Input normalization by global feedforward inhibition expands cortical dynamic range

Transcription

1 Input normliztion y glol feedforwrd inhiition expnds ortil dynmi rnge Frédéri Pouille 1,2, Antoni Mrin-Burgin 1,2, Hillel Adesnik 1, Bssm V Atllh 1 & Mssimo Snzini 1 29 Nture Ameri, In. All rights reserved. The ortex is sensitive to wek stimuli, ut responds to stronger inputs without sturting. The mehnisms tht enle this wide rnge of opertion re not fully understood. We found tht the mplitude of exittory synpti urrents neessry to fire rodent pyrmidl ells, the threshold exittory urrent, inresed with stimulus strength. Consequently, the reltive ontriution of individul fferents in firing neuron ws inversely proportionl to the totl numer of tive fferents. Feedforwrd inhiition, ting homogeneously ross pyrmidl ells, ensured tht threshold exittory urrents inresed with stimulus strength. In ontrst, heterogeneities in the distriution of exittory urrents in the neuronl popultion determined the speifi set of pyrmidl ells reruited. Together, these mehnisms expnd the rnge of fferent input strengths tht neuronl popultions n represent. A hrteristi of ortil exittory neurons is their widely divergent xonl projetion. This property enles ortil neurons to ontt lrge numer of postsynpti ells nd llows eh postsynpti ell to reeive inputs from mny presynpti neurons. In iruit onstruted with this exittory divergene lone, the numer of tive presynpti neurons (input strength) tht is suffiient to reruit ll neurons in the postsynpti popultion is only slightly lrger thn the input strength required to reruit ny postsynpti neuron t ll. In other words, the input rnge tht n e fithfully represented y the postsynpti popultion is restrited. For exmple, if presynpti neurons onnet to postsynpti popultion with proility of 15% 1 nd eh postsynpti ell requires 4 tive inputs to e reruited 2, then 2% of the postsynpti ells would e reruited y the tivity of 2 presynpti neurons nd lmost ll (>99%) would e reruited y simply douling the numer of tive presynpti neurons (s determined y inomil sttistis). Thus, in the sene of ontrol mehnisms, smll flututions in the frtion of presynptilly tive neurons results in ll-or-none reruitment of the postsynpti popultion 3 6 (this ll-or-none ehvior is qulittively similr for wide rnge of onnetivity vlues nd numer of inputs neessry to reh threshold). However, oth spontneous nd sensory-evoked ortil tivity involves lrge flututions in the frtion of tive neurons (for exmple, refs. 7 9). Wht mehnisms does the ortex use to expnd the rnge of input strengths over whih it fithfully responds? One ould imgine t lest two distint mehnisms. Reduing the gin of individul neurons (tht is, the hnge in spiking proility s funtion of input strength) would llow eh neuron in the popultion to respond over wider rnge of input strengths; this gin modultion ould e hieved through GABA A reeptor medited ondutnes Alterntively, stggering the reruitment of individul neurons over wide rnge of input strengths would llow the popultion s whole, rther thn individul neurons, to represent wider input rnge. This ould e hieved y vrying the mplitude of the exittory postsynpti urrents (EPSCs) neessry to reh threshold for spike genertion s funtion of input strength. We found tht hippompl nd neoortil feedforwrd inhiitory iruits stggered the reruitment of individul pyrmidl ells over wide rnge of input strengths. Feedforwrd inhiition (FFI) ted homogeneously ross the postsynpti popultion of pyrmidl ells to rpidly djust their exitility to the strength of inoming presynpti tivity. As result, the mplitude of the EPSC neessry for pyrmidl ell to reh spike threshold ws dynmi nd vried with the strength of the input. Heterogeneities in the mplitudes of EPSCs ross the postsynpti popultion determined the speifi suset of pyrmidl ells tht would spike in response to the presynpti input. Through this oordinted tion of diret exittion nd FFI, pyrmidl ell popultions n remin sensitive to wek inputs, ut will not sturte in response to stronger tivity. RESULTS EPSC neessry to spike pyrmidl ell is dynmi We estlished the rnge of stimulus strengths over whih the CA1 pyrmidl ell popultion responds, tht is, the dynmi rnge. We reorded from individul pyrmidl ells in the loose-pth onfigurtion nd stimulted Shffer ollterls over rnge of intensities, from those tht filed to trigger ny spike to those tht triggered spikes on every tril (Fig. 1). The reltionship etween spiking proility of individul pyrmidl ells nd input strength (input strength is proportionl to the numer of tivted Shffer ollterl; for detils see Online Methods nd Supplementry Fig. 1) ws fitted with sigmoid to interpolte the threshold input strength, where pyrmidl ells spiked in 5% of the trils (Fig. 1). The umultive distriution of threshold input strengths for ll reorded pyrmidl ells 1 Howrd Hughes Medil Institute nd Neuroiology Setion, Division of Biology, University of Cliforni Sn Diego, L Joll, Cliforni, USA. 2 These uthors ontriuted eqully to this work. Correspondene should e ddressed to M.S. (mssimo@usd.edu). eived 5 August; epted 25 Septemer; pulished online 1 Novemer 29; doi:1.138/nn.2441 nture NEUROSCIENCE VOLUME 12 NUMBER 12 deemer

2 29 Nture Ameri, In. All rights reserved. Figure 1 The stronger the stimulus, the lrger the exittion neessry to reruit pyrmidl ell. () Left, spiking proility plotted ginst input strength for one CA1 pyrmidl ell (, sigmoidl fit, dshed lines indite 5% spiking proility). Inset, loose-pth reording t two different input strengths, five onseutive sweeps. Suesses re shown in lk nd filures in gry. Right, lk dt points represent the tivtion urve (tht is, the umultive distriution of input strengths eliiting 5% spiking, n = 39). Dshed lines represent the input strengths reruiting 5% nd 95% of the pyrmidl ell popultion. Gry sigmoids indite the spiking proility of the 39 pyrmidl ells mking up the tivtion urve. Their reruitment ws stggered long the rnge of input strengths. Red sigmoid indites the experiment shown on the left. () Top left, reording onfigurtion., reording eletrode. Top tres represent the responses of two CA1 pyrmidl ells simultneously reorded in loose-pth to threshold stimultion of Shffer ollterl inputs (five superimposed sweeps; suesses re shown in lk, filures in gry). The pyrmidl ell on the left ws reruited t weker stimulus thn the pyrmidl ell on the right. Bottom tres represent threshold EPSCs (tht is, EPSCs evoked t threshold input strength, verge of ten tres) reorded in the sme two ells voltge lmped t 8 mv. The pyrmidl ell on the left neessitted less exittion to reh threshold. Right, summry grph of s (lk, n = 32, spike threshold determined in loose pth for n = 15 ells nd in whole-ell urrent lmp for n = 17 ells) plotted ginst input strength t threshold (in width.1). Dshed line represents the verge for the.1 in. Error rs re s.e.m. () Top left, reording onfigurtion. Top tres represent the response of single CA1 pyrmidl ell reorded in loose pth to threshold stimultion of two different Shffer ollterl inputs (stimuli nd, five superimposed sweeps eh). Bottom tres represent threshold EPSCs (verge of ten tres) reorded in the sme ell voltge lmped t 9 mv. The differene in mplitude of the two threshold EPSCs should e noted. Right, summry grph (n = 19). There ws no orreltion etween s evoked y input nd input (liner regression, R 2 =.7; spike threshold determined in loose pth for ll ells, red dt point indites the experiment shown on the left). (n = 39) represents the frtionl reruitment of the CA1 pyrmidl ell popultion, or tivtion urve (Fig. 1). The dynmi rnge of the pyrmidl ell popultion (tht is, the rtio of the input strength neessry to tivte 95% versus 5% of the pyrmidl ell popultion) ws pproximtely 34 (Fig. 1), mening tht the pyrmidl ell popultion n differentilly represent 34-fold inrese in the numer of tive Shffer ollterl inputs efore sturting. This is muh lrger thn the dynmi rnge of n individul pyrmidl ell (1.6 ±.7, n = 37; Fig. 1; invrint etween pyrmidl ells reruited t different input strengths, R 2 =.34, P =.27; Supplementry Fig. 2) nd is the result of stggered reruitment of CA1 pyrmidl ells over wide rnge of stimulus strengths (Fig. 1). Why re some pyrmidl ells reruited t low input strength, wheres others require muh stronger stimuli? We ompred the exittory postsynpti ondutne (EPSG) evoked t threshold input strength 2 of pyrmidl ells reruited over the rnge of input strengths ( refers to the EPSG evoked t threshold). Figure 1 illustrtes n exmple of two pyrmidl ells, simultneously reorded in the loose-pth onfigurtion, tht required different stimulus strengths to spike. Whole-ell, voltge-lmp reording from the sme two ells showed tht the in the pyrmidl ell reruited y the stronger stimulus ws muh lrger thn in the pyrmidl ell reruited with weker stimulus (Fig. 1). Over ll of the experiments, we oserved steep inrese in s with inresing input strength (.3-nS inrese per perentile input strength, n = 32, P =.24; Fig. 1). The inrese in with input strength ws not unique to pyrmidl ells reruited y single-shok stimultion of the Shffer ollterl. Even when pirs of pyrmidl ells were reruited y repetitive high-frequeny stimultion (2 6 stimuli t.2.5 khz), mimiking ursting tivity in CA3, s reorded in vivo 14, s were 1 Spiking proility (%) 5 Single ruited s (%) 1 95 Popultion tivtion urve 3 pa 5 ms Stimulus 5 pa 5 ms Loose pth Wek input Strong input Stimulus Whole ell Stimulus Loose pth Input (.43) Input (.66) Whole ell 5 pa 25 pa 5 pa 5 ms (ns) Input (ns) signifintly lrger in the ell reruited with higher input strength (1.7 ±.2 fold lrger, n = 12 pirs, P =.12; Supplementry Fig. 3). This held true for even higher stimulus frequenies (4 6 stimuli t 1 khz, 1.6 ±.2 fold lrger, n = 6 pirs, P =.45; Supplementry Fig. 3). Thus, pyrmidl ells reruited t higher input strengths need lrger EPSGs to reh spike threshold. Are differenes in mplitudes the results of vriility in intrinsi pyrmidl ell properties? Input resistne, memrne time onstnt, resting potentil nd threshold potentil did not signifintly differ etween pyrmidl ells reruited t different input strengths (Supplementry Fig. 4). To further rule out the influene of intrinsi vriility etween pyrmidl ells, we ompred s etween two independent Shffer ollterl inputs onverging onto single pyrmidl ell (Fig. 1). s were unorrelted etween the two inputs (Fig. 1). Furthermore, in n individul pyrmidl ell, the evoked y the stronger input ws invrily lrger thn the evoked y the weker input (1.5 ±.2 fold lrger; P =.2, n = 19; Fig. 1). Finlly, there ws no signifint differene in the rise nd dey kinetis of s evoked y the wek nd strong inputs (1 9% rise time: strong stimulus, 2.8 ±.2 ms; wek stimulus, 2.6 ±.3 ms; P =.67, n = 19; dey time onstnt: strong stimulus, 7. ±.4 ms; wek stimulus, 6.8 ±.4 ms; P =.74, n = 19), ruling out differenes resulting from the distriution of the exittory inputs long the somtodendriti xis. Thus, even in n individul pyrmidl ell, the vried depending on the tivted input, inditing tht the sme pyrmidl ell n e reruited t oth the low or high end of the stimulus rnge. The inrese in implies tht the ontriution of eh individul fferent in firing the neuron dereses with inresing input strength. By how muh does this derese? Over the rnge of R 2 = Input (ns) 1 5 Spiking proility (%) 1578 VOLUME 12 NUMBER 12 deemer 29 nture NEUROSCIENCE

3 29 Nture Ameri, In. All rights reserved. input strengths from to.5, the mplitude of the inresed pproximtely linerly (Fig. 1; see model elow) suh tht EPSGTN = Nk + EPSGT where N is the numer of tive fferents, N is the when N fferents re tive, is the EPSG neessry to reh threshold t miniml input strength (under our ondition, it ws ~6 ns; Fig. 1) nd k is the proportionlity ftor. Given g, the synpti ondutne produed y n individul fferent, the reltive ontriution g of eh fferent towrd firing ell, N, is g. ( Nk + EPSGT) Thus, the reltive ontriution of individul fferents in firing ell is normlized y the numer of tive fferents. FFI expnds popultion s dynmi rnge Wht determines the mplitude of the nd why does it vry with input strength? Stimultion of Shffer ollterls triggers powerful FFI in CA1 pyrmidl ells through the reruitment of GABAergi interneurons There ws strong orreltion etween the mplitude of the nd the mplitude of the onomitntly triggered feedforwrd inhiitory postsynpti ondutne (IPSG; Fig. 2; see Online Methods nd Supplementry Fig. 5). Furthermore, onsistent with the orreltion etween nd input strength (Fig. 1), FFI inresed with input strength, efore sturting t input vlues ove ~.5 (Fig. 2). These dt suggest tht my vry with input strength euse of prllel inrese of FFI. We diretly tested this possiility y either olishing GABAergi trnsmission or y imposing fixed mount of inhiition (Fig. 2). Aolishing FFI with the GABA A reeptor ntgonist Figure 2 Feedforwrd inhiition expnds the dynmi rnge of the pyrmidl ell popultion. () Top tres represent whole-ell urrent-lmp reording from two CA1 pyrmidl ells reruited t threshold y wek (left) or strong (right) Shffer ollterl stimultion (five superimposed sweeps; lk indites suesses nd gry indites filures to trigger spike). Bottom tres, represent threshold EPSC (lk, verge of five tres reorded in the voltge lmp, 88 nd 92 mv for left nd right, respetively) nd onomitntly evoked feedforwrd IPSC (lue, reorded t 52 nd 59 mv for left nd right, respetively, nd isolted y sutrtion from verge of ten sweeps). Insets represent expnded timesle of the sweeps. The size of the two insets hs een sled to mth EPSC mplitudes. Bottom left, threshold feedforwrd IPSG (IPSG T ) plotted ginst (in width of 2.5 ns, n = 3, spike threshold determined in loose pth for n = 19 ells nd in whole-ell urrent lmp for n = 11 ells, dotted line is the liner regression fit of the inned dt, R 2 =.61, slope of.82). Bottom right, feedforwrd IPSG plotted ginst input strength (in width of.1, n = 5, ontinuous lue line is Boltzmnn fit of the inned dt). Error rs re s.e.m. () Summry grph of s plotted ginst input strength in the presene of gzine (6 µm, n = 3, spike threshold determined in loose pth for n = 2 ells nd in whole-ell urrent lmp for n = 1 ells) or under toni inhiition (1 µm musimol nd.5 1 µm DAMGO, n = 14, spike threshold determined in loose-pth for n = 11 ells nd in whole-ell urrent lmp for n = 3 ells). Dotted nd dshed horizontl lines represent the verge during toni inhiition or gzine tretment, respetively. In ontrst with ontrol onditions (lk line from Fig. 1), the reorded in gzine or toni inhiition hnged little with inresing input strength. For ll input strengths, the during toni inhiition ws lrger thn during gzine tretment. () Ativtion urves (umultive distriution of input strengths eliiting 5% spiking) in ontrol onditions (lk symols from Fig. 1) nd fter gzine tretment (n = 28, spike threshold determined in loose pth for ll ells). Dshed lines indite input strengths reruiting 5% nd 95% of the pyrmidl ell popultion. Error rs re s.e.m. gzine eliminted the inrese in with input strength (nonsignifint inrese of.7 ns per perentile input strength, n = 3, P =.22; Fig. 2), demonstrting ruil role of GABA A reeptors. To impose fixed mount of inhiition, irrespetive of input strength (Fig. 2), we first inhiited GABA relese with the µ-opioid reeptor gonist DAMGO (.5 1 µm, 77.9 ± 7.3% redution, n = 4; Supplementry Fig. 6) 18 nd then produed toni tivtion of GABA A reeptors y perfusing the seletive gonist musimol (1 µm; verge hyperpolriztion, 2.4 ±.5 mv; verge ondutne inrese, 4.7 ±.3 ns; n = 4). In the presene of toni inhiition, no longer vried with input strength (nonsignifint derese of.3 ns per perentile input strength, n = 14, P =.4; Fig. 2). Thus, the progressive inrese in GABA A reeptor tivtion ounts for the inrese in. This is supported y the ft tht t high input strengths (>.5), when the mplitude of FFI no longer inresed (Fig. 2), remined onstnt (nonsignifint derese of.2 ns per perentile input strength, P =.4; Supplementry Fig. 7). By how muh does the dynmi inrese the rnge of inputs tht the pyrmidl ell popultion responds to? We ompred the tivtion urve of the CA1 pyrmidl ell popultion under ontrol onditions nd in the presene of gzine, where the is fixed (Fig. 2). The numer of Shffer ollterls neessry to reruit the lowest frtions of the pyrmidl ell popultion ws omprle in oth onditions (for exmple, 5% reruitment: ontrol,.28 input strength; gzine,.3 input strength; Fig. 2). The sitution ws, however, rdilly different when lrger numers of Shffer ollterls were tivted. In the presene of gzine, n pproximtely eightfold inrese in the numer of tivted Shffer ollterls redily led to the sturtion of the pyrmidl ell popultion (95% reruitment with.26 input strength), wheres the sme inrese in stimulted IPSG T (ns) (ns) Threshold input strength:.3 Current-lmp Voltge-lmp (ns) Gzine Toni inhiition Control 1 R 2 = IPSG (ns) ruited s (%) Threshold input strength:.7 Current-lmp Voltge-lmp Ativtion urves 5 mv 6 ms 25 pa 5 pa 2 ms Control Gzine nture NEUROSCIENCE VOLUME 12 NUMBER 12 deember

4 Figure 3 Pyrmidl ells spike fter the onset of feedforwrd inhiition. () Top tres re loose pth reordings from CA1 pyrmidl ell in response to threshold stimultion of Shffer ollterls (five superimposed sweeps). Bottom tres re voltge-lmp reordings from the sme neuron; the EPSC (lk line, verge of five tres) ws reorded t 85 mv nd the feedforwrd IPSC (lue line, isolted y sutrtion from verge of ten tres) ws reorded t 6 mv. The vertil dshed lines mrk the onset of the IPSC nd the verge timing of the spike in the pyrmidl ell. The IPSC onset ourred efore the spiking of the pyrmidl ell. Bottom, summry of 3 similr experiments (spike threshold determined in loose pth for n = 18 ells nd in wholeell urrent lmp for n = 12 ells). The open symol represents the verge. () The net threshold hrge (EPSC minus IPSC) entering the ell from the onset of the EPSC to the time of the spike nd the re Loose pth 4 pa Whole ell 1 na 1 ms t t (ms) Stimulus shown for two different rnges of threshold input strengths. The threshold hrge did not inrese signifintly with inresing input strength (P =.3). In ontrst, the pek ondutne of the reorded in the sme ells ws signifintly lrger for lrger input strengths (P =.3). Error rs represent s.e.m. () Simultneous reording from two neighoring pyrmidl ells in whih Shffer ollterls stimultion ws suffiiently strong to reh threshold in one ell (lk), ut not in the other (gry, n = 15 pirs). The net threshold hrge entering in the spiking pyrmidl ells ws signifintly lrger thn the net hrge entering the nonspiking pyrmidl ells (P =.4). Cirles represent individul experiments nd horizontl lines represent verges. Threshold hrge (pc) 1 5 (ns) Spiking Nonspiking Chrge (pc) 29 Nture Ameri, In. All rights reserved. Shffer ollterls in ontrol onditions reruited only 19.9% of the popultion (Fig. 2). Thus, gzine inresed the slope of the tivtion urve without produing mjor hnges in the offset (Fig. 2). Hene, dynmi leds to fourfold expnsion of the rnge of inputs tht the CA1 pyrmidl ell popultion n respond to. FFI rrives efore spike By wht mehnism does the feedforwrd inhiitory postsynpti urrent (IPSC) ontrol the size of the? We ompred the timing of the spike eliited in pyrmidl ells y Shffer ollterl stimultion with the onset of the feedforwrd IPSC. When stimulted t threshold for spike genertion, the spike ourred 5. ±.4 ms fter the onset of the EPSC nd 3.3 ±.4 ms fter the onset of the feedforwrd IPSC (n = 3; Fig. 3). The lteny etween the onset of the EPSC nd of the IPSC ws 1.65 ±.8 ms (n = 3), onsistent with previous dt 15, nd did not hnge with stimulus strength (R 2 =.7, P =.4, n = 3). Thus, in response to threshold Shffer ollterl stimultion, FFI rehed pyrmidl ells efore the memrne potentil of the neuron rehed threshold for spike genertion. Over the period preeding the spike, synpti inhiition overlpped with the EPSC, therey reduing the exittory hrge entering the ell y 28.1 ± 4.3% (n = 3). Speifilly, lthough the integrl of the EPSC from its onset to the time of the spike (exittory hrge) verged 2.2 ±.2 pc (n = 3), the net synpti hrge (exittory-inhiitory hrge, see Online Methods) entering pyrmidl ells ws 1.4 ±.1 pc (n = 3). In ontrst to the, this net threshold hrge ws onstnt nd independent of input strength (threshold hrge, 1.1 ±.1 pc, (n = 4) t.25 input strength versus 1.4 ±.1 pc, (n = 14) t.25.5, P =.3;, 7.1 ±.9 ns (n = 4) t.25 input strength versus 13.1 ± 1.3 ns (n = 14) t.25.5, P =.3; Fig. 3). Furthermore, t ny given input strength, the threshold hrge ws signifintly lrger in pyrmidl ells tht rehed threshold for spike genertion s ompred with the hrge entering over the sme time intervl in simultneously reorded ells tht did not reh threshold (.8 ±.1 pc, P =.4, n = 15; Fig. 3). In the ells tht did not spike, the threshold hrge (tht is, ~1.5 pc) would hve een rehed 5. ±.3 ms fter the onset of the EPSC if inhiition hd not een present. Thus, y overlpping with exittion efore spike ourrene, FFI ontrols the mplitude of the EPSC neessry to reh spike threshold. Heterogeneous exittion nd homogeneous inhiition Wht determines whih pyrmidl ells in the popultion re reruited in response to Shffer ollterl stimultion? We reorded from two neighoring pyrmidl ells simultneously (somt seprted y 5 µm) nd inresed the numer of tivted Shffer ollterls until one of the two ells spiked (Fig. 4). We then ompred the EPSGs nd feedforwrd IPSGs in the two ells. Although the EPSG ws, on verge, 1.6 ±.1 fold lrger in the ell tht spiked (P =.1, n = 15; Fig. 4), the IPSG ws, on verge, not signifintly different etween the two neurons (1.1 ±.1 fold differene, P =.3, n = 15; Fig. 4). Furthermore, the lteny of the feedforwrd IPSC (with respet to the onset of the EPSC) did not differ signifintly etween spiking (1.75 ±.9 ms) nd nonspiking neurons (1.59 ±.6 ms, P =.9, n = 15). Thus, differenes in the mplitude of synpti exittion, rther thn in the mplitude or timing of inhiition, govern whih neuron will spike in response to Shffer ollterl stimultion. To determine whether inhiition is more homogeneously distriuted ross pyrmidl ells s ompred with exittion, we omputed the spred, tht is, the solute differene in mplitude of simultneously reorded EPSGs or IPSGs normlized y the verge of the mplitudes nd divided y two. Although the spred of EPSGs etween two simultneously reorded pyrmidl ells ws 21 ± 3% (n = 15, sme pired vlues s ove; Fig. 4), the spred of the onomitnt IPSGs ws only 11 ± 2% (P =.3, n = 15). We lso lulted how well the mplitude of inhiition in one ell orrelted with the mplitude of inhiition in its neighor nd did the sme for exittion. For this, we used the sme pired vlues s desried ove (Fig. 4), ut we rndomly lloted the spiking ell to either one of the two xes (Fig. 4). This rndomiztion removes the orreltion is used y systemtilly hving the lrger mplitude on the sme xis. The orreltion etween IPSGs (R In =.79) ws signifintly lrger thn the orreltion etween EPSGs (R Ex =.3, P <.2, see Online Methods; Fig. 4). Thus, inhiition is more homogeneous thn exittion ross the pyrmidl ell popultion. To test whether the reltive homogeneity of inhiition with respet to exittion lso holds true for individul synpti events, we ompred tril-y-tril flututions of the mplitude of EPSC nd feedforwrd IPSC etween two simultneously reorded pyrmidl ells (Fig. 4). Using esium-sed internl solution, we isolted feedforwrd IPSCs nd monosynpti EPSCs y voltge lmping the ell t the EPSC or IPSC reversl potentil, respetively (Fig. 4). The mplitude of the feedforwrd IPSC ovried etween the two reorded neurons (verge orreltion, R 2 =.26 ±.6, n = 5; Fig. 4). This orreltion ws signifintly less pronouned for monosynpti EPSCs (verge orreltion, R 2 =.6 ±.5, n = 5, P =.33; Fig. 4). These results indite tht, lthough FFI is 158 VOLUME 12 NUMBER 12 deemer 29 nture neuroscience

5 29 Nture Ameri, In. All rights reserved. Figure 4 Homogeneous inhiition nd heterogeneous exittion ontrol the reruitment of pyrmidl ells. () Top, reording onfigurtion. Left tres represent simultneous loose-pth reording from two neighoring CA1 pyrmidl ells. Shffer ollterls stimultion ws suffiiently strong to reh threshold in one ell (lk tres, five superimposed sweeps), ut not in the other (gry tres). Right tres represent wholeell voltge-lmp reording from the sme two neurons (top, feedforwrd IPSCs reorded t 6 mv nd isolted y sutrtion from verge of ten tres; ottom, EPSCs reorded t 8 mv, verge of ten tres). The mplitude of the feedforwrd IPSC ws similr in oth ells, wheres the EPSC ws lrger in the ell tht spiked. () Left, summry grph of 15 similr experiments in whih the EPSG in the nonspiking ell is plotted ginst the EPSG in the spiking ell. The mjority of the dt points re elow the unity line (red dt point indites the experiment shown on top). Right, summry grph of the sme 15 experiments in whih the feedforwrd IPSG in the nonspiking ell is plotted ginst the feedforwrd IPSG in the spiking ell. In ontrst with the EPSG, ll of the dt points re sttered round the unity line (spike threshold determined in loose pth for n = 8 pirs nd in whole-ell urrent lmp for n = 7 pirs; red dt point indites experiment shown in ; sme set of experiments illustrted in Fig. 3). Error rs re s.e.m. Insets hve the sme dt points s re shown in the min grphs, ut the spiking ell is rndomly lloted to either one of the two xes. Note the lrger spred of EPSGs s ompred with IPSGs. () Tril-y-tril flutution of EPSGs (left) nd IPSGs (right) simultneously reorded in two pyrmidl ells ( 1 nd 2) voltge lmped t the reversl potentil of IPSCs (left) or EPSCs (right, sme Shffer ollterl stimultion intensity for oth holding potentils, esium internl). Left, single-tril EPSGs reorded in 1 re plotted ginst the EPSGs reorded simultneously in 2. Upper tres re five exmple EPSCs reorded in 1 ordered ording to mplitude. Lower tres re the orresponding five EPSCs reorded simultneously in 2. Right, single-tril IPSGs reorded in 1 re plotted ginst the IPSGs reorded simultneously in 2. Upper tres re five exmple IPSCs reorded in 1 ordered ording to mplitude. Lower tres re the orresponding five IPSCs reorded simultneously in 2. Note the mrked ovrition in mplitude of IPSGs reorded in the two pyrmidl ells s ompred with EPSGs. reltively homogenous ross the pyrmidl ell popultion, nd thus sets glol threshold for pyrmidl ell reruitment y Shffer ollterls, heterogeneities in exittion determine whih pyrmidl ells in the popultion overome this threshold. Bsket ells expnd dynmi rnge of pyrmidl ell popultion Severl types of hippompl inhiitory interneurons re tivted in feedforwrd mnner Wht type of interneuron ontrols the mplitude of the in pyrmidl ells? Beuse the onset of FFI ours efore the spiking of pyrmidl ells (see ove; Fig. 3), these interneurons must spike efore pyrmidl ells in response to fferent stimultion. We ompred the spike timing of different types of interneurons in response to Shffer ollterl stimultion with the timing of the pek of the popultion spike (reorded in the strtum pyrmidle; Fig. 5,). Pyrmidl ells fired simultneously with the popultion spike (.2 ±.2 ms, P =.52, n = 21; Fig. 5,), wheres regulr-spiking interneurons 19,2 were reruited fter the popultion spike (.7 ±.3 ms, P =.4, n = 34). In ontrst, fst-spiking inhiitory interneurons 19,2 fired 1 ±.2 ms efore the popultion spike (P =.3, n = 18; Fig. 5,), onsistent with the erly onset of FFI. We omputed the tivtion urve of fst-spiking nd regulrspiking interneurons (Fig. 5), s we did for pyrmidl ells (Fig. 1). The tivtion urve of fst-spiking interneurons ws muh steeper tht the one for regulr-spiking interneurons nd pyrmidl ells (Fig. 5), onsistent with the strong nd fst exittion reeived y fst-spiking interneurons 21,23,24 (hlf mximl reruitment of fstspiking interneurons ourred t n input strength of.11 ompred with.37 nd.4 for regulr-spiking interneurons nd pyrmidl ells, Nonspiking (ns) 1 (ns) Loose pth Spiking Nonspiking EPSG 1 2 Spiking (ns) 25 pa 5 ms 5 pa 5 ms 5 5pA 2 (ns) Stimulus 1 Whole ell IPSC respetively). Notly, the tivtion urve of fst-spiking interneurons (Fig. 5) provided good mth for the inrese in the mplitude of FFI over the sme rnge of stimuli (hlf mximl reruitment of FFI ourred t n input strength of.13; Fig. 2). Antomil identifition of these fst-spiking interneurons reveled tht 63% were sket ells (for xonl nd dendriti distriution, see Supplementry Fig. 8) nd the rest were omposed of dendrite trgeting interneurons exhiiting xonl roriztion onsistent with i-strtified nd tri-lminr ells 19,2 (n = 11 fst-spiking ells; Fig. 5). Thus, these dt indite tht the dynmi in CA1 pyrmidl ells is primrily enfored y fst-spiking interneurons, the mjority of whih re sket ells. Model Is the oserved hnge in suffiient to ount for the expnsion of the rnge of inputs the pyrmidl ell popultion responds to? We reted simple quntittive model of Shffer ollterl exittion onto popultion of CA1 pyrmidl ells (Fig. 6; see Online Methods). Shffer ollterl inputs ontted pyrmidl ells with proility of.6 (ref. 25). We omputed the frtion of reruited pyrmidl ells s funtion of the numer of stimulted Shffer ollterl nd ompred the resulting tivtion urves under two onditions: with either fixed or dynmi (Fig. 6). The threshold to reruit pyrmidl ell ws set t 1 Shffer ollterl inputs 2 nd remined onstnt in the fixed ondition. In the dynmi ondition, the numer of Shffer ollterls neessry to reruit pyrmidl ell inresed linerly with inresing numer of stimulted Shffer ollterls, up to pproximtely threefold, Nonspiking (ns) 1 (ns) Spiking (ns) EPSC IPSG T IPSG 1 na 1 ms (ns) 5 pa 25 ms nture NEUROSCIENCE VOLUME 12 NUMBER 12 deember

6 29 Nture Ameri, In. All rights reserved. Figure 5 Fst-spiking interneurons enfore dynmi. () Top, reording onfigurtion. Left, simultneous field (from pyrmidl ell lyer) nd loose-pth reording from three types of neurons in response to Shffer ollterl stimultion (three different experiments): pyrmidl ell (top, five superimposed sweeps), fst-spiking interneuron (FS, middle, five superimposed sweeps) nd regulr-spiking interneuron (RS, ottom, five superimposed sweeps). Although the spike in the pyrmidl ell ws onomitnt with the pek of the popultion spike reorded with the field eletrode, the tion potentil in the fst-spiking ell preeded, nd in the regulr-spiking ell followed, the popultion spike. Right, mer luid reonstrution of the three neurons on the left (dshed lines mrk the mrgins of the pyrmidl ell lyer, strtum rditum is ove the dshed lines; d, dendrite; x, xon). Insets, spiking pttern from sme neurons reorded in whole-ell urrent-lmp onfigurtion in response to depolrizing nd hyperpolrizing urrent steps. () Reltive timing of spikes eliited in response to Shffer ollterl stimultion in pyrmidl ells (lk, n = 21), fst-spiking interneurons (lue, n = 18) nd regulrspiking interneurons (gry, n = 34) with respet to the pek of popultion spike (spike threshold determined in loose pth for ll ells). On verge, fst-spiking interneurons fired efore nd regulr-spiking interneurons fired fter the popultion spike. () Ativtion urves (umultive distriution of threshold input strengths) for fst-spiking (lue, n = 18), regulr-spiking (gry, n = 34) nd pyrmidl ells (lk, n = 39, from Fig. 1). The input strength tht reruited 5% of fst-spiking interneurons eliited spikes in only 13% of pyrmidl ells (spike threshold determined in loose pth for ll ells). The ontinuous light lue line is the fit of the feedforwrd IPSG s funtion of input strength (right ordinte from Fig. 2, symptote sled to 1%). There ws good mth with the tivtion urve of fst-spiking neurons. nd then remined onstnt to simulte experimentl oservtion (Fig. 6). The ext inrement of the modeled dynmi ws hosen to yield n tivtion urve tht est pproximted the experimentlly oserved tivtion urve (Fig. 6). With fixed, the tivtion urve ws steep nd hd nrrow dynmi rnge (Fig. 6). With dynmi, on the other hnd, the tivtion urve hd n onset similr to the fixed threshold tivtion urve, ut rose muh less steeply, resulting in wider dynmi rnge (Fig. 6; the sensitivity of the slope of the tivtion urve to the rte of inrese of the dynmi is illustrted in Supplementry Fig. 9). The dynmi tivtion urve (Fig. 6) is the synthesis of fmily of fixed tivtion urves, eh hving progressively lrger ( suset re illustrted in Fig. 6). An intersetion numer (%) ruited s (%) Fixed Dynmi 16 Input strength: 8 Are N ove T: 3% 3% T T N 8 33% 8% T 1.2T EPSG mplitude T N 1.25T 1.75T 2.2T 4T 3T 2T T 24 Dynmi ESPG T Fixed ESPG T T 3.1T Model Fixed ESPG T Dynmi ESPG T Experiment Gzine Control.4.6 Model (u) Experimentl (ns) FS RS FS Field Loose pth RS Stimulus 3 mv 2 pa 3 mv 2 pa 3 mv Dely reltive to popultion spike (ms) ours t the speifi input strength t whih dynmi nd fixed tivtion urves hve equl. This simple model ptures the si experimentl finding, nmely tht the dynmi expnds the rnge of inputs tht the CA1 pyrmidl ell popultion responds to y mintining sensitivity to wek inputs nd preventing sturtion to stronger stimuli. It should e noted, however, tht the modeled fixed nd dynmi tivtion urves oth fil to fully ount for the experimentlly oserved tivtion urves, where more thn ~8% of pyrmidl ells re tive (see Disussion). Dynmi in somtosensory ortex Feedforwrd inhiitory iruits involving fst-spiking interneurons hve een desried long severl ortil projetions 24, Is the dynmi generl mehnism y whih the ortex expnds the rnge of inputs it n respond to? We tested this hypothesis t one of the min projetions in the neoortil nonil iruit, the IN lyer lyer d x 4 mv 5 ms 4 pa 5 ms 2 µm ruited neurons (%) 1 Ativtion urves FS RS Figure 6 Model of tivtion urve with dynmi. () Modeled distriution of EPSG mplitudes in the popultion of pyrmidl ells when N (top row) nd 1.5N (ottom) fferent fiers re tive. The re ove spike threshold T, under the urve (lk shded), is the frtion of pyrmidl ells reruited with either fixed (left olumn) or dynmi (right). For dynmi, smller frtion of pyrmidl ells ws reruited when inresing the numer of tive fferent fiers from N to 1.5N. () s funtion of input strength used in the model; dynmi (ontinuous line), fixed (dshed line) nd experimentlly mesured dynmi (gry olumns, sme dt s in Fig. 1, ut for the entire rnge of stimulus strengths) re shown. () Modeled pyrmidl ell tivtion urves with fixed or dynmi. Experimentlly mesured tivtion urve in gzine nd ontrol onditions re superimposed. Dotted gry lines represent set of fixed tivtion urves (for eh tivtion urve, the threshold is given in multiples of T, the threshold t miniml input strength). The dynmi tivtion urve intersets eh of the fixed tivtion urves t the speifi input strength t whih the threshold of the two urves mthes. The dynmi tivtion urve thus results from the synthesis of fmily of fixed tivtion urves. IPSG (ns) 1582 VOLUME 12 NUMBER 12 deemer 29 nture neuroscience

7 29 Nture Ameri, In. All rights reserved. Figure 7 The stronger the photostimultion of L2/3 pyrmidl ells, the lrger the exittion neessry to reruit L5 pyrmidl ells. () Left, overly of low-mgnifition right-field nd red nd green fluoresent imges of representtive ortil rin slie from 27-dy-old mouse eletroported while in utero with ChR2-Venus (green). The red ells re pir of L5 pyrmids filled with Alex Fluor 568. Note the green nd of ChR2-Venus expressing ells in L2/3 nd the finter nd in L5 representing the xonl roriztion of the L2/3-L5 projetion. Sle r represents 2 µm. Right, two-photon stk of the sme slie t higher mgnifition. Sle r represents 5 µm. Nerly ll of the ChR2-Venus expressing neurons were in L2/3. Individul xons from L2/3 ells n e seen rossing L4 nd rorizing in L5. () Top, simultneous urrent-lmp reording from two lyer 5 pyrmidl ells exited y 5-ms squre light pulse. The pyrmidl ell on the left ws reruited t threshold y lower-intensity photostimultion s ompred with the pyrmidl ell on the right. Suesses in triggering spike re shown in lk nd filures in gry (3 4 superimposed tres). The lue tres indite the durtion nd reltive mplitudes of the photostimultions. Bottom tres, threshold EPSCs (verge of 5 1 tres) reorded in the sme two ells voltge lmped t 7 mv. The pyrmidl ell on the left neessitted less exittion to reh threshold. A summry grph of s reorded in ells reruited y the higher-intensity photostimulus plotted ginst the simultneously reorded in ells reruited y the lower-intensity photostimulus (n = 15) is shown. Most of the dt points lie ove the unity line (red dt point indites the experiment shown on top). () The sme experimentl onfigurtion s in ws used, ut photostimultion is 1-ms light rmp. The stter plot shows the threshold exittory hrge (threshold EPSC time integrl) reorded in ells reruited y the higher-intensity photostimulus plotted ginst the threshold exittory hrge simultneously reorded in ells reruited y the lower-intensity photostimulus (n = 14). As in, most of the dt points lie ove the unity line (red dt point indites the experiment shown on top). (d) Top, plot of spike ltenies of L5 pyrmidl ells to stimultion of L2/3 with 5-ms pulses of lue light (lk symols). Bottom, EPSC (lk tre, reorded t IPSC reversl potentil) nd IPSC (lue tre, reorded t EPSC reversl potentil) in lyer 5 pyrmidl ell with esium-sed internl solution. The onset of disynpti inhiition (5% of pek, vertil dotted line) preeded the verge spike lteny (open symol). Error rs re s.e.m. (e) Stter plot of the spred of EPSCs plotted ginst the spred of IPSCs reorded in pirs of L5 neurons. Most dt-point re elow the unity line (n = 13 pirs). lyer 2/3 to lyer 5 exittory projetion of the somtosensory ortex 1 (Fig. 7). Beuse ortil rhiteture does not permit seletive stimultion of the xons of lyer 2/3 pyrmidl ells with n extrellulr stimultion eletrode, we eletroported mie in utero with hnnelrhodopsin-2 (ChR2) to trget lyer 2/3 pyrmidl ell progenitors 3,31 (Fig. 7; see Online Methods). Phototivtion of ChR2-expressing lyer 2/3 pyrmidl ells in slies of juvenile rins triggered oth diret exittion nd FFI in lyer 5 pyrmidl ells (Supplementry Fig. 1). We reorded from two postsynpti lyer 5 pyrmidl ells simultneously nd inresed the intensity of the photostimultion of the lyer 2/3 xons until one of the two reorded lyer 5 pyrmidl ells rehed spike threshold (low-intensity stimultion). We then further inresed the intensity of the photostimultion until the seond lyer 5 pyrmidl ell rehed threshold (high-intensity stimultion). Similr to the hippompus, lyer 5 pyrmidl ells reruited t high intensity required lrger s s ompred with pyrmidl ell reruited t low intensity (1.3 ±.1 fold more, n = 15, P =.19; Fig. 7). This differene in ws not result of differenes in memrne properties of lyer 5 pyrmidl ells (input resistne: low-intensity ells, 92 ± 14 MΩ; high-intensity ells, 94 ± 14 MΩ; P =.9, n = 15; memrne time onstnt: low-intensity ells, 19 ± 2 ms; high-intensity ells, 21 ± 2 ms; P =.41, n = 15). Does the lso vry when fferent tivity is distriuted in time, similr to wht ours under more physiologil onditions 7? L2/3 L4 L5 L6 Current-lmp Voltge-lmp High intensity (ns) d Voltgelmp Light pulse 4 mv 25 pa 5 ms Current-lmp Voltge-lmp Low intensity (ns) Low intensity (pc) Time (ms) na 5 pa 1 ms To desynhronize the tivtion of the input popultion, we progressively rmped up the intensity of the light stimulus over period of 1 ms (Fig. 7; see Online Methods). When reording from two lyer 5 pyrmidl ells simultneously, s desried ove, the exittory hrge neessry to reruit the lyer 5 pyrmidl ell with the highintensity stimulus ws 1.8 ±.4 fold lrger thn for pyrmidl ells reruited t low intensity (n = 14, P =.21; Fig. 7). This indites tht the lso vries under onditions in whih the tivity of presynpti neurons is desynhronized. Thus, the is dynmi for lyer 5 ortil pyrmidl ells, suh tht pyrmidl ell reruited when mny presynpti lyer 2/3 pyrmidl ells re tive neessittes signifintly lrger EPSGs thn pyrmidl ell reruited when few lyer 2/3 pyrmidl ells re tive. Is the dynmi in the somtosensory ortex sed on the sme mehnism s the one identified in the hippompus? As in the hippompus, the onset of FFI preeded the spiking of lyer 5 pyrmidl ells (y 2. ±.4 ms, n = 24; Fig. 7d) nd inresed with stimulus strength (Supplementry Fig. 1). The overlp etween EPSC nd IPSC efore spike genertion ws even more pronouned in response to desynhronized tivtion of the inputs. In ft, lthough the lteny etween EPSC nd feedforwrd IPSC ws similr to wht we oserved in response to synhronized stimulus (1.5 ±.1 ms, n = 3), the spiking ourred only 6.2 ±.1 ms (n = 26) fter the onset of the EPSC, resulting in n lmost 5-ms overlp. Finlly, similr to the hippompus, High intensity (pc) e IPSC spred (%) Light rmp 4 mv Threshold exittory hrge EPSC spred (%) nture NEUROSCIENCE VOLUME 12 NUMBER 12 deember

8 29 Nture Ameri, In. All rights reserved. FFI ppered to e more homogenously distriuted ross lyer 5 pyrmidl ells s ompred with exittion. Although the spred of the EPSCs simultneously reorded in two pyrmidl ells ws 36 ± 5% (n = 13), the spred of the onomitnt IPSCs ws 23 ± 4% (P =.7, n = 13; Fig. 7e). DISCUSSION Our dt suggest tht the EPSC mplitude neessry to reh spike threshold in hippompl nd neoortil pyrmidl ells is dynmi nd inreses when the numer of tive neurons in the presynpti lyer inreses. Aordingly, the frtionl ontriution of n individul fferent input in firing neuron is not fixed, ut insted is ontinuously normlized y the totl numer of tive fferents. Through this simple mehnism, the pyrmidl ell popultion n smoothly operte over wide rnge of stimulus strengths. This mehnism is proly importnt for enling ortil strutures suh s the hippompus to e responsive to wek stimuli, ut remin sprsely tive even when onfronted with stronger inputs. Although FFI sets glol threshold for reruitment of pyrmidl ells, lol differenes in fferent exittion determine whih pyrmidl ell is reruited. The frtion of neurons tive t ny given moment in ortil res strongly flututes s result of either vrying sensory stimuli or ongoing intrinsi tivity. The proility of spiking in lyer 2/3 pyrmidl ells in the somtosensory ortex, for exmple, rpidly flututes in response to nturlisti stimuli pplied to the whiskers 7. Similrly, tivity levels in the hippompus n vry from sprse tivity in the exploring niml 8 to synhronous tivtion of lrge frtion of neurons during ripples 9. As onsequene, downstrem trgets of these neuronl popultions experienes sustntil flututions in the frtion of fferents tht re tive t ny given time point. The onnetivity ptterns of ortil exittory projetions, however, re ill-suited to llow postsynpti popultions of neurons to operte over wide rnge of fferent tivity 5 ; fferent xons typilly form very divergent projetions to ontt lrge numer of postsynpti trgets through reltively wek ontts, suh tht the simultneous tivity of severl fferents is neessry to reruit trget neuron 2,32. Beuse of this divergene, grdul inreses in the numer of tive fferents produe very steep, or explosive, inreses in the frtion of reruited trgets 5, resulting in limited rnge of input strengths tht n e differentilly represented y the postsynpti popultion. Our dt indite tht pleo- nd neoortil iruits expnd the rnge of fferent input strengths tht the ells n respond to y ensuring tht, when the input is strong, pyrmidl ells neessitte lrger EPSCs to reh thresholds. The dynmi rnge of the popultion is severl-fold wider thn the dynmi rnge of n individul pyrmidl ell (34 versus 2). Neither the dynmi rnge nor the gin (slope of the sigmoidl fit, R 2 =.46, P =.2) of individul pyrmidl ells vried etween pyrmidl ells reruited long the entire input rnge. Thus, lthough GABA A reeptor medited ondutnes n regulte the gin of individul neurons 1 12, our dt suggest tht pyrmidl ell popultions n funtion over wide rnge of fferent intensities without requiring gin hnges in individul neurons. Individul exittory fferent inputs to ortil res diverge to ontt fst-spiking sket ells nd prinipl neurons. The mth etween the inrese in FFI nd the tivtion urve of fst-spiking ells (Fig. 5) suggests tht the inrese in FFI results from the inresed frtion of reruited fst-spiking ells. Beuse fst-spiking sket ells reeive lrger nd fster EPSPs 21,24,26, they were reruited efore pyrmidl ells in response to fferent tivity (Fig. 3) 33. This led to sustntil temporl overlp etween EPSCs nd FFI efore spike genertion in pyrmidl ells (Fig. 3). Only those pyrmidl ells tht reeived lrge enough EPSCs to overome the onomitntly ourring inhiition rehed spike threshold. With inresing stimulus strength, the mplitude of FFI inresed (Fig. 2) nd lrger EPSCs eme neessry for pyrmidl ells to reh spike threshold. The onset of inhiition efore spike genertion in pyrmidl ells lso mens tht this erly phse of inhiition is unlikely to e of feedk origin, s feedk inhiition is onsequene of pyrmidl ell spiking. The ontrol of the mplitude of the EPSC neessry to reh threshold y FFI is likely to e even more mrked in response to repetitive, urst-like 14 or synhronous fferent tivity, s result of the lrge temporl overlp etween fferent exittion nd FFI generted y the previous stimulus. Whether the tivtion urve of fstspiking ells (Fig. 5) is lso ontrolled y inhiitory inputs remins to e estlished. If so, reiprol inhiition of fst-spiking ells my influene the dynmi rnge of pyrmidl ell popultions. Our model ptures the initil 8% of the tivtion urve; tht is, until the plteus ove input strength of ~.5. At these greter input strengths, the is fixed nd the model predits tht the tivtion urve ehves ordingly. However, the top 2% of the experimentlly determined tivtion urve extends eyond this predition. It is possile tht the oserved onset of the plteu is inurte euse of n error in mesurement (for exmple, the lk of proper voltge lmp) nd tht the rel ontinues to grow with inresing input strength. Alterntively, smll portion of pyrmidl ells my reeive Shffer ollterl inputs with very low proility s ompred with the rest of the popultion (resulting from heterogeneity in the popultion or dmge to their dendrites) suh tht they neessitte muh lrger stimulus strength to e reruited. Our model lso illustrtes the ft tht the tivtion urve is sensitive to how steeply the vries with input strength. Beuse the inrese in is, t lest in prt, determined y the inrese in FFI, ny prmeter tht ontrols the exitility of GABAergi interneurons, suh s neuromodultors, will proly ffet the slope of the tivtion urve, nd thus the rnge of input strength tht n e represented y the postsynpti pyrmidl ell popultion. Wht determines the speifi pttern of pyrmidl ells reruited y stimulus? We found distint roles for exittion nd FFI; reltively homogeneous inhiition ross pyrmidl ells sets glol threshold 34 tht is proportionl to stimulus strength nd heterogeneities in exittion determine the preise pttern of pyrmidl ells reruited in tht popultion. The sitution in vivo my rete dditionl ises; through the lol tion of neuromodultors, suh s etylholine, some pyrmidl ells my e more depolrized thn others nd reh spike threshold even if they reeive less exittion thn their neighors. The lol tion of presynpti inhiitors of GABA relese, suh s opites, nninoids or GABA itself, my rete sptil heterogeneities in inhiition tht were not present in the slie. Severl synpti nd onnetivity properties my ontriute to homogeneous distriution of FFI, inluding the strong divergene of individul fst-spiking ells onto the pyrmidl ell popultion 21,35 37, the lrge numer of ontts mde y individul fst-spiking ells onto eh pyrmidl ell 38 nd the reltively lrge numer of reruited fst-spiking ells even t low stimulus strength (Fig. 5). In summry, euse the is ontrolled in feedforwrd mnner, the sensitivity of pyrmidl ells is virtully instntneously djusted to mth the strength of the fferent stimulus. This instntneous djustment differs from dpttion euse it does not rely on the previous history of the network through negtive feedk mehnism, suh s feedk inhiition, spike dpttion, synpti depression or presynpti inhiition. The presene of feedforwrd inhiitory iruits long severl mjor exittory pthwys 1584 VOLUME 12 NUMBER 12 deemer 29 nture neuroscience

9 29 Nture Ameri, In. All rights reserved. in the rin 15,16,29,39,4 suggests tht the expnsion of the dynmi rnge y instntneously vrying the mplitude of the EPSC neessry to reh threshold my not e unique to hippompus nd somtosensory ortex. Methods Methods nd ny ssoited referenes re ville in the online version of the pper t Note: Supplementry informtion is ville on the Nture Neurosiene wesite. Aknowledgments We thnk P. Aelkop for ntomil reonstrutions of ioytin filled neurons, F. Fröhlih for developing the initil versions of model, M. Crndini nd J. Isson for omments nd suggestions during the entire ourse of the projet, C. Poo nd F. Bertso for inputs on the mnusript, nd ll of the memers of the Snzini lortory for their input on the projet nd the mnusript. M.S. thnks C. Stu for the originl disussions leding to the projet. This work ws funded in prt y the US Ntionl Institutes of Helth (MH7141 to M.S. nd NS61521 to B.V.A.). H.A. is fellow of the Helen Hy Whithney Foundtion. M.S. is n investigtor of the Howrd Hughes Medil Institute. AUTHOR CONTRIBUTIONS F.P. nd A.M.-B. onduted the experiments in the hippompus; H.A. onduted the experiments in the somtosensory ortex; B.V.A. mde the model; nd M.S. supervised the projet nd wrote the mnusript. Pulished online t Reprints nd permissions informtion is ville online t reprintsndpermissions/. 1. Lefort, S., Tomm, C., Floyd Srri, J.C. & Petersen, C.C. The exittory neuronl network of the C2 rrel olumn in mouse primry somtosensory ortex. Neuron 61, (29). 2. Otmkhov, N., Shirke, A.M. & Mlinow, R. Mesuring the impt of proilisti trnsmission on neuronl output. Neuron 1, (1993). 3. Mrr, D. A theory of ereellr ortex. J. Physiol. (Lond.) 22, (1969). 4. Vogels, T.P. & Aott, L.F. Signl propgtion nd logi gting in networks of integrte-nd-fire neurons. J. Neurosi. 25, (25). 5. Shdlen, M.N. & Newsome, W.T. The vrile dishrge of ortil neurons: implitions for onnetivity, omputtion nd informtion oding. J. Neurosi. 18, (1998). 6. Diesmnn, M., Gewltig, M.O. & Aertsen, A. Stle propgtion of synhronous spiking in ortil neurl networks. Nture 42, (1999). 7. Arzdeh, E., Zorzin, E. & Dimond, M.E. Neuronl enoding of texture in the whisker sensory pthwy. PLoS Biol. 3, e17 (25). 8. Wilson, M.A. & MNughton, B.L. Dynmis of the hippompl ensemle ode for spe. Siene 261, (1993). 9. Csisvri, J., Hirse, H., Mmiy, A. & Buzski, G. Ensemle ptterns of hippompl CA3 CA1 neurons during shrp wve ssoited popultion events. Neuron 28, (2). 1. Shu, Y., Hsenstu, A., Bdoul, M., Bl, T. & MCormik, D.A. Brrges of synpti tivity ontrol the gin nd sensitivity of ortil neurons. J. Neurosi. 23, (23). 11. Mithell, S.J. & Silver, R.A. Shunting inhiition modultes neuronl gin during synpti exittion. Neuron 38, (23). 12. Chne, F.S., Aott, L.F. & Reyes, A.D. Gin modultion from kground synpti input. Neuron 35, (22). 13. Crvlho, T.P. & Buonomno, D.V. Differentil effets of exittory nd inhiitory plstiity on synptilly driven neuronl input-output funtions. Neuron 61, (29). 14. Tropp Sneider, J., Chrok, J.J., Quirk, M.C., Oler, J.A. & Mrkus, E.J. Differentil ehviorl stte-dependene in the urst properties of CA3 nd CA1 neurons. Neurosiene 141, (26). 15. Pouille, F. & Snzini, M. Enforement of temporl fidelity in pyrmidl ells y somti feedforwrd inhiition. Siene 293, (21). 16. Buzsáki, G. Feed-forwrd inhiition in the hippompl formtion. Prog. Neuroiol. 22, (1984). 17. Alger, B.E. & Nioll, R.A. Feed-forwrd dendriti inhiition in rt hippompl pyrmidl ells studied in vitro. J. Physiol. (Lond.) 328, (1982). 18. Nioll, R.A., Alger, B.E. & Jhr, C.E. Enkephlin loks inhiitory pthwys in the verterte CNS. Nture 287, (198). 19. Somogyi, P. & Kluserger, T. Defined types of ortil interneurone struture spe nd spike timing in the hippompus. J. Physiol. (Lond.) 562, 9 26 (25). 2. Freund, T.F. & Buzsáki, G. Interneurons of the hippompus. Hippompus 6, (1996). 21. Glikfeld, L.L. & Snzini, M. Distint timing in the tivity of nninoid-sensitive nd nninoid-insensitive sket ells. Nt. Neurosi. 9, (26). 22. Mferri, G. & Dingledine, R. Control of feedforwrd dendriti inhiition y NMDA reeptor dependent spike timing in hippompl interneurons. J. Neurosi. 22, (22). 23. Geiger, J.R., Luke, J., Roth, A., Frotsher, M. & Jons, P. Sumilliseond AMPA reeptor medited signling t prinipl neuron-interneuron synpse. Neuron 18, (1997). 24. Cruikshnk, S.J., Lewis, T.J. & Connors, B.W. Synpti sis for intense thlmoortil tivtion of feedforwrd inhiitory ells in neoortex. Nt. Neurosi. 1, (27). 25. Syer, R.J., Friedlnder, M.J. & Redmn, S.J. The time ourse nd mplitude of EPSPs evoked t synpses etween pirs of CA3/CA1 neurons in the hippompl slie. J. Neurosi. 1, (199). 26. Gernet, L., Jdhv, S.P., Feldmn, D.E., Crndini, M. & Snzini, M. Somtosensory integrtion ontrolled y dynmi thlmoortil feed-forwrd inhiition. Neuron 48, (25). 27. Helmstedter, M., Stiger, J.F., Skmnn, B. & Feldmeyer, D. Effiient reruitment of lyer 2/3 interneurons y lyer 4 input in single olumns of rt somtosensory ortex. J. Neurosi. 28, (28). 28. Dw, M.I., Ashy, M.C. & Is, J.T. Coordinted developmentl reruitment of ltent fst spiking interneurons in lyer IV rrel ortex. Nt. Neurosi. 1, (27). 29. Mittmnn, W., Koh, U. & Husser, M. Feed-forwrd inhiition shpes the spike output of ereellr Purkinje ells. J. Physiol. (Lond.) 563, (25). 3. Sito, T. & Nktsuji, N. Effiient gene trnsfer into the emryoni mouse rin using in vivo eletroportion. Dev. Biol. 24, (21). 31. Petrenu, L., Huer, D., Sozyk, A. & Svood, K. Chnnelrhodopsin-2 ssisted iruit mpping of long-rnge llosl projetions. Nt. Neurosi. 1, (27). 32. Bruno, R.M. & Skmnn, B. Cortex is driven y wek, ut synhronously tive, thlmoortil synpses. Siene 312, (26). 33. Porter, J.T., Johnson, C.K. & Agmon, A. Diverse types of interneurons generte thlmus-evoked feedforwrd inhiition in the mouse rrel ortex. J. Neurosi. 21, (21). 34. Poo, C. & Isson, J.S. Odor representtions in olftory ortex: sprse oding, glol inhiition nd osilltions. Neuron 62, (29). 35. Holmgren, C., Hrkny, T., Svennenfors, B. & Zilerter, Y. Pyrmidl ell ommunition within lol networks in lyer 2/3 of rt neoortex. J. Physiol. (Lond.) 551, (23). 36. Beierlein, M., Gison, J.R. & Connors, B.W. Two dynmilly distint inhiitory networks in lyer 4 of the neoortex. J. Neurophysiol. 9, (23). 37. Thomson, A.M., West, D.C., Wng, Y. & Bnnister, A.P. Synpti onnetions nd smll iruits involving exittory nd inhiitory neurons in lyers 2 5 of dult rt nd t neoortex: triple intrellulr reordings nd ioytin leling in vitro. Cere. Cortex 12, (22). 38. Buhl, E.H., Hlsy, K. & Somogyi, P. Diverse soures of hippompl unitry inhiitory postsynpti potentils nd the numer of synpti relese sites. Nture 368, (1994). 39. Blitz, D.M. & Regehr, W.G. Timing nd speifiity of feed-forwrd inhiition within the LGN. Neuron 45, (25). 4. Agmon, A. & Connors, B.W. Thlmoortil responses of mouse somtosensory (rrel) ortex in vitro. Neurosiene 41, (1991). nture NEUROSCIENCE VOLUME 12 NUMBER 12 deember