof complex networks. Science. 2002; 298: 824–827. https://doi.org/10.1126/science.298.5594.824PMID: 12399590