Fig 1.
Examples of human CA1 pyramidal reconstructions that were cut in the same plane during tissue sectioning.
Example of a 3D-reconstructed human CA1 pyramidal cell shown on the XY A, and YZ B, planes, to illustrate that, due to technical limitations, part of the dendritic arbour closest to the surface of the slice from which the cell soma is injected (typically at a depth of ∼ 30μm from the surface) is lost. Axon, main apical, collateral and basal dendrites are shown in green, black, blue and orange, respectively. Scale bar (in panel B) = 100μm. C, Three human CA1 pyramidal neuron reconstructions (yellow, pink and blue) from the same preparation viewed from the side. Raw data from [26].
Fig 2.
Reproduction of biological regrowth of severed class IV Drosophila neurons.
A, The impact of the balancing factor bf demonstrated by repairing a 2D artificially created morphology (left with cut dendrites in magenta). Repairs with different balancing factors on the right (repaired dendrites in green). bf = 0 favours the minimisation of the overall cable length whereas bf = 1 optimises the direct path length towards the soma. B and C, Left, Reference Drosophila larva class IV morphology in which the branches that will be severed deliberately are marked in magenta. B, Right, Example of repaired dendrite where invasion has occurred from adjacent branches marked (green). C, Right, Sample repair where the severed branch regenerated from the cut end (green). D, Morphological statistics of the regrown dendrites from B and C (green) and 498 other random cuts. The repaired morphologies were compared to the original reference neuron (black + magenta in B and C) shown here as the black dashed line, as well as to the cut dendrites (magenta). The root mean square errors (RMSE) indicate the deviation from the reference value as a percentage of the reference value. The examples shown in B and C are represented by the darker square data points. E, Average Sholl distribution for the cut and repaired morphologies from D with standard deviation as shaded area. The reference distribution is shown in black. F, Histogram for 500 regrown dendrites using our repair function, showing the percentage of regenerated branches. G, Percentage of regenerated branches as a function of the size of the removed branch. The points show the actual data as an overlay to the binned normalised histograms of the data points. The bin boundaries are indicated by the black vertical lines. H, Histograms for 500 regrown morphologies as in F but for a higher number of target points N and a higher balancing factor bf.
Fig 3.
Repair algorithm successfully restores removed dendrites of different neuron types with high, low and intermediate balancing factors.
A Left, Reference dentate granule cell morphology [Morphology from 60] with cut dendrites in magenta, Right, Repaired morphology with restored dendrites in green. The area enclosed by the dashed black line indicates the volume into which the dendrite has grown. B, Morphological statistics of 20 different cuts and repairs of the neuron shown in A. Total number of branches (Top left), Total dendritic length (Top right) and Mean segment length (Bottom left). The RMSE shows the deviation from the reference value as a percentage of the reference value. C, Average Sholl distribution for the cut and repaired morphologies from B with standard deviation as shaded area. The reference distribution is shown in black. D, Histogram of 500 regrown morphologies using our repair function fix_tree, with the percentage of the repair regrowing from the cut branch similar to Fig 2F. E-H, Same layout as in A-D but using a repaired mouse cerebellar Purkinje cell [Morphology from 61].
Fig 4.
A new software tool for the repair of dendrites with a graphical user interface (GUI).
Top, Example screenshot of fix_tree_UI (Neuron Repair Graphical User Interface). The numbers 1–8 represent the steps of successfully uploading a morphology and background image stack and repairing a missing region. Bottom, Showcase of the output of fix_tree_UI with the repaired neuron and two example statistics (the output contains more statistics than shown). Step one appears when launching the GUI asking the user to upload a morphology and step two can be initiated by clicking the “Load stack” button.
Fig 5.
The repair algorithm successfully recovered artificially removed dendrites from mouse CA1 pyramidal cells and restored their Sholl profiles.
A, Three example repairs of apical and basal dendrites of mouse CA1 pyramidal neurons [Reconstructions from 26]. For each repair, the left morphology is the reconstructed reference with cut branches in magenta and the right morphology is the repaired tree with regrown branches in green. The area enclosed by the dashed black line represents the 3D volume into which the artificial dendrites grew, corresponding to the convex hull of the severed dendrites (see Methods). The graphs below each repair show the distributions of Sholl intersections for the Cut, Repaired and Reference morphologies. B, Each graph shows the value of the repaired morphology (green dots) plotted against the value of the original morphology in black on the identity line. For comparison, the data points in magenta show the values for the cut morphologies. Top, left to right Total number of branch points, Total dendritic length, Average dendritic length per segment in the apical dendrite (one segment is measured from one branch point to the next). Bottom, left to right Average dendritic length per segment in the basal dendrite, Average diameter per segment in the apical dendrite, Average diameter per segment in the basal dendrite. The root mean square errors (RMSE) show the deviation from the reference values as a percentage of the reference value.
Fig 6.
Growth algorithm extends incomplete human CA1 pyramidal cell morphologies.
A, Confocal microscope image of the human hippocampal CA1 region (DG: dentate gyrus; SLM: stratum lacunosum moleculare; SR: stratum radiatum; SP: stratum pyramidale; SO: stratum oriens) with stained pyramidal cells and ROI (region of interest). B, Morphology reconstructions. C, ROI enlarged from A. D, ROI with overlays of originally reconstructed pyramidal cell morphologies by Ref. [26], which are incomplete due to experimental limitations (see text). Scale bar = 460 μm in A, B. E, ROI showing morphologies from D that have been artificially extended in the apical and basal arbour, showing plausible completion of incomplete dendrites based on known layer-specific target growth regions. Each individual neuron has been given a different colour to distinguish the morphologies.
Fig 7.
The repair algorithm restores the electrophysiological behaviour of cut and repaired mouse pyramidal cells and allows for better predictions of neuronal function in human neurons.
A, CA1 pyramidal cell of the mouse. Left, Reference morphology, Middle, Repaired morphology with growth volume indicated by the black dashed line. Cut dendritic sections in magenta, repaired dendritic sections in green. Right, Somatic voltage traces induced by current injections in the soma of reference (black), cut (magenta) and repaired (green) morphology with resting membrane potentials (Top) and current clamp increments (Bottom). The inset on the left shows the F-I curves of the data on the right (same colour scheme). B, Human CA1 pyramidal cell. Same arrangement as in A. Repair restores the electrophysiological behaviour of the reference neuron. C, Prediction of the electrophysiological behaviour of an extended human CA1 pyramidal cell. Arrangement as in A but the reference morphology on the left is the full but incomplete reconstruction as provided in Ref. [26], which has been extended using the repair algorithm (c.f. Fig 6).
Fig 8.
Repairs restore the detailed electrophysiological behaviour of neurons.
Top, Example voltage traces of intentionally cut and subsequently repaired mouse CA1 pyramidal neuron current clamp simulations using the model by [65, 66] (neuron data by [68]). Reference in black, cut in magenta, repair in green (same colour scheme for the entire figure). A zoomed in version for the spiking traces is shown below for clearer visibility (zoom shows the beginning of the traces where spiking commences). Middle (left to right), Firing frequency, sag voltage and half width of the first spike plotted against stimulation current increments. Bottom (left to right), Interspike interval between the first and second spike, spike adaptation index and difference between the peak of the first and second spike plotted against stimulation current increments.
Fig 9.
Repairing neuronal dendrites is likely to improve simulations of NMDA spikes, which are reduced in extended human neurons compared to mice.
A, Mouse CA1 pyramidal cell with basal dendrites in blue. Stimulation and recording sites are indicated on the basal dendrite. Right, Human and human extended morphology with basal dendrites in green and black with the growth volumes indicated by the black dashed lines. The human extended version was created by extending the human reconstuction in A, Middle. (continued) B, Example dendritic NMDA spikes for a mouse (blue), human (black) and human extended (green) morphology at 85.19% of the maximum possible Euclidean distance in the basal tree away from the soma for each morphology, respectively. C, Peak NMDA spike voltage measured for different numbers of synapses at different distances from the soma in the basal dendrite, given as a percentage of the maximum possible distance in the basal tree (colour scheme as in B). For each distance, 10 different locations at that distance were tested (transparent dashed coloured lines). The average is shown as a solid line. The synapses were distributed over 20μm sections. D, Dendritic diameters for the locations described in C, with mean and standard deviation.