Conditional dendritic spike propagation following distal synaptic activation of hippocampal CA1 pyramidal neurons

Transcription

1 COMPUTATION AND SYSTEMS ARTICLES Nture Pulishing Group Conditionl dendritic spike propgtion following distl synptic ctivtion of hippocmpl CA pyrmidl neurons Tim Jrsky,4, Alex Roxin 4, Willim L Kth, & Nelson Spruston The perfornt-pth projection to the hippocmpus forms synpses in the picl tuft of CA pyrmidl neurons. We used computer modeling to exmine the function of these distl synptic inputs, which led to three predictions tht we confirmed in experiments using rt hippocmpl slices. First, ctivtion of CA neurons y the perfornt pth is limited, result of the long distnce etween these inputs nd the som. Second, ctivtion of CA neurons y the perfornt pth depends on the genertion of dendritic spikes. Third, the forwrd propgtion of these spikes is unrelile, ut cn e fcilitted y modest ctivtion of Schffer-collterl synpses in the upper picl dendrites. This gting of dendritic spike propgtion my e n importnt ctivtion mode of CA pyrmidl neurons, nd its modultion y neurotrnsmitters or long-term, ctivity-dependent plsticity my e n importnt feture of dendritic integrtion during mnemonic processing in the hippocmpus. Excittory synpses on distl dendrites re common in the nervous system. For exmple, corticl pyrmidl neurons receive corticocorticl inputs in lyer, often hundreds of microns from the som. Mitrl cells of the olfctory ul receive input from olfctory receptor neurons on tuft of picl dendrites, similrly fr from the som. Purkinje cells of the cereellum receive inputs from mossy cells, mny of which terminte on distl dendrites. Synpses like these re intriguing, ecuse their long distnce from the spike initition zone, thought to reside in the xon, suggests tht specil mechnisms my e required for these synpses to trigger ction potentils. One possile mechnism to preserve the efficcy of distl synpses is through the genertion of locl dendritic spikes. Indeed, ll dendrites studied to dte express numerous voltge-gted chnnels,ndsustntil evidence now supports the notion tht spikes cn e initited in dendrites 3. Unlike xonl ction potentils, however, dendritic spikes do not propgte relily over long distnces in dendrites. A remining chllenge, therefore, is to determine wht types of synptic inputs cn trigger dendritic spikes nd how the spikes propgte long dendrites of vrious morphologies expressing unique comintions of chnnels. For such questions, computtionl pproch cn plce the ville dt in n integrted, quntittive frmework nd provide testle predictions concerning the function of vrious types of synpses nd dendrites. In the hippocmpus, CA pyrmidl neurons receive two distinct excittory synptic inputs 4 : the perfornt pth (or temporo-mmonic pth) provides direct input from lyer 3 of entorhinl cortex to the picl dendritic tuft, nd the Schffer collterls provide input from CA3 pyrmidl neurons to sl nd picl dendrites in CA. Although ll of the Schffer-collterl synpses re closer to the som thn re the perfornt-pth synpses, some of the Schffer-collterl synpses re nevertheless hundreds of microns from the som. The perfornt-pth projection to CA termintes in excittory, glutmtergic synpses,6, ut it is controversil whether these synpses cn depolrize the somtic memrne to ction-potentil threshold 7 9. Distl perforntpth nd Schffer-collterl synpses elicit dendritic spikes, oth in vitro nd in vivo,, ut some of these dendritic spikes do not propgte relily to the som, rising questions out the function of dendritic spikes nd the conditions necessry for them to rech the som nd trigger n ction potentil in the xon 3. Here we explored these questions y exmining the impct of distl perfornt-pth nd Schffer-collterl synptic inputs to CA pyrmidl neurons, using comined computtionl nd experimentl pproch. We found tht strong perfornt-pth ctivtion resulted in dendritic spikes tht could fil to propgte to the som. Modest ctivtion of Schffer-collterl synpses in the upper picl dendrites, however, fcilitted the forwrd propgtion of tuft dendritic spikes, thus llowing ction-potentil output in the xon following distl synptic ctivtion. RESULTS Modeling distl synptic ctivtion in excitle CA dendrites To ssess the ility of perfornt-pth synpses to ctivte CA pyrmidl neurons, we used comprtmentl models of two reconstructed neurons. Ech neuron included four ctive conductnces: voltgegted N + conductnce (G N ), delyed rectifier K + conductnce (G Kdr ) Institute for Neuroscience, Deprtment of Neuroiology nd Physiology, Northwestern University, Tech Drive, Evnston, Illinois 68, USA. Deprtment of Engineering Science nd Applied Mthemtics, Northwestern University, 4 Sheridn Rod, Evnston, Illinois 68, USA. 3 Unité Mixte de Recherche 89, Centre Ntionle de l Recherche Scientifique, Neurophysics nd Physiology, Université René Descrtes, 4 Rue des Sints Pères, 77 Pris Cedex 6, Frnce. 4 These uthors contriuted eqully to this work. Correspondence should e ddressed to N.S. (spruston@northwestern.edu). Received 3 August; ccepted Octoer; pulished online Novemer ; doi:.38/nn99 NATURE NEUROSCIENCE VOLUME 8 [ NUMBER [ DECEMBER 667

2 Nture Pulishing Group Mx. conductnce (S cm ) mv Strong G N G Kdr G KAp G KAd Distnce from som (µm) µm nd two A-type K + conductnces (G KA ). Consistent with experimentl reports, G N nd G Kdr were modeled with moderte conductnce long the somto-dendritic xis, ut higher G N inthexonmdethisthe preferentil site of ction-potentil initition 4.G KA ws modeled with the reported sixfold increse in conductnce long the somto-dendritic xis nd lower hlf-inctivtion voltge in proximl (G KAp ) versus distl (GK Ad ) dendrites,6. We used two versions of the model: wek dendritic excitility model with uniform G N in the som nd dendrites, nd strong dendritic excitility model with slight grdient of incresing G N with distnce from the som (Fig. ). These two models re simple compred to the full repertoire of voltgegted chnnels known to e expressed in pyrmidl cells,7 ; nevertheless, these models reproduce two popultions of CA neurons with distinct profiles of ction-potentil ckpropgtion 4.Weusedthese models to mke experimentlly testle predictions out the ility of perfornt-pth inputs to ctivte CA neurons. The two models exhiited mrkedly different ehvior (Fig. ). In the strong dendritic excitility model, simultion of strong perfornt-pth stimulus (% of ville synpses in the distl picl tuft) triggered the initition of dendritic spike, which propgted to the som nd triggered n ction potentil (Fig., leftndsupple- mentry Video online). Similr ehvior ws oserved in the model of second reconstructed neuron (Supplementry Fig. online nd Supplementry Video online). In the wek dendritic excitility model, the sme strong perfornt-pth input triggered dendritic spike, ut the spike filed to propgte to the som, thus resulting in suthreshold somtic depolriztion (Fig., right nd Supplementry Video 3 online). With high-frequency urst of synptic inputs ( t Hz), dendritic spikes sometimes propgted to the som, ut successful propgtion required ctivting lrge numer of synpses (for exmple, 3% success t % of perfornt-pth synpses; dt not shown). These results suggest tht the only wy perfornt-pth inputs cn trigger n ction potentil in the xon is if dendritic spike is Mx. conductnce (S cm ).3.. Wek Distnce from som (µm) 4 mv ms Figure Strong nd wek dendritic excitility models of CA pyrmidl neurons respond differently to perfornt-pth ctivtion. () Chnnel distriutions s function of distnce from the som in the two models. G N, voltgegted N + conductnce; G Kdr, delyed rectifier K + conductnce; G KAp, proximl A-type K + conductnce; G KAd, distl A-type K + conductnce. () Perfornt-pth () ctivtion (% of ville synpses) triggered dendritic spikes in the picl tuft. These spikes propgte to the som nd trigger n ction potentil in neurons with strong (left), ut not wek (right), dendritic excitility. Scles pply to oth pnels. Color mps re of mximl (pek) voltge in ech comprtment of the model. Trces re voltge versus time plots t the three dendritic loctions indicted y the electrodes. Animtions of these simultions in this cell nd nother reconstructed cell re provided in Supplementry Videos 3 nd 6. initited in the tuft nd propgtes down the picl dendrite. In the wek excitility model, forwrd propgtion ws unrelile nd did not ensure n ction-potentil output. This finding is consistent with our previous experimentl oservtion tht dendritic spikes cn occur in the sence of somtic ction potentils,8. Interction etween perfornt-pth nd Schffer-collterl inputs In the models, when dendritic spikes filed to propgte from the distl picl tuft to the som, they usully filed in the upper picl dendrite (Fig., right; Supplementry Figs. nd 3 online; Supplementry Videos 3 8 online). We therefore investigted how perfornt-pth synptic responses were influenced y the ctivtion of synpses in the upper picl dendrites, just elow the picl tuft. These synpses correspond to distl Schffer-collterl inputs. When perfornt-pth nd Schffer-collterl inputs re ctivted together, ction-potentil firing is proilistic, depending on the exct numer nd loction of the rndomly selected synpses ctivted in ech region. Actionpotentil proility ws determined from more thn 7, simultions nd plotted s function of the percentge of perfornt-pth nd Schffer-collterl synpses ctivted (Fig. ). In the strong dendritic excitility model, two modes of ctivtion were pprent. In the first mode of the strong excitility model (Fig., left, region ), ctivtion of the Schffer collterls (43%) ws on its own sufficient to trigger ction potentils on t lest some trils. With little or no ctivtion of perfornt pth, ction-potentil proility incresed from to over nrrow rnge of Schffercollterl ctivtion (out 3 %). As the percentge of perfornt-pth inputs ws incresed from % to %, the percentge of Schffercollterl inputs necessry to trigger n ction potentil % of the time (green in Fig., left) decresed from out 4% to 3%. In this mode, the Schffer-collterl input ws stronger thn the perforntpth input nd dendritic spikes lmost lwys egn in the upper picl dendrites, elow the tuft nd ner the Schffer-collterl inputs (green in Fig. c, left, region ). In the second mode of the strong excitility model (Fig., left, region ), perfornt pth ws the dominnt input, nd ction potentils could e elicited on t lest some trils with 4% of the synpses ctivted. With little or no ctivtion of the Schffer collterls, ctionpotentil proility incresed from to over nrrow rnge of 668 VOLUME 8 [ NUMBER [ DECEMBER NATURE NEUROSCIENCE

3 Nture Pulishing Group SC.. c d Percentge Schffer collterl Percentge Schffer collterl Tuft spike proility Percentge perfornt pth Percentge perfornt pth Percentge perfornt pth perfornt-pth ctivtion (out 6%). In this mode, ction potentils were driven y dendritic spikes originting in the picl tuft (red in Fig. c, left, region ). With perfornt-pth stimultion lone, the proility of dendritic spike initition in the tuft incresed from ner to over the sme rnge (out 6%, Fig. d, left).over this rnge of perfornt-pth ctivtion, modest ctivtion of the Schffer collterls ( 3%) reduced the percentge of perfornt-pth inputs necessry to trigger propgting dendritic spike, ut the spikes lmost lwys egn in the picl tuft (innervted y perfornt pth) nd propgted to the som (Fig. c, left). Stronger ctivtion of either or oth pthwys led to relile dendritic spike initition, with the spikes tending to egin ner the strongest synptic input. In summry, in the strong excitility model, ech input ws le to trigger dendritic spikes on its own, nd ech input could cooperte with the other to trigger propgting dendritic spike. The ehvior of the wek dendritic excitility model ws drmticlly different (Fig., right). Importntly, single coincident ctivtion of perfornt-pth synpses ws lmost never le to produce n ction potentil on its own. Only when perfornt-pth ctivtion ws incresed to very high levels (4% of ll synpses) did n ction potentil occur in the som (dt not shown). Even then, ction Percentge perfornt pth Percentge perfornt pth Percentge perfornt pth Figure Modes of ctivtion of CA pyrmidl neurons y perfornt-pth nd Schffer-collterl ctivtion. () Schemtic of model cell with perfornt-pth () nd Schffer-collterl (SC) ctivtion sites indicted. Electrodes t indicted loctions were used to determine whether spikes egn in the som (vlue, lue), mid picl dendrite (vlue., green) or picl tuft (vlue., red). () Proility of ction-potentil initition (in the xon) in response to ctivtion of different percentges of ville (picl tuft) nd SC synpses (upper picl dendrites). (c) Loctions of spike initition s descried in. Intermedite colors re determined y the frctions of trils yielding initition t one or nother site. In oth nd c, regions nd correspond to modes of ctivtion where SC nd inputs dominte, respectively (see text for detils). (d) Proility of dendritic spike in the picl tuft s function of the percentge of synpses ctivted (no SC ctivtion). Columns in, c nd d correspond to the strong nd wek dendritic excitility models, respectively. Proilities were determined from trils of rndomly plced synpses for ech %, %SC pir. In nd c, horizontl guide lines correspond to the miniml percentge of SC inputs necessry to trigger n ction potentil; in, c nd d, verticl guide lines correspond to the percentge of inputs necessry to trigger spike in the picl tuft on % of trils Action potentil proility Averge site of spike initition potentils only occurred on some trils (out % t 6 7% perfornt-pth ctivtion), nd even greter ctivtion of perfornt-pth inputs decresed the proility of ction-potentil firing (owing to the ctivtion of K + chnnels nd the inctivtion of N + chnnels y lrge excittory post synptic potentils (EPSPs)). Becuse such lrge ctivtion of perfornt pth is required, interction of perfornt-pth nd Schffer-collterl inputs is more plusile explntion for how perfornt pth influences ction-potentil firing in CA pyrmidl neurons with wekly excitle dendrites. In the wek dendritic excitility model, perfornt-pth nd Schffer-collterl inputs intercted in two importnt wys. First, low levels of perfornt-pth ctivtion slightly reduced the percentge of Schffer-collterl inputs necessry to trigger n ction potentil (Fig., right, region ). Activtion of etween % nd 6% of perfornt-pth synpses reduced the percentge of Schffercollterl synpses necessry to trigger n ction potentil % of the time (green in Fig., right) from out 6.% to.% (slope ¼.). In this mode, most of the spikes egn in the picl dendrite (elow the tuft) nd propgted to the som (green in Fig. c, right, region ), ut in some cses, the Schffer-collterl nd perfornt-pth inputs summed in the som to trigger n ction potentil first in the xon (lue in Fig. c, right, region ). Within this ctivtion mode, there ws little difference etween perforntpth inputs tht triggered dendritic spikes nd those tht did not. Even when perfornt-pth ctivtion triggered dendritic spike (for exmple, % of the time t % ctivtion: Fig. d, right), the spikes did not spred into the upper picl dendrites, elow the tuft. Insted, the min effect in this mode of ctivtion ws tht perfornt-pth inputs reduced the mount of Schffer-collterl input necessry to trigger dendritic spike in the picl dendrites elow the tuft. By contrst, stronger ctivtion of perfornt pth (ove out 6%) ctivted dendritic spikes in the tuft more frequently. These spikes tended to e lrger nd thus spred throughout the picl tuft. In this mode (Fig., right, region ), ctivtion of etween 6% nd % of perfornt-pth synpses reduced the percentge of Schffer-collterl synpses necessry to trigger n ction potentil % of the time from out.% to % (slope ¼.8). Furthermore, in this mode, ction potentils lmost lwys egn s dendritic spikes in the picl tuft (red in Fig. c, right, region ). This ctivtion mode thus reveled n especilly interesting interction etween perfornt pth nd Schffer collterls. Dendritic spikes were triggered y strong perfornt-pth input, ut their propgtion to the som ws gted y Schffercollterl inputs. In the sence of Schffer-collterl inputs, the picl tuft dendritic spikes filed to rech the som nd did not produce n ction potentil. However, even modest levels of Schffer-collterl input were le to fcilitte the forwrd propgtion of these dendritic spikes, such tht they could trigger full ction potentil in the xon NATURE NEUROSCIENCE VOLUME 8 [ NUMBER [ DECEMBER 669

4 Nture Pulishing Group c mv SC SC nd som (Fig. 3) This gting of perfornt pth evoked dendritic spikes y Schffer-collterl synptic inputs ws oserved in oth of the CA cell models (Supplementry Figs. nd 3; Supplementry Videos 3 8 online). One notle feture of the ehvior of oth the strong nd the wek dendritic excitility models is tht the numer of perfornt-pth inputs necessry to trigger dendritic spike in the tuft ws ctully lower thn the numer of Schffer-collterl inputs necessry to trigger dendritic spike in the upper picl dendrites. In the strong excitility model, ctivtion of minimum of B% of perfornt-pth inputs resulted in dendritic spikes on t lest some trils, compred to t lest 3% for the Schffer-collterl input (Fig. c,d, left). Aout 4% of ech input ws necessry to produce n ction potentil on hlf of the trils (Fig., left), ut ecuse there were fewer synpses in the perforntpth pool (Methods), fewer perfornt-pth synpses were needed to produce hlf ctivtion. In the wek excitility model, 43% of perfornt-pth synpses produced tuft spikes (Fig. d, right), compred to 4% of Schffer-collterl synpses required to produce 4 mv ms Figure 3 Gting of perfornt-pth evoked dendritic spikes y Schffercollterl evoked EPSPs. () Color mp of pek depolriztion nd voltge versus time plots t three dendritic loctions for ctivtion of % of synpses (one tril) in the wek dendritic excitility model. () Response of the sme model to ctivtion of 3% of SC synpses in the upper picl dendrites. (c) Response of the sme model to coincident ctivtion of % nd 3% SC synpses. Animtions of these simultions in this nd the other reconstructed cell re provided in Supplementry Videos 3 8. Scles in pply to nd c s well. upper picl dendritic spikes (Fig., right). The lower numer of perfornt-pth synpses required to produce dendritic spikes, in oth the strong nd wek excitility models, is ttriutle to the high input impednce of the smll-dimeter tuft rnches. A possile limittion of our models, which were originlly designed to reproduce ction potentil ckpropgtion 4,isthttheyused miniml repertoire of voltge-gted conductnces. Although these models produced dendritic spikes without chnges from their originl forms, they re clerly very simple models. Nevertheless, the models mke some interesting predictions concerning the response to distl synptic ctivtion. To determine if these predictions were roust, we used more complex model, which reproduces mny spects of CA pyrmidl neuron excitility 7. This model contins vriety of conductnces missing from our models, including H current, C + currents nd C + -ctivted K + currents. Neither the presence of these dditionl conductnces nor the longer dendrites in this model ltered the sic ehvior we hve descried. Strong ctivtion of distl synptic inputs triggered dendritic spikes tht filed to propgte to the som unless synpses on the picl dendrites elow the tuft were lso ctivted (Supplementry Fig. 4 online). A potentilly importnt spect of the interction etween perforntpth nd Schffer-collterl synptic inputs is their timing. We used our models to explore the timing reltionship necessry to otin gting of forwrd-propgting dendritic spikes. Strong perfornt-pth ctivtion (8% sufficient to induce dendritic spikes on most trils) ws simulted in the wek dendritic excitility model, nd oth the strength nd timing of the simulted distl Schffer-collterl input were vried. Action-potentil proility ws gretest when the two inputs were coincident, nd the efficiency of gting decresed s the dely etween the two inputs incresed (Fig. 4). The exct nture of the interction etween the Schffer-collterl nd perfornt-pth inputs depended on synptic strength nd timing in complex wy. For 3.% Schffer-collterl ctivtion, timing could vry y few milliseconds, ut Schffer-collterl input preceding perfornt-pth input ws slightly more effective thn the reverse. For % Schffer-collterl ctivtion, the interction ws stronger nd the rnge of effective timing ws roder. In contrst to the weker Schffer-collterl stimulus, however, the % Schffer-collterl input ws most effective t gting the dendritic spikes if it cme slightly fter the perfornt-pth input. Mechnism of dendritic spike gting We explored the mechnism y which the Schffer-collterl input fcilittes the forwrd propgtion of distlly evoked dendritic spikes, y exmining the voltge in the dendritic tree for Schffer-collterl stimuli just elow nd just ove threshold for successful propgtion (Fig. ). Perfornt-pth inputs were simulted on ech picl rnch nd set to conductnces lrge enough to evoke dendritic spikes in the picl tuft. A wek Schffer-collterl input (3 ns) ws too smll to fcilitte propgtion of the dendritic spike into the min picl dendrites (Fig. c, left). In this cse, the voltge ner the min ifurction in the picl dendrites peked out 4 ms fter the onset of 67 VOLUME 8 [ NUMBER [ DECEMBER NATURE NEUROSCIENCE

5 Nture Pulishing Group Action potentil proility % SC 3.% SC.% SC first t (ms) SC first Figure 4 Dependence of -evoked dendritic spike propgtion on the strength nd timing of SC ctivtion. Action-potentil proility (som or xon) is plotted s function of the time difference etween ctivtion (8%) nd SC ctivtion (., 3. or %). Proilities were determined from trils of rndomly plced synpses for ech vlue of %SC nd ltency ( ms intervls). the synptic inputs (6.3 ms from the eginning of the simultion). A slightly lrger Schffer-collterl input (4 ns) ws sufficient to successfully fcilitte propgtion through the min ifurction nd into the primry picl dendrite (Fig. c, right). In this cse, the spike in the rnch point occurred out ms lter thn the pek of the filed spike in the previous cse. We found tht the spike did not propgte eyond the lrge rnch point tht seprtes the primry picl dendrite from the picl tuft (Fig., left; note shrp drop in mximum voltge round 4 mm in Fig., left; see lso Figs. nd 3). The low sfety fctor for propgtion of spikes through rnch points is known consequence of the decresed impednce t rnch points 9,. In our models of CA neurons, reductions in spike mplitude were oserved t multiple rnch points, ut filures of propgtion were most evident t the lrge rnch point tht divided the min picl dendrite from the picl tuft. Bckpropgting ction potentils fil t the sme loction in the CA dendritic tree 4. The mechnism y which Schffer-collterl inputs rescued the propgtion of dendritic spikes through this rnch point cn e inferred y exmining the currents flowing in the mjor picl rnch, just efore the spike (4.3 ms fter the onset of the synptic stimulus). Compred to the smller Schffer-collterl input, which resulted in filure of the dendritic spike, the lrger Schffer-collterl synptic input pproximtely tripled the N current t this criticl time nd loction (Fig. d). The extr N current ws the key to the successful forwrd propgtion of the dendritic spike. A similr mechnism underlies enhncement of ction-potentil ckpropgtion y EPSPs in corticl pyrmidl neurons. Experimentl tests of model predictions We performed experiments to test three importnt predictions of the model. First, we tested the prediction tht, on their own, perforntpth inputs hve limited ility to trigger ction potentils in CA neurons. Second, we tested the prediction tht when perfornt-pth inputs do trigger ction potentils, they do so y eliciting dendritic spikes. Third, we tested the prediction tht the propgtion of perfornt-pth evoked dendritic spikes from the picl tuft to the som is fcilitted y modest ctivtion of Schffer-collterl synptic inputs. To test whether perfornt-pth stimultion could trigger ction potentils, we plced lrge ipolr stimulting electrode in strtum lcunosum-moleculre of hippocmpl slices. The perfornt-pth projection to CA is visile in this region under infrred, differentilinterference contrst optics. In response to stimultion of the perfornt pth in the presence of GABA-receptor ntgonists (Methods), we recorded monosynptic EPSPs using whole-cell ptch-clmp recordings from CA somt. As the stimulus intensity ws incresed, EPSPs incresed in mplitude, ut were never lrge enough to evoke n ction potentil (Fig. 6; verge mximum EPSP ¼ 6.7 ±. mv, n ¼ ). These results re consistent with the model s prediction tht it is difficult to produce n ction potentil using single stimulus of perfornt pth lone. In fct, it ws even more difficult to get ction potentils thn predicted y the model, ecuse single stimuli of the perfornt pth were never effective, even though some of the recorded neurons should hve hd reltively strong dendritic excitility 4. To further explore the ility of perfornt pth to evoke ction potentils, we stimulted perfornt pth using high-frequency ursts ( Hz) of five or ten pulses. In most cells, it ws possile to evoke ction potentils using high-intensity stimultion with five pulses; moreover, in ll cells, ten pulses were effective (Fig. 6). When spikes were driven y perfornt-pth synptic stimultion in this wy, ctionpotentil threshold ws lower thn the threshold for evoking ction potentils with current injection t the som (Fig. 7). Specificlly, we used somtic current injections resemling excittory post synptic currents (EPSCs) in the sme ten-pulse pttern used for synptic stimultion. Synpticlly evoked ction potentils hd significntly lower thresholds thn those evoked y synptic current injection (Fig. 7). In mny cses, even suthreshold responses to current injection were lrger thn the threshold for synpticlly evoked ction potentils (Fig. 7). These oservtions re consistent with the notion tht perfornt-pth evoked ction potentils re driven y spikes tht egin in the dendrites. In some cses, suthreshold synptic responses exhiited spikelets (Fig. 7c). We hve shown previously tht these smll spikes re the somticlly recorded counterprts of lrge dendritic spikes 8.Spikelets were identified more frequently during locl ppliction of tetrodotoxin (TTX) ner the som (Fig. 7d; Methods), suggesting tht spikelets re often msked y ction-potentil firing. In two cells, we otined enough trils to compre the occurrence of ction potentils (control) nd spikelets (locl TTX) t the sme perforntpth stimulus intensity; in these cells, the rtio of trils exhiiting spikelets to those exhiiting spikes ws close to unity (.4 nd.79, trils ech). These results indicte tht spikelets reflect dendritic spikes tht hve spred effectively to the som nd would usully trigger n xonl ction potentil. To test whether the forwrd propgtion of perfornt-pth evoked dendritic spikes cn e gted y Schffer-collterl stimultion of the upper picl dendrites, we used second stimulting electrode to ctivte Schffer-collterl inputs (single pulse) t the end of urst of perfornt-pth inputs where the spikelets were usully oserved (Fig. 8). Stimultion strength ws set to produce single Schffercollterl EPSPs of 3 mv in the som nd single perfornt-pth EPSPs of -4 mv, which summted in ursts to mv (t or just elow threshold for spikelets). Schffer-collterl nd perfornt-pth responses were lso simulted using EPSC-like current injections. We pplied TTX ner the som so s to lock ction-potentil firing nd revel spikelets. At the stimulus intensities used in these experiments, spikelets were never oserved in response to Schffer-collterl stimultion nd were only occsionlly evoked y perfornt-pth stimultion lone. When the two inputs were coctivted, however, the frequency of spikelets incresed drmticlly (Fig. 8c). To test whether this enhncement of spikelets ws cused y summtion of the somtic EPSPs, s opposed to dendritic integrtion of the events, we sustituted somtic, EPSC-like current injection for the Schffer-collterl NATURE NEUROSCIENCE VOLUME 8 [ NUMBER [ DECEMBER 67

6 mv ARTICLES Nture Pulishing Group c d Voltge (mv) Voltge (mv) Current (na) mv I syn Time (ms) Chrge (pc) µm V (6.3 ms) 4 V (6.3 ms) V mx V Distnce (µm) mx Distnce (µm) Tuft Apicl Q syn Current (ma cm ) Voltge (mv) Voltge (mv) N mv Current (µa cm ) Time (ms) K dr Current (ma cm ) Tuft Apicl..... K A Figure Mechnisms of conditionl dendritic spike propgtion. (,) Mximum voltge in upper hlf of the picl dendritic tree. A powerful synpse (3 ns) ws plced midwy long ech reconstructed section of the dendritic tuft (, 9 synpses, red dots). A single synpse ws plced ner the end of the primry picl dendrite (SC, red dot with lck rrow). On the left, the SC input hs the sme conductnce s the tuft synpses nd the dendritic ction potentils initited y the input in the tuft fil to invde the picl dendrites elow the mjor ifurction. On the right, the SC input hs slightly lrger conductnce (4 ns) nd spike propgtion is successful. Prmeters s given in Figure, right. () Plots showing the voltge (nd mximum voltge, dshed lines) t 6.3 ms (time just efore the successful dendritic spike in the min picl dendrite in (right); green nd red rrows in c long the primry picl dendrite nd the tuft rnch contining the synpse mrked y lue rrow. Blck nd lue rrows correspond to loctions indicted in. (c) Voltge versus time t the tuft (lue) nd picl (lck) synpses. Green nd red rrows indicte the picl voltge trces t 6.3 ms. (d) Comprison etween dendritic currents t the picl synpse t 6.3 ms (time of mximum depolriztion t the picl synpse for the cse of propgtion filure nd just efore the spike in the upper picl dendrite when propgtion is successful). Green nd red rs correspond to the points indicted y rrows in c. I syn, instntneous synptic current; Q syn, totl synptic chrge trnsferred up to tht time; N, sodium current; K dr, delyed rectifier potssium current; K A, A-type potssium current. increse the frequency of spikelets ove tht oserved y perfornt-pth stimultion lone (Fig. 8c), suggesting tht spikelet enhncement y coincident stimultion of perfornt pth nd Schffer collterls indeed reflects gting of dendritic spikes eginning in the picl tuft nd propgting towrd the som. stimulus. The mplitude of the current injection ws set to yield simulted EPSPs lrger thn the Schffer-collterl evoked EPSPs (Schffercollterl EPSP ¼ 4. ±.3 mv, simulted EPSP ¼ 7.7 ±.8 mv, perfornt-pth EPSP ¼ 3. ±.6 mv, n ¼ ). This protocol did not DISCUSSION The long distnce etween the perfornt-pth input nd the xon of CA neurons hs rised questions concerning the function of these synpses. Erly reports suggested tht the perfornt pth ws minly c ms Pek depolriztion (mv) 4 6 Stimulus intensity (µa) Cells with stimulus evoked APs (%) 7 n = n = n = pulse pulse pulse Stimulus numer Figure 6 Experimentl perfornt-pth stimulus-response plot. () Representtive trces of EPSPs evoked y perfornt pth stimultion cross rnge of intensities. Left: EPSPs in response to single. ms stimulus. Middle: EPSPs in response to five. ms stimuli t Hz. Right: EPSPs in response to ten. ms stimuli t Hz. Action potentils re truncted. () Stimulus response plot of EPSP mplitude for representtive neuron (men ± s.e.m.; closedsqures, single stimulus; open-squres, five stimuli t Hz; tringles, stimuli t Hz). (c) Summry r grph indicting the percent of neurons in which perfornt-pth stimultion induced n ction potentil recorded t the som in response to high-intensity stimultion. 67 VOLUME 8 [ NUMBER [ DECEMBER NATURE NEUROSCIENCE

7 mv mv mv Threshold (mv) mv mv ARTICLES Nture Pulishing Group c * ms ms ms inhiitory,7,3, result of strong feedforwrd inhiition,8,9,4 8. There is considerle evidence, however, for n excittory influence of the perfornt pth on CA cells,4,7,9,3. Furthermore, plce-field firing occurs in CA cells deprived of Schffer-collterl inputs y CA3 or dentte gyrus lesions 3,3 ; this suggests tht perfornt pth is cple of driving firing of CA neurons in vivo. Our findings suggest tht the excittory influence of the perfornt pth depends on oth dendritic spikes nd n interction with Schffer-collterl inputs. Previous reports on the interction etween perfornt-pth nd Schffer-collterl inputs re seemingly contrdictory. One report shows tht ursts of perfornt-pth ctivtion cn increse the proility of Schffer-collterl evoked ction potentils in CA 7, wheres nother study showed tht perfornt-pth ctivtion did not increse the response to Schffer-collterl ctivtion. These two findings cn oth e understood in the context of dendritic excitility. Perfornt-pth stimuli tht re elow threshold for dendritic spikes my hve little effect in the som, ecuse of the enormous ttenution of perfornt-pth EPSPs long the dendrites. Furthermore, perforntpth inputs my hve negligile effect when Schffer-collterl inputs re strong enough to evoke spikes on their own. However, perfornt-pth inputs tht trigger dendritic spikes on their own or in comintion with milder Schffer-collterl input cn contriute to ction-potentil firing. I inj Syn Syn d Synptic * * ms ** Current injection Figure 7 Dendritic spikes underlie perfornt-pth evoked ction potentils. () Threshold for ction potentils evoked y perfornt-pth stimultion (Syn: pulses t Hz) occurs t more hyperpolrized potentils thn those evoked y somtic current injection (I inj ). Top nd ottom left: ction potentils evoked y ten-pulse stimultion. Dshed line indictes ction-potentil threshold (identified s the voltge t which dv/dt first exceeds mv ms ). Top right: depolriztion induced y current injection mimicking -evoked EPSP urst, which filed to elicit n ction potentil. Dshed line indictes pek depolriztion. Bottom right: current injection evoked ction potentil. Dshed line indictes ction-potentil threshold. Insets contin mgnified views of the ction potentils. () Threshold of synpticlly evoked ction potentils is significntly hyperpolrized reltive to currentinjection evoked ction potentils. **P o., two-tiled Student s pired t-test, n ¼ 8. Dt from individul cells (connected points) nd men ± s.e.m. re shown. (c) Suthreshold spikelet (sterisk, mgnified in inset) in response to miniml (sufficient to elicit single ction potentil) high-frequency stimultion (Syn: three -pulse, -Hz ursts t 3 Hz). (d) Locl ppliction of TTX to the som revels spikelets (sterisks, second one mgnified in inset) in response to high-frequency stimultion (Syn: three -pulse, -Hz ursts t 3 Hz). ms Syn Our findings indicte tht perfornt-pth excittory synptic inputs cn e effective in two wys. They cn reduce the numer of Schffer-collterl inputs required to trigger dendritic spike. In this cse, the Schffercollterl synpses re the dominnt input nd the perfornt pth cn e considered modultory. Alterntively, perfornt pth is the dominnt input, ut even smll Schffer-collterl input cn e crucil in fcilitting the propgtion of dendritic spikes from the picl tuft towrd the som nd xon. This ltter scheme indictes tht perfornt pth cn serve s the mjor input to CA neurons. In models with strongly excitle dendrites, the perfornt-pth input is ctully more efficcious thn distl Schffer-collterl inputs, owing to the genertion of relily propgting dendritic spikes in the smll, high-impednce rnches of the picl tuft. In models with wekly excitle dendrites, dendritic spikes re lso redily triggered y perfornt-pth input, ut their propgtion towrd the som is unrelile. Activtion of Schffer-collterl synptic inputs, however, promotes the forwrd propgtion of perfornt-pth evoked distl dendritic spikes. This synptic gting of dendritic spike propgtion is consistent with previous reports tht dendritic spike propgtion is fcilitted y the depolriztion of the picl dendrites 33 nd resemles similr effect predicted to occur stochsticlly during the ckground synptic ctivtion of CA dendrites 34. In our models, the gte seems to exist ner the order etween perfornt-pth nd Schffer-collterl synptic inputs. It exists, in lrge mesure, ecuse of the mjor ifurction of the picl dendrites t this loction in mny CA pyrmidl neurons. Both the vilility of voltge-gted chnnels nd the types of synptic inputs will influence whether the gte is open or closed. When the gte is closed y defult s result of low rtio of N + to K + chnnels (wek dendritic excitility 4 ) it cn e opened y modest Schffercollterl synptic input. Our experiments lso suggest tht ursts of perfornt-pth input could led to successful forwrd propgtion of dendritic spikes n effect tht cn e reproduced in the models (dt not shown). Thus, perfornt-pth inputs my open the gte on their own, without the help of Schffer-collterl inputs. To determine whether this is likely to hppen in vivo, it will e importnt for future work to determine the firing ptterns of lyer 3 pyrmidl neurons of entorhinl cortex in wke, ehving nimls. In our models, we lso show tht the gte cn exist in n open stte y defult (strong dendritic excitility 4 ). Whether neurons operte in wek or strong dendritic excitility mode in vivo is likely to e influenced y numer of fctors, including ionic conditions 33, neuromodultory stte nd ctivity-dependent plsticity. Chnges in the vilility of A-type K + chnnels, which re undnt in CA dendrites, is one mechnism y which trnsitions etween the two defult sttes (gted closed or open) could occur,3,36. Feedforwrd inhiition is likely to further enrich synptic integrtion in distl dendrites. In the perfornt pth, inhiition is likely to NATURE NEUROSCIENCE VOLUME 8 [ NUMBER [ DECEMBER 673

8 Spikelet frequency (% trils) mv na ARTICLES Nture Pulishing Group TTX c SC + SC + simsc SC 4 3 simsc sim * * n = + simsc ms sim sim sim simsc temporlly limit the excittory influence of the perfornt pth; in contrst, in the Schffer collterls, inhiition my regulte the gting of distlly evoked dendritic spikes. Our simultions indicte tht ppropritely timed Schffer-collterl inhiition cn prevent the spred of perfornt-pth evoked dendritic spikes (tht is, close the gte), even in strongly excitle dendrites. In wekly excitle dendrites, feedforwrd inhiition nrrowed the time window for fcilittion of dendritic spike propgtion (tht is, opening the gte) y the Schffercollterl EPSP (see Supplementry Fig. online). Modultion of inhiition is likely to e n importnt mens of regulting synptic integrtion vi these mechnisms. When TTX ws loclly pplied ner the som, the spikelets we oserved in response to perfornt-pth stimultion or comined perfornt-pth nd Schffer-collterl stimultion resemled the D-spikes or fst prepotentils oserved in erly intrcellulr recordings from hippocmpl pyrmidl neurons 3,37,38. Tht these events re somtic indictions of lrger dendritic spikes is consistent with the erly reports nd our previous somtic nd dendritic recordings 8. More thn forty yers go, it ws suggested tht the propgtion of dendritic spikes is required for the ctivtion of CA neurons y distl synptic inputs 37. A few yers lter, reserchers hypothesized tht the propgtion of dendritic spikes might hve low sfety fctor, requiring interctions t points of confluence in the dendrites to trigger somtic ction potentil 3. Still lter, it ws suggested tht spikes originting in the picl tuft of neocorticl pyrmidl neurons could e modulted y synptic inputs long the picl trunk 39. Our findings vlidte these predictions y demonstrting such interctions in the distl dendrites of CA pyrmidl neurons. Whether such interctions occur in other neurons remins to e determined. In lyer pyrmidl neurons, distl input cn increse the efficcy (gin) of more proximl input 4. This resemles the ** + SC Figure 8 Coincident Schffer-collterl stimultion increses spikelet frequency in response to perfornt-pth stimultion. () Experimentl schemtic (left): TTX ws pplied loclly ner the cell ody throughout the experiment to lock somtic ction potentils. Stimulus protocols (right): perfornt pth () three -pulse, -Hz ursts t 3 Hz; Schffer collterl (SC) single stimuli timed to occur in coincidence with stimultion; simulted Schffer collterl (simsc) somtic current injection designed to elicit depolriztions with identicl timing, similr kinetics, nd equl or greter mplitude thn the SC stimultion (difference of exponentils: t rise ¼. ms, t decy ¼ ms); simulted perfornt-pth (sim) somtic current injection with kinetics nd mplitude similr to single urst of stimultion, followed y na, ms current injection to test TTX efficcy. Action potentils were never evoked with this stimulus during TTX ppliction, even though the stimulus ws well ove threshold under control conditions. () Responses to vrious comintions of the stimuli represented in. + SC elicited spikelets (sterisks nd inset) in representtive neuron. + simsc ws unle to mimic the effect tht SC stimultion hd on spikelet production (inset). (c) Summry grph (men ± s.e.m.) of spikelet frequency in response to, + simsc nd + SC stimultion. **P o. (repeted-mesures nlysis of vrince followed y Dunn s test with Bonferroni correction, P o.). modultory effect we descrie, ut the mechnism my e different. In the neocortex, this effect occurs t lest in prt through the enhncement of ction potentil ursting, vi n interction etween ckpropgting ction potentils nd distl synptic inputs 4,4,whichhs not yet een descried in CA pyrmidl neurons. The sitution where the distl input cn serve s the dominnt input nd e modulted y modest proximl input hs not een studied in corticl pyrmidl neurons. Thus, it will e importnt for future studies to determine whether this gting of distl dendritic spikes is generl feture of pyrmidl neurons in vrious corticl regions. METHODS Computtionl modeling. All simultions were performed using the NEU- RON simultion environment 4 using 64-processor Beowulf cluster. Dt presented in Figures through 4 re from one reconstructed CA pyrmidl cell. An dditionl reconstructed cell ws used to repet the numericl experiments, yielding qulittively similr results. The model cells, long with ll code for our simultions, re freely ville on the we ( edu/dendrite). A third reconstructed cell, using seprte model of ctive dendritic chnnels ws otined from B. Mel (Univ. Southern Cliforni, 7 nd used to verify the results of the other cells. Pssive nd ctive properties of the models. Our models were prepred from CA neurons reconstructed fter stining following intrcellulr iocytin filling in hippocmpl slices from 7-d-old Wistr rts. The resulting comprtmentl models included pssive memrne properties (R m ¼ 4, Ocm, C m ¼.7 mf cm, R i ¼ Ocm) nd three voltge-gted conductnces: N + conductnce, delyed rectifier K + conductnce nd two vrints of n A-type K + conductnce, implemented s descried 4,6. Additionl detils regrding the models used here re provided in the Results section nd Figure. Synpse distriution. Synpses were distriuted throughout the dendritic ror of the reconstructed cells, with totl of out 7, synpses in ech of the first two reconstructed cells. The numer nd distriution of these synpses ws modeled in ccordnce with experimentl dt 43.Inourstudy, only synpses in the distl picl dendrites were ctivted. Densities of excittory synpses were higher in distl strtum rditum (tht is, distl Schffer collterls) thn in strtum lcunosum-moleculre (tht is, perfornt pth). The third cell ws lrger thn the others (stright-line distnce from som to distl tip pproximtely, mm s opposed to 7 mm for the other two cells) nd therefore included more synpses (44,39 totl). As result, lower percentges of synpses were needed to elicit dendritic spikes nd somtic ctions potentils in the third model. 674 VOLUME 8 [ NUMBER [ DECEMBER NATURE NEUROSCIENCE

9 Nture Pulishing Group Synpse properties. AMPA receptor medited synptic conductnces were modeled s the difference of two exponentils (t rise of. ms nd t decy of. ms) with reversl potentil of mv. The conductnce of single AMPA synpse ws set to.8 ns, vlue chosen ecuse it produced somtic EPSP of. mv when ctivted t synpse mm from the som 44. Schffer-collterl nd perfornt-pth inputs. Schffer-collterl input ws simulted s the ctivtion of percentge of totl synpses present in the distl picl dendrites, pproximtely mm from the som. On ech simultion tril, percentge of synpses were chosen rndomly from the ville pool. For the first two model cells, the totl numer of synpses (% ctivtion) ville for distl Schffer-collterl input were 3,838 nd 4,76. The third cell ws lrger nd hd, synpses ville for distl Schffercollterl input. Perfornt-pth input ws modeled s the ctivtion of percentge of ville synpses in the picl tuft. For the first two cells, these synpses were situted pproximtely 7 mm from the som wheres for the third cell, this distnce ws pproximtely 7, mm. As for the Schffer-collterl input, perfornt-pth synpses were chosen rndomly from the ville pool on ny individul tril. The numers of synpses for the three cells were,,,47 nd,968. Hippocmpl slices preprtion. All niml procedures were pproved y the Northwestern University Animl Cre nd Use Committee. Trnsverse hippocmpl slices were otined from 4 -week-old mle Wistr rts. Rts were nesthetized with hlothne (Sigm-Aldrich), perfused trnscrdilly with icecold rtificil cererospinl fluid (CSF) nd decpitted. The rin ws then removed from the skull nd trnsverse hippocmpl slices (3 mm thick)were prepred using virtome. The extrcellulr solution used in slice preprtion nd incution ( 3 min t 3 C) ws either stndrd CSF ( mm NCl, mm dextrose, mm NHCO 3,. mm KCl,. mm NH PO 4, mmccl nd mm MgCl ) or sucrose-sed solution (7 mm sucrose, 7 mm NCl, mm dextrose, mm NHCO 3, 7 mm MgCl,. mm KCl,. mm NHPO4 nd. mm CCl ). All solutions were uled with 9% O nd % CO to mintin ph of 7.4. Following incution, ut efore recording, slices were mintined t room temperture ( C). Three of the experiments shown in Figure 8 were repeted in CSF contining higher K + (3 mm) nd lower C + (.3 mm with.7 mm Mg + ), which enhnces dendritic excitility 33. Despite this increse in dendritic excitility, the results shown in Figure 8 were not ffected y ionic conditions. Ptch-clmp recording. Individul slices were held in smll chmer perfused with CSF t 3 ml min (37 C) nd visulized with n upright, fixed-stge microscope (Zeiss Axioscop FS plus) using differentil interference-contrst, infrred video microscopy. Whole-cell current-clmp recordings were mde with BVC-7 mplifier (Dgn Instruments) nd ptch electrodes with n open tip resistnce of 3 4 MO. Series resistnce (3 MO) nd cpcitnce were compensted using the mplifier. The intrcellulr solution contined mm potssium gluconte, mm KCl, mm sodium phosphocretine, mm HEPES, mm Mg-ATP,.3 mm N-GTP nd.% iocytin. Only cells with memrne potentils less thn mv t the onset of the whole-cell recording were used for experiments. In most experiments, cells were mintined t 67 mv throughout the experiment with DC current injection s needed. In three of the experiments shown in Figure 8, however, cells were held t the resting potentil (tht is, no holding current); no significnt differences in the results were noted. Synptic stimultion. Activtion of the perfornt pth nd Schffer collterls ws otined with two-contct cluster electrodes (CED, FHC). The Schffercollterl stimulting electrode ws plced mm to the side of the cell ody of the recorded neuron nd pproximtely hlfwy etween the cell ody lyer nd the perfornt pth (visully identified y the high density of xons). The second stimulting electrode ws plced in the perfornt pth, mmlterlto the som of the recorded neuron. Synptic stimultion trils were repeted t intervls of s, to minimize effects of ctivity-dependent plsticity. In initil experiments, to ssess the locliztion of synptic stimultion, efore trnsferring slice to the recording chmer, we mde longitudinl cut long the order etween strtum rditum nd the strtum lcunosum-moleculre. In seprte experiment, to rule out the involvement of disynptic ctivtion, we mde cut from the CA3/CA order to the CA3/grnule cell order, therey removing CA3. Cuts did not influence the efficcy of synptic stimultion nd were therefore not used in susequent experiments. Drugs. In ll experiments, mm SR93 (Sigm-Aldrich) nd 3 mm CGP43 (Tocris Bioscience) were dded to the CSF to lock GABA A nd GABA B receptors, respectively. In some experiments, mm TTX (Sigm- Aldrich) ws dissolved in CSF nd pplied loclly using positive pressure (.4. psi) pplied to the ck of glss pipette (tip resistnce of 3 4 MO) positioned mm from the som. Dt cquisition nd nlysis. Dt were trnsferred to computer during experiments y n ITC-8 digitl-nlog converter (Instrutech). Igor Pro softwre (Wvemetrics) ws used for cquisition nd nlysis. Electrophysiologicl records were filtered t khz nd digitlly smpled t khz. Sttisticl tests were performed using Excel softwre (Microsoft) or GB-STAT (Dynmic Microsystems). All results re reported s men ± s.e.m., nd significnce ws determined t the P o. level. Note: Supplementry informtion is ville on the Nture Neuroscience wesite. ACKNOWLEDGMENTS We thnk memers of the Spruston nd Kth ls nd B. Mel for discussions. This work ws supported y the US Ntionl Institutes of Helth (NS-38 to N.S., NS-4664 to N.S. nd W.L.K., nd NS-4437 to T.J.) nd y the Ntionl Science Foundtion (DGE to A.R.). COMPETING INTERESTS STATEMENT The uthors declre tht they hve no competing finncil interests. Pulished online t Reprints nd permissions informtion is ville online t reprintsndpermissions/. Sturt, G., Spruston, N., Skmnn, B. & Häusser, M. Action potentil initition nd ckpropgtion in neurons of the mmmlin CNS. Trends Neurosci., 3 (997).. Johnston, D., Mgee, J.C., Colert, C.M. & Christie, B.R. Active properties of neuronl dendrites. Annu. Rev. Neurosci. 9, 6 86 (996). 3. Häusser, M., Spruston, N. & Sturt, G.J. Diversity nd dynmics of dendritic signling. Science 9, (). 4. Amrl, D.G. Emerging principles of intrinsic hippocmpl orgniztion. Curr. Opin. Neuroiol. 3, 9 (993).. Colert, C.M. & Levy, W.B. Electrophysiologicl nd phrmcologicl chrcteriztion of perfornt pth synpses in CA: medition y glutmte receptors. J. Neurophysiol. 68, 8 (99). 6. Desmond, N.L., Scott, C.A., Jne, J.A., Jr. & Levy, W.B. Ultrstructurl identifiction of entorhinl corticl synpses in CA strtum lcunosum-moleculre of the rt. Hippocmpus 4, 94 6 (994). 7. Levy, W.B., Colert, C.M. & Desmond, N.L. Another network model ites the dust: entorhinl inputs re no more thn wekly excittory in the hippocmpl CA region. Hippocmpus, 37 4 (99). 8. Soltesz, I. Brief history of cortico-hippocmpl time with specil reference to the direct entorhinl input to CA. Hippocmpus, 4 (99). 9. Yeckel, M.F. & Berger, T.W. Monosynptic excittion of hippocmpl CA pyrmidl cells y fferents from the entorhinl cortex. Hippocmpus, 8 4 (99).. Golding, N.L. & Spruston, N. Dendritic sodium spikes re vrile triggers of xonl ction potentils in hippocmpl CA pyrmidl neurons. Neuron, 89 (998).. Kmondi, A., Acsdy, L. & Buzski, G. Dendritic spikes re enhnced y coopertive network ctivity in the intct hippocmpus. J. Neurosci. 8, (998).. Colert, C.M. & Johnston, D. Axonl ction-potentil initition nd N + chnnel densities in the som nd xon initil segment of suiculr pyrmidl neurons. J. Neurosci. 6, (996). 3. Spruston, N., Schiller, Y., Sturt, G. & Skmnn, B. Activity-dependent ction potentil invsion nd clcium influx into hippocmpl CA dendrites. Science 68, 97 3 (99). 4. Golding, N.L., Kth, W.L. & Spruston, N. Dichotomy of ction-potentil ckpropgtion in CA pyrmidl neuron dendrites. J. Neurophysiol. 86, ().. Hoffmn, D.A., Mgee, J.C., Colert, C.M. & Johnston, D.K. K + chnnel regultion of signl propgtion in dendrites of hippocmpl pyrmidl neurons. Nture 387, (997). 6. Migliore, M., Hoffmn, D.A., Mgee, J.C. & Johnston, D. Role of n A-type K + conductnce in the ck-propgtion of ction potentils in the dendrites of hippocmpl pyrmidl neurons. J. Comput. Neurosci. 7, (999). 7. Poirzi, P., Brnnon, T. & Mel, B.W. Arithmetic of suthreshold synptic summtion in model CA pyrmidl cell. Neuron 37, (3). NATURE NEUROSCIENCE VOLUME 8 [ NUMBER [ DECEMBER 67