Functional strengthening through synaptic scaling upon connectivity disruption in neuronal cultures

RESEARCH Func ional s eng hening h ough synap ic scaling upon connec i i y dis up ion in neu onal cul u es Es e anía Es é ez-P iego1,2, Sa a Telle 1,2, Cla a G anell3,4, Alex A enas5, and Jo di So iano 1,2 1Depa amen de Física de la Ma è ia Condensada, Uni e si a de Ba celona, Ba celona, Spain 2Uni e si a de Ba celona Ins i u e o Complex Sys ems (UBICS), Ba celona, Spain 3GOTHAM Lab – Ins i u e o Biocompu a ion and Physics o Complex Sys ems (BIFI), Uni e si y o Za agoza, Za agoza, Spain 4Depa men o Condensed Ma e Physics, Uni e si y o Za agoza, Za agoza, Spain 5Depa amen d’Enginye ia In o mà ica i Ma emà iques, Uni e si a Ro i a i Vi gili, Ta agona, Spain Keywo ds: Neu onal cul u es, Calcium imaging, CNQX, E ec i e connec i i y, Global e iciency, Func ional o ganiza ion, Synap ic scaling ABSTRACT An elusi e phenomenon in ne wo k neu oscience is he ex en o neu onal ac i i y emodeling upon damage. He e, we in es iga e he ac ion o g adual synap ic blockade on he e ec i e connec i i y in co ical ne wo ks in i o. We use wo neu onal cul u es con igu a ions—one o med by abou 130 neu onal agg ega es and ano he one o med by abou 600 indi idual neu ons—and moni o hei spon aneous ac i i y upon p og essi e weakening o exci a o y connec i i y. We epo ha he e ec i e connec i i y in all cul u es exhibi s a i s phase o ansien s eng hening ollowed by a second phase o s eady de e io a ion. We quan i y hese phases by measu ing GEFF, he global e iciency in p ocessing ne wo k in o ma ion. We e m hype e iciency he sudden s eng hening o GEFF upon ne wo k de e io a ion, which inc eases by 20–50% depending on cul u e ype. Relying on nume ical simula ions we e eal he ole o synap ic scaling, an ac i i y–dependen mechanism o synap ic plas ici y, in coun e ac ing he pe u ba i e ac ion, nea ly ep oducing he obse ed hype e iciency. Ou esul s demons a e he impo ance o synap ic scaling as esilience mechanism. AUTHOR SUMMARY Neu onal ci cui s exhibi homeos a ic plas ici y mechanisms o cope wi h pe u ba ions o damage. A cen al mechanism is ‘synap ic scaling,’ a sel -o ganized esponse in which he s eng h o neu ons’ exci a o y synapses is adjus ed o compensa e o ac i i y a ia ions. He e we p esen expe imen s in which he exci a o y connec i i y o in i o co ical ne wo ks is p og essi ely weakened h ough chemical ac ion. The spon aneous ac i i y and e ec i e connec i i y o he whole ne wo k is moni o ed as deg ada ion p og esses, and he capaci y o he ne wo k o b oad in o ma ion communica ion is quan i ied h ough he global e iciency. We obse ed ha he ne wo k esponded o he pe u ba ion by s eng hening he e ec i e connec i i y, eaching a hype e icien s a e o mode a e pe u ba ions. The s udy p o es he impo ance o ‘synap ic scaling’ as a d i e o unc ional eo ganiza ion and ne wo k-wide esilience. a n o p e n a c c e s s j o u n a l Ci a ion: Es é ez-P iego, E., Telle , S., G anell, C., A enas, A., & So iano, J. (2020). Func ional s eng hening h ough synap ic scaling upon connec i i y dis up ion in neu onal cul u es. Ne wo k Neu oscience, 4(4), 1160–1180. h ps://doi.o g/10.1162 /ne n_a_00156 DOI: h ps://doi.o g/10.1162/ne n_a_00156 Recei ed: 16 Janua y 2020 Accep ed: 15 July 2020 Compe ing In e es s: The au ho s ha e decla ed ha no compe ing in e es s exis . Co esponding Au ho : Jo di So iano [email p o ec ed] Handling Edi o : Ola Spo ns Copy igh : ©2020 Massachuse s Ins i u e o Technology Published unde a C ea i e Commons A ibu ion 4.0 In e na ional (CC BY 4.0) license The MIT P ess Func ional s eng hening h ough synap ic scaling INTRODUCTION Response o pe u ba ions o damage in li ing neu onal ci cui s is a cen al ye un esol ed p oblem in neu oscience. A simple ne wo k desc ip ion o damage in ui i ely sugges s ha he ailu e o connec i i y pa hways p ecipi a es a p og essi e de e io a ion o ne wo k ac i i y ha ul ima ely comp omises he unc ionali y o he sys em. Realis ic desc ip ions o li ing neu- onal ci cui s, howe e , expose he complex in e play be ween he physical a chi ec u e o he ne wo k, neu onal dynamics, and plas ici y mechanisms in go e ning ci cui beha io , aming scena ios in which esponse o damage is no only possible, bu as and ne wo kwide. This is pa icula ly impo an in he con ex o b ain and neu ological diso de s, since adequa e p ognosis o unc ional al e a ions may help o elucida e ac ions o s op o e e damage. S udies in s oke, o ins ance, ha e poin ed ou he impo an ole o plas ici y in p e en - ing a cascade o deg ada ion and os e ing eco e y (Calab esi, Cen onze, Pisani, Cupini, & Be na di, 2003;Mu phy & Co be ,2009). Cu en mul idisciplina y app oaches o p ognosis a e making an e o o in eg a e acqui ed expe imen al da a wi h nume ical models o cha ac- e ize neu onal damage and subsequen ne wo k esponses (Basse & Spo ns,2017;Fo ni o, Zalesky, & B eakspea , 2015). Thus, unde s anding he di e en ac o s a play—connec i i y, dynamics, and plas ici y—and hei in e ela ion has become a cen al goal in medical ne wo k neu oscience. In he las ew yea s, subs an ial e o s ha e been in es ed in linking s uc u al cha ac e - is ics o neu onal ne wo ks wi h hei capaci y o cope wi h pe u ba ions and damage (Ae s, Fias, Caeyenbe ghs, & Ma inazzo, 2016;Fa ooq, Chen, Geo giou, Tannenbaum, & Lengle , 2019;Majdandzic e al.,2013). Speci ically o b ain unc ional ne wo ks, s udies un eiled he impo ance o modula and hie a chical o ganiza ions (Meunie , Lambio e, Fo ni o, E sche, & Bullmo e, 2009), hubs ( an den Heu el & Spo ns,2013), and o he node s uc u al aspec s (Basse & Spo ns,2017;Basse , Zu n, & Gold,2018;Spo ns,2014;S am,2014). Despi e he p og ess and enligh enmen in his di ec ion, he inco po a ion o plas ici y and sel - egula o y mechanisms as addi ional ac o s is opening new a enues o unde s anding esilience o in- Resilience: The capaci y o a neu onal ci cui o main ain ac i i y and unc ional cha ac e is ics upon loss o weakening o neu ons o connec ions. sul . Indeed, he b ain and o he li ing neu onal ci cui s a e elian on di e se adap i e mech- anisms o p ope pe o mance and s abili y ha include synap ic plas ici y and s uc u al Synap ic plas ici y: The s eng hening o weakening o synapses acco ding o neu onal ac i i y, o ins ance, o po en ia e speci ic connec i i y pa hways. plas ici y (Bu z, S eenbuck, & an Ooyen,2014;Fau h & Te zla ,2016;Fau h, Wö gö e , & Te zla , 2015). These mechanisms a e cen al o egain ci cui s’ ope abili y upon se e e pe - u ba ions o inju y (Ba al & Reyes,2016;Cos a, Mizusaki, Sjös öm, & an Rossum,2017; Ma de ,2011;Mu phy & Co be ,2009;Telle e al.,2019). One o he mos impo an synap ic plas ici y mechanisms is synap ic scaling, in which he s eng h o exci a o y synapses is ad- Synap ic scaling: A egula o y mechanism o neu onal ci cui s by which synap ic s eng h is adjus ed o compensa e o ch onic changes in ac i i y. jus ed o compensa e o a ia ions in ac i i y (Desai, Ru he o d, & Tu igiano,1999;Fong, Newman, Po e , & Wenne , 2015;G. G. Tu igiano, Leslie, Desai, Ru he o d, & Nelson, 1998). The capaci y o synap ic scaling o egula e neu onal ac i i y was demons a ed bo h in i o (Echegoyen, Neu, G abe , & Sol esz,2007;G. Tu igiano,2012;G. G. Tu igiano,2008) and in i o (Ba al & Reyes,2016;Hanes e al.,2020) a a synapse le el. Al hough di e - en in es iga ions poin ed ou he ole o synap ic scaling in co ec ing neu onal ac i i y a e damage, o ins ance in ela ion o Alzheime ’s pa ien s cogni i e de ici s (Bo ge-Hol hoe e , Mo eno, & A enas, 2011) and e inal lesions (Keck e al.,2013), he e a e no s udies demon- s a ing i s impac in he unc ional o ganiza ion o a neu onal ci cui . Func ional o ganiza ion: A se o global ne wo k ai s, such as modula i y o global e iciency, ha imp in dis inc o ganiza ional cha ac e is ics o a ne wo k. The aim o his s udy is o analyze he ac ion o synap ic scaling om a ne wo k neu osci- ence pe spec i e and expose i s po en ial as a esilience mechanism. To his end, we designed an expe imen al pipeline in which small co ical ci cui s in i o we e exposed o a mal unc ion- like e en ha g aduallyweakened he connec i i y among in e connec ed neu onal assemblies Ne wo k Neu oscience 1 161 Func ional s eng hening h ough synap ic scaling (clus e s), and analyzed simul aneously he ac i i y o he en i e ne wo k. The pe u ba i e ac- ion was deli e ed h ough inc easing doses o CNQX, an AMPA-glu ama e ecep o an agonis CNQX (6-cyano-7-ni oquinoxaline- 2,3-dione): An agonis o AMPA-glu ama e ecep o s in exci a o y neu ons, equen ly used o ei he weaken exci a o y synap ic communica ion o block i en i ely. in exci a o y synapses (So iano, Ma ínez, Tlus y, and Moses,2008;Tibau, Valencia, & So iano, 2013a). Al hough CNQX a ge ed synap ic connec i i y, we measu ed solely i s impac on he spon aneous ac i i y, ha is, he spon aneous dynamic in e ac ions among clus e s. Func- ional cha ac e is ics o he cul u ed ne wo ks we e he e o e quan i ied by measu ing he al e - a ions in ac i i y as he blockade p og essed and by compu ing he global e iciency (La o a & Global e iciency: A measu e o ne wo k capaci y o in o ma ion exchange. The highe he o e all connec i i y among nodes, he highe he global e iciency. Ma chio i, 2001;Rubino & Spo ns,2010), a measu e o he in eg a ion o he low o in o ma- ion in a ne wo ked sys em. Ou expe imen s showed ha , as CNQX was applied, he e was an ini ial phase in which he global e iciency aised, e ealing an inc eased in o ma ion low capaci y in he ne wo k. This hype e iciency was ep oducible among ne wo k ealiza ions, Hype e iciency: In he con ex o ou wo k, an ab up inc ease o global e iciency esul ing om a s eng hening o e ec i e connec ions. al hough he le el o impac depended on cul u e de ails. Beyond his phase, global e iciency p og essi ely decayed un il ac i i y ceased. By using nume ical simula ions we showed ha synap ic scaling, simula ed as a local s eng hening o emaining e ec i e connec ions upon damage, was su icien o explain he expe imen al beha io . Simple models based on bond pe cola ion (Ki kpa ick,1973), ha is, he deg ada ion o dele ion o connec ions wi hou any Pe cola ion: Beha io o in o ma ion o o he measu e as nodes o connec ions a e emo ed, ypically sepa a ing a ‘connec ed’ and a ‘disconnec ed’ phase. scaling, ailed a cap u ing his hype e iciency phenomenon. To ou knowledge, he expe i- men s and modeling p esen ed he e is he i s s ong e idence o he impac o synap ic scaling in p ese ing ne wo k e iciency and whole-sys em pe o mance, hus exposing synap ic scal- ing as an in insic mechanism o esilience in li ing neu onal ne wo ks. Ou s udy s eng hens he ision o synap ic scaling as an eme gen and gene al p ope y o li ing neu onal ci cui s, and pa es he way owa d a be e p edic ion o he ac ion o pe u ba ions and damage in such ci cui s. RESULTS Weakening o Synap ic Communica ion Al e s Func ional O ganiza ion We in es iga ed spon aneous ac i i y in clus e ed neu onal cul u es p epa ed in 6-mm-diame e PDMS wells on glass, as illus a ed in Figu e 1A. These cul u es eme ged as a sel -o ganized p ocess in which he absence o adhesi e p o eins in he glass subs a e a o ed neu onal mobili y and agg ega ion (Sege , Ben enis e, Shapi a, & Ben-Jacob,2003;Telle e al.,2014; Telle , Tahi begi, Mi , Sami ie , & So iano,2015). By day in i o (DIV) 7 upon p epa a ion, agg ega ion shaped highly compac neu onal islands e med clus e s ha emained s able in posi ion and size. In ou expe imen s, abou 130 neu onal clus e s ypically o med in he 6-mm-diame e wells (Figu e 1B). Table 1 summa izes he se o ne wo ks in es iga ed and hei cha ac e is ics. The spon aneous ac i i y in he clus e ed ne wo ks was moni o ed a DIV 8–12 h ough calcium luo escence imaging (Figu e 1B), which allowed us o ollow he e o- lu ion o all clus e s in he cul u e wi h high spa ial and empo al esolu ion. Ac i e clus e s appea ed as b igh objec s whose luo escence in ensi y amped up upon ac i a ion. The ad- an age o clus e ed neu onal cul u es is ha hey exhibi spon aneous ac i i y wi h a ich spa io empo al s uc u e (Telle e al.,2014,2015), in which clus e s coac i a e in g oups o a ying size and iming (Figu e 1B–C). Since he s uc u e o coac i a ion pa e ns, and in u n ne wo k unc ional cha ac e is ics, is g ounded on he coupling among clus e s, hese cul u es s and ou as a pa icula ly sui ed expe imen al p epa a ion o in es iga e he impac o chemi- cal pe u ba ions o damage (Telle e al.,2014 ,2015,2019). The ichly s uc u ed dynamics o clus e ed ne wo ks—and sis e designs in he o m o enginee ed ci cui s (Kanne e al.,2015; Yamamo o e al.,2018)—is enligh ening since dynamics shape ne wo ks wi h a unc ional o - ganiza ion quali a i ely simila o ha o he b ain, wi h modula ai s and hubs (Telle e al., 2015;Yamamo o e al.,2018). Thus, ou clus e ed cul u es combine he con ollabili y and Ne wo k Neu oscience 1162 Func ional s eng hening h ough synap ic scaling Figu e 1. Clus e ed neu onal cul u es and expe imen al p ocedu e. (A) Schema ic p epa a ion o clus e ed neu onal ne wo ks. Dissocia ed a co ical neu ons we e pla ed on a 6-mm-diame e PDMS wells. The absence o adhesi e p o eins acili a ed neu onal agg ega ion, gi ing ise o an assembly o neu onal islands (clus e s) by day in i o (DIV7). (B) Fluo escence image o a ypical neu onal cul u e. Ci cula objec s a e neu onal clus e s. Labels highligh ou he loca ion o ou clus e s whose luo escence aces a e shown in panel C. (C) Rep esen a i e luo escence aces o ou clus e s o g adually highe le els o he AMPA glu ama e ecep o an agonis CNQX. Sha p inc eases o he signal co espond o ac i i y e en s. Yellow boxes illus a e episodes o coo dina ed ac i i y. The o e all ac i i y o he clus e s dec eases wi h CNQX as a esul o he educed exci a o y d i e. DFF indica es no malized luo escence signal, gi en by DFF(%) = 100.(F−F0)/F−0, wi h F0 he luo escence signal a es . accessibili y o an in i o sys em wi h some o he dynamical ichness and unc ional com- plexi y o na u ally o med neu onal ci cui s. In ou expe imen s we we e in e es ed in analyzing he dynamics o a clus e ed ne wo k as he connec i i y s eng h among clus e s, and in u n hei dynamic coupling, was educed. The abili y o ou expe imen al sys em o simul aneously moni o he en i e ne wo k made i unique o in es iga e he ac ion o pe u ba ions and hei impac on ne wo k unc ional o ganiza ion. Since spon aneous coo dina ed ac i i y is mos ly media ed by he exci a o y d i e o AMPA-glu ama e ecep o s, bo h he capaci y o he ne wo k o i e and clus e - o-clus e synap ic coupling could be al e ed in a con olled manne h ough he blockade o AMPA ecep o s. Thus, ollowing So iano, Rod íguez Ma ínez, Tlus y, and Moses (2008), we a ge ed he exci a o y AMPA-glu ama e ecep o s wi h g adually highe doses o he an agonis CNQX, causing a p og essi e educ ion o spon aneous ac i i y un il i ceased. Fo cla i y in he analysis o he da a and i s in e p e a ion, NMDA exci a o y ecep o s (which accoun o abou 20% o Ne wo k Neu oscience 1 163 Func ional s eng hening h ough synap ic scaling Table 1. Summa y o he s uc u al and dynamical cha ac e is ics o he eigh clus e ed ne wo ks in es iga ed. Expe imen DIV Nφ(µm)d(µm)a(bu s s/min) hki hwi 1 8 157 156 ±56 406 ±16 7.10 6.0 ±3.1 2.8 ±0.5 2 8 125 170 ±55 437 ±15 2.52 5.9 ±2.8 3.2 ±0.6 3 8 116 177 ±60 455 ±14 1.47 6.3 ±2.7 2.9 ±0.5 4 9 152 137 ±34 282 ±8 5.51 5.3 ±2.1 3.0 ±0.4 5 9 144 177 ±53 411 ±15 1.71 4.4 ±1.9 3.3 ±0.7 6 9 124 160 ±52 452 ±15 7.14 5.0 ±1.7 2.9 ±0.6 7 9 118 215 ±68 469 ±18 3.85 4.9 ±1.9 3.1 ±1.1 8 12 128 193 ±70 462 ±15 2.91 5.7 ±2.2 2.9 ±0.5 No e. DIV is he day in i o o he cul u e, N he numbe o clus e s, φ hei a e age diame e , d he ypical clus e s’ in e dis ance, a he a e age collec i e ac i i y, hki he a e age e ec i e connec i i y pe clus e , and hwi he a e age weigh o e ec i e connec ions. φ,d,hki, and hwi a e epo ed as he means ±s anda d de ia ions. he o al exci a o y d i e in co ical neu ons) and GABA inhibi o y ecep o s we e ully blocked wi h he co esponding an agonis s. An example o he impac o CNQX on ne wo k spon aneous ac i i y is shown in Figu e 1C. He e, he luo escence aces o ou ep esen a i e clus e s a e depic ed o g adually highe CNQX le els. Wi hou blockade, all ou clus e s exhibi ed a ich ac i i y wi h s ong coo di- na ion. As CNQX was applied, he le el o ac i i y and coo dina ion swi ly dec eased, and o ull synap ic blockade he clus e s ei he i ed weakly and independen ly o became silen . The all o spon aneous ac i i y and he b eakdown o clus e s’ coac i a ion pa e ns o he en i e ne wo k a e shown as as e plo s in Figu e 2A. In he plo s, he do s ma k he iming o clus e s’ ac i a ion, da a ha a e di ec ly ob ained om he analysis o he luo escence signal. In he sequence o as e plo s, [CNQX] =0(unpe u bed ne wo k) shapes a s ongly cohe en dynamics in which mos o he clus e s coac i a e a unison, a ea u e ha e lec s he s ong coupling among clus e s. The s eng h o spon aneous ac i i y was quan i ied h ough he col- lec i e ac i i y a, de ined as he equency o ac i i y episodes (bu s s) in which a leas i e clus- e s pa icipa ed (abou 5% o he ne wo k). Ac i i y awas high o [CNQX] =0, wi h clus e s exhibi ing abou 7 bu s s/min. Ac i i y se e ely d opped o abou 4.4 bu s s/min o he small pe u ba ion wi h [CNQX] =60 nM, which was accompanied by a s ong dis up ion o he unison beha io , ende ing an o e all dynamic in which clus e s ac i a ed in g oups o a ying size. The decay in ac i i y accen ua ed as CNQX g ew, al hough some g oups o clus e s s ill main ained s ong coo dina ed ac i i y e en o ela i ely la ge doses o [CNQX] =600 nM. Spon aneous ac i i y inally s opped by [CNQX] &1, 000 nM. In ou s udy we in es iga ed a o al o eigh clus e ed ne wo ks. All o hem exhibi ed a simila quali a i e end, al hough he s uc u e o coo dina ed ac i i y and i s changes on deg ada ion depended on he clus e - o- clus e wi ing de ails in he cul u e (Table 1). Ac i i y da a was nex analyzed in de ail o compu e he e ec i e connec i i y o he ne - E ec i e connec i i y: Causal s a is ical ela ionship among he ac i i y pa e ns o neu ons in a ne wo k, leading o di ec ed and weigh ed adjacency connec i i y ma ices. wo ks, allowing us o quan i y whole-ne wo k al e a ions on CNQX ac ion. The e ec i e ne - wo ks (weigh ed and di ec ed) we e ex ac ed using ans e en opy, an app oach ha we showed app op ia e o neu onal cul u es in p e ious s udies (S e e , Ba aglia, So iano, & Geisel, 2012;Telle e al.,2014;Tibau, Ludl, Rüdige , O landi, & So iano,2020). F om he in e ed e ec i e ne wo ks we ex ac ed he global e iciency GEFF, which in o ms abou he capabili y o he ne wo k o ope a e as a whole. Fo cla i y, he global e iciency was always Ne wo k Neu oscience 1164 Func ional s eng hening h ough synap ic scaling Figu e 2. Ne wo k ac i i y and unc ional o ganiza ion upon deg ada ion. (A) Rep esen a i e as e plo s o spon aneous ac i i y o g adually highe exci a o y blockade. Blue do s ma k he clus e s’ ac i a ions. The concen a ion o CNQX and he a e age collec i e spon aneous ac i i y aa e in- dica ed abo e each panel. Ras e s a e limi ed o 10 min o cla i y. (B) Co esponding e ec i e connec i i y ma ices. The b igh e he colo , he highe he connec i i y weigh among wo clus- e s. Ma ices a e o de ed acco ding o he unc ional communi ies a [CNQX] = 60 nM. The scaled global e iciency GEFF inc eases a [CNQX] = 60 nM o s eadily dec ease a e wa ds. (C) E ec i e connec i i y maps o pe u bed and con ol expe imen s. Clus e s’ and links’ colo s iden i y di e - en unc ional communi ies. Con ol expe imen s expe ienced he same expe imen al pipeline as pe u bed ones, bu only GABAAand NMDA ecep o s we e blocked. No signi ican al e a ions in he e ec i e connec i i y we e obse ed. The pe u bed expe imen co esponds o he da a shown in (A). scaled ela i e o i s alue in he unpe u bed condi ion. We no e ha he global e iciency is based on he calcula ion o he sho es opological pa hs among connec ed nodes in a ne - wo k, and ha o weigh ed ne wo ks he sho e pa hs a e hose wi h s onge weigh . Thus, an inc ease in GEFF can a ise om ei he a highe numbe o links, s onge weigh s, o bo h. Figu e 2B p o ides an example o he in e ed e ec i e connec i i y ma ices. A summa y o he a e age connec i i y hkiand a e age weigh hwi o he eigh s udied cul u es is p o ided Ne wo k Neu oscience 1 165 Func ional s eng hening h ough synap ic scaling in Table 1. Func ional communi ies we e al eady p esen o [CNQX] =0, a ea u e ha in- dica es ha he obse ed coo dina ed ac i i y was shaped by s ongly in e ac ing g oups o clus e s. This is clea in he co esponding connec i i y map o Figu e 2C, in which he majo - i y o he clus e s belonged o he same communi y, as calcula ed by modula i y op imiza ion Modula i y: Tendency o he nodes in a ne wo k o connec wi hin a g oup mo e s ongly han wi h nodes in ano he g oup. using he Lou ain algo i hm (Blondel, Guillaume, Lambio e, & Le eb e,2008). Well-de ined communi ies eme ged o [CNQX] =60 nM and highe concen a ions, as indica ed by he iche s uc u e o he connec i i y maps, which show g adually smalle and mo e isola ed communi ies. Fo [CNQX] &600 nM ac i i y p ac ically ceased and no unc ional cha ac e - is ics could be ende ed. The connec i i y maps o Figu e 2C also helped o pic u e he CNQX deg ada ion p ocess in e ms o he spa ial a angemen o he clus e s. The maps showed ha he unc ional commu- ni ies we e physically compac , ha is, hey we e cons i u ed by g oups o adjacen clus e s. This indica es ha he clus e s ypically connec ed o hei immedia e neighbo s upon ne - wo k o ma ion, and sugges s ha he o e all dynamics o he ne wo k a e media ed by local in e ac ions. In his di ec ion, he maps also show ha he physical loca ion o he unc ional communi ies was p ese ed, bu ha hey educed in size as CNQX g ew. Eme gence o Hype e iciency in CNQX-Pe u bed Ne wo ks The abo e g adual agmen a ion o he ne wo k wi h CNQX was no accompanied by a s eady loss o clus e s’ in e communica ion. Su p isingly, o he mode a e CNQX concen a ion o 60 nM, he global e iciency GEFF inc eased by 8%— o he pa icula expe imen o Figu e 2— wi h espec o he ini ial condi ion, o la e all as expec ed. We no e ha he inc ease in GEFF s ongly con as s wi h he all o spon aneous ac i i y, which is educed by 60%. We e m hype e iciency his ab up inc ease in GEFF and, as we elabo a e la e , i s p esence indica es he nonlinea ela ionship be ween he s uc u al al e a ions caused by CNQX and he e ec i e connec i i y ones ha eme ge om dynamical in e ac ions among he clus e s. In o he wo ds, while CNQX ac ion on exci a o y synapses is s eady and ollows a pe cola ion-like silencing p ocess (So iano, Rod íguez Ma ínez, e al., 2008), he e ec i e connec i i y obeys mo e complex mechanisms in which swi ching dynamic in e ac ions and plas ici y-d i en adap a- ion play a cen al ole. Be o e in es iga ing he possible hype e iciency mechanisms, we e i ied ha his phe- nomenon was no an expe imen al a i ac by compa ing he e olu ion o CNQX-dosed ne - wo ks wi h con ol ones ha we e p epa ed and manipula ed iden ically. Figu e 2C compa es he connec i i y maps o he ep esen a i e pe u bed ne wo k wi h a con ol one. Along he expe imen , which las ed 75 min, he con ol cul u e e ained he same unc ional ai s ac oss he ne wo k wi h small a ia ions, while he deg aded one signi ican ly changed. The da a o he ep esen a i e expe imen o Figu e 2 was a gene al end ha consis en ly epea ed ac oss eigh di e en cul u es, despi e hei a iabili y in numbe o clus e s, spon a- neous ac i i y, and e ec i e connec i i y ai s. In all cases, CNQX caused a agmen a ion o he ne wo k in o communi ies oge he wi h a sudden inc ease o GEFF o mode a e CNQX le els. To a e age o e cul u e ealiza ions, howe e , we obse ed ha each cul u e esponded sligh ly di e en ly upon pe u ba ion. This esul ed in speci ic CNQX concen a ions o he beginning o unc ional al e a ions and he end o ac i i y. Thus, o each cul u e, we consid- e ed a e e ence alue [CNQX] e abo e which dynamic al e a ions we e de ec able, and a maximum [CNQX]max a which ac i i y ully ceased (Figu e 3A). Da a was hen escaled ac- co ding o hese poin s o ob ain he global e iciency as a unc ion o an adequa e con ol Ne wo k Neu oscience 1 166 Func ional s eng hening h ough synap ic scaling Figu e 3. T ea men o expe imen al cu es and ela ionship be ween spon aneous ac i i y and global e iciency. (A) De ini ion o [CNQX] e (concen a ion be o e any ne wo k al e a ion occu ed) and [CNQX]max (no ne wo k ac i i y). (B) Alignmen o he eigh expe imen s (g ay cu es), hei a e age (blue), and compa ison wi h ou a e aged con ols (o ange). (C) E olu ion o GEFF as a unc ion o he a e age global ac i i y a o e e y CNQX concen a ion. Da a is scaled ela i e o [CNQX] e o pool all expe imen s oge he and highligh de ia ions om he unpe u bed condi ion (do ed lines). The plo shows ha hype e iciency s a es (GEFF >1) a e compa ible wi h low spon aneous ac i i y a es (g ay box). The hick blue line is a nonlinea sca e plo smoo hing o indica e he gene al end o he da a. I s shading is he con idence in e al se as 80%. pa ame e ha we e m deg ada ion le el d. The deg ada ion le el d=0means ha no CNQX is applied and he e o e he cul u e exhibi s i s na u al dynamics, while d=1co esponds o he CNQX dose ha is able o ully silence he cul u e’s ac i i y. The esul ing GEFF(d)cu es o he eigh expe imen s and hei a e age a e shown Figu e 3B. All analyzed cul u es exhib- i ed hype e iciency, which anged om a minimum o 3% o a maximum o 50%, and ha on a e age p ocu ed abou 20% GEFF inc ease. This clea peak in global e iciency was inexis en in con ols. An aspec ha is impo an o poin ou is he complex ela ionship be ween clus e s’ collec i e ac i i y aand global e iciency GEFF. The da a o Figu e 2A and B shows ha he hype e iciency obse ed a [CNQX] =60 nM coincides wi h a p ominen decay in ne wo k’s collec i e ac i i y. The eason o his appa en con adic ion is ha he global ac i i y jus in o ms abou he equency o collec i e beha io , bu does no cap u e he local dynamic in- e ac ion among clus e s. This in o ma ion is p o ided by he e ec i e connec i i y, om which he global e iciency is compu ed. Thus, we hypo hesize ha he clus e s s eng hen hei cou- pling despi e coac i a ing oge he less o en. The complex ela ionship be ween aand GEFF is illus a ed in Figu e 3C, in which we plo ed one agains he o he o each CNQX concen- a ion, bu scaled he alues ela i e o [CNQX] e o pool all eigh expe imen s in a single plo . In e es ingly, se e al o he poin s co esponding o a d op in ac i i y, o example, in he ange 0.5–1, a e associa ed o GEFF alues abo e 1 (shaded a ea). These poin s a e p ecisely he ones ha a e behind he obse ed hype e iciency, and ha we associa e o a s eng hen- ing o he e ec i e coupling among clus e s. This analysis hus sugges s ha a d op in global ac i i y caused by CNQX is esponded o by an inc ease in clus e s’ e ec i e coupling. A Deg ada ion Model wi h Synap ic Scaling Rep oduces he Hype e iciency Phenomenon We ca ied ou nume ical simula ions o unde s and he o igin o hype e iciency in he CNQX pe u bed expe imen s. In he simula ions we conside ed he same e ec i e ne wo ks as in he expe imen s o [CNQX] e and applied wo connec i i y deg ada ion schemes, namely pe cola i e deg ada ion (PD) and synap ic scaling deg ada ion (SSD). As ske ched in Figu e 4A, on he one hand, PD co esponds o a s anda d bond pe cola ion in which links a e g adually Ne wo k Neu oscience 1 167 Func ional s eng hening h ough synap ic scaling Figu e 4. Nume ical models and hype e iciency. (A) Ske ch o he models used o simula e he impac o a g adual loss o e ec i e connec ions. In his example, he wo models a ge an e ec i e connec ion wi h weigh H=2. Fo pe cola i e deg ada ion (PD, le ) he bo om connec ion is simply emo ed upon a ack. Fo synap ic scaling deg ada ion (SSD, igh ) he same connec ion is emo ed, bu he su i ing ou going connec ions a e upscaled o dis ibu e he los weigh p opo - ionally among hei s. All cu es a e a e ages o e eigh ealiza ions, and shadings indica e s anda d de ia ion. (B) Compa ison o he deg ada ion ac ion be ween expe imen s (blue) and nume ical models (PD, pink; SSD, g een). PD po ays a as -decaying beha io and he ne wo k becomes dis- connec ed by d≃0.4. SSD ep oduces he hype e iciency (HE) peak and he gen le decay o GEFF wi h dup o he end o he disin eg a ion p ocess. The inse shows he dis ibu ion o maximum hy- pe e iciency alues o expe imen s and model. Despi e he dispe sion in he da a, SSD p ocu ed o e all simila alues. emo ed om weake o s onge acco ding o a h eshold H. This model e lec s he si ua ion in which clus e s’ dynamic in e ac ion p og essi ely anish as deg ada ion g ows. On he o he hand, SSD consis s in he same link emo al bu wi h he c ucial inclusion o synap ic scaling, which is simula ed as he s eng hening o he su i ing links’ weigh s p opo ionally o he los link’s weigh . This model cap u es a compensa o y esponse mechanism o he loss in ac i i y. In bo h models, he h esholds a e ecalcula ed a e he emo al p ocess o a ge he nex weakes link, yielding a na u ally inc easing sequence o h esholds. To compa e he models wi h he expe imen s, he h esholds Hwe e escaled in he ange [0, 1] o conco d wi h he deg ada ion le el d. Ne wo k Neu oscience 1 168 Func ional s eng hening h ough synap ic scaling a e o 50 ames pe second ( ps), a size o 1,024 ×1,024 pixels, and a spa ial esolu ion o 5.76 µm/pixel. Spon aneous Ac i i y and Ne wo k Bu s s Clus e ed cul u es we e measu ed a DIV 8–12, a de elopmen al s age in which he clus e s exhibi ed spon aneous ac i i y wi h a leas 1 collec i e ac i a ion (bu s ) pe minu e (Table 1). Since, as obse ed p e iously by Telle e al. (2014,2015), spon aneous ac i i y in clus e ed cul u es is o en o a modula na u e, in which he size o collec i e ac i a ions encompasses om ew clus e s o he en i e ne wo k, we conside ed as bu s hose ac i i y e en s in which a leas i e clus e s coac i a ed in a window o 200 ms (Telle e al.,2014). Homgeneous cul u es we e measu ed a DIV 12 and exhibi ed s ongly cohe en dynamics, in which he whole ne wo k ac i a ed oge he du ing a bu s ing episode, emaining p ac- ically silen in be ween bu s s. A e age bu s ing equency in he ou s udied cul u es was 9.6 ±4.9 bu s s/min. Expe imen al P ocedu e and Pha macology Bo h exci a o y and inhibi o y connec ions we e p esen in he clus e ed cul u es. To in es i- ga e he changes in ac i i y induced solely by p og essi ely weake AMPA glu ama e exci a ion in neu ons, NMDA exci a o y ecep o s, which accoun o 20% o exci a o y d i e in co ical neu ons, we e blocked wi h 20 µM o he ecep o an agonis APV (Sigma A5282), and GABAA inhibi o y ecep o s we e blocked wi h 40 µM o he an agonis bicuculline (Sigma B7561). Bo h an agonis s we e applied simul aneously and he cul u e was le 5 min in da kness o biochemical s abiliza ion. A 15-min eco ding was hen conduc ed and swi ly analyzed o e alua e he ini ial s a e o he ne wo k. Cul u es exhibi ing poo o agmen ed ac i i y we e disca ded. Fo he highly ac i e cul u es, AMPA glu ama e exci a o y connec i i y was hen g adually weakened by applica ion o he ecep o an agonis CNQX, wi h p ese concen a- ions o [CNQX] (nM) =30, 60, 120, 300, 600, 1,000. Fo each CNQX concen a ion, spon a- neous ac i i y was eco ded o 15 min. Cul u es we e kep in da kness o 5 min in be ween eco dings o he nex d ug applica ion o ake e ec . To in es iga e whe he he neu onal cul u es we e able o eco e he ac i i y pa e ns p io o CNQX ac ion, he s udied clus e ed cul u es we e washed o wi h esh RS a he end o he eco ding session, APV and bicuculine applied again, and ac i i y moni o ed o addi ional 20 min. Con ol expe imen s we e ca ied ou o e i y ha he en i e p ocedu e and du a ion o he eco dings did no a ec he heal h o he neu ons. Con ol eco dings consis ed iden ical manipula ions as s anda d expe imen s bu wi hou addi ion o CNQX. A p ac ically iden ical p ocedu e was ollowed wi h homogeneous cul u es, wi h he only di e ence ha he CNQX concen a ions we e adjus ed ollowing Tibau e al. (2020) and Tibau e al. (2013b), and used [CNQX] (nM) =100, 200, 400, 800, 1,000, 2,000. Da a Analysis Fluo escence aces. Fo each expe imen , neu onal clus e s o indi idual neu ons we e man- ually iden i ied on a highly con as ed luo escence image and asc ibed as egions o in e es (ROIs). The a e age luo escence (g ay scale le el) wi hin each ROI was hen compu ed, and he aw luo escence aces Fi( ) o each clus e iex ac ed. Each ace was hen co ec ed Ne wo k Neu oscience 1 175 Func ional s eng hening h ough synap ic scaling om global d i s and a i ac s and no malized as DFFi(%)≡100 ·(Fi( )−Fi,0)/Fi,0, whe e Fi,0 i he basal luo escence o each clus e . The onse imes o ac i a ion o clus e s o indi id- ual neu ons we e de e mined as hose e en s in which he luo escence ace c ossed a p ese h eshold, se as he mean + 2 imes he s anda d de ia ion o he en i e luo escence ace (Telle e al.,2014,2015). T ains o ac i i y we e hen buil by asc ibing as ‘1’ he exis ence o an ac i a ion a a gi en ime poin , and ‘0’ o he wise. Ne wo k Analysis E ec i e connec i i y. I was in e ed using a modi ied e sion o ans e en opy (TE) (Sch eibe , 2000). TE was compu ed o each pai o nodes Xand Y(ei he clus e s o neu ons) wi h signals xmand ymindexed by 0≤m≤mmax, whe e mmax is he o al numbe o ime s eps in he da a, as TEY→X=−∑ 0≤n≤nmax 0≤i≤iM pxm+1,x(i) m,y(i) m×log2 pxm+1  x(i) m,y(i) m pxm+1  x(i) m,(1) whe e iis he index o he pas ime s ep conside ed, ha is, he leng h o he ec o s {x(i) m}, and iM=2is he Ma ko o de o he model. He e, ins an aneous eedback was assumed, meaning ha Xand Ycould in e ac wi hin a ime bin, as in gene alized ans e en opy (O landi, S e e , So iano, Geisel, & Ba aglia,2014;S e e e al.,2012). E ec i e connec i i y was in e ed o he 15-min-long as e plo s (mmax =45, 000 poin s wi h an acquisi ion a e o 50 ps). TE =0 o an iden ical o andom pai o signals. Since expe imen al eco dings a e sho , i may occu ha andom ac i a ions in nea by ime bins a e aken as causal in e ac ions. Thus, o ejec spu ious connec ions, signi icance zwas es ablished o any connec ion X o Yby compa ing he TEY→Xes ima e wi h he join dis ibu ion o TE o all inpu sco es X′ o Yand ou pu sco es X o Y′( o any X′and Y′), as z=TEY→X− hTEjoin i σjoin ,(2) whe e hTEjoin iis he a e age alue o he join dis ibu ion and σjoin is i s s anda d de ia- ion. Signi ican connec ions we e hen se as hose wi h z≥2. This h eshold was consid- e ed op imal since i cap u ed he low o neu onal communica ion du ing ac i i y a bo h local and global scales. A lowe h eshold o z=1yielded ne wo ks ha excessi ely em- phasized whole-ne wo k coo dina ed ac i i y, e ec i ely shaping andom g aphs in all s udied cases. Th esholds z&3emphasized he s onges neu on- o-neu on in e ac ions only and o - en yielded emp y ma ices. All esul s p esen ed he e we e consis en wi h h esholds in he ange 1.5 ≤z≤2.5. By cons uc ion, he adjacency ma ix A={aij}o signi ican connec- ions was di ec ed and weigh ed, concep ually cap u ing he s eng h o dynamic in e ac ions be ween clus e s o neu ons, and wi h he weigh o he connec ion gi en in s anda d de ia ion uni s o he join dis ibu ion. Densi y o links D. Fo a ne wo k wi h Nnodes, i was de ined as he ac ion o o al exis ing e ec i e links ( ega dless hei weigh ) o all possible links, ha is, D=∑ij bij/(N(N−1)), whe e B={bij}is he bina ized adjacency ma ix o A. Single ealiza ion global e iciency G. The analysis o he da a o each in es iga ed cul u e p o ided a se e ec i e adjacency ma ices {aij}k, whe e kis he s ep o CNQX applica ion. Ne wo k Neu oscience 1 176 Func ional s eng hening h ough synap ic scaling Fo each o hese ma ices, he global e iciency Go a ne wo k wi h Nnodes was de e mined ollowing he s anda d de ini ion (La o a & Ma chio i,2001;Rubino & Spo ns,2010) G=1 N(N−1)∑ i6=j 1 d(i,j),(3) whe e d(i,j)is he sho es weigh ed pa h calcula ed on he ma ix {aij}, ha is, he smalles sum o weigh s h oughou all possible pa hs be ween iand j. In his cons uc ion, highe connec ion weigh s co espond o sho e leng hs, and he e o e he global e iciency inc eases wi h he weigh o he connec ions. I no pa h exis s be ween nodes iand j hen d(i,j) = ∞, so ha he e is no con ibu ion o he summa ion in Equa ion 3. The e m N(N−1)in his equa ion is a no maliza ion ac o ha accoun s o all possible di ec ed links ha can be o med by he Nnodes. Global e iciency was calcula ed using he B ain Connec i i y Toolbox in Ma lab. Deg ada ion Le el dand Alignmen o Expe imen al Cu es The CNQX concen a ions we e con e ed in o a well-de ined pa ame e ha cap u ed in a ange [0, 1] he p og essi e damage o a neu onal ne wo k. Thus, o each expe imen n, he CNQX concen a ions we e ans o med in o s eps o a deg ada ion le el dby de ining d(n) k=[CNQX](n) k−[CNQX](n) e [CNQX](n) max ,(4) whe e kis he s ep o CNQX applica ion along he expe imen , [CNQX]k he co esponding concen a ion alue, [CNQX] e he e e ence concen a ion abo e which he dynamic al e - a ions in he ne wo k a e measu able, and [CNQX]max he concen a ion ha ully silenced he ne wo k. Wi h his cons uc ion, he deg ada ion le el d a ied be ween 0 (no al e a ion o he ne wo k) o 1 ( ull silencing o ac i i y and comple e e ec i e ne wo k b eakdown). Da a ou - side hese bounds was dis ega ded. Since [CNQX] e and [CNQX]max a ied ac oss cul u es, he use o dp o ided a obus manne o a e age among cul u es. Scaled Global E iciency GEFF and A e aged GEFF(d)Cu es The se o global e iciency alues o each expe imen , G(n)(d), was scaled wi h espec o i s e e ence alue p io damage, as G(n) EFF(d) = G(n)(d)/G(n) e , whe e G(n) e is he global e iciency measu ed a d=0. The G(n) EFF(d)cu es we e hen in e pola ed as 100 equidis an poin s by using a cubic in e pola ion ha p ese ed he shape o he cu es, and a e aged o e n=8 expe imen al ealiza ions (clus e ed cul u es) o n=4(homogeneous cul u es). Nume ical Models o Ne wo k Deg ada ion Two nume ical models we e p oposed o elucida e he impac o g adual deg ada ion on ne - wo k global e iciency. Bo h models conside ed he same ini ial e ec i e ne wo ks as in he expe imen s o d=0, deno ed as {a(0) ij }, and applied a se ies o ules o he dele ion o links. Since he e ec i e connec i i y ma ices we e ob ained om ans e en opy wi h a signi i- cance h eshold z=2, he alues {a(0) ij }we e bounded be ween amin =2and a gi en amax ha depended on each ne wo k. Pe cola i e deg ada ion (PD). This model p og essi ely dele ed he e ec i e ou pu links whose weigh s we e equal o lowe han a h eshold H, s a ing a H=amin and p og essi ely Ne wo k Neu oscience 1 177 Func ional s eng hening h ough synap ic scaling a ge ing s onge links (Figu e 4A, le ). Thus, a a gi en simula ion s ep, he dele ed links we e hose ha ul illed H=aij, inc easing Ha e wa d o a ge he nex weak link in he ne wo k, un il all links o he ne wo k we e dele ed. The global e iciency Gwas compu ed a each h eshold, and he inal GPD( H) alues s o ed o u he analysis. Synap ic scaling deg ada ion (SSD). This model also p og essi ely dele ed he weakes ou - pu links, bu wi h he c ucial addi ional ule ha , upon link emo al (wi h weigh H), all su i ing ou pu links sha ed he weigh o he emo ed one, inc easing hei s p opo ionally (Figu e 4A, igh ). Thus, i he e we e a leas wo links and he o al ou pu s eng h be o e dele ion was sbe i=∑jabe ij , he new weigh s we e gi en by aij =abe ij ·sbe i/(sbe i− H). I only one ou pu link was p esen , i was dele ed as in PD. The new global e iciency o he ne wo k GSSD( H)was hen compu ed, he h eshold upda ed o seek o he new weakes links, and he p ocess s a ed again. Simula ion ended when all links in he ne wo k we e emo ed, and he inal GSSD( H) alues s o ed. The GPD( H)and GSSD( H)da a ob ained om he models was inally ea ed as in he expe imen s. Fi s , o each simula ed ne wo k, he H alues (which anged om amin o amax in he PD model, and om amin o a ce ain a∗ max ≥amax o SSD) we e scaled as d∈[0, 1]. The new GPD(d)and GSSD(d)cu es we e hen scaled ela i e o hei alues a d=0and a e aged among ealiza ions o ob ain he scaled global e iciencies GEFF(d), which could be di ec ly compa ed wi h he expe imen s. E hics S a emen All p ocedu es we e app o ed by he E hical Commi ee o Animal Expe imen a ion o he Uni e si y o Ba celona, unde o de DMAH-5461, and in acco dance o he egula ions o he Gene ali a de Ca alunya (Spain). AUTHOR CONTRIBUTIONS Es e anía Es é ez-P iego: Da a cu a ion; Fo mal analysis; In es iga ion; Me hodology; So - wa e; Visualiza ion; W i ing - O iginal D a . Sa a Telle : Da a cu a ion; Me hodology; Re- sou ces; So wa e; Supe ision; Valida ion. Cla a G anell: Da a cu a ion; Fo mal analysis; Me hodology; Valida ion; Visualiza ion; W i ing - Re iew & Edi ing. Alex A enas: Concep ual- iza ion; Fo mal analysis; Funding acquisi ion; In es iga ion; P ojec adminis a ion; Resou ces; Supe ision; W i ing - Re iew & Edi ing. Jo di So iano: Concep ualiza ion; Funding acqui- si ion; In es iga ion; Me hodology; Resou ces; So wa e; Supe ision; Valida ion; W i ing - O iginal D a ; W i ing - Re iew & Edi ing. FUNDING INFORMATION Jo di So iano, Fundació Bancà ia “La Caixa” (h p://dx.doi.o g/10.13039/100000778), Awa d ID: LCF/PR/HR19/52160007. Jo di So iano, H2020 Fu u e and Eme ging Technologies (h p://dx.doi.o g/10.13039/100010664), Awa d ID: 713140 MESOBRAIN. Es e anía Es é ez- P iego, H2020 Fu u e and Eme ging Technologies (h p://dx.doi.o g/10.13039/100010664), Awa d ID: 713140 MESOBRAIN. Jo di So iano, Sec e a ía de Es ado de In es igación, Desa ollo e Inno ación (h p://dx.doi.o g/10.13039/501100007136), Awa d ID: FIS2013- 41144-P, FIS2016- 78507-C2-2-P, FIS2017-90782-REDT. Sa a Telle , Sec e a ía de Es ado de In es igacíón, Desa ollo e Inno ación (h p://dx.doi.o g/10.13039/501100007136),Awa d ID: FIS2013-41144-P, FIS2016- 78507-C2-2-P, FIS2017-90782-REDT. Jo di So iano, Depa amen Ne wo k Neu oscience 1 178 Func ional s eng hening h ough synap ic scaling d’Inno ació, Uni e si a s i Emp esa, Gene ali a de Ca alunya (h p://dx.doi.o g/10.13039 /501100002943), Awa d ID: 2017-SGR-1061. Alex A enas, Depa amen d’Inno ació, Uni e - si a s i Emp esa, Gene ali a de Ca alunya (h p://dx.doi.o g/10.13039/501100002943), Awa d ID: 2017-SGR-896. Alex A enas, Sec e a ía de Es ado de In es igación, Desa ollo e Inno- ación (h p://dx.doi.o g/10.13039/501100007136),Awa d ID: PGC2018-094754-B-C21. Alex A enas, Uni e si a Ro i a i Vi gili (h p://dx.doi.o g/10.13039/501100007512), Awa d ID: 2017PFRURV-B2-41. Alex A enas, ICREA Academia . Alex A enas, James S. McDonnell Foun- da ion (h p://dx.doi.o g/10.13039/100000913), Awa d ID: 220020325. Cla a G anell, James S. McDonnell Founda ion (h p://dx.doi.o g/10.13039/100000913), Awa d ID: 220020457. Cla a G anell, Sec e a ía de Es ado de In es igación, Desa ollo e Inno ación (h p://dx.doi.o g /10.13039/501100007136), Awa d ID: Juan de la Cie a. REFERENCES Ae s, H., Fias, W., Caeyenbe ghs, K., & Ma inazzo, D. (2016). B ain ne wo ks unde a ack: Robus ness p ope ies and he impac o lesions. B ain,139(12), 3063–3083. Ba al, J., & Reyes, A. (2016). Synap ic scaling ule p ese es exci a o y–inhibi o y balance and salien neu onal ne wo k dy- namics. Na u e Neu oscience,19, 1690. Basse , D. S., & Spo ns, O. (2017). Ne wo k neu oscience. Na u e Neu oscience,20(3), 353–364. Basse , D. S., Zu n, P., & Gold, J. I. (2018). On he na u e and use o models in ne wo k neu oscience. Na u e Re iews Neu oscience, 19(9), 566–578. Blondel, V. D., Guillaume, J.-L., Lambio e, R., & Le eb e, E. (2008). Fas un olding o communi ies in la ge ne wo ks. Jou - nal o S a is ical Mechanics: Theo y and Expe imen ,2008(10), P10008. Bo ge-Hol hoe e , J., Mo eno, Y., & A enas, A. (2011). Modeling abno mal p iming in Alzheime ’s pa ien s wi h a ee associa ion ne wo k. PLoS ONE,6(8), e22651. Bu z, M., S eenbuck, I., & an Ooyen, A. (2014). Homeos a ic s uc- u al plas ici y inc eases he e iciency o small-wo ld ne wo ks. F on ie s in Synap ic Neu oscience,6, 7. Calab esi, P., Cen onze, D., Pisani, A., Cupini, L. M., & Be na di, G. (2003). Synap ic plas ici y in he ischaemic b ain. The Lance Neu ology,2(10), 622–629. Cos a, R., Mizusaki, B., Sjös öm, P., & an Rossum, M. (2017). Func ional consequences o p e- and pos synap ic exp ession o synap ic plas ici y. Philosophical T ansac ions o he Royal Socie y B: Biological Sciences,372, 20160153. h ps://doi.o g/10 .1098/ s b.2016.0153 Desai, N. S., Ru he o d, L. C., & Tu igiano, G. G. (1999). Plas ici y in he in insic exci abili y o co ical py amidal neu ons. Na u e Neu oscience,2, 515–520. Echegoyen, J., Neu, A., G abe , K. D., & Sol esz, I. (2007). Homeos a ic plas ici y s udied using in i o hippocampal ac i i y-blockade: Synap ic scaling, in insic plas ici y and age- dependence. PLoS ONE,2(8), 1–9. Fa ooq, H., Chen, Y., Geo giou, T. T., Tannenbaum, A., & Lengle , C. (2019). Ne wo k cu a u e as a hallma k o b ain s uc u al connec i i y. Na u e Communica ions,10(1), 1–11. Fau h, M., & Te zla , C. (2016). Opposing e ec s o neu onal ac- i i y on s uc u al plas ici y. F on ie s in Neu oana omy,10, 75. Fau h, M., Wö gö e , F., & Te zla , C. (2015). The o ma ion o mul i-synap ic connec ions by he in e ac ion o synap ic and s uc u al plas ici y and hei unc ional consequences. PLOS Compu a ional Biology,11, 1–29. Fong, M., Newman, J. P., Po e , S. M., & Wenne , P. (2015). Up- wa d synap ic scaling is dependen on neu o ansmission a he han spiking. Na u e Communica ions,6, 6339. Fo ni o, A., Zalesky, A., & B eakspea , M. (2015). The connec omics o b ain diso de s. Na u e e iews. Neu oscience,16, 159–72. Hanes, A. L., Koes e s, A. G., Fong, M.-F., Al imimi, H. F., S ellwagen, D., Wenne , P., e al. (2020). Di e gen synap ic scaling o minia u e exci a o y synap ic cu en s ollowing ac i - i y blockade in dissocia ed neu onal cul u es. Jou nal o Neu o- science, JN-RM-1393-19. He nández-Na a o, L., O landi, J. G., Ce u i, B., Vi es, E., & So iano, J. (2017). Dominance o me ic co ela ions in wo- dimensional neu onal cul u es desc ibed h ough a andom ield ising model. Physical Re iew Le e s,118, 208101. Kanne , S., Bisio, M., Cohen, G., Goldin, M., Tedesco, M., Hanein, Y., . . . Boni azi, P. (2015). Design, su ace ea men , cellula pla ing, and cul u ing o modula neu onal ne wo ks composed o unc ionally in e -connec ed ci cui s. Jou nal o Visualized Ex- pe imen s : JoVE(98). Keck, T., Kelle , G. B., Jacobsen, R., Eysel, U. T., Bonhoe e , T., & Hübene , M. (2013). Synap ic scaling and homeos a ic plas ici y in he mouse isual co ex in i o. Neu on,80, 327–334. Ki kpa ick, S. (1973). Pe cola ion and conduc ion. Re iews o Mode n Physics,45(4), 574. La o a, V., & Ma chio i, M. (2001). E icien beha io o small-wo ld ne wo ks. Physical Re iew Le e s,87, 198701. Majdandzic, A., Podobnik, B., Buldy e , S. V., Kene , D. Y., Ha lin, S., & Eugene S anley, H. (2013). Spon aneous eco e y in dy- namical ne wo ks. Na u e Physics,10, 34. Ma de , E. (2011). Va iabili y, compensa ion, and modula ion in neu ons and ci cui s. P oceedings o he Na ional Academy o Sciences o he Uni ed S a es o Ame ica,108(3), 15542–15548. Ne wo k Neu oscience 1 179 Func ional s eng hening h ough synap ic scaling Meunie , D., Lambio e, R., Fo ni o, A., E sche, K. D., & Bullmo e, E. T. (2009). Hie a chical modula i y in human b ain unc ional ne wo ks. F on ie s in Neu oin o ma ics,3, a icle 37. Mu phy, T. H., & Co be , D. (2009). Plas ici y du ing s oke eco - e y: F om synapse o beha iou . Na u e Re iews Neu oscience, 10(12), 861–872. O’B ien, R. J., Kamboj, S., Ehle s, M. D., Rosen, K. R., Fischbach, G. D., & Hugani , R. L. (1998). Ac i i y-dependen modula- ion o synap ic ampa ecep o accumula ion. Neu on,21(5), 1067–1078. O landi, J. G., So iano, J., Al a ez-Lacalle, E., Telle , S., & Casademun , J. (2013). Noise ocusing and he eme gence o cohe en ac i i y in neu onal cul u es. Na u e Physics,9(9), 582–590. O landi, J. G., S e e , O., So iano, J., Geisel, T., & Ba aglia, D. (2014). T ans e en opy econs uc ion and labeling o neu onal connec ions om simula ed calcium imaging. PLoS ONE,9(6), e98842. h ps://doi.o g/10.1371/jou nal.pone.0098842 Rubino , M., & Spo ns, O. (2010). Complex ne wo k measu es o b ain connec i i y: Uses and in e p e a ions. Neu oimage,52(3), 1059–1069. Sch eibe , T. (2000). Measu ing in o ma ion ans e . Physical Re- iew Le e s,85(2), 461–464. Sege , R., Ben enis e, M., Shapi a, Y., & Ben-Jacob, E. (2003). Fo - ma ion o elec ically ac i e clus e ized neu al ne wo ks. Physical Re iew Le e s,90, 168101. Shein Idelson, M., Ben-Jacob, E., & Hanein, Y. (2011). Inna e syn- ch onous oscilla ions in eely-o ganized small neu onal ci cui s. PLoS ONE,5, 1–9. Slomowi z, E., S y , B., Ve kin, I., Milsh ein-Pa ush, H., Nelken, I., Slu sky, M., & Slu sky, L. (2015). In e play be ween popula ion i ing s abili y and single neu on dynamics in hippocampal ne - wo ks. eLi e,4, e04378. So iano, J., Ma ínez, M. R., Tlus y, T., & Moses, E. (2008). De elop- men o inpu connec ions in neu al cul u es. P oceedings o he Na ional Academy o Sciences o he Uni ed S a es o Ame ica, 105(37), 13758–13763. So iano, J., Rod íguez Ma ínez, M., Tlus y, T., & Moses, E. (2008). De elopmen o inpu connec ions in neu al cul u es. P oceed- ings o he Na ional Academy o Sciences o he Uni ed S a es o Ame ica,105(37), 13758–13763. h ps://doi.o g/10.1073/pnas .0707492105 Spo ns, O. (2014). Con ibu ions and challenges o ne wo k mod- els in cogni i e neu oscience. Na u e Neu oscience,17(5), 652. S am, C. J. (2014). Mode n ne wo k science o neu ological dis- o de s. Na u e Re iews Neu oscience,15(10), 683. S e e , O., Ba aglia, D., So iano, J., & Geisel, T. (2012). Model- ee econs uc ion o exci a o y neu onal connec i i y om calcium imaging signals. PLoS Compu a ional Biology,8(8), e1002653. h ps://doi.o g/10.1371/jou nal.pcbi.1002653 Telle , S., Es é ez-P iego, E., G anell, C., To ne o, D., Andilla, J., Ola e, O. E., . . . So iano, J. (2019). Spon aneous unc ional eco e y a e ocal damage in neu onal cul u es. eNeu o,9, 0254–19. Telle , S., G anell, C., De Domenico, M., So iano, J., Gómez, S., & A enas, A. (2014). Eme gence o asso a i e mixing be ween clus e s o cul u ed neu ons. PLoS Compu a ional Biology,10, e1003796. Telle , S., Tahi begi, I. B., Mi , M., Sami ie , J., & So iano, J. (2015). Magne i e-amyloid-βde e io a es ac i i y and unc ional o gan- iza ion in an in i o model o Alzheime ’s disease. Scien i ic Repo s,5. Tibau, E., Ludl, A., Rüdige , S., O landi, J. G., & So iano, J. (2020). Neu onal spa ial a angemen shapes e ec i e connec i i y ai s o in i o co ical ne wo ks. IEEE T ansac ions on Ne wo k Science and Enginee ing,7(1), 435–448. Tibau, E., Valencia, M., & So iano, J. (2013a). Iden i ica ion o neu- onal ne wo k p ope ies om he spec al analysis o calcium imaging signals in neu onal cul u es. F on ie s in Neu al Ci cui s, 7, 199. Tibau, E., Valencia, M., & So iano, J. (2013b). Iden i ica ion o neu- onal ne wo k p ope ies om he spec al analysis o calcium imaging signals in neu onal cul u es. F on ie s in Neu al Ci cui s, 7, 199. Tu igiano, G. (2012). Homeos a ic synap ic plas ici y: Local and global mechanisms o s abilizing neu onal unc ion. Cold Sp ing Ha bo Pe spec i es in Biology,4(1), a005736. Tu igiano, G. G. (2008). The sel - uning neu on: Synap ic scaling o exci a o y synapses. Cell,135, 422–435. Tu igiano, G. G., Leslie, K. R., Desai, N. S., Ru he o d, L. C., & Nelson, S. B. (1998). Ac i i y-dependen scaling o quan al am- pli ude in neoco ical neu ons. Na u e,391(6670), 892–896. Tu igiano, G. G., & Nelson, S. B. (2004). Homeos a ic plas ici y in he de eloping ne ous sys em. Na u e Re iews Neu oscience, 5(2), 97. an den Heu el, M. P., & Spo ns, O. (2013). Ne wo k hubs in he human b ain. T ends in Cogni i e Sciences,17(12), 683–696. Yamamo o, H., Mo iya, S., Ide, K., Hayakawa, T., Akima, H., Sa o, S., . . . Hi anel-wa a, A. (2018). Impac o modula o ganiza ion on dynamical ichness in co ical ne wo ks. Science Ad ances, 4(11), eaau4914. Ne wo k Neu oscience 1 180