Cell reconstructions reveal expanded dendritic size and compartmentalization of human PCs.

A-B) Manually reconstructed human and mouse PCs (same scale). C) Meta-analysis of historical measures of mean±sd total dendritic length in human (dark grey) and mouse (light grey) compared with that reported here for human (blue) and mouse (green) (values in D). D-I) Total dendritic length, number of branches, maximum terminal branch order, maximum width and height of the dendritic arbor, and numerical index for the shape of the dendritic area by morphological category (N—Normative, S—Split, P—Poly; positive values indicate arbors with greater width than height; n = 5,5 cells, 3,2 individuals). Cells pictured in A-B are highlighted by colored outlines according to morphology (Normative=Yellow, Split=Green, Poly=Purple). J) Schematic calulation of branch segment eccentricity relative to cortical depth and the PC layer (PCL) plane (top). Eccentricity data for all branches on quarter-radial plots (middle; black arrows are population mean) and as distributions with cell averages (bottom) by categories of distance from the soma (proximal, intermediate and distal thirds of ML thickness) in human and mouse (ANOVA, Tukey’s HSD post-hoc; n = 2749,7595,2541 and 877,1247,1056 branches). K) Maximum diameter of the primary dendrite (or dendrites for Poly PCs). L) Dendrite diameter normalized to the thickest primary dendrite by order. M) Distribution of all branch diameters. N) Cumulative distribution of branch segments by their diameter relative to the thickest dendrite. (Kolmogorov-Smirnov test; n = 5,5 cells). O) Percent of the total dendritic length classified as thin caliber using a 1.31um threshold (log-normal mean + 1sd of all branch thicknesses across species (Student’s t-test; n = 5,5 cells). P) Branch segment lengths (Mann-Whitney U test; n = 12885,3180 branches). Q) Cumulative (top) and non-cumulative (bottom) normalized distributions of dendritic length by centrifugal Sholl distance from the soma (Kolmogorov-Smirnov test, n = 5,5 cells). R) Total dendritic length at each 1μm Sholl radius without normalization.

Human PCs host expanded sites for putative input and spines with complex morphology.

A) Example 3D reconstructions of human and mouse spiny dendrites (background removed for clarity) with inset schematic of spine types: thin (solid arrowhead), mushroom (open arrowhead), branched (back-to-back arrowheads), cluster (star). B) Density of all spines (left) and by spine type in human and mouse (n = 15,12 branches). C) Spine densities by branch location relative to the soma (Student’s paired one way t-test for promixal vs distal compartments; n = 5,4 cells) D) Spine head diameters of thin and mushroom spines (Student’s t-test; n = 2066,1246 spines). E) Spine neck length by species (Student’s t-test; n = 2323,1376 spines). F) Ratio of spine head volume to volume surrounding the dendrite (Student’s t-test; n = 15,12 branches). G) Spine head diameter by branch location (ANOVA, Tukey’s HSD post-hoc; n = 345,478,424 spines in human, 676,767,812 in mouse). H) Spine to surround ratios by branch location (n = 4,4 branches). I) Example spine cluster with enlarged neck diameter and spine head. J) Spine cluster density, included in the total density in (B, left). K) Spine cluster density by branch location, included in the total density in (C, left) (Student’s one way t-test for promixal vs distal compartments; n=5 individuals). L) Number of puncta per cluster by branch location (n = 46,67,76 spines). M) Spine cluster head volume by branch location (n = 46,67,76 spines). N) Spine cluster diameter as a function of puncta number (n = 189 spines). Blue points are means by puncta number. O) Total puncta across spine clusters on each branch segment by location and normalized to the proximal branch (Student’s one way t-test for promixal vs distal compartments; n = 5 individuals).

Peripherin and calbindin dual-labeling reveals non-canonical CF multi-innervation of adult human PCs.

A) Example reconstruction from human of a PC and peripherin-labeled fiber as originally traced (left) and with masks drawn for visualization (right) to exemplify putative mono-innervation. B) Example PC and peripherin-fiber masks to exemplify absence multi-innervation. C) Example of putative multi-innervation. Examples of fully labeled multi-innervation. E) A Poly PC with multiple peripherin fibers of varying thickness approaching separate dendrite compartments from distinct locations in the granule cell layer. Untraced composite PC and peripherin-fiber images (top, center) with separate (top, outside) and combined (bottom) masks of each primary dendrite. F-G) Distribution of cell types (F) and orientations (G) by peripherin fiber classification (n = 2 individuals, 44 cells). Numbers above bars indicate absolute counts.

Regional and locally clustered PC demographics in the vermis of human and mouse.

A) Example exhaustive reconstruction of PC morphological distributions in a parasagittal section of human vermis. B) Morphological orientation (top) and type (bottom) demographics across lobules (Chi-squared test; n = 3 individuals, 6346 cells). C) Example exhaustive reconstruction of a parasagittal section from mouse vermis. D) Morphological type (top) and orientation (bottom) demographics across lobules in mouse (Chi-squared test; n = 3 individuals, 2284 cells). E) Observed and shuffled rates of adjacent PC clustering in human and mouse (within species ANOVA, Tukey’s HSD post-hoc; n = 3 individuals and 20 shuffles of each). Numbers between graphs indicate the difference between observed and shuffled mean. * p < 0.05, ** p < 0.01, *** p < 0.005.

Adjacent and local non-random cell type clustering in human and mouse vermis.

A) Schematic of cluster score assignment based on complete morphological match (left) and score use to identify the number of cells and parasagittal length of clusters (right). B-C) Number of cells (B) and length (C) of PC clusters in mouse and human compared to shuffled data (Mann-Whitney U test; n = 288,1295 observed clusters; n = 263,1122 shuffled clusters). D) Schematic of local population demographics measured over variable distances (top). Radii in mouse are 20% the length in human, matching the difference in dendritic width. Rates of matching (dark points) and non-matching (light points) morphologies in a shuffled population are subtracted from unshuffled data to measure the percent elevation of clustering in human and mouse. Points mark the average demographic difference between observed and shuffled rates across all five cell groups (i.e. one dark point represents the average for matching cell rates of Normative to Normative, Split to Split, etc.). E) To measure the absolute spatial scale of elevated PC clustering, we measure the observed vs shuffled rate for 500μm increments of a shell region around a growing core. By ignoring the morphologies of the core region, we exclude local clustering from measurements of distant clustering.

Inter-hemisphere similarity of PC demographics is congruent with functional lateralization.

A) Schematic of human cerebellum. B) Parasagittal reconstructions of PC morphological distributions in left and right mid-hemisphere lobules L5-Crus II within the same individual. C) Morphological demographics across lobules in left (top) and right (bottom) hemispheres by individual. D) Absolute differences of morphological demographics across hemispheres by lobule and individual. E) Normalized mean inter-hemisphere demographic difference from L6 by lobule and individual (Student’s one-way t-test; n = 3 individuals).

All manual cell reconstructions.

A) Example confocal tile scan (top left) and resolution of each individual image (below) with manually reconstructed human PCs. B) Manually reconstructed mouse PCs. Cells throughout are at the same scale.

Additional morphological data from digital reconstructions.

A) Number of branch points by species. B) Maximum number of branch points, regardless of whether the branching is symmetrical or reduces diameter such as would define a change in dendrite caliber, between the terminal segments and the soma. C) Number of primary dendrite compartments, defined as the number of thick dendritic sections giving rise to only thin caliber branches, by the morphological category of each reconstructed PC. D) Distribution of branch eccentricities as a function of branching order for each reconstructed PC. E) Somatic diameters. F) Mean branch diameter by order. G) Distribution of total branch segments in each branching order. A reference symmetrical fractal pattern is represented by the function y=2x+1 (black line). The right panel zooms in on the early branching orders. H) Distribution of mean dendritic length across branches of each order. I) As in (F) but normalizing for maximum branch order. J) Distribution of terminal branch segment lengths (Mann-Whitney U test; n = 6883,1773 branches). K) Branch distances from the soma (Mann-Whitney U test; n = 12885,3180 branches). L) Distribution of branch segments by branch order as in Fig. 1M but without normalizing segment number across species. M) Distribution of branch points by normalized Sholl distance from the soma.

Additional data from digitally reconstructed spines.

A-B) Example 3D reconstructions of mouse (A) and human (B) spiny dendritic branches. Images in B1 are from one 93yo specimen while images in B2 are from one 37yo. All images on the same scale. C) Diameter of all dendritic banches reconstructed in this study. Lines connect the compartments across each cell (n = 4,5 individuals). D) Distributions of spine neck lengths by location and species. E) Distribution of spine neck lengths by spine type and species. F) Spine protrusion distance, measured from the edge of the dendritic shaft to the distal tip of the spine head, in human and mouse (ANOVA, Tukey’s HSD; n = 720,818,862 spines, n = 376,528,476 spines). G) Spine head diameter by spine type and location in mouse (top) and human (bottom). H) Spine head volume by spine type and location in mouse (top) and human (bottom).

Human PCs with multi-branched and horizontally oriented morphology can avail themselves of additional branch specific capacities.

A) Example masks from a digitally reconstructed PC (grey) and peripherin fiber (blue) on which two unconnected calbindin(+) PC axons (dark and light green) converge to innervate distinct dendritic compartments. B) Example z-projections of two horizontally oriented Poly PCs hosting axons that emerge from the dendrite or an intermediate zone between somatic and dendritic compartments (orange arrowheads).

Additional reconstructions and analysis of PC demographics in parasagittal slices of vermis from human and mouse.

A-B) Parasagittal human and mouse vermis reconstructions demonstrating the spatial distributions of each morphological type. C) Schematic of foliar subdivisions into gyrus, bank, and sulcus (top) and the demographics of PC types by foliar area (bottom). D) Demographics of PC orientations by foliar area. D) Observed versus shuffled rate of PCs having the same morphologyical classification by cell type (as if Fig. 4E) when the shuffled data are constrained within foliar subdivision. Numbers between graphs indicate the difference between observed and shuffled mean. F-G) Matrices of all cell type-by-cell type clustering likelihoods observed in the original data (top) and in the shuffled dataset (bottom) and separated by foliar area in human and mouse (N—Normative, S—Vertical Split, P—Vertical Poly, SS—Horizontal Split, SP— Horizontal Poly). * p < 0.05, ** p < 0.01, *** p < 0.005.

Adjacent PC demographics reveal the spatial scale of non-random morphological clustering among local PCs.

A) Schematic of cluster score assignment for each PC based on whether the morphology of adjacent cells (<1000um away) matches (left) and the use of these scores to identify the number of PCs and parasagittal length of clusters of redundant cell types (right). B) Number of cells (top) and length (bottom) of PC clusters with identical morphology (complete match with five cell categories: Normative, vertical Split, vertical Poly, horizontal Split, horizontal Poly) in mouse and human with that of shuffled data in which morphologies are randomly resorted while cell locatiuon is held constant. C-D) As in (B), but for PC clusters with only a partial match based on cell type (Normative, Split, or Poly; for example, vertical and horizontal Split PCs are a match) or cell orientation (Vertical or Horizontal; for example, horizontal Split PCs and horizontal Poly PCs are a match). E-G) As in (B-D) for PCs in slices from mid-hemisphere.

Population PC demographics reveal the spatial scale of non-random morphological clustering among parasagittal PC circuits.

A) Schematic of how various radius distances are used to survey the demographics of all cells within a given distance from a PC (top). Radii used in the mouse are set to 20% that of the human, matching the difference in the width of a mouse vs human PC. The rate of PCs with matching morphology (using a complete match across five categories) is compared with the same measurement of shuffled data to generate a difference between the observed vs shuffled percents of cells with a matching morphology (dark points) compared to non-matching morphologies (light points) in human and mouse (bottom). Points mark the average demographic difference between observed and shuffled rates across all five cell groups (i.e. one dark point represents the average for matching cell rates of Normative to Normative, Split to Split, etc.). B) To measure the absolute spatial scale of elevated PC clustering, we measure the observed vs shuffled rate for 500um increments of a shell region around a growing core. By ignoring the morphologies of the core region, we control for the effect of highly local clustering on the measurement of more distant PCs. This reveals that non-random morphological clustering appears to drop off at distances of ∼1500um and 300um in human and mouse, respectively.

Additional reconstructions and analysis of PC demographics in parasagittal slices of the mid-hemisphere of human.

A-B) Parasagittal reconstructions of lobules 5-Crus II of the opposing hemispheres of human specimens used previously, demonstrating the spatial distributions of each morphological type. C) Distribution of PCs across lobules of the mid-hemisphere by cell orientation (as in Fig. 4B). D) Observed versus shuffled rate of adjacent PCs having the same morphologyical classification by cell type in human (top) and mouse (bottom). Numbers between graphs indicate the difference between observed and shuffled mean. E) As in (D), but with shuffled morphologies being constrained within foliar subdivision (e.g. PCs in the sulcus are only shuffled among each other, not also with PCs in the bank or gyrus). F-G) Matrices of all cell type by cell type clustering likelihoods observed in the original data (top) and in the shuffled dataset (bottom) and separated by foliar area in human and mouse (N—Normative, S—Vertical Split, P—Vertical Poly, SS—Horizontal Split, SP—Horizontal Poly). * p < 0.05, ** p < 0.01, *** p < 0.005.

Human cerebellar regional borders as defined by morphological demographics or functional imaging.

The top and bottom rows provide four regional sub-divisions according to either involvement in behavioral tasks49 or PC morphological demographics. The middle row indicates the lobules corresponding to both the above functional and below morphological sub-divisions.