<p> ) % (</p><p> e 200 Proximal 200 Distal p o l s</p><p>P</p><p>S 100 100 P E</p><p> d e z i l</p><p> a 50 50 m r o -200 -150 -100 -50 0 50 100 -200 -150 -100 -50 0 50 100 N Pre/post interval (ms)</p><p>Supplementary Figure 1. Spike-timing-dependent synaptic depression at proximal and distal dendrites in the presence of 10 M picrotoxin (black circles), superimposed on the STDP window measured in normal bath solution (gray circles, same as Fig. 1d). At 100<t<50 ms, significant LTD was induced at distal (–25.2±6.1%, n=6, p<0.01) but not at proximal dendrite (6.8±5.6%, n=4, p>0.8). a 35 m 75 m 105 m</p><p> b 4AP c</p><p>) Control V</p><p>150 ) 12 s m (</p><p> m ( e</p><p> d</p><p>100 h 8 t u t d i i l p w 50 4 m e a k i</p><p> p e k S i 0 0 p S 0 50 100 150 0 50 100 150 Distance ( m) Distance ( m) d e )</p><p>100 ) 60 % s (</p><p> m e (</p><p> d 75 h u</p><p> t 40 t i d l i</p><p> p 50 w</p><p> m</p><p>. a r 20</p><p>. p</p><p> r 25 p p u p S u 0 0 S 0 50 100 150 0 4 8 12 Spike amplitude (mV) Spike width (ms)</p><p>Supplementary Figure 2. Back-propagating APs and AP-induced suppression of EPSPs recorded from apical dendrites of L2/3 pyramidal cells. a, Examples of back-propagating APs recorded at 35, 75, and 105 m from the soma (from different cells). APs were evoked antidromically with a small stimulating electrode placed at the axon initial segment. Scale bars: 40 mV, 5 ms. b, Spike amplitude as a function of dendritic distance from the soma. Red symbols indicate experiments performed with 4-AP (3 mM) in the recording electrode. c, Spike width (measured at half-height of the AP) as a function of dendritic distance. d, Amplitude of EPSP suppression window as a function of spike amplitude. Each suppression window was fit with a single exponential 1 Aet / ; A is used to represent amplitude. e, Width of EPSP suppression window (measured at half-height of the exponential fit, which equals  log2) as a function of spike width. AP. Same as Fig. 2e, f, except that the recordings were made in the dendrite near the stimulating the near to 6(AMPAR-EPSPs)7(NMDAR-EPSPs). n=2 and 3to electrode. dendrite the in made were recordings the that except f, 2e, Fig. as Same AP. 3 Figure Supplementary Normalized amplitude (%) 100 120 40 60 80 -150 Pre/post interval(ms) prox dist AMPAR-EPSPs -100 Dendritic . Suppression of AMPAR-EPSP and NMDAR-EPSP by back-propagating by NMDAR-EPSP and AMPAR-EPSP of Suppression . -50 0 100 120 40 60 80 -150 Pre/postinterval (ms) NMDAR-EPSPs -100 dist Dendritic -50 prox 0 not significantly different from age-matched control neurons (p>0.3). neurons control from age-matched significantly different not were neurons infected of resistance input and potentials membrane Resting labeling. GFP on based 2003, al., et (Takahashi cut were slices brain before days two for recover to allowed were Rats diameter). tip μm (15 pipette small a with cortex superficial the into injected was media SFV-containing and cortex visual the over performed were mm (1 craniotomies Small mg/kg). (10 xylazine and mg/kg) (80 ketamine with anesthetized were California, of University the at Committee Use and Care Animal the of the guidelines with accordance in rats old 4-week in performed were Surgeries temperature). room min, (45 0.5mg/ml with activated was virus and harvested was media culture hours, 48 After 2000. lipofectamine using pSCAhelper and pSCA3-GFP-NR2A with cells HEK293 co-transfecting Neuropharmacology 2002, al., et (Krupp desensitization NMDAR calcineurin-dependent for important C-terminus the of region serine-containing a NMDAR, the of subunit NR2A the of 871-950 residues to corresponding Methods. (p<0.001, significant p>0.4, n=3; (–5.1±3.9%; neurons post 60 by induced was LTD (p<10 ms –50 (p<0.001), ms –30 (p<0.01), ms Data s.e.m. bar: Error neurons. pyramidal L2/3 12) to n=7 line, (gray control and n=4) line, (black infected of induction. 4 Figure Supplementary</p><p>Normalized amplitude (%) a 100 120 80 40 60</p><p> a were significantly different over the following pre/post intervals: –10 ms (p<0.05), –20 (p<0.05), ms –10 intervals: pre/post following the over different significantly were Using standard molecular biology techniques, GFP techniques, biology molecular standard Using , Temporal windows of AP-induced suppression of NMDAR-EPSPs at distal dendrite distal at NMDAR-EPSPs of suppression AP-induced of windows Temporal , -150 Pre/post interval (ms)Pre/post NR2A Control</p><p> suppression(dist) NMDAR-EPSP t 42 test). :593). Semliki Forest Virus (SFV) containing this construct was made by made was construct this containing (SFV) Virus Forest Semliki :593). . 871-950 -100 Expression of NR2A of Expression  pre spike pairs for control (–40.4±4.3%; n=18) but not for infected for not but n=18) (–40.4±4.3%; control for pairs spike pre t test). The difference between control and infected cells was cells infected and control between difference The test). -50  6 ), –75 ms (p<0.01), and –100 ms (p<0.05); (p<0.05); ms –100 and (p<0.01), ms –75 ), Science 871-950 871-950 0 blocks NMDAR-EPSP suppression and LTD and suppression NMDAR-EPSP blocks</p><p>299 b</p><p>:1585). Infected cells were identified were cells Infected :1585). Synaptic modification (%) 100 150 50 0 was inserted before a sequence a before inserted was Control LTD NR2A</p><p> Berkeley. Rats Berkeley. -chymotrypsin 871-950 t test. test. b 2  ) , Supplementary Figure 5. Dendritic Ca2+ signal evoked by back-propagating AP. a, Two-photon fluorescence image of a layer 2/3 pyramidal neuron filled with 200 M Oregon Green BAPTA 1. Scale bar: 20 m. b, Upper traces: Fluorescence Ca2+ signals at distal (108 m) and proximal (22 m) dendrites (white boxes in a) in response to single back-propagating APs evoked by somatic current injection, each averaged from 11 to 18 trials. Lower plot: distal vs. proximal Ca2+ signals evoked by single APs, measured by either peak amplitude () or peak area (integral, , in arbitrary units). Each symbol represents data from one cell (n=10); open symbol: with 100 M Oregon Green BAPTA 1; filled symbol: 200 M. c, 4-AP enhances the AP-induced Ca2+ signal at proximal dendrite. Upper traces: Ca2+ signals at proximal dendrite of the cell shown in (a) in the presence of extracellular 4-AP and after washout. Lower plot: AP-induced Ca2+ signals at proximal dendrite after vs. before bath application of 4-AP (n=6).</p><p>Method. Oregon Green BAPTA 1 (100-200 M, Molecular Probes) was loaded into the cell via the somatic whole-cell pipette (loading time before imaging: ~15 min). Imaging was done using a two- photon laser scanning system consisting of an upright microscope (DM LFS, Leica Microsystems, Heidelberg) with 40×W0.8IR objective and a “Tsunami” titanium-sapphire laser pumped by a “Millennia” neodymium yttrium vanadate solid state laser (SpectraPhysics, Mountain View, CA), which provides <100 fs pulses at 890 nm at 80 MHz. Line-scan mode was used to record single AP-induced Ca2+ transients (temporal resolution: 1 ms). At each dendritic location, 5 to 71 sweeps were measured (interleaved between distal and proximal sites), and results were averaged. Change in Ca2+ concentration is given in the change of fluorescence normalized by baseline fluorescence (ΔF/F) after background correction. Data were analyzed in Matlab (MathWorks). a b - + 30 ms</p><p>15 ms Stimulus y t i</p><p> l 7.5 ms i</p><p> b 3 ms a b o r</p><p>Presynaptic P 9 ms spike trains 0 18 ms c 37.5 ms 50 ) % (</p><p>67.5 ms c i n t o i p t a</p><p> a 0 n c i y f i S d</p><p>Postsynaptic o spike train m -50 -100 0 100 50 ms Pre/post interval (ms)</p><p> d</p><p>...... </p><p>Before training</p><p> c 1 1 1 1 i t t h p g i a e n y w 0 0 0 0 S 0 15 30 45 60 75 0 15 30 45 60 75 0 15 30 45 60 75 0 15 30 45 60 75  (ms)  (ms)  (ms)  (ms) After training</p><p> c 1 1 1 1 i t t h p g i a e n y w 0 0 0 0 S 0 15 30 45 60 75 0 15 30 45 60 75 0 15 30 45 60 75 0 15 30 45 60 75  (ms)  (ms)  (ms)  (ms)</p><p>Supplementary Figure 6. Further analyses of the model on location-dependent input selection. a, Sensory stimulus used in simulation and spike trains of example model neurons. Number on the right of each presynaptic spike train indicates time constant of impulse response function of the cell</p><p>(isee Supplementary Methods). To effectively illustrate the relative spike timing, spike trains are shown for a selected period with high-frequency bursts from many presynaptic neurons; the overall mean firing rate of the cells were much lower. b, Cross-correlogram between pre- and postsynaptic spike trains. Warmer colors represent inputs with more transient responses (numbers on the right indicate i). The number of pre/post spike pairs contributing to LTP and LTD can be roughly estimated from the dark and light shaded areas under each curve; dark shading: width of LTP window in c (+=12.5 ms), light shading: width of LTD window (=103.4 ms). c, Single exponential fit of the STDP window measured at distal dendrites. d, Simulation results of a 4- compartment model. Shown are diagram of synaptic inputs (upper), and synaptic weight vs. i before (middle) and after (lower) training. The location dependence of input selection is similar to that found in the 2-compartment model (Fig. 5b).</p>