Na + imaging reveals little difference in action potential evoked Na + influx between axon and soma

Transcription

1 N + imging revels little difference in ction potentil evoked N + influx etween xon nd som Ily A Fleidervish 1 3, Nechm Lsser-Ross 2,4, Michel J Gutnick 2,3 & Willim N Ross 2,4 21 Nture Americ, Inc. All rights reserved. In corticl pyrmidl neurons, the xon initil segment (AIS) is pivotl in synptic integrtion. It hs een sserted tht this is ecuse there is high density of N + chnnels in the AIS. However, we found tht ction potentil ssocited N + flux, s mesured y high-speed fluorescence N + imging, ws out threefold lrger in the rt AIS thn in the som. Spike-evoked N + flux in the AIS nd the first node of Rnvier ws similr nd ws eightfold lower in sl dendrites. At ner-threshold voltges, persistent N + conductnce ws lmost entirely xonl. On time scle of seconds, pssive diffusion, nd not pumping, ws responsile for mintining trnsmemrne N + grdients in thin xons during high-frequency ction potentil firing. In computer simultions, these dt were consistent with the known fetures of ction potentil genertion in these neurons. In neocorticl pyrmidl cells, s in mny CNS neurons, the AIS hs pivotl integrtive role ecuse it hs lower threshold for ction potentil genertion thn the som nd dendrites 1 nd thus controls oth the trnsformtion of dendrosomtic synptic input into spike output nd the ckpropgtion of ction potentils into the dendrites. The properties of N + chnnels in the AIS re reported to differ from those in other regions of the cell. For exmple, severl studies hve found tht the ctivtion voltge of N + chnnels is 6 14 mv more hyperpolrized in the xon thn in the som 2 4. Moreover, the AIS hs een implicted s the primry source of persistent N + current (I NP ) 5,6. One mjor suject of disgreement is whether the density of N + chnnels is sustntilly greter in the AIS thn in the som. An erly computtionl study 7 found tht, in n ntomiclly correct comprtmentl model of pyrmidl neuron with identicl N + chnnel properties in ll regions, N + chnnel density in the AIS must e orders of mgnitude higher in the AIS thn in the som to simulte the lower xonl threshold. However, susequent ptch recordings from the AIS suggested tht the density in the two regions is similr 2,3,8. More recently, in the study in which ptch recordings indicted equl densities, the uthors 3 presented severl rguments why this result might not e correct nd sserted tht the true N + chnnel density in the AIS is much higher thn in the som or dendrites. A similr conclusion ws reched on the sis of recordings from les tht form when corticl xons re cut 4. We used high-speed fluorescence imging of the N + indictor sodium-inding enzofurn isophthlte (SBFI) 9,1 to quntittively descrie the N + dynmics tht ccompny suthreshold depolriztions nd single nd multiple ction potentil genertion in xons, dendrites nd somt of lyer 5 neocorticl neurons. Our mesurements of N + flux in xon, som nd sl dendrites suggest tht the rtios of N + chnnel densities in these regions is pproximtely 3:1:.3. These results provide evidence for the xonl loction of the suthreshold persistent N + conductnce nd indicte tht diffusion is the min removl mechnism following N + entry in the AIS nd first node of Rnvier. RESULTS N + trnsients in the xon decy fster thn in the som In 197 lyer 5 pyrmidl neurons, we recorded chnges in SBFI fluorescence during single or multiple ction potentils elicited y rief somtic current injections in the som nd nery xon nd sl dendrites. All the fluorescence trnsients were locked y thpplied tetrodotoxin (1 M, n = 4). During experiments, xons were distinguished from the other fine processes ecuse they emerged from the som opposite the picl dendrite nd hd distinctive N + trnsients (see elow). We exmined 36 neurons live in two-photon microscope fter the physiologicl experiment. In these cells, xons were redily distinguished from sl dendrites y their lck of spines (Fig. 1), confirming the determintion tht hd een mde during recording. These reconstructions were used to mesure the dimensions of the xon, som nd dendrites, which were lter used to estimte the reltive N + fluxes in the different comprtments. In representtive neuron, verged N + trnsients (n = 2) elicited y single ction potentils were prominent only in the xons nd were poorly or not t ll detectle in som nd dendrites (Fig. 1; see lso Supplementry Movie 1). The gretest chnge in xonl SBFI fluorescence occurred 1 3 m from the som. The intrcellulr N + concentrtion ([N + ] i ) grew rpidly (1 9% rise time less thn 1 ms) nd decyed rpidly with time constnt of 2 6 ms. At distnces eyond 35 5 m, in the presumed myelinted region of the xon 11, N + trnsient mplitudes were smller nd their peks were progressively delyed (Fig. 1). The mplitudes nd time courses of the fluorescence trnsients re expected to ccurtely reflect the entry nd removl of N + from the cell, s N + is not sustntilly uffered y either components of the cytoplsm or the N + indictor (see Online Methods). 1 Deprtment of Physiology, Fculty of Helth Sciences, Ben-Gurion University of the Negev, Beer Shev, Isrel. 2 Mrine Biologicl Lortory, Woods Hole, Msschusetts, USA. 3 Koret School of Veterinry Medicine, The Herew University of Jeruslem, Rehovot, Isrel. 4 Deprtment of Physiology, New York Medicl College, Vlhll, New York, USA. Correspondence should e ddressed to I.A.F. (ily@gu.c.il). Received 25 Jnury; ccepted 1 My; pulished online 13 June 21; doi:1.138/nn VOLUME 13 NUMBER 7 JULY 21 NATURE NEUROSCIENCE

2 Figure 1 Time course of ction potentil induced [N + ] i chnges is different in different comprtments. () Reconstruction of 58 opticl sections tken t 1- m intervls through prt of lyer 5 pyrmidl neuron filled with 2 mm SBFI. () Left, the sme neuron s seen during the fluorescence imging experiment with NeuroCCD-SMQ cmer. The rectngles nd rrows indicte the regions from which fluorescence mesurements were otined. Middle, verged N + trnsients (n = 2) elicited y single ction potentil were only prominent in the xon. Between 3 m, the trnsients peked shrply t the time of the spike, wheres the rise ws more grdul etween 3 nd 4 m. Right, verged N + Axon 2 m Som Axon 5 mv.5 s m 5 m Apicl dendrite trnsients (n = 1) elicited y trin of ten spikes (33 Hz) were detected in som, sl nd picl dendrites, lthough they were lrger in the xon. In the proximl xon, [N + ] i grew throughout the durtion of the trin; immeditely fter the trin, the [N + ] i declined rpidly. In som nd dendrites, [N + ] i styed t nerly stedy level fter the end of the spike trin m Axon Som 21 Nture Americ, Inc. All rights reserved. When trins of three or more ction potentils were generted, we lso detected N + signls in the som nd sl nd picl dendrites, lthough they remined lrgest in the proximl xon (Fig. 1). The most notle difference ws in the rtes of N + clernce. In the som nd dendrites, [N + ] i either styed t plteu level or decyed slowly fter the end of the spike trin; [N + ] i recovery ws often not complete even following the cesstion of stimuli. In contrst, [N + ] i rpidly declined immeditely on termintion of the stimuli in the proximl xon. Rpid N + removl in xons is not due to ctive trnsport We used two experimentl mnipultions to decrese the ctivity of the pump. First, we cooled the slices from 33 C to 23 C. This hd only miniml effect on the decy time constnt of the N + trnsients in the proximl xon ( = s t 33 C versus.54. t 23 C, P >.5, n = 4; Fig. 2). 22 C 32 C =.69 s =.68 s Control Ouin 22 C Wshout =.65 s 4% =.47 s =.52 s =.67 s 1% Figure 2 Active trnsport cnnot ccount for the rpid N + clernce in the xon. Left, the decy of xonl N + trnsients ~2 m from the som elicited y trin of ten ction potentils ws not ffected y heting of the slice from 22 C to 32 C. Thin trces re superimposed est fits to single exponentil t 22 C, 32 C nd on return of the temperture ck to 22 C. The difference ws undetectle. Right, lockde of N + /K + - ATPse with th-pplied ouin (1 M) hd little effect on xonl N + clernce. Thin trces re superimposed est fits of decy of the N + trnsients in control, during ouin ppliction nd following wshout. Second, th-pplied ouin (1 M), which locks the N + /K + - ATPse 12, hd little or no effect on xonl N + clernce ( = s in control versus.37.2 with ouin, P >.5, n = 3). Therefore, in the time frme of our experiments, ctive trnsport is not responsile for the dynmics of N + clernce from the xons. N + dynmics reflects loclized N + influx nd diffusion We next exmined whether spike-evoked N + entry nd susequent xil diffusion is sufficient to ccount for the time course nd oserved rpid N + clernce in the xons y compring experimentlly oserved N + trnsients with the results of numericl simultions. We uilt simplified comprtmentl model tht encompssed the fundmentl morphologicl fetures of lyer 5 pyrmidl neuron. In the model, the first 35 5 m of the xonl length were ssumed to e uncovered y the myelin sheth 11 nd to possess uniform N + chnnel density out threefold higher thn tht of the som (see Online Methods). The susequent, myelinted segment possessed no N + chnnels. In ddition to Hodgkin-Huxley type N + nd K + conductnces, the model incorported longitudinl diffusion of N + ions etween neighoring comprtments with diffusion coefficient of.6 m 2 ms 1 (ref. 13); no mechnism for N + extrusion ws included. Thus, the sptio-temporl ptterns of the simulted [N + ] i trnsients solely reflected trnsmemrne influx into the ctive comprtments tht possess N + chnnels (som nd AIS) nd N + diffusion into comprtments tht either possess no N + chnnels (myelinted segments) or re lrge sinks ecuse they hve reltively smll surfce-to-volume rtio (som). In ll of the ctive loctions in the model, [N + ] i grew throughout the ction potentil trin; it continued to grow fter the termintion of the trin in the pssive myelinted internodes, reflecting diffusion of N + from the djcent AIS (Fig. 3). The initil prts of the decy time constnts of the simulted N + trnsients were fstest t the most distl nd proximl extremes of the AIS nd incresed towrd its center. Within severl tens of milliseconds, however, the rte of N + diffusion from the AIS into the myelinted internodes slowed, reflecting the progressive decrese in the steepness of the N + concentrtion grdient. In contrst, the steep xo-somtic N + grdient persisted, s [N + ] i in the reltively immense som remined low nd nerly constnt. The model closely mimicked experimentlly oserved chnges in fluorescence during single ction potentil (Fig. 3) s well s ction potentil trins (Fig. 3,c). Amplitudes of [N + ] i elevtions were lrgest in the AIS, where the time-to-pek ws fstest nd uniform throughout the segment (Fig. 3,). In the first myelinted segment, NATURE NEUROSCIENCE VOLUME 13 NUMBER 7 JULY

3 Experiment Model Experiment Model 2 m 2 m 1%.4 mm 3% 2 mm 2 m 5 mv c Experiment Model d 1,5 e 1,2 3 L5 * 21 Nture Americ, Inc. All rights reserved. 3 ms 1 ms 1 ms 1% 3 ms 1, ms 1, ms mm 2 2 1, in which simulted N + trnsients only reflected longitudinl N + diffusion from the AIS, time-to-pek grew nd mplitude declined s function of the distnce from the site of N + influx (Fig. 3,,d). Finlly, we compred the N + trnsients in corticl neurons to those in cereellr Purkinje cells (Supplementry Fig. 1), which possess much shorter AIS of only 15 m (ref. 14). Both the sptil distriution of the times-to-pek of the N + trnsients long the xonl length nd the hlf-width t distnce of 1 m from the hillock differed significntly etween the two types of neurons (P <.5; Fig. 3e). The difference ws completely predicted y diffusion models y chnging the AIS length of 38 m in corticl neurons to 15 m for Purkinje cells (Fig. 3e). Reltive spike-evoked N + influx in different comprtments Becuse the dynmics of the N + trnsients for short time periods cn e entirely descried in terms of influx nd diffusion, we cn use the fluorescence mesurements to estimte the reltive fluxes of N + ions into the different comprtments during ction potentil genertion. However, inference of spike-evoked N + influx per unit memrne re (Q N ) from the imging dt requires creful considertion of differences in the surfce-to-volume rtio, ckground tissue fluorescence nd other technicl fctors. We used three different pproches to evlute reltive Q N. Our first pproch focused on the shpes of the N + trnsients in the AIS, the som nd the sl dendrites. We elicited N + trnsients with trins of 1 nd 1 ction potentils (Fig. 4). Under oth protocols, [N + ] i grew stedily throughout the trin in ll neuronl Time-to-pek (ms) Time-to-pek (ms) 8 4 L5 comprtments. In the AIS, [N + ] i declined rpidly fter reching its pek, chieving stedy level t out 2 s. In the som nd sl dendrites, the [N + ] i lso stopped rising t the end of the trin, ut did not fll over the next few seconds. The stedy level in ll three comprtments ws out the sme, indicting tht diffusionl equilirium ws chieved. The rpid recovery of the AIS signl shows tht the pek [N + ] i chnge in tht comprtment ws much higher thn the end-of-trin [N + ] i in the som. But it is not immeditely cler whether this results from higher Q N cross the memrne or ecuse the xon is thinner thn the som. Tht the dendritic signl did not decline fter the trin indictes tht the [N + ] i chnge in the dendrite ws close to the [N + ] i chnge in the som. As the surfce-to-volume rtio of the dendrite is much greter thn tht of the som, this closeness suggests tht Q N into the sl dendrite must e less thn Q N into the som. To otin more quntittive estimte of the Q N rtio mong the three comprtments, we mesured the pek nd stedy-stte vlues nd then constructed simple comprtmentl model to reproduce these vlues. The stedy stte to end-of-trin rtio of the AIS N + trnsients ws.22.7 (men s.d., n = 7) following trin of ten ction potentils nd.47.8 following trin of 1 ction potentils; in the sl dendrites, the rtios were nd.92.13, respectively. We next performed computer simultions to find the rtio of chnnel densities consistent with the mesured time courses nd reltive mgnitudes. The somtic N + chnnel density ws kept constnt, wheres N + chnnel density in the process (either xon, consisting PC t 1/2 (ms) PC 4 AIS length ( m) Figure 3 Axonl N + trnsients reflect loclized N + influx into the AIS followed y diffusion to the som nd first myelinted internode. () Experimentlly oserved (left) nd simulted (right) chnges in [N + ] i elicited y single ction potentil. () Similr [N + ] i chnges elicited y ten ction potentils t the indicted loctions. The chnges peked fter the end of the spike trin (dshed line) t the drk lue nd pink loctions. (c) The sme [N + ] i chnges plotted s function of distnce from the som for three different times fter the lst spike. Dots re verge vlues (n = 9). (d) Time-to-pek of the N + trnsients elicited y single ction potentil versus distnce from the xon hillock (n = 5). Red continuous line clculted from simplified model, ssuming 5- m AIS. (e) [N + ] i chnges elicited y ten ction potentils in pyrmidl neurons (L5) nd Purkinje neurons (PC). Left, time-to-pek of the N + trnsients in pyrmidl neurons (lck, n = 9) nd in Purkinje neurons (lue, n = 5; see Supplementry Fig. 1) versus distnce from the som. Continuous lines re fitted from the model, which ssumed n AIS length of 38 m for pyrmidl cells (red) nd 15 m for Purkinje cells (lue). Right, effect of AIS length on width of the xonl N + trnsients 1 m from the hillock of pyrmidl (lck, n = 8) nd Purkinje neurons (lue, n = 5). Points re t the AIS lengths for the two cell types (15 m for Purkinje cells nd 4 m for pyrmidl cells). The red continuous line is the hlf-width of the simulted trnsients. 854 VOLUME 13 NUMBER 7 JULY 21 NATURE NEUROSCIENCE

4 Figure 4 The shpe of spike-evoked N + trnsients constrins the rtio of N + chnnel densities in different comprtments. () N + trnsients elicited y 1 (left) nd 1 (right) ction potentils t the som (lck) in sl dendrite (2 m from the edge of the som, lue) nd in the AIS (2 m from the hillock, red) of representtive neuron. Arrowheds indicte the points where the stedy stte nd the end-of-trin fluorescence vlues were mesured nd the numers re the rtios of the stedy-stte to end-of-trin vlues. () N + trnsients in models with different process-to-som rtios of N + chnnel density. Somtic N + chnnel density ws kept constnt nd N + chnnel density in the process ws vried over the rnge of 3-fold of the somtic density (numers long the left side of the figure). The numers to the right of the trces re the clculted rtios of the stedy-stte to end-of-trin signl. The model required the sl dendrite N + chnnel density to e.1.3-fold lrger thn the somtic density nd the AIS density to e one- to threefold lrger thn the somtic density to mtch the experimentlly determined signls. 4% Experiment.1.9 Model BsD Som AIS 2 s 5 mv Nture Americ, Inc. All rights reserved. of 4- m-long AIS followed y myelinted internode, or sl dendrite) ws systemticlly vried. To mtch the experimentlly determined stedy stte to end-of-trin rtios, the N + chnnel density in the AIS hs to e one- to threefold lrger thn the somtic density nd the N + chnnel density in the sl dendrite hs to e.1.3-fold lrger thn the somtic density (Fig. 4). The presence of myelinted segment mde only smll difference in the predicted recovery curves (dt not shown). A second pproch to estimting reltive Q N from the imging dt relies on the following considertions. First, [N + ] i, where [N + ] i is the N + concentrtion chnge per spike nd is the frctionl chnge in SBFI fluorescence. This reltionship holds if the chnge in fluorescence is liner, which should e true ecuse the K D for the N + -SBFI interction is high (26 mm) 15 nd vlues re Q << 1. Second, [ N ] N i V, where Q N is the trnsmemrne N+ influx per spike nd V is the volume of the region of interest, ssuming the removl rte of N + is slow compred with the rte of entry. Assuming tht the indictor concentrtion is uniform throughout ll comprtments, which is resonle for the peri-somtic region fter 15 min of dilysis, then V F nd Q N F, s the fluorescence of SBFI in region should e proportionl to the volume of the region. This reltionship llows us to directly compre the chrge entries in different regions. Additionl technicl considertions re descried in the Online Methods. We mesured F in the som, AIS nd sl dendrites following trins of five ction potentils t 1-ms intervls (Fig. 5) or ten ction potentils t 3-ms intervls. We found tht F AIS / F som = for the fster trin (men s.d., n = 11) nd for the slower trin (n = 19), mesured in region round the som nd rectngle over the xon. The rtios were sensitive to chnges in the res of the circles or rectngles (Supplementry Fig. 2), ut the resulting error ws not greter thn. If we correct the mplitude determined from the slower trin for the reduction resulting from xil diffusion (increse y ~4%) then tht rtio ecomes , which is close to the rtio determined from the fster trin. The verge is Applying n dditionl correction fctor of.78 for the difference in surfce to cross section for the xon nd som (see Online Methods), we estimted tht the rtio of Q N in the two comprtments, Q AIS /Q som, to e We mde similr comprison etween the xon nd sl dendrites (Fig. 5). We found tht F xon / F sl = for spikes t 1-ms intervls (men s.d., n = 11) nd for spikes t 3-ms intervls (n = 9). We still hd to pply correction for diffusion for the slower trin, ut there ws no need to correct for differences in the surfce to cross section, s oth comprtments were ssumed to e cylinders. After correcting nd verging, we found tht Q AIS / Q sl = The third pproch for estimting Q N strts from the sme equtions. As there is lmost no uffering of entering N + ions (see Online F volume Methods), QN k CFrdy, where k is the chnge F surfcere in [N + ] i tht cuses = 1%, C Frdy is Frdy s constnt, F is the ction potentil evoked chnge in fluorescence nd F is the resting fluorescence of SBFI in the volume (equl to the totl mesured fluorescence minus F, the tissue utofluorescence). Typiclly, F ws out 5% of F in the som nd 35% of F in the xon nd sl dendrites. The volume-to surfce re rtio ws ssumed to e dimeter divided y 6 for qusisphericl som nd dimeter divided y 4 for the cylindricl AIS nd sl dendrites. The dimeters of the som, xon nd sl dendrites for ech neuron were estimted from twophoton reconstructions. In the som, AIS nd sl dendrites, the men vlues of F F volume were.3.6 (n = 14), (n = 14) nd surfcere.8.1% m (n = 9), respectively (Fig. 5). Q N in som, AIS nd in sl dendrites, ssuming k =.4 mm per 1% (ref. 15) ws , nd fc m 2, respectively. Using these numers, we found tht Q AIS /Q som = (n = 14) nd Q AIS /Q sl = (n = 9), close to the estimtes otined y the first two pproches (Fig. 5c) NATURE NEUROSCIENCE VOLUME 13 NUMBER 7 JULY

5 F F 1 ms 4 mv F volume (% m) F re c Q N rtio Som AIS AIS to som AIS to sl dendrite Bsl dendrite Figure 5 Reltive spike-evoked N + flux in different neuronl comprtments. () Top, nd F chnges t the indicted loctions (colored trces) elicited y trin of five ction potentils. Bottom, pseudocolor mps of the chnges etween the times mrked y the rrowheds. () Action potentil evoked N + chrge trnsfer derived from the mplitude of N + trnsients nd morphologicl dt. Dots represent individul mesurements of F volume clculted for som (n = 14), F re AIS (n = 14) nd sl dendrites (n = 9). Dshed lines re the men vlues. (c) AIS to som nd AIS to sl dendrite N + chrge trnsfer rtios clculted from F (n = 11) nd F volume (n = 9 14) vlues. F re 4 21 Nture Americ, Inc. All rights reserved. In these experiments, the [N + ] i chnges were elicited y trins of ction potentils. We lso evluted the SBFI fluorescence chnges ssocited with single ction potentils, lthough the smll signl size nd the inility to completely eliminte virtions mde the nlysis more difficult. In two neurons in which the single spike chnges could e clerly mesured in oth the som nd the xon, the Q AIS /Q som rtio for one ction potentil ws similr to tht determined with spike trins (Supplementry Fig. 3). Becuse the rise time of the single spike signls ws much fster thn the fll time there ws no need to correct for diffusion. Sodium dynmics t nodes of Rnvier In eight corticl neurons, nodes were identified y the chrcteristics of their N + trnsients (see elow). In four of these neurons, susequent two-photon reconstructions reveled thin collterl xonl rches tht originted from the xonl trunk t the presumed node loction (Fig. 6). An dditionl node ws found in the xon of cereellr Purkinje cell (dt not shown). In oth cell types, the first nodes were found within m of the som. In the two xons in which second nodes of Rnvier were identified (oth in N + -imging experiments nd Figure 6 Action potentil evoked N + chrge trnsfer in nodes of Rnvier is comprle to the trnsfer in the AIS. () Left, reconstruction of stck of 61 opticl sections through prt of lyer 5 pyrmidl neuron filled with 2 mm SBFI. The first xonl rnching point is ~1 m from the hillock (lue rrowhed). Middle, the sme neuron s seen during the fluorescence imging experiment with NeuroCCD-SMQ cmer. Right, verged N + trnsients (n = 2) elicited y trin of five ction potentils (2 Hz). See lso Supplementry Figure 4. () Simulted chnges in [N + ] i elicited y single ction potentil plotted ginst distnce from the xon hillock. The node ws ssumed to e 1 m long nd the AIS to e 45 m. Dshed lines re [N + ] i chnges in model with no N + diffusion. Blck continuous lines re [N + ] i vlues 1 ms following the pek of the spike in the AIS nd.7, 2 nd 1 ms following the pek in the region round the node of Rnvier. (c) Experimentlly oserved (left) nd simulted (right) chnges in [N + ] i elicited y trin of five ction potentils plotted ginst distnce from the xon hillock. Left, dots re verged vlues from ech m-long pixel long the xon during the time intervl 2 1 ms following the pek of the lst spike in trin. Red line, AIS; lue rrowhed, the first node of Rnvier. Right, chnges in [N + ] i in models with different node-to-ais rtios of N + chnnel density. F F volume F re in two-photon reconstructions), they were locted 22 nd 28 m from the first node. Physiologiclly, nodes were identified s isolted loctions where the time course of the N + trnsient ws in synchrony with the spike trin (Fig. 6). When these N + elevtions occurred, they were evident for distnces of 1 2 m long the xonl length, oviously longer thn the dimensions of centrl node of Rnvier 16. The mximl mplitudes of nodl N + trnsients elicited y five ction potentils t 1-ms intervls ( =.5 to ) were <1% of those in the AIS of the sme neuron (Fig. 6). The mplitude ws mximl in the center of segment nd decyed in oth directions (Supplementry Fig. 4). c.7 ms 2 ms Experiment 2 m Node 1 ms 4.4 mm 1 mm AIS g N node/ais Model ms ms 1% 856 VOLUME 13 NUMBER 7 JULY 21 NATURE NEUROSCIENCE

6 21 Nture Americ, Inc. All rights reserved. Figure 7 At suthreshold voltges, persistent N + current is generted predomintely in the AIS. () A 1-s, 7-pA current step to just suthreshold voltge elicited lrge [N + ] i increse in the AIS (red trce). A rief (5 ms) pulse tht generted single ction potentil generted N + trnsient tht rose shrply. () Suthreshold pulses of.3, 1. nd 3. s ech generted [N + ] i increse tht lsted the durtion of the pulse. The rpid recovery t the end of the pulses indictes tht the current ws ctive throughout. A shrply rising N + trnsient elicited y trin of five ction potentils t 5 Hz is shown for comprison. (c) A just suthreshold, 1-s stimulus elicited lrge [N + ] i increse in the AIS (red trce), wheres the increse in the som (lck trce) ws much smller. The trin of ten ction potentils (4 Hz) cused sizle N + trnsients in oth loctions. (d) A 2-s voltge rmp from 7 to 4 mv elicited N + trnsients only in the xon. Interpolting long the rmp indictes tht I NP nd xonl [N + ] i oth egn to chnge t the sme voltge ( 57 6 mv, n = 5). (e) A 2-s-long Becuse the nodes re smll, N + ions diffused lterlly lmost s quickly s they enter cross the nodl memrne. Computer simultions showed tht diffusion-medited [N + ] i equilirtion etween the nodl nd internodl volumes should e nerly complete in milliseconds (Fig. 6) nd tht more thn 9% of N + ions tht enter 1- m-long node were lredy in the internodl region within 1 ms of the pek of the ction potentil. In contrst, 1 ms fter n ction potentil in the 48- m-long AIS, dissiption of the N + grdient ws miniml. We cnnot mesure the pek [N + ] i in the node even if we record the fluorescence chnge t 5 frmes per s, s N + ions diffuse wy from the node too quickly. However, ecuse diffusion is the primry clernce mechnism, we expect tht ll of the N + tht entered t the node will e contined for short time in region extending 2 m to either side. Thus, integrtion of the [N + ] i chnge over this region 1 ms fter the spike trin estimtes the totl chrge trnsfer vi nodl N + chnnels. This mesure ws similr to the chrge trnsfer density in the AIS (Fig. 6c). This conclusion ws supported y computer simultions in which the AIS N + chnnel density ws held constnt while N + chnnel density in the first node ws vried (Fig. 6c). Comprison of the experimentlly determined chnges in this cell with those predicted y the model indictes nodl N + chnnel density of the sme order of mgnitude s in the AIS (Supplementry Fig. 4). [N + ] i chnges ssocited with persistent N + conductnce To exmine [N + ] i chnges resulting from ctivtion of I NP, we pplied suthreshold pulses of vrying durtions. These depolriztions evoked lrge chnge in SBFI fluorescence in the xon tht persisted throughout the pulse nd were completely nd reversily locked y tetrodotoxin (dt not shown, n = 3), indicting tht they reflect N + influx through voltge-sensitive N + chnnels. In ll of the neurons studied, the mplitude of the xonl N + trnsients elicited y 1-s suthreshold depolriztions ws comprle to or lrger thn the trnsients elicited y single ction potentils, even though the signl ssocited with the prolonged suthreshold pulse must hve included N + diffusion out of the AIS (Fig. 7). Even with c voltge rmp from 7 to mv elicited N + signls tht were clerly detectle in som, sl nd proximl picl dendrites. With this lrger rmp, the memrne current nd AIS opticl signls still egn to chnge t 57 5 mv (n = 21), ut the signls in the som nd sl dendrites egn to chnge t 41 5 mv (n = 21). 3 s 1% 4 mv d e 7 mv 7 mv 1% suthreshold depolriztions s long s 3 s, the signl ws mintined throughout the durtion of the pulse, reveling the truly persistent nture of the underlying N + current (Fig. 7). I NP undergoes some slow inctivtion 17, ut this ws not evident in the imging dt. In most experiments, there ws no detectle persistent N + entry in the djcent sl dendrite (Fig. 7) nd som (Fig. 7c) following suthreshold depolriztion. Extensive signl verging reveled smll somtic signl in only 3 of the 12 cells. We determined the AIS-to-som rtio of the mplitudes of the signl evoked y 1-s suthreshold depolriztion nd found tht it ws much lrger thn the rtio of mplitudes evoked y trin of ten ction potentils (Fig. 7c). This result ws found even though N + diffusion out of the AIS should hve much greter effect on pek xonl during the prolonged pulse thn on the rpid, spike-evoked signl. To evlute the contriution of I NP t suprthreshold potentils, we exmined the N + trnsients generted during rmps under voltge clmp. Slow voltge rmps from 7 mv to 4 mv or to mv (Fig. 7d,e) were pplied to somt with K + currents locked using Cs + s the min intrcellulr ction nd C 2+ currents locked y dding 2 M Cd 2+ to the th 6,18. Anlysis of rmp-induced N + trnsients in 21 neurons showed tht voltge onset of I NP genertion ( 57 5 mv) ws ccompnied y prllel onset of the chnge in AIS SBFI fluorescence ( 57 6 mv), wheres somtic nd sl dendritic N + signls were first detected t more depolrized potentils ( 41 6 mv). The signls in the som nd dendrites correspond primrily to N + entry into those regions nd not diffusion from the xon, s the signls egn to decline s soon s the rmp ended. This difference in onset voltges suggests tht I NP hs shifted voltge dependence in the xon compred with the other comprtments. Implictions of N + flux distriution for spike genertion In xonl recordings, ction potentil mximl rtes of rise re greter thn 1, V s 1 (ref. 3) nd the preferred loction for spike initition is ner the distl end of the AIS 11,19. Some studies 3,7 suggest tht high xon-to-som rtio of N + conductnces is required to ccount for these oservtions. As our imging experiments found tht 4 mv mv 25 pa 4 pa NATURE NEUROSCIENCE VOLUME 13 NUMBER 7 JULY

7 m (µs) m scling fctor: V m (mv) 1 mv 114 µs 23 µs m = 6 µs mv.5 ms Uniform g N density AIS g N density ms Distnce from hillock (µm) Distnce from hillock (µm) V m (mv) ms V m (mv) 21 Nture Americ, Inc. All rights reserved AIS g N = 2 mv 2 V s 1 5 ps µm 2 1 ms 5 1, g N (ps µm 2 ) the conductnce rtio etween these comprtments ws not high, we sked whether model with resonle set of prmeters could e constructed tht would mtch our oservtions nd still produce ction potentils with these chrcteristics. We found tht we could simulte the fst upstroke despite the much lower AIS N + chnnel density y ssigning fster, more relistic vlues to the kinetic prmeters of the N + chnnels (Fig. 8). The velocity of the ction potentil upstroke is given y dv I N, dt Clocl where I N is the current flowing through the N + chnnels nd C locl is the memrne cpcitnce (ssuming tht I N is sustntilly lrger thn other currents). Thus, the smllest I N tht could underlie dv of 1, V s 1 is ~1 pa m 2 (ssuming C = 1 F cm 2 ), which dt corresponds to g N of 2 ps m 2 (ssuming tht t the point t which dv is mximl the driving force for N + is 5 mv). This is the dt vlue if N + chnnel ctivtion is instntneous; it will e lrger if the time constnt of N + chnnel ctivtion ( m in the Hodgkin-Huxley formlism) is slow. The vlue of m t physiologicl temperture is not known nd is difficult to mesure. Generlly, the ndwidth used for N + chnnel recordings in CNS neurons is 2 khz ( 3 db, 8-pole Bessel filter) (for exmple, refs. 2,3), which optimizes the signl-to-noise rtio nd permits ccurte cpcitive trnsient sutrction. However, recording t 2 khz t 23 C distorts nd slows the onset kinetics of currents if m is fster thn out 12 s (Fig. 8). The most ccurte estimte of N + chnnel ctivtion kinetics in centrl neurons ws mde from mossy fier outons 2. Recording t 23 C, with 1-kHz low-pss filter, the uthors reported m of 38 s nd 14 s for mv nd +4 mv, respectively. At 32 C, if Q 1 = 3 nd the rte is nonsturting, m will e13 s nd 5 s, respectively. Mx. rte-of-rise (V s 1 ) 2, 1,5 1, 5 m scling fctor: Dendrite AIS Node Dendrite AIS Node AIS g N density 3, m ms AIS g N density 3, m.2, G NP 5% Distnce from hillock (µm) Distnce from hillock (µm) Figure 8 Comprtmentl model of n ction potentil tht mtches the fst mximl rtes of rise of recorded spikes nd initites in the xon without requiring high AIS N + chnnel density. () Top left, lue line indictes m versus V m in model sed on pulished recordings 2. Light lue nd red lines represent the sme reltionships using scling fctors of.2 nd.5. Top right, single comprtment simultions of N + currents produced y voltge steps from 1 mv to mv s if recorded with n open ndwidth mplifier (continuous line) or filtered t 2 khz. Bottom left, xonl ction potentils (left) nd the first derivtive of ction potentil voltge (right) in the models with m curves s indicted. The AIS N + conductnce ws 5 ps m 2. Bottom right, mximl rte of rise versus xonl N + chnnel density (g N ) if m is (lck), using scling fctors of.5,.2 nd 1 reltive to the model 2. Dshed line indictes 1,13 V s 1, the mximl mesured rte of rise 3. () Effect of AIS chnnel density nd properties on ction potentil initition. Top left, with AIS g N s in the som, the ction potentil initited simultneously in the som nd in the AIS. Top right, with threefold higher AIS g N nd with shifted voltge dependence, initition shifted to the xon. Bottom left, scling N + chnnel m y.2 in ll comprtments increses the rte of rise, shifts the threshold nd enhnces the xo-somtic delys nd voltge grdients. Bottom right, with G NP eing 5% of the totl AIS N + conductnce, the shifts in threshold nd the xo-somtic delys nd voltge grdients re greter V m (mv) ms When m is in this rnge, very fst ction potentils re possile with considerly lower current density (Fig. 8). The ction potentils re shown t the distl end of the AIS, where they re most isolted from the som y the thin AIS. Using simultions, we exmined the effects of density nd properties of the N + chnnels in the AIS t the site of ction potentil initition (Fig. 8). In ll models, the voltge dependence of xonl g N ws shifted y 6 mv compred with somtodendritic chnnels 2 4,21. When N + conductnce ws equl in the som nd AIS (25 ps m 2 ), somtic current injection (2 ms, 1 na) produced reltively slow nd uniform regenertive response cross som, proximl xon nd dendrites (Fig. 8; see lso Supplementry Movie 2). A threefold increse in the AIS conductnce, however, ws sufficient to chnge this pttern towrd xonl spike initition (Fig. 8). The ction potentil voltge threshold ws shifted to more negtive potentils nd xo-somtic delys nd voltge grdients were greter in models with fster m ( m.2; Fig. 8) nd the xo-somtic difference ws even more evident when persistent N + conductnce ws dded to the AIS (Fig. 8; see lso Supplementry Movie 3). In ll of these models, ction potentils ckpropgted over the dendrites following initition in the AIS. DISCUSSION N + imging llowed us to quntittively evlute the chrcteristics nd density of sodium chnnels in thin neuronl processes, providing complementry pproch to electrophysiologicl nd immunocytochemicl techniques, which hve sometimes een controversil. For exmple, immunocytochemicl mesurements 3 might overestimte chnnel density y leling chnnel proteins tht re not functionl nd my not e relevnt to excitility. We concluded tht N + chnnel density (or more precisely, N + current density per ction V m (mv) 858 VOLUME 13 NUMBER 7 JULY 21 NATURE NEUROSCIENCE

8 21 Nture Americ, Inc. All rights reserved. potentil) in the som is out threefold lower thn in the AIS nd threefold greter thn in sl dendrites. In nodes of Rnvier, ction potentil induced N + influx ws of the sme order of mgnitude s in the AIS. At functionlly criticl suthreshold rnge of voltges, I NP ws primrily generted y the AIS. Pssive diffusion, not ctive pumping, ws responsile for rpid clernce of N + from eneth the memrne of the AIS nd node of Rnvier during repetitive ction potentil genertion. N + chnnel density in different neuronl comprtments The finding tht ction potentil induced N + flux density is out threefold greter in the AIS thn in the som would indicte the sme rtio of chnnel densities if the mplitude nd shpe of the ction potentil nd the temporl kinetics of the chnnels were the sme in the two regions. In the proximl AIS, the mximl rte of rise of the ction potentil is out twice tht of the somtic ction potentil, wheres spike mplitudes re similr in oth comprtments 3. Becuse the time to pek is out 2 3% riefer thn in the som, we would underestimte the rtio of pek g N y out tht mount if we use the xon-to-som flux rtio tht we mesured to estimte the conductnces. On the other hnd, the N + current is entirely inctivted y the time of the pek of the ction potentil in the isolted pyrmidl cell som 22. Becuse the pek is reched erlier in the xon, there my e some N + current during repolriztion. As this is n dditionl current going through the sme chnnels tht re open on the rising phse, the xon-to-som flux rtio of 3:1 tht we recorded proly reflects n even smller rtio of pek g N, nd chnnel densities. In fct, severl studies 2,3,8 hve estimted chnnel densities on the sis of electrophysiologicl recordings from cell-ttched or excised ptches in the AIS nd rrived t estimtes in the AIS tht re less thn threefold higher thn their estimtes for the som. One suggestion 3 is tht the rtios determined from ptch recordings do not reflect the effective physiologicl chnnel density ecuse N + chnnels in the AIS re nchored to the cytoskeleton nd do not completely revel themselves in ptch recordings. This nlysis is sed on experiments tht disrupted the connection of N + chnnels to nkyrin G nd therey incresed the pprent chnnel density. However, even if tht interprettion is correct, there ws only threefold increse in density, which is still closer to the rnge of cell-ttched mesurements nd our determintion thn to the fctor of ~4 5 tht hs een proposed on the sis of other considertions 3. The reltive ction potentil evoked [N + ] i chnges tht we recorded in different regions were qulittively similr to those reported in pyrmidl neurons 3 nd in other cell types 23,24. All experiments gree tht the frctionl fluorescence chnge () following depolrizing stimulus is much lower in the som thn in the AIS. However, reflects concentrtion chnge nd not Q N, which is the prmeter closest to chnnel density. When we used F, which is more closely relted to current density, the fluorescence chnges in the two comprtments were similr. In our other pproch, where we compred [N + ] i chnges, ut corrected for the difference in the surfce-to-volume rtio, we cme to the sme conclusion. Another experimentl oservtion tht suggested 3 high density of chnnels in the AIS is the extremely fst rise time of the ction potentil in AIS. The model used in tht nlysis 3 proly required high chnnel density to compenste for high estimte of the N + chnnel ctivtion time constnt m. We were le to simulte the sme fst rise time with much lower N + conductnce in the AIS, provided we ssigned fster vlues (sed on other experiments 2,25 ) to the kinetic prmeters of the N + chnnels (Fig. 8). Another rgument 3 for high N + chnnel density in the AIS is tht it is required to support the experimentl finding tht ction potentils initited t the distl portion of the AIS nd not in the first node of Rnvier. This rgument, however, depends criticlly on the reltive N + chnnel density in the nodes nd in the AIS. The model referred to in tht rgument 3 used estimted vlues sed on recordings in rt peripherl nerve 2 (21, chnnels per node). Our results (Fig. 6) indicte tht, in the xon of the pyrmidl neuron, where nodes re smller nd closer together, the current density is smller. There is therefore no need to postulte high N + chnnel density in the AIS. Our findings out the reltive sl dendrite/xon current densities re consistent with previously reported imging dt 3. As with the xonl dt, the interprettion of these results in terms of chnnel densities depends on the reltive spike mplitudes in the different comprtments. There is still some disgreement out ction potentil properties in the sl dendrites. Direct ptch recordings from the sl dendrites suggest ttenuted ck-propgtion of ction potentils 26, wheres experiments using voltge-sensitive dyes found less ttenution 27. If the spikes re ttenuted, we would expect slightly higher reltive chnnel density thn tht estimted y Q N. As our mesurements were mde less thn 3 m from the som, where ttenution is miniml, we expect the correction to e smll. Persistent N + current Our imging dt confirm tht persistent N + conductnce in lyer 5 pyrmidl cells ws generted predomintely in the xon in the functionlly criticl suthreshold voltge rnge, s previously reported 6. The voltge-clmp experiments indicte tht persistent N + conductnce ws lso ctivted in the somtic nd dendritic memrnes t more depolrized potentils. It remins uncler whether different voltge dependence of the persistent conductnce in these comprtments prllels the reported difference in voltge dependence for the trnsient N + current 2 4,21. It is generlly ccepted tht the ctivtion voltge of I NP is more negtive thn tht of the trnsient current, lthough the extent of the leftwrd shift is not known (for review, see ref. 28). Thus, the xonl I NP undoutedly contriutes to the lower threshold of the AIS 1,8. Diffusion nd clernce of N + ions Mny types of centrl neurons fire high-frequency ursts tht propgte vi myelinted xons to trget cells with timing precision tht is criticl oth for ongoing communiction etween neurons 29 nd for synptic plsticity 3. The structurl reltionships etween the xonl memrne nd the myelin sheth re optimized for initition nd rpid, slttory propgtion of ction potentils. Our results suggest n dditionl function for this orgniztion, tht the geometricl dimensions of myelinted nd nonmyelinted segments of the xon permit rpid, energy efficient restortion of trnsmemrne N + grdients eneth the excitle memrne. We found oth experimentlly nd using numericl simultions tht, in myelinted xons, ction potentils re ssocited with mrked heterogeneity in [N + ] i over short length of AIS-som nd node-internode ssemlies. We did find smll voltge-dependent N + nd C 2+ entry in the myelinted region (Supplementry Fig. 5), consistent with previous oservtions 31,32, ut the spike-evoked N + entry in this region ws much smller thn in the AIS or first node. The resultnt shrp N + grdients fcilitted rpid recovery of [N + ] i y diffusion. Of course, N + ions tht enter the xoplsm during ctivity will eventully hve to e extruded y the N + /K + -ATPse, which uses out hlf of the rin energy udget for this purpose 33. However, in the time frme of milliseconds, it is N + clernce medited y lterl diffusion tht prevents N + ccumultion in the NATURE NEUROSCIENCE VOLUME 13 NUMBER 7 JULY

9 21 Nture Americ, Inc. All rights reserved. functionlly criticl, nonmyelinted segments. Moreover, s N + influx is predomintely restricted to the initil segment nd to the periodic nodes of Rnvier, the N + extrusion mechnism utilizes the whole neuronl memrne, including the myelinted internodes nd som where the pump is present 34. METHODS Methods nd ny ssocited references re ville in the online version of the pper t Note: Supplementry informtion is ville on the Nture Neuroscience wesite. ACKNOWLEDGMENTS We thnk S. Mnit for excellent help with the preprtion of slices. This work ws supported y US-Isrel Bintionl Science Foundtion grnt (2382), grnt from the Isrel Science Foundtion (1376-6), Grss Fculty grnt from the Mrine Biologicl Lortory, US Ntionl Institutes of Helth grnt (NS16295), Multiple Sclerosis Society grnt (PP1367) nd fellowship from the Gruss Lipper Foundtion. AUTHOR CONTRIBUTIONS I.A.F., N.L.-R., M.J.G. nd W.N.R. designed the study, performed the corticl experiments nd wrote the pper. N.L.-R. nd W.N.R. performed the cereellr experiments. I.A.F. constructed the computtionl models. COMPETING FINANCIAL INTERESTS The uthors declre no competing finncil interests. Pulished online t Reprints nd permissions informtion is ville online t reprintsndpermissions/. 1. Kole, M.H. & Sturt, G.J. Is ction potentil threshold lowest in the xon? Nt. Neurosci. 11, (28). 2. Colert, C.M. & Pn, E. Ion chnnel properties underlying xonl ction potentil initition in pyrmidl neurons. Nt. Neurosci. 5, (22). 3. Kole, M.H. et l. Action potentil genertion requires high sodium chnnel density in the xon initil segment. Nt. Neurosci. 11, (28). 4. Hu, W. et l. Distinct contriutions of N(v)1.6 nd N(v)1.2 in ction potentil initition nd ckpropgtion. Nt. Neurosci. 12, (29). 5. Sturt, G. & Skmnn, B. Amplifiction of EPSPs y xosomtic sodium chnnels in neocorticl pyrmidl neurons. Neuron 15, (1995). 6. Astmn, N., Gutnick, M.J. & Fleidervish, I.A. Persistent sodium current in lyer 5 neocorticl neurons is primrily generted in the proximl xon. J. Neurosci. 26, (26). 7. Minen, Z.F., Joerges, J., Huguenrd, J.R. & Sejnowski, T.J. A model of spike initition in neocorticl pyrmidl neurons. Neuron 15, (1995). 8. 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. 16, (1996). 9. Mint, A. & Tsien, R.Y. Fluorescent indictors for cytosolic sodium. J. Biol. Chem. 264, (1989). 1. Cllwy, J.C. & Ross, W.N. Sptil distriution of synpticlly ctivted sodium concentrtion chnges in cereellr Purkinje neurons. J. Neurophysiol. 77, (1997). 11. Plmer, L.M. & Sturt, G.J. Site of ction potentil initition in lyer 5 pyrmidl neurons. J. Neurosci. 26, (26). 12. Lelievre, L., Zchowski, A., Chrlemgne, D., Lget, P. & Prf, A. Inhiition of (N + + K + ) ATPse y ouin: involvement of clcium nd memrne proteins. Biochim. Biophys. Act 557, (1979). 13. Kushmerick, M.J. & Podolsky, R.J. Ionic moility in muscle cells. Science 166, (1969). 14. Clrk, B.A., Monsivis, P., Brnco, T., London, M. & Husser, M. The site of ction potentil initition in cereellr Purkinje neurons. Nt. Neurosci. 8, (25). 15. Rose, C.R., Kovlchuk, Y., Eilers, J. & Konnerth, A. Two-photon N + imging in spines nd fine dendrites of centrl neurons. Pflugers Arch. 439, (1999). 16. Peters, A. The node of Rnvier in the centrl nervous system. Q. J. Exp. Physiol. Cogn. Med. Sci. 51, (1966). 17. Fleidervish, I.A. & Gutnick, M.J. Kinetics of slow inctivtion of persistent sodium current in lyer V neurons of mouse neocorticl slices. J. Neurophysiol. 76, (1996). 18. Alzheimer, C., Schwindt, P.C. & Crill, W.E. Modl gting of N + chnnels s mechnism of persistent N + current in pyrmidl neurons from rt nd ct sensorimotor cortex. J. Neurosci. 13, (1993). 19. Kole, M.H., Letzkus, J.J. & Sturt, G.J. Axon initil segment Kv1 chnnels control xonl ction potentil wveform nd synptic efficcy. Neuron 55, (27). 2. Engel, D. & Jons, P. Presynptic ction potentil mplifiction y voltge-gted N + chnnels in hippocmpl mossy fier outons. Neuron 45, (25). 21. Royeck, M. et l. Role of xonl N V 1.6 sodium chnnels in ction potentil initition of CA1 pyrmidl neurons. J. Neurophysiol. 1, (28). 22. Crter, B.C. & Ben, B.P. Sodium entry during ction potentils of mmmlin neurons: incomplete inctivtion nd reduced metolic efficiency in fst-spiking neurons. Neuron 64, (29). 23. Bender, K.J. & Trussell, L.O. Axon initil segment C 2+ chnnels influence ction potentil genertion nd timing. Neuron 61, (29). 24. Lsser-Ross, N. & Ross, W.N. Imging voltge nd synpticlly ctivted sodium trnsients in cereellr Purkinje cells. Proc. Biol. Sci. 247, (1992). 25. Alle, H., Roth, A. & Geiger, J.R. Energy-efficient ction potentils in hippocmpl mossy fiers. Science 325, (29). 26. Nevin, T., Lrkum, M.E., Polsky, A. & Schiller, J. Properties of sl dendrites of lyer 5 pyrmidl neurons: direct ptch-clmp recording study. Nt. Neurosci. 1, (27). 27. Acker, C.D. & Antic, S.D. Quntittive ssessment of the distriutions of memrne conductnces involved in ction potentil ckpropgtion long sl dendrites. J. Neurophysiol. 11, (29). 28. Ben, B.P. The ction potentil in mmmlin centrl neurons. Nt. Rev. Neurosci. 8, (27). 29. Sugihr, I., Lng, E.J. & Llins, R. Uniform olivocereellr conduction time underlies Purkinje cell complex spike synchronicity in the rt cereellum. J. Physiol. (Lond.) 47, (1993). 3. Mrkrm, H., Luke, J., Frotscher, M. & Skmnn, B. Regultion of synptic efficcy y coincidence of postsynptic APs nd EPSPs. Science 275, (1997). 31. Shrger, P. The distriution of sodium nd potssium chnnels in single demyelinted xons of the frog. J. Physiol. (Lond.) 392, (1987). 32. Zhng, C.L., Wilson, J.A., Willims, J. & Chiu, S.Y. Action potentils induce uniform clcium influx in mmmlin myelinted optic nerves. J. Neurophysiol. 96, (26). 33. Attwell, D. & Idecol, C. The neurl sis of functionl rin imging signls. Trends Neurosci. 25, (22). 34. Mt, M., Fink, D.J., Ernst, S.A. & Siegel, G.J. Immunocytochemicl demonstrtion of N +,K + -ATPse in internodl xolemm of myelinted fiers of rt scitic nd optic nerves. J. Neurochem. 57, (1991). 86 VOLUME 13 NUMBER 7 JULY 21 NATURE NEUROSCIENCE