Figures and data
Paired recordings from synaptically connected layer 2/3 rat and human pyramidal cells.
A Representative reconstructions of electrophysiologically recorded and biocytin filled rat (left, gray soma and dendrites) and human (right, blue soma and dendrites) synaptically connected pyramidal cell pairs. The presynaptic soma and the axon are in red; the postsynaptic dendritic path from the synapse to the soma is highlighted in green. Minimal intersomatic distance was calculated as the sum of the shortest presynaptic axonal (red) and postsynaptic dendritic (green) paths. Boxed region is magnified on the bottom. Scale bars for insets are 20 µm. B Synaptic latency was determined as the time difference between the peak of the presynaptic AP (pink dot) and the onset of the postsynaptic excitatory postsynaptic potential (red dot). Straight lines indicate baseline and rise phase fitting. C Summary of synaptic latencies in rat (red) and human (blue) cell pair recordings. Each dot represents the average latency in a cell measured from the AP peak to EPSP onset as illustrated in panel B. The darker colors represent the paired recordings with full reconstruction. For these data points there was no significant difference between the two species (Mann-Whitney test: P = 0.931). The extended dataset with cell pairs without reconstruction shows no significant difference between the two species (Mann-Whitney test: P = 0.949). D Minimal intersomatic distance of the recorded cell pairs. Intersomatic distance was calculated through every putative synapse and the shortest was taken into account. The minimal intersomatic distance was significantly longer in the human dataset compared to rats (Mann-Whitney test: P = 0.009). **P < 0.01.
Propagation velocity of dendritic and axonal signals in rat and human cortical pyramidal cells.
A Left, Human pyramidal cell simultaneously recorded with a somatic (red pipette) and axonal (green pipette) electrode. Right, Somatic depolarizing current (Isoma) evoked action potentials (Vsoma) and their propagation to the axonal recording site (Vaxon). B Path distances and AP latencies measured between the soma and axon bleb. AP propagation speed measured along the axon showed no significant difference (two sample t test: P = 0.986). All recordings were made at resting membrane potential. C Left, Two-photon image of a rat pyramidal cell recorded simultaneously with a somatic (red pipette) and dendritic (green pipette) electrode. Top, Dendritic stimulation (Idend) with simulated EPSP waveform (Vdend) and somatic response (Vsoma). Bottom, Somatic stimulation (Isoma) triggers an AP (Vsoma) detected in the dendrite as bAP (Vdend). D Left, simulated EPSP propagation speed in rat and human cells (rat: 0.074 ± 0.018 m/s vs. human: 0.093 ± 0.025 m/s, two sample t test: P = 0.004). Top right, simulated EPSP dendritic propagation speed was lower than bAP propagation speed (sEPSP: 0.084 ± 0.023 m/s vs. bAP: 0.337 ± 0.128 m/s, Wilcoxon signed ranks test: P = 1.631×10-9). Bottom right: there was a significant correlation in the forward propagating sEPSP speed and the speed of bAPs. Darker dot is the data for the cell shown on panel C. E Left, Two-photon image and reconstruction of a human pyramidal cell recorded simultaneously with a somatic (red pipette) and dendritic (green pipette) electrode. Right, Somatic current (Isoma) evoked APs (Vsoma) and their backpropagation into the dendritic recording site (Vdend). F Top left, recording distance. Lower left, bAP latency was shorter in human cells (Mann-Whitney test: P=0.005). Right, bAP propagation speed was significantly higher in human dendrites (Mann-Whitney test: P = 6.369×10-6). Darker dot indicate the data for the cell shown on panel E. Scale bars A and C: 10 µm, E: 20 µm.*P<0.05, **P<0.01, ***P<0.001
Contribution of HCN, Ca2+, Na+ and NMDA channels to bAP propagation speed in rat and human dendrites.
A Representative recording from layer 2/3 pyramidal cell of a rat. Two-photon maximum intensity projection image of Alexa 594 and biocytin filled neuron on the left, representative somatic AP (red) and dendritic bAP (green) on the upper right in the control condition (left) and after 20 µM ZD7288 application (right). The light green represents the dendritic signal scaled to the amplitude of the somatic signal for better visibility. Effect of ZD7288 on bAP propagation speed. Darker color represents the example cell. B Same as in panel A but for human cells. C Changes in bAP propagation speeds from control to drug application. The blockage of HCN channels changed bAP speeds more in human compared to the rat (two-sample t test: P = 0.048). The darker colors represent the example cells in panel A and B. D-E Same as A and B but the ACSF contained 1 µM TTX, 200 µM CdCl2, and 20 µM AP5 in the drug application condition. F Comparison of bAP velocities measured in the cocktail of TTX/CdCl2/AP5 blockers reveals higher speed of propagation in human. Black scale bars 20 mV and 0.3 ms. Green scale bars 5 mV on A and B, 2.5 mV on D and E. Scale bars on microphotographs 20 µm. All recordings were done on resting membrane potential. *P < 0.05, **P < 0.01, ***P < 0.001.
Comparative analysis of membrane capacitance and thickness in rat and human cortex
A Representative capacitive transient of a nucleated patch pulled from layer 2/3 neocortical pyramidal cell (black). A single exponential function was fitted on the measured signal (red) for the calculation of the time constant of the membrane. Scale bar: 100 pA, 20 µs. B Differential interference contrast microscope image of a neuronal nucleus. The shortest (a) and longest (b) diameter values were used to calculate the membrane surface. Scale bar 5 µm. C Specific membrane capacitance of rat (red) and human (blue) neocortical pyramidal cells. D Electron micrographs of dendritic membranes used for membrane thickness measurements. Yellow lines indicate measuring profiles. Scale bar 40 nm. Boxed region magnified on the right. The two red dots on the green line show the edges of the membrane (see methods). Inset scale bar 10 nm. E Membrane thickness of rat (red, n = 151 from n = 3 rat) and human (blue, n = 213 from n = 3 patient) neocortical cell dendrites (Mann-Whitney test: P = 0.212).
Dendritic thickness reconstructions and comparison of layer 2/3 pyramidal cells in the human and rat cortex.
A Left, Maximum intensity projection of Alexa594 and biocytin filled human pyramidal cell imaged in two-photon microscope. Right, model of 3D reconstructed apical dendrite. Middle, overlay of the two-photon image and the model. B Apical dendrite thickness measurements on the sample shown in A. Left, The center of the dendrite is tracked by a thick green line while the perpendicular thin lines show measured profiles. Right, Stacked thickness measurements with micrometer scale. C Same as in B with a rat sample. Scale bars 20 µm. D Comparison of rat and human apical dendrite averaged thickness. The mean dendritic diameter of human dendrites was significantly thicker than rat ones (two sample t test: P = 0.019). Darker dots indicate data measured on image stacks shown in panel B and C. E bAP propagation speed correlates significantly with dendrite thickness. Pearson correlation coefficient (r) values for fitted lines are shown on the upper left corner of the plot. The shaded area around the regression line shows the 0-100 % confidence interval for the bootstrapped data. *** P < 0.001.
Modeling explains the enhanced EPSPs velocity in the apical dendrites of human L2/3 PCs.
A Latency and B, velocity of EPSP in models of 5 human (blue) and 4 rat (red) reconstructed L2/3 PCs. Insets show the respective averages for the zoom-in region (box), which brackets the experimental range of dendritic recordings. Note the smaller latency and larger velocity in human PCs. C. Dendritic radius as a function of distance from the soma. Note the larger radius of human dendrites in the outlined region. D,E As A and B, but now distance is normalized in cable units (thus incorporating the diameters differences between cells) and time is normalized in units of membrane time constant. F Sum of radii of basal dendrites as a function of distance from the soma (blue – human, red – rat), in 20µm bins. Dashed lines are the respective averages. G-H As D and E but for ‘hybrid cells’, computed for the 5 modeled human neurons, all having the basal tree of ‘Rat4’ (blue) and for the 4 modeled rat cells, all with the basal tree of ‘Rat4’ (red). Note that the differences in latency and velocity between human and rat diminished (insets). I Two examples of a color-coded “latency-gram”, visualizing the effect of replacing the basal tree of human L2/3 PC with the basal tree of rat L2/3 PC and vice versa. Top left: “Human1” apical tree with basal tree (in black) of “Rat4” PC. Lower left: “Rat4” apical tree with basal tree of “Human1”. Color-coded difference in latency was calculated by subtracting the respective values of the original cells from those calculated for the “hybrid cells”. The blueish apical tree of the human apical tree indicates deceleration whereas the reddish apical tree of the rat PC indicates acceleration of the EPSPs. Inset shows examples of a somatic EPSP’s in these two cases. Shown are the original EPSPs (black lines) and the EPSPs computed for the respective hybrid cases (dashed line in blue – deceleration; dashed red line – acceleration) both for synaptic inputs at 288 μm from soma. Specific cable properties in all cells were: Cm = 1 µF/cm2, Rm = 15,000 Ωcm2, Ra = 150 Ωcm.
Modeling EPSPs latency and velocity in dendrites of human and rat L2/3 pyramidal cells based on experimentally-fitted cable parameters.
A Exemplar modeled (“Human5”) L2/3 PC, also showing the locations of the two recording/stimulating electrodes. B Top (D→S): step hyperpolarizing current (-100 pA) injected to the dendrite of the modeled cell (cyan). Lower trace: Model fit (dark purple line) to the voltage response at the soma (noisy light purple line). The resultant fit to the local dendritic voltage response is also shown (in cyan). Bottom (S→D): as is the case at top, but with current step injected to the soma (purple step current). This fitting procedure resulted with the following passive parameters: Cm = 0.63 µF/cm2, Rm = 15,570 Ωcm2, Ra = 109 Ωcm. C Latency and D Velocity of EPSPs for the 9 model cells as in Figure 6A,B, but now with specific cable parameters fitted to the individual modeled neurons (see Table 1). E-F As in C and D, with distance normalized in cable units and time normalized by the membrane time constant (see Table 2). Note the smaller latency and larger velocity for the human PCs, which is now more significant as compared to the case where the cable parameters were uniform for all modeled cells (compare to Figure 6D and E).
Passive cable parameters fitted to experimental data.
Cm and Rm are the specific membrane capacitance and resistivity, respectively; Ra is the specific axial resistance.
Morphological and cable parameters, and model prediction, of the average EPSPs latency and velocity within experimental range of dendritic recordings per modeled cell.
Cable parameters were fitted per cell as in Table 1. 𝑙𝑎𝑣𝑔 𝑎𝑛𝑑 𝑑𝑎𝑣𝑔 - the average physical distance and diameter respectively from which the respective experiments (per cell) were performed (zoom-in region in Fig. 7C,D). 𝐿𝑎𝑣𝑔 is the respective distances in cable units 
Impact of conductance load of the basal tree on EPSPs velocity and latency.
A Equivalent cable for the apical tree (in blue) and the basal tree (in red) for the 9 L2/3 cells modeled in this study. Note the relatively large conductance load (sink) imposed by the large basal tree in human cells. B EPSP velocity and C latency as a function of distance of the (apical) synapse from the soma. The synapse was located along the “apical” cable (blue cylinder, inset). The respective 5 cases are shown in the inset. Velocity and latency were computed as in Figs. 6 and 7. Note the enhanced velocity and reduced latency for larger basal dendrites. Cable parameters were: Cm = 1 µF/cm2, Rm = 15,000 Ωcm2, Ra = 150 Ωcm. The apical cylinder is of infinite length with diameter of 3 μm; the basal tree (color cables) have linearly increasing diameter (d) and length (L), approximating the increment from rat to human basal trees (Fig 6F): red (l =800 μm, d = 20 μm), yellow (l = 700 μm, d = 18 μm); green (l = 600 μm, d = 16 μm); light blue (l = 500 μm, d = 14 μm); dark blue (l = 400 μm, d = 12 μm). Soma diameter was 20μm in all cases.
Size comparison of layer 2/3 pyramidal cells in the human and rat cortex.
A Sample reconstructions of fully recovered rat and human cortical pyramidal cells. Left horizontal line indicates the location of pia mater. B Comparison of dendritic length, number of nodes, maximum vertical and horizontal extension, and the number of primary dendrites respectively of all reconstructed dendritic arborization. C Comparison of length, number of nodes, maximum vertical and horizontal extension and the number of maximum branch order respectively of the apical dendrites. Boxes represent median and IQR, whiskers represent outlier range (±1.5 IQR); mean is indicated by open square, crosses denote minimum and maximum values. ** denotes significant difference P < 0.01.
Properties of dendro-somatic recording and measured membrane parameters.
A Resting membrane potential of the recorded cells (rat: -70.49 ± 5.78 mV, human: -64.30 ± 7.28 mV, Mann-Whitney U test: P = 7.37× 10-6) were different in human and in rat pyramidal cells. Input resistance of recorded cells (rat: 59.56 ± 21.86 MΩ, human: 71.37 ± 65.48 MΩ, Mann-Whitney U test: P = 0.3466). B Resting membrane potential of recorded cells after ZD7288 application (red, rat control: -70.98 ± 5.04 mV vs rat ZD7288: -72.88 ± 9.75 mV, Wilcoxon signed ranks test: P = 0.40694; blue, human control: -70.43 ± 6.28 mV vs human ZD7288: -75.47 ± 6.991 mV, paired sample t test: P = 0.02682). C The input resistance changed significantly in rat (red, rat control: 86.95 ± 26.34 MΩ vs rat ZD7288: 98.18 ± 28.53 MΩ, paired sample t test: P = 0.00488) and human (blue, human control: 54.38 ± 28.8 MΩ vs human ZD7288: 70.21 ± 26.09 MΩ, paired sample t test: P = 0.02434) after the application of 20 µM ZD7288. D Effect of voltage gated ion channel blockage on bAP amplitude. The amplitudes of the bAPs were significantly decreased upon the application of voltage gated ion channel blockers (rat control: 46.32 ± 25.78 mV vs rat TTX, CdCl2, AP5: 6.26 ± 3.47 mV, paired sample t test: P = 0.00188, human control: 51.95 ± 22.81 mV vs. human TTX, CdCl2, AP5: 7.52 ± 2.84 mV, Wilcoxon signed ranks test: P = 0.0156).
Latencies and propagation speed measured at different points of the propagating waveforms.
A The presynaptic AP peak and EPSP latency were measured at different points. Left: latency at onset, middle: latency at half amplitude, right: latency at EPSP peak. B: Same as A but for bAP speed values. C Same as A but for AP axonal speed values. D Upper: Same as A but for sEPSP speed values. Lower: comparison of sEPSP and bAP speed. E: Pharmacological experiments with ZD7288. F: Same as E but for a cocktail of TTX, CdCl2, and AP5.
Velocity of EPSP peak as a function of distance from the synapse input site for the case of an infinite passive cylindrical cable with sealed end at the recording site (X = 0).
Note the high velocity of the EPSP peak when the synapse is near the recording site; the velocity converges to 2λ/τ for electrotonically distant synapses (horizontal dotted line). Cyan and red vertical lines show the maximal mean cable distance L_max (Table 2) measured experimentally in human and in rat neurons. Cable parameters and diameter are as in Table1 and Table 2 respectively. Note that because, on average, the location of the experimentally-recorded human synapses is closer (in cable units) to the recording site (“soma”), the EPSP velocity in human falls at a higher velocity compared to that of the rat.
Morphological irregularities affect EPSP latency and velocity.
A Cable with a single branch, with symmetrical (top left) or asymmetrical (top right) branches. Thick branches diameter is 4µm, while thin branches’ diameter is 1µm. Latency and velocity were calculated as explained in the text and in Figs. 7 and 8; synaptic inputs were activated at different sites along the structure. The recording site (“soma”) is at left (dark blue rectangle), with diameter of 13µm. B As in A, with normalized space and time constants. For symmetrical branches, both latency and velocity overlap for the two branches (left column in both A and B), while in asymmetrical case, the latency from the thick branch is smaller as it is electrotonically closer to the soma and, therefore, for the same physical distance the initial velocity of the EPSP at its site of origin is larger (right column in B, red branch compared with green). However, there is a small increase in latency (decrease in velocity at these daughter branches) due to local impedance mismatch. C. Cable with diameters replicating the apical main-branch of ‘Human2’ (left column) and ‘Rat1’ (right column) PCs. Note the local irregularities shifts the velocity above (left column) or below (right column) 2λ/𝜏 despite having identical lengths across all sections. Moreover, velocity pattern changes due to the proximity of the synapse to the soma, as a function of the cable diameters. Cable parameters are identical for all morphologies (Cm = 1.5 µF/cm2, Rm = 10,000 Ωcm2, Ra = 150 Ωcm).
Morphology of the nine modeled cells.
Each dendritic branch is marked by a different color.
Quantifying the effect of switching the basal tree between rat and human (and vice versa-the ‘hybrid cells’ on mean latency, uniform parameters.
Average latency as a result of using each of the nine modeled cells basal trees as the basal tree of all other cells (e.g. a “hybrid cell”), compared with original models latencies (e.g. “Uniform”). Average latency was calculated similar to Suppl. Table 2-4 Note the acceleration due to most of the human basal trees versus the deceleration due to rat basal trees. Mann-Whitney U test. * denotes significant difference P < 0.05.
‘hybrid cells’ effect on latency and velocity for the experimentally-fitted cable parameters. A,B
Same as Fig 7E,F but for ‘hybrid cells’, computed for the 5 human neurons, all having the basal tree of ‘Rat4’ (in blue) and for the 4 rat cells, all with the basal tree of ‘Rat1’ (in red). Note that the differences in latency and velocity between human and rat were diminished (insets).
Paired recordings EPSP latency distributions.
A EPSP latency distributions from all the cell pairs shown in Fig. 1. B EPSP latency distributions for the fully reconstructed cell pairs. Blue: human cell pairs, red: rat cell pairs. Each dot represents a latency value measured on a single sweep.
Comparison of sEPSP and EPSP features.
Each dot represents the mean of all the recorded values on individual trials for a given cell. Blue: human, red: rat. The example cell in Figure 2 is highlighted with darker red, to give an intuition of how representative it is.
Effect of dendritic branching points on signal propagation velocity.
A Dendritic branching point counts between the putative synapse and the soma of the postsynaptic cells of the fully reconstructed cell pairs. We could not find significant correlation between synaptic latency and branching point counts (Red: rat, Blue: human). B Branching point counts between the dendritic recording site and the soma during sEPSP recordings. We could not find significant correlation between branching point count and sEPSP propagation speed. C Branching point counts between the dendritic recording site and the soma during bAP recordings. We found a significant correlation between branching point count and bAP propagation speed in the rat dataset (red) but not in the human dataset (blue). D Simulation of the effect of a branching point on the signal propagation velocity. Adding a branch point (yellow versus red, marked with a circle) to the dendrite did not affect the velocity and the latency of the simulated signal. E Same as D but for cable units.
Effect of series resistance of the dendritic electrode on measurement of EPSP latency.
A. Top: simulated EPSPs in Human 1 neuron as recorded at the injected point in the apical dendrite, located 150 m from the soma. Simulated synaptic current is shown by the dashed line. Bottom: the resultant EPSP at the soma. Simulation was performed for a range of series resistance (Rs) values (shown at right). B. As in A but for Rat 3 neuron. C. EPSPs latency as a function of Rs for the 9 modeled neurons. Electrode capacitance was 6pF with variable series resistance, Rs.
‘Hybrid cells’ effect on latency and velocity for the experimentally-uniform vs fitted cable parameters.
Values as in Table 2 and Suppl.Tables 2-4. Note that the differences in latency and velocity between human and rat are significant in fitted and uniform parameters and diminished in “Hybrid” case. * denotes significant difference P < 0.05. Mann-Whitney test, with Bonferroni correction.
Basal load effect.
Significant difference between human and rat’s basal load, despite no significant difference in real and cable distance of basal tree.
Passive cable parameters fitted to experimental data.
Cm and Rm are the specific membrane capacitance and resistivity, respectively; Ra is the specific axial resistance. Note the similar time-constant between human and rat.
SHAP analysis feature importance result.
Cumulative SHAP values (main and interaction effects) for the top five variables influencing signal propagation latency. The model used for this analysis is the Gradient Tree Boosting model.
Morphological scaling factors due to fixation.
Model prediction of the average EPSPs latency within experimental recording distance range per modeled cell for the case of identical cable parameters for all cells.
𝑙𝑎𝑣𝑔 is the average physical distance from which the respective experiments (per cell) were performed (zoom-in region in Fig. 6A,B). 𝑑𝑎𝑣𝑔 is the average diameter at the same range. 𝐿𝑎𝑣𝑔 is the respective distances in cable units 
Model prediction of the average EPSPs latency within experimental recording distance range per modeled cell for the case of identical cable parameters and “hybrid cell” whereby all modeled cells consist of the basal tree of “Rat4”.
𝑙𝑎𝑣𝑔 is the average physical distance from which the respective experiments (per cell) were performed (zoom-in region in Fig. 6E,F). 𝑑𝑎𝑣𝑔 is the average diameter at the same range. 𝐿𝑎𝑣𝑔 is the respective distances in cable units 
Model prediction of the average EPSPs latency within experimental recording distance range per modeled cell for the case of fitted cable parameters per cell and “hybrid cell” where all modeled cells consist of “Rat4” basal tree.
𝑙𝑚𝑎𝑥 is the maximal physical distance from which the respective experiments (per cell) were performed (zoom-in region in Fig. 8S A,B). 𝑑𝑚𝑎𝑥 is the (average) diameter at 𝑙𝑚𝑎𝑥 . 𝐿𝑚𝑎𝑥 is the respective distances in cable units 
Examining factors influencing EPSP latency via GLM model without interaction terms.
Note the ranking (1-5) of the factors affecting EPSP latency (in ms units). Factors were ranked (1-5) based on the magnitude of their absolute coefficients, which provide insight into their relative contribution to the model. The model was fit using the Gamma family and included continuous factors that were standardized prior to fitting, as well as categorical factors (see comments above for the reference values). Each factor used in the model’s formula is highlighted in bold. The formula used for the model: latency ∼ species + distance + hybrid + radius + fitted.
Examining factors influencing EPSP latency via GLM model with interaction terms.
This table presents evidence of significant interactions between the main factors affecting EPSP latency (see Suppl. Table 5 for the non-interaction model). The model was fit using the Gamma family and included continuous factors that were standardized prior to fitting, as well as categorical factors (see Supp. Table 5 for the reference value). The formula used for the model (names match the factors from Suppl. Table 5): latency ∼ species + distance + radius + (species×fitted×distance) + (species×hybrid×distance) + (species×radius).
