Dendritic Spine Quantification Using an Automatic Three-Dimensional Neuron Reconstruction Software

Kevin M. Keary III; Ellen Sojka; Melissa Gonzalez; Zheng Li

doi:10.3791/66493

JoVE Journal > Neuroscience

Neurociencias

Dendritic Spine Quantification Using an Automatic Three-Dimensional Neuron Reconstruction Software

Published: September 27, 2024

doi:

10.3791/66493

Kevin M. Keary III¹, Ellen Sojka^1,2, Melissa Gonzalez^1,2, Zheng Li¹

¹Section on Synapse Development Plasticity, National Institute of Mental Health,National Institutes of Health, ²Colgate University

Summary

Dendritic spines are post-synaptic compartments of most excitatory synapses. Alterations to dendritic spine morphology occur during neurodevelopment, aging, learning, and many neurological and psychiatric disorders, underscoring the importance of reliable dendritic spine analysis. This protocol describes quantifying dendritic spine morphology accurately and reproducibly using automatic three-dimensional neuron reconstruction software.

Abstract

Synaptic connections allow for the exchange and processing of information between neurons. The post-synaptic site of excitatory synapses is often formed on dendritic spines. Dendritic spines are structures of great interest in research centered around synaptic plasticity, neurodevelopment, and neurological and psychiatric disorders. Dendritic spines undergo structural modifications during their lifespan, with properties such as total spine number, dendritic spine size, and morphologically defined subtype altering in response to different processes. Delineating the molecular mechanisms regulating these structural alterations of dendritic spines relies on morphological measurement. This mandates accurate and reproducible dendritic spine analysis to provide experimental evidence. The present study outlines a detailed protocol for dendritic spine quantification and classification using Neurolucida 360 (automatic three-dimensional neuron reconstruction software). This protocol allows for the determination of key dendritic spine properties such as total spine density, spine head volume, and classification into spine subtypes thus enabling effective analysis of dendritic spine structural phenotypes.

Introduction

Dendritic spines are protrusions of dendrites often comprising the post-synaptic site of glutamatergic synapses¹^,². Dendritic spines are of particular interest in the field of synaptic plasticity. Spines are often altered when synaptic strength changes, becoming larger and stronger in long-term synaptic potentiation or smaller and weaker in long-term synaptic depression³^,⁴^,⁵^,⁶^,⁷. Beyond synaptic plasticity, the profile of dendritic spines changes throughout the lifespan. In early development, there is a period of dendritic spine formation and growth, followed by dendritic spine pruning until reaching a steady state⁸^,⁹^,¹⁰. In the aging brain, spine loss accompanies brain shrinkage and cognitive decline¹¹. Additionally, many neurological, neurodegenerative, and psychiatric disorders are characterized by aberrant dendritic spines. Multiple brain regions in individuals affected with schizophrenia have fewer dendritic spines, likely resulting from altered synaptic pruning¹². Autism spectrum disorders are also characterized by dendritic spine pathologies¹³. Dendritic spine loss is a hallmark of both Alzheimer's and Parkinson's disease¹⁴^,¹⁵. Given the wide array of research topics encompassing investigations into dendritic spine properties, techniques for accurate spine quantification are of paramount importance.

Staining, i.e., the Golgi method, or labeling neurons via dye filling or expressing fluorescent proteins are common methods for dendritic spine visualization¹⁶^,¹⁷^,¹⁸. Once visualized, spines can be analyzed with a variety of free and commercially available software clients. The desired output of the analysis is an important factor in determining which software will be of the most use. Fiji is a viable software option for questions centered around dendritic spine density. However, this technique largely relies on time consuming manual counting that can introduce the potential for bias. New plugins such as SpineJ allow for automatic quantification, additionally allowing for more accurate spine neck analysis¹⁹. A drawback of these approaches is the loss of a three-dimensional analysis for determining spine volume, as SpineJ is limited to two-dimensional image stacks. Additionally, obtaining spine subtype information becomes challenging via these processes. The four predominant spine subtypes, thin, mushroom, stubby, and filopodia, all connote individual functions and are largely classified via morphology²⁰. Thin spines are characterized by an elongated neck and defined head²¹. Mushroom spines have a much larger and pronounced spine head²². Stubby spines are short and have little variance between head and neck²³. Filopodia are immature spines with a long, thin neck and no obviously observable head²⁴. While classification provides valuable information, spines exist on a continuum of dimensions. Classification into categories is based on ranges of morphological measurements²⁵^,²⁶. Manually measuring spines for classification compounds the logistical burden for researchers in this approach.

Other software options focusing specifically on three-dimensional dendritic spine analysis are better suited for investigations into spine volume and subtype properties²⁷^,²⁸^,²⁹^,³⁰^,³¹. Despite the difficulty presented by three-dimensional analysis, such as poor z-plane resolution and smear, these software options allow for reliable three-dimensional reconstruction of dendrites and dendritic spines in a user-guided semi-automated fashion. Automatic classification of identified spines into their subtypes is also a feature present in some of these spine analysis software packages. This can ameliorate concerns of potential workload and experimental bias. Neurolucida 360 is one commercially available software allowing for reliable and reproducible three-dimensional dendritic spine identification and classification³². Here, we present a comprehensive protocol to effectively prepare fixed tissue, acquire images, and ultimately quantify and classify dendritic spines using this software.

Protocol

All animal procedures followed the US National Institutes of Health Guidelines Using Animals in Intramural Research and were approved by the National Institute of Mental Health Animal Care and Use Committee.

1. Preparation of fixed hippocampal slices

Anesthetize mice with an intraperitoneal injection of Ketamine/Xylazine (Ketamine: 100 mg/kg; Xylazine: 8 mg/kg). Validate anesthesia via tail pinch and affix mouse to perfusion plate.
Using large surgical scissors, remove the skin and fur from the chest, allowing for easier visualization of the underlying ribcage.
Make a horizontal cut below the width of the ribcage, avoiding the liver and diaphragm. Using fine forceps, pull the xiphoid process up and cut each lateral side of the rib cage. Flip up the rib cage to the neck region and clamp in place using hemostat forceps.
Insert a 21 G butterfly needle into the left ventricle of the heart and begin to perfuse with room temperature 1x PBS at approximately 5 mL/min. Make a small cut in the right atrium with small surgical scissors. Perfuse with PBS until the solution leaving the atrium runs clear.
Turn off the perfusion pump to ensure no bubbles enter the tubing. Place the tubing from PBS into ice-cold 4% paraformaldehyde (PFA) in PBS. Perfuse with PFA at a rate of 5 mL/min until the animal is fully stiffened, approximately 25 mL.
NOTE: Ensure the PFA is fresh (no more than 1 week old if stock is stored at 4° C) for optimal fixation.
Remove the skin from the surface of the skull with small surgical scissors. Make a midsagittal cut with small surgical scissors along the central fissure of the skull. Make lateral cuts rostral to the olfactory bulb and over the cerebellum.
Open the skull with fine forceps to expose the brain. Using a spatula, scoop out the brain gently from the olfactory bulbs and place in 4% PFA in PBS overnight.
NOTE: For a more comprehensive protocol of rodent perfusion, please refer to Gage et al.³³.
Cryoprotect the fixed brain by replacing the 4% PFA with 15% sucrose in PBS for 1 day. Following this, replace the 15% sucrose with 30% sucrose in PBS solution for 1 day until the brain sinks in the solution.
Remove the brain from the sucrose solution and place it in a Petri dish with PBS. Cut off the cerebellum and olfactory bulb using a scalpel blade.
Place a small amount, 1-2 cm in diameter, of optimal cutting temperature compound (OCT) on the specimen holder surface. Mount the brain coronally to the specimen holder with the caudal cut surface in the OCT. Quick freeze the brain by placing the specimen holder in pulverized dry ice until visibly frozen, approximately 5-7 min.
Ensure that the blade angle of the cryostat is set between 0° and 5° to produce uniform sections. Adjust the roll plate angle for optimal slice flattening. Please refer to the equipment manual for specific instructions.
Place the specimen holder in the cryostat with the ventral surface of the brain closest to the blade. Cut the brain into 30 µm dorsal hippocampal sections, discarding all slices rostral to the hippocampus.
NOTE: This part of the protocol can be adapted to any desired brain region of choice. Steps 1.9-1.12 would change depending on the region of interest.
Transfer dorsal hippocampal slices to PBS. Using a paintbrush, gently mount the hippocampal sections onto a microscope slide. Remove any excess solution with cotton swabs or delicate task wipes.
Apply 100 µL of hard-set mounting medium to the microscope slide covering all brain slices. To prevent bubbles, lower the coverslip slowly using forceps onto the mounting medium. If bubbles form, gently tap the coverslip with forceps to allow them to escape. Let the slides set overnight before imaging.

2. High-resolution confocal imaging

Use low magnification eyepieces to identify fluorescent cells. Switch to a 63x (NA = 1.4) or higher objective, applying proper immersion medium to the objective.
NOTE: For the best results, utilize a laser scanning confocal microscope with a 63x or higher objective.
Identify well-labeled dendritic segments with limited overlap for image acquisition. Set the laser power and gain to ensure the fluorescent dendrites are not saturated. Additionally, reducing the scanning speed can provide better image resolution.
Acquire z-stacks encompassing the full dendritic segments for future analysis. Z-stacks larger than 10 µm are undesirable due to the added potential for dendritic overlap in the z-plane.
NOTE: Utilize the smallest z-step size available (0.2-0.7 µm)and 1 airy unit pinhole size. The smaller step size results in more images in the z-stack, compensating for the limited Z resolution of many microscopes.
Optional: If available, utilize the microscope's respective deconvolution software functionality to deconvolve images. This will allow for higher-resolution images.

3. Dendritic spine quantification

Open Neurolucida 360 (v2022.1.1 or later). Open the image file in the spine analysis software (File > Open > Image). Ensure the image file is visible in the main window and the 3D Environment. If the 3D environment window does not appear, left-click on 3D Environment in the top toolbar of the main window in the Trace tab (Supplementary Figure 1)
NOTE: This section of the protocol can be adapted for any dendritic images, not exclusively dendrites from mouse tissue.
While in the Change Image Display tab of the 3D Environment window, ensure the image is displayed as 3D Volume in the Display Image As box. In the Image Stack Settings box of the Image tab, select Max Projection on the Show Surface As drop-down menu. (Supplementary Figure 2)
Identify a suitable dendritic segment for tracing.
1. Left-click the Move Pivot Point tool in the top toolbar of the 3D Environment window. Left-click on the desired dendrite to set a new pivot point. This will change the orientation to enable effective zooming in.
2. Re-establish the original orientation by left-clicking the Reset Orientation icon. After setting the pivot point, left-click the Move Pivot Point tool to begin tracing the dendrite. (Supplementary Figure 2).
  NOTE: The ideal dendrite is one with limited overlap with other dendrites in any of the coordinate planes and not intersecting with another dendrite or superficial to another underneath. Dendrites in low proximity to others in the XY plane are also preferable to prevent inappropriate assignment of neighboring spines to the traced dendrite. It must also be noted that dendrites of differing thicknesses, orders, and distances from soma have different dendritic spine densities³⁴^,³⁵. This needs to be accounted for in the experimental design. Secondary order dendrites <1.5 µm in thickness are ideal candidates for tracing (Figure 1).
Left-click the Tree tab of the 3D Environment window. Left-click User-Guided for the tracing mode and Directional Kernels as the method in User-Guided Tracing Options.
Left click on the dendrite when a circular kernel appears to begin the tracing. Move the cursor along the dendritic segment. This will populate the kernels automatically. If the kernels are not populating automatically, see step 3.51.
1. Gently move the cursor back and forth on the dendrite until kernels populate. Left click to preserve the existing detected kernels. If kernels stop populating, left-click when a kernel populates further down the dendrite to place one manually. Right-click to end the tracing. Ensure the traced dendrite is a minimum of 7 µm in length (Supplementary Figure 3).
Verify the accuracy of the dendritic tracing using all three directions of the 3D Environment, pitch, yaw, and roll, by left-clicking and dragging the 3D Environment window. Identify points where the dendritic tracing is outside of the appropriate location on the dendrite. There can be instances where it looks accurately traced from the top down, but in the z-dimension, the points are not on the dendrite (Figure 2).
To correct improperly traced dendritic segments, left-click the Edit tab within the Tree menu. Left-click the dendrite of interest, then left-click Points.
Move improperly placed points back onto the dendrite segment via click and drag. Delete extraneous points by left-clicking the point and clicking the Delete button. Alter the size of the points if the dendrite is not adequately filled. To alter the size of the point, left-click a point and adjust the thickness slider to change the size (Supplementary Figure 4).
NOTE: Inadequately filled dendrites can result in identifying false spines that are components of the dendritic segment. Conversely, overfilling dendrites can obscure true spines.
Repeat steps 3.3-3.8 for multiple dendrites in the image before proceeding to spine identification in step 3.10.
Left-click the Spine tab in the 3D Environment window. Set Detection Settings for Outer Range, Minimum Height, and Minimum Voxel Count. Depending on the preparation, the preset values may need to be altered in the case of clear and specific justification for changes. The preset conditions are Outer Range: 2.5 µm, Minimum Height: 0.3 µm, Minimum Voxel Count: 10 Voxels.
NOTE: Different preparations, such as cell cultures vs. acute tissues, as well as different developmental time points will require different criteria that must be derived from existing literature. It is also vital to note that altering the detection settings can significantly alter results. For example, a higher minimum height can exclude short spines. Detection settings must remain consistent throughout the entire course of the experiment.
Set the Detector Sensitivity to 70% and left-click Detect All. This will populate the spines identified by this detector sensitivity on all dendrites. If selecting spines in a dendrite-specific manner is desired, left-click the box Click Image to Detect All on Nearest Branch, and left-click on each dendrite manually with different detector sensitivities.
NOTE: At this stage it is normal that not all of the dendritic spines will populate. Similarly, non-spines may improperly populate. The initial 70% sensitivity is also flexible; this may change depending on the preparation.
Examine the spines selected by this detector sensitivity by clicking and dragging the dendrite in all three directions. If the majority of detected spines are not fully filled, proceed to step 3.12.1. If the spines that have been detected are overfilled, proceed to step 3.12.2. If the detected spines appear to be adequately filled, proceed to step 3.13.
1. Increase the Detector Sensitivity by 5%-10% and left-click Detect All again. This will replace all previously detected spines with new ones at a higher sensitivity. Repeat as needed until the detected spines are adequately filled.
2. Decrease the Detector Sensitivity by 5%-10% and left-click Detect All again. This will replace all previously detected spines with new ones at lower sensitivity . Repeat as needed until the detected spines are adequately filled.
Left-click Keep Existing Spines in the Spine tab of the 3D Environment. If Click Image to Detect All on Nearest Branch has been selected, deselect it.
NOTE: By checking Keep Existing Spines ensures that newly identifying dendritic spines manually will not overwrite previously identified spines. Ensure this box is selected before proceeding so as not to overwrite the previous work.
Left click Move Pivot Point and left click on the dendrite requiring further spine detection to set the pivot point.
1. Deselect Move Pivot Point. Identify an unfilled dendritic spine. Increase Detector Sensitivity 10%-20% beyond the previous detection and left-click on the spine. If the detected spine is under or overfilled, proceed to step 3.14.3 or 3.14.4. If the spine does not populate, the message Unable to detect a spine at the selected location will appear. In this case, proceed to step 3.14.2.
2. Increase the Detector Sensitivity incrementally, possibly above 100%, until the spine has been detected and adequately filled. If the spine is detected but inadequately filled, proceed to step 3.14.3. If the spine is overfilled, proceed to step 3.14.4 (Figure 3).
3. Left-click the Edit tab and left-click on the underfilled spine. Left click Remove. Deselect the Edit tab. Increase the sensitivity by 5%-10% and left-click on the spine. Repeat this step if the spine is still underfilled.
4. Left-click the Edit tab and left-click on the overfilled spine. Left click Remove. Deselect the Edit tab. Decrease the sensitivity by 5%-10% and click on the spine. Repeat this step if the spine is still overfilled.
Repeat steps 3.14-3.14.4 until all spines identified by visual identification have been detected. Double check the dendrite for any spines belonging to neighboring dendrites, false spines corresponding to no true signal, or potential segments of dendrite mislabeled as a spine. Delete these false spines with the Remove function.
Examine the identified spines on the dendrite. In some instances, multiple spines may appear as one conglomerate spine. If a spine appears to encompass two, left-click the Edit tab. Left-click the spine and left-click Hide Selection. After confirming a conglomerate spine, in the Edit tab, left-click Show Selection and select Split. If more than two spines are in one conglomerate, this step may need to be repeated (Figure 4).
NOTE: If the conglomerate spine does not split after step 3.16, remove the spine . Then select the more intense spine of the conglomerate at a lower sensitivity. Once the more intense spine is filled, increase the sensitivity to select the other unfilled spine. Alternatively, deleting the conglomerate spine and then increasing the sensitivity may allow for proper splitting.
With all visually identifiable spines detected and filled appropriately, in the Spine tab, left-click Classify All to classify the spines into four subtypes: thin, mushroom, stubby, and filopodia (Figure 5).
NOTE: Spine classification parameters may be changed in the Settings window of the Classification box in the Spine tab. As with detector settings, a clear rationale for altering the existing parameters is strongly encouraged. The preset values are head-to-neck ratio: 1.1, length-to-head ratio: 2.5, mushroom head size: 0.35 µm, Filopodium Length- 3 µm.
In the top toolbar of the 3D Environment window, select Save and View in Neurolucida Explorer. Neurolucida Explorer is where the data is gathered from the tracings. The work will be saved as a .dat file containing all tracings and spines.
Within the Explorer window in the View tab, left-click Select All to highlight all dendrites and spines.
Left-click the Analyze tab in the upper toolbar. Left-click the Structure drop-down menu. Left click Branched Structure Analysis.
Depending on the variables of interest, any of the analyses can be selected. The two most useful for questions centered around spine density and average spine volume are Each Tree > Each Dendrite and Spines > Spine Details. Select OK, and the data will appear in two separate windows.
Copy the data to a spreadsheet for further compilation and analysis.
NOTE: The individual tree will be separated by dendrite, but the spine volume will not be. Using the sort function in the spreadsheet, the spine details can be filtered by features.

Representative Results

Effectively utilizing this analysis method begins with the selection of dendritic segments for tracing. As described in Figure 1, the ideal dendrites for tracing are not in close proximity to other dendrites. Dendrites running in parallel can result in improperly identifying spines from a neighboring dendrite. Dendrites directly intersecting or running perpendicular in a different z-plane add significant difficulty to accurate dendritic tracing as well. It is also important to note the differences in dendrite thickness. As previously reported, there are key differences in spine density with dendrites of varying thickness³⁶. There can also be differences in the same dendrite with increased distance from the branch point³⁷. Tracing dendrites of the same order and thickness, ideally with similar branch point origins, can control for the existing heterogeneity of dendritic spine density. Identifying the branch point in some preparations may prove unfeasible, but the thickness of the dendrite should always be a controllable factor in dendrite tracing. The accurate tracing of dendritic segments is vital to obtaining accurate results from this analysis. It is necessary to ensure that all points of the traced dendrite are truly within the dendrite. Viewing the three-dimensional dendrite from different directions can assist with this process. As demonstrated in Figure 2A,B, the top-down view shows what appears to be a properly traced dendrite. In the side view; however, numerous points are not located on the dendrite itself. These issues are not present in the side view of Figure 2C. It is also vital to ensure dendrites are properly filled during tracing. A dendrite that is underfilled can result in pieces of dendrites being inappropriately identified as spines. A dendrite that is overfilled can prevent true spines from being identified due to the minimum height threshold. This manual assessment of the user-guided tracing is critical to allow for accurate dendritic spine analysis.

The identification of dendritic spines also requires a user-guided approach. Using the "Detect All" function to set the uniform detector sensitivity threshold is inadequate for numerous reasons. Using the "Detect All" feature is useful for identifying the most blatantly obvious spines, but the filling of these spines must be checked to verify. The identified spines with the initial "Detect All" may be underfilled. To correct this, the identified spine must be individually deleted and then reidentified manually at a higher detector sensitivity (Figure 3A–C). This ensures that the spine is adequately filled. There is substantial heterogeneity in the required detector sensitivity for spines that must be accounted for manually. Increasing the detector sensitivity to detect all may result in overly filled spines, which require manual correction as well (Figure 3D). An additional issue with improper detector sensitivity is the inappropriate creation of a conglomerate spine, one filled dendritic spine that encompasses multiple spines. Two spines in close proximity to each other can be improperly merged into one conglomerate spine (Figure 4A,B). The spine detection software has a "Split" feature, which can be used to separate spines that have been merged by overfilling. The "Split" feature allows for the individual spines to be readily generated from the conglomerate spine (Figure 4C). Accurate dendrite tracing and dendritic spine filling allow for accurate classification into spine subtypes. Spine classification relies on morphology from the filled spines and distance from dendrites, so every step in the process plays a role in the morphological classification (Figure 5).

Due to the necessity for manual selection and thresholding, it is crucial to follow a uniform standard for all analyses. This is especially pertinent if multiple users contribute to data analysis. To ensure that all investigators performing analysis are following the same standard, investigators should compare data from the same traced dendrites. This can reduce the potential for experimenter bias by ensuring that each researcher is identifying spines based on shared, uniform criteria in a blinded fashion. There is also the possibility for bias from a single researcher between days or even on the same day due to fatigue. This should be monitored throughout the process of data analysis. To further ensure the validity of the analysis, comparing initial results to those published in the literature ensures that the protocol is being effectively followed. It is critical to note that this comparison will only be effective if the preparation and parameters are shared. Differences in staining, acquisition of fluorescent signals, order and thickness of dendrites, or brain region can contribute to different results⁸^,³⁶. In the case of missing published results, using multiple researchers to validate spine identification allows for increased confidence in the reliability and reproducibility of the analysis. A supplemental analysis folder has been included in this manuscript. This folder contains files of sample images of dendritic segments, traced dendrites, traced dendrites with identified and classified spines, and data output (Supplementary Table 1, Supplementary File 1, Supplementary File 2, Supplementary File 3, and Supplementary File 4). New users can train on this data set to practice the procedures described in this paper. User-generated results within 10% of the provided sample dataset are considered acceptable for reproducing the standard of analysis. Due to the potentially subjective criteria of a fully filled spine and the need for manual examination of automatically detected spines, variance between and within researchers is a normal part of the analysis. Should the generated results exceed that threshold; however, a side-by-side comparison should be conducted to determine instances of different spine volumes as well as improperly included or excluded spines. The sample dataset can then be reanalyzed until the acceptable threshold is reached.

Figure 1: Selecting dendrites for dendritic spine analysis. (A) 3D-volume display of z-stack confocal images taken from CA1 proximal dendrites in the THY1-YFP transgenic mouse line. Note the heterogeneity of dendrite order with thicker primary dendrites in blue ovals and thinner, secondary and tertiary dendrites in pink ovals. (B) Ideal candidates for dendrite tracing are denoted by green ovals. Note the thickness and limited intersections, overlaps, and proximity to other dendrites. The red oval denotes dendritic segments to be avoided for dendritic tracing due to high intersections, overlaps, and proximity to other dendrites. Thicker, primary dendrites are also not candidates suitable for tracing. Scale bar = 25 µm. Please click here to view a larger version of this figure.

Figure 2: Accurately tracing dendritic segments. (A) 3D-volume display of z-stack confocal images taken from CA1 proximal dendrites in the THY1-YFP transgenic mouse line to be traced via the user-guided directional kernel method. Scale bar = 10 µm. (B) Example of poor dendrite tracing. The dendrite appears to be properly traced in the top-down view. The side view shows the dendrite is improperly filled with points deviating from the dendrite. (C) Example of a proper dendrite tracing. The top-down view appears similar to B, but the side view differs substantially. The dendrite in C is properly traced as indicated by being fully filled with no deviations from the dendrite. Please click here to view a larger version of this figure.

Figure 3: Accurately filling dendritic spines using manual selection. (A) 3D-volume display of z-stack confocal images taken from CA1 proximal dendrites in the THY1-YFP transgenic mouse line of a spine awaiting manual detection. Scale bar = 0.5 µm. (B) Example of an underfilled dendritic spine. There is a substantial fluorescent signal still visible due to incomplete filling. (C) Example of a properly filled dendritic spine. The presence of a "corona" of signal just barely visible around the exterior of the filling is the standard for accurately filling dendritic spines. (D) Example of an overfilled dendritic spine. The detector sensitivity is too high, resulting in an overfilled spine. The filling has gone beyond the borders of the fluorescence and has an almost imperceptible corona. Please click here to view a larger version of this figure.

Figure 4: Splitting conglomerate dendritic spines. (A) 3D-volume display of z-stack confocal images taken from CA1 proximal dendrites in the THY1-YFP transgenic mouse line with two spines in close proximity. Scale bar = 0.15 µm. (B) An example of two independent spines improperly filled as one conglomerate dendritic spine. (C) Following the use of the "Split" feature, the conglomerate spine is split into two distinct properly filled dendritic spines. Please click here to view a larger version of this figure.

Figure 5: Dendritic spine identification and classification into subtypes. (A) 3D-volume display of z-stack confocal images taken from CA1 proximal dendrites in the THY1-YFP transgenic mouse line of a traced dendritic segment isolated for dendritic spine quantification and classification. Scale bar = 5 µm. (B) Traced dendritic segment with all dendritic spines identified and examined to ensure proper filling and splitting. The software arbitrarily assigns colors to identified spines in this step. (C) Classification of all identified dendritic spines into subtypes using defined parameters in the software. Blue = mushroom, yellow = thin, and green = stubby. Filopodia are not present due to the age of this tissue. (D) Representative images of mushroom, thin, and stubby spines unfilled (top) and filled (bottom). Scale bar = 0.3 µm. Please click here to view a larger version of this figure.

Supplementary Figure 1: Accessing the 3D Environment. Z-stack of confocal images viewed in the software interface. 3D Environment navigation from the Trace tab in the main viewer has been highlighted in yellow. Please click here to download this File.

Supplementary Figure 2: Image parameters and orientation settings for 3D Environment. 3D Environment viewer for confocal z-stack images. Parameters in the highlighted Change Image Display tab denoted by yellow arrows are set to Display Image As: 3D Volume and Show Surface As: Max Projection. Move Pivot Point and Reset Orientation are identified by yellow arrows. Please click here to download this File.

Supplementary Figure 3: Dendrite segment tracing. (A) 3D-volume of z-stack confocal images for dendrite tracing. With the tree tab, user-guided, and directional kernels all selected, tracing begins by placing the initial kernel on the dendrite with a left click. (B) Propagation of directional kernels down the dendrite following cursor movement. (C) Left-clicking further down the dendrite fills the directional kernels. (D) Example of directional kernels not populating down dendrite. Instead, a lone kernel is present further down the segment. (E) Left-clicking at the lone kernel fills the dendrite between the two points. Right-clicking ends the tracing. Please click here to download this File.

Supplementary Figure 4: Adjusting points in traced dendrites. (A) Traced dendrite segment pending point adjustment. Dendrite editing requires the "Tree" tab and "Edit" tab to be selected. Both are highlighted in yellow. Dendrite has been selected for editing with a left click. (B) Selecting the points tab, highlighted in yellow, allows for the selection of individual points on the dendrite segment. The green point has a thickness of 1.2 µm. (C) Adjusted point to fill the dendrite more accurately. The new thickness value of the green point is 0.6 µm. Please click here to download this File.

Supplementary Table 1: Sample image analysis results. Please click here to download this File.

Supplementary File 1: Sample image tracings with dendrites and spines.dat Please click here to download this File.

Supplementary File 2: Sample tracings with dendrites.dat Please click here to download this File.

Supplementary File 3: Sample dendrite image file.czi Please click here to download this File.

Supplementary File 4: Sample dendrite image file.jpx Please click here to download this File.

Discussion

This protocol details the specific steps of sample preparation, imaging, and the process of dendritic spine quantification and classification using three-dimensional reconstruction software. This software is a powerful tool capable of producing robust structural data that contributes to a diverse array of investigations. Throughout the process, there are some critical steps that make this protocol less of a methodological burden and enhance the overall output of the data. The method for labeling dendritic spines is one of the first things researchers should consider before embarking on this protocol. Issues with spine quantification can arise from insufficient labeling methods. Staining for certain proteins expressed at low levels in spines can result in signals that are too low for the software. It must also be noted that bias can be introduced by tracing the brightest fluorescing dendrites. While it is unclear if different fluorescing dendrites have different physiological properties, it is still a limitation of the protocol to consider. Additionally, in some transgenic lines, such as the THY1-YFP-H line, fluorescence in dendritic spines does not appear until around P21. This makes this line unsuitable for investigations into younger developmental time points. Consideration of the method utilized to label spines for the avenue of investigation is not a trivial aspect, as without sufficient fluorescence, the software has diminished usability. Similarly, the image acquisition hardware requires consideration. There are some file types that prove to be less compatible with the analysis software than others. Specifically, ND2 files have been identified as problematic file types for effective use of the software. The software providers recommend conversion to file types such as JPEG2000 should issues occur.

Tissue preparation and image acquisition are also important steps for high-quality spine quantification. Proper fixation, slicing, and mounting of tissues ensures a long-lasting sample with minimal artifacts that can interfere with the data analysis. Imaging the tissue is also not simply a matter of taking z-stack images of the entire brain slice. During imaging, the intent should always be to acquire stacks containing dendrites for spine quantification. An emphasis should be put on acquiring z-stacks incorporating dendrites that will be easy to trace. A thicker z-stack typically results in more background dendrites. This makes it more difficult to effectively trace dendrites with the software. Taking extra time during imaging to find better candidates for tracing will save more time during the spine quantification analysis. Additionally, ensure that the z-stack images incorporate the entire dendrites to be traced. If the dendrites are only partially viewable in the z-stack, dendrite tracing, and spine identification will prove to be difficult and inaccurate due to incomplete 3D rendering.

The process of tracing dendrites and obtaining an accurate profile of identified dendritic spines can be an arduous process. There is a degree of nuance to it. During dendrite tracing the user-guided function can occasionally not function as intended. On occasions, the directional kernels will not populate over a certain segment or begin populating over an undesired segment. One way to circumvent this is to start with a smaller typical process width. This makes the dendrite more detectable in the software, allowing for easier tracing. Should the directional kernel fail to populate entirely, left-clicking on the dendrite will place one manually. It will come to a very small point and not fill the dendrite, but that can be corrected with the thickness adjustment as described in step 3.8 of the protocol. While dendrites can be manually traced should the software prove inadequate, manually tracing spines is not a capability of the software. There will be instances where a clear spine appears to be visible, but no matter how high the detector sensitivity goes, the software will fail to detect it. One thing to check is if the suspected spine is out of range. If it is in range but still not being identified by the software, then this spine will be excluded from the analysis. While this may rarely occur, it is a limitation to consider. As with any analysis requiring thresholding and a manual classification component, there is the possibility of introducing bias. This issue can be further compounded when comparing data generated by multiple users. The semi-automated nature of this analysis seeks to minimize the introduction of this bias, but it is not wholly eliminated. In our lab, a 10% variance between researchers on a standard dataset was reasonable with sufficient practice and training. While efforts have been made to minimize bias, it is still important to consider inter-researcher bias when evaluating data generated through this protocol.

Taking into consideration the minor drawbacks of the software, the output of dendritic spine analysis using this technique is very robust. As previously described, there are a myriad of metrics that can be extrapolated from accurate dendritic segment tracing and spine identification. The ability to obtain spine subtype information provides valuable insight at a deeper level than the basic metrics. This data is important due to the interconnectivity between structure and function. Each spine subtype connotes a function. Thin spines are the predominant subtype undergoing abundant turnover²¹. Thin spines also have the potential to develop into mushroom spines³⁸. This is consistent with mushroom spines being strongly associated with learning and memory³⁸^,³⁹. Stubby spines are additionally believed to be a component of learning, potentially as remnants of mushroom spines⁴⁰. Filopodia, while not prevalent in many adult tissues, are spine precursors that are of key interest in development⁴¹^,⁴². 3D electron microscopy remains the gold standard for the most accurate classification of spine subtypes. While valuable, this technique is limited by arduous manual sorting and classification that are prone to human errors. The semi-automated design of this analysis reduces the instances where subjective bias could be introduced. While there may be drawbacks in terms of the absolute resolution and fluorescence intensity required to make perfect classifications in this protocol, it still provides a methodologically less taxing alternative to 3D electron microscopy and manual classification. Furthermore, full dendrite analysis of multiple dendritic branches from a wider swatch of brain regions is possible using the analysis outlined in this work. This is not the case with electron microscopy. Through the use of this protocol, it is possible to address structure-centered questions in multiple disciplines, including but not limited to synaptic plasticity, development, and neurological and psychiatric disorders, in a reliable and reproducible manner.

Divulgaciones

The authors have nothing to disclose.

Acknowledgements

We would like to acknowledge Carolyn Smith, Sarah Williams Avram, Ted Usdin, and the NIMH SNIR for technical assistance. We would additionally like to acknowledge the Colgate University Bethesda Biomedical Research Study Group. This work is supported by the NIMH Intramural Program (1ZIAMH002881 to Z.L.).

Materials

518F Immersion Oil	Zeiss	444960-0000-000
Cryostat	Leica	CM3050S	For slice preparation
Fine Forceps	FST	11150-10
Hemostat Forceps	FST	13020-12
Large Surgical Scissors	FST	14002-16
LSM 880 Confocal Microscope	Zeiss	LSM 880
Microscope Cover Glass	Fisherbrand	12-541-035
Mini-Peristaltic Pump II	Harvard Apparatus	70-2027	For perfusions
Neurolucida 360	MBF Bioscience	v2022.1.1	Spine Analysis Software
Neurolucida Explorer	MBF Bioscience	v2022.1.1	Spine Analysis Software
OCT Compound	Sakura Finetek	4583	For cryostat sectioning
Paraformaldehyde (37%)	Fisherbrand	F79-1
Plan-Apochromat 63x/1.40 Oil DIC	Zeiss	440762-9904-000
Scalpel Blade	FST	10022-00
Small Surgical Scissors	FST	14060-09
Spatula	FST	10091-12
Sucrose	FIsherbrand	S5-500
Superfrost Plus Microslides	Diagger	ES4951+
Vectashield HardSet Mounting Medium	Vector Laboratories	H-1400-10

Referencias

Gray, E. G. Electron microscopy of synaptic contacts on dendrite spines of the cerebral cortex. Nature. 183, 1592-1593 (1959).
Ramón Y Cajal, S. Sobre la fibras nerviosas de la capa molecular del cerebelo. Rev Trim Histol Norm. 1, 33-49 (1888).
Desmond, N. L., Levy, W. Changes in the numerical density of synaptic contacts with long-term potentiation in the hippocampal dendate gyrus. J Comp Neurol. 253, 466-475 (1986).
Engert, F., Bonhoeffer, T. Dendritic spine chances associated with hippocampal long-term synaptic plasticity. Nature. 399, 66-70 (1999).
Yang, Y., Wang, X. B., Frerking, M., Zhou, Q. Spine expansion and stabilization associated with long-term potentiation. J Neurosci. 28 (22), 5740-5751 (2008).
Oh, W. C., Parajuli, L. K., Zito, K. Heterosynaptic structural plasticity on local dendritic segments of hippocampal ca1 neurons. Cell Rep. 10 (2), 162-169 (2015).
Shinoda, Y., Tanaka, T., Tominaga-Yoshino, K., Ogura, A. Persistent synapse loss induced by repetitive ltd in developing rat hippocampal neurons. PLoS One. 5 (4), e10390 (2010).
Markus, E. J., Petit, T. L. Neocortical synaptogenesis, aging, and behavior lifespan development in the motor-sensory system of the rat. Exp Neurol. 96 (2), 262-278 (1987).
Duan, H., Wearne, S. L., Rocher, A. B., Macedo, A., Morrison, J. H., Hof, P. R. Age-related dendritic and spine changes in corticocortically projecting neurons in macaque monkeys. Cereb Cortex. 13 (9), 950-961 (2003).
Chen, C. C., Lu, J., Zuo, Y. Spatiotemporal dynamics of dendritic spines in the living brain. Front Neuroanat. 8, 28 (2014).
Dickstein, D. L., Weaver, C. M., Luebke, J. I., Hof, P. R. Dendritic spine changes associated with normal aging. Neurociencias. 251, 21-32 (2013).
Glausier, J. R., Lewis, D. A. Dendritic spine pathology in schizophrenia. Neurociencias. 251, 90-107 (2013).
Phillips, M., Pozzo-Miller, L. Dendritic spine dysgenesis in autism related disorders. Neurosci Lett. 601, 30-40 (2015).
Dorostkar, M. M., Zou, C., Blazquez-Llorca, L., Herms, J. Analyzing dendritic spine pathology in alzheimer’s disease: Problems and opportunities. Acta Neuropathol. 130 (1), 1-19 (2015).
Villalba, R. M., Smith, Y. Loss and remodeling of striatal dendritic spines in parkinson’s disease: From homeostasis to maladaptive plasticity. J Neural Transm (Vienna). 125 (3), 431-447 (2018).
Cheng, C., Trzcinski, O., Doering, L. C. Fluorescent labeling of dendritic spines in cell cultures with the carbocyanine dye "dii&#34. Front Neuroanat. 8, 30 (2014).
Feng, G., et al. Imaging neuronal subsets in transgenic mice expressing multiple spectral variants of gfp. Neuron. 28 (1), 41-51 (2000).
Baloyannis, S. J. Staining neurons with golgi techniques in degenerative diseases of the brain. Neural Regen Res. 10 (5), 693-695 (2015).
Levet, F., Tonnesen, J., Nagerl, U. V., Sibarita, J. B. SpineJ: A software tool for quantitative analysis of nanoscale spine morphology. Methods. 174, 49-55 (2020).
Peters, A., Kaiserman-Abramof, I. R. The small pyramidal neuron of the rat cerebral cortex. The perikaryon, dendrites and spines. Am J Anat. 127 (4), 321-356 (1970).
Pfeiffer, T., et al. Chronic 2p-sted imaging reveals high turnover of dendritic spines in the hippocampus in vivo. Elife. 7, e34700 (2018).
Harris, K. M. Structure, development, and plasticity of dendritic spines. Curr Opin Neurobiol. 9 (3), 343-348 (1999).
Hering, H., Sheng, M. Dendritic spines: Structure, dynamics, and regulation. Nat Rev Neurosci. 2 (12), 880-888 (2001).
Jontes, J. D., Smith, S. J. Filopodia, spines, and the generation of synaptic diversity. Neuron. 27 (1), 11-14 (2000).
Pchitskaya, E., Bezprozvanny, I. Dendritic spines shape analysis-classification or clusterization? Perspective. Front Synaptic Neurosci. 12, 31 (2020).
Berry, K. P., Nedivi, E. Spine dynamics: Are they all the same. Neuron. 96 (1), 43-55 (2017).
Rodriguez, A., Ehlenberger, D. B., Dickstein, D. L., Hof, P. R., Wearne, S. L. Automated three-dimensional detection and shape classification of dendritic spines from fluorescence microscopy images. PLoS One. 3 (4), e1997 (2008).
Swanger, S. A., Yao, X., Gross, C., Bassell, G. J. Automated 4D analysis of dendritic spine morphology: Applications to stimulus-induced spine remodeling and pharmacological rescue in disease model. Mol Brain. 4, 38 (2011).
Basu, S., et al. Quantitative 3-D morphometric analysis of individual dendritic spines. Sci Rep. 8 (1), 3545 (2018).
Ekaterina, P., Peter, V., Smirnova, D., Vyacheslav, C., Ilya, B. Spinetool is an open-source software for analysis of morphology of dendritic spines. Sci Rep. 13 (1), 10561 (2023).
Li, B. Z., Sumera, A., Booker, S. A., Mccullagh, E. A. Current best practices for analysis of dendritic spine morphology and number in neurodevelopmental disorder research. ACS Chem Neurosci. 14 (9), 1561-1572 (2023).
Dickstein, D. L., et al. Automatic dendritic spine quantification from confocal data with Neurolucida 360. Curr Protoc Neurosci. 77, 1-21 (2016).
Gage, G. J., Kipke, D. R., Shain, W. Whole animal perfusion fixation for rodents. J Vis Exp. (65), e3564 (2012).
Parajuli, L. K., Koike, M. Three-dimensional structure of dendritic spines revealed by volume electron microscopy techniques. Front Neuroanat. 15, 627368 (2021).
Ferreira, J. S., et al. Distance-dependent regulation of NMDAR nanoscale organization along hippocampal neuron dendrites. Proc Natl Acad Sci U S A. 117 (39), 24526-24533 (2020).
Megias, M., Emri, Z. s., Freund, T. F., Gulyas, A. I. Total number and distribution of inhibitory and excitatory synapses on hippocampal ca1 pyramidal cells. Neurociencias. 102 (3), 527-540 (2001).
Katz, Y., et al. Synapse distribution suggests a two-stage model of dendritic integration in ca1 pyramidal neurons. Neuron. 63 (2), 171-177 (2009).
Bourne, J., Harris, K. M. Do thin spines learn to be mushroom spines that remember. Curr Opin Neurobiol. 17 (3), 381-386 (2007).
Runge, K., Cardoso, C., De Chevigny, A. Dendritic spine plasticity: Function and mechanisms. Front Synaptic Neurosci. 12, 36 (2020).
Tonnesen, J., Katona, G., Rozsa, B., Nagerl, U. V. Spine neck plasticity regulates compartmentalization of synapses. Nat Neurosci. 17 (5), 678-685 (2014).
Mattila, P. K., Lappalainen, P. Filopodia: Molecular architecture and cellular functions. Nat Rev Mol Cell Biol. 9 (6), 446-454 (2008).
Grutzendler, J., Kasthuri, N., Gan, W. Long-term dendritic spine stability in the adult cortex. Nature. 420 (6917), 812-816 (2002).

Play Video

PDF

DOI

DOWNLOAD MATERIALS LIST

Citar este artículo

Keary III, K. M., Sojka, E., Gonzalez, M., Li, Z. Dendritic Spine Quantification Using an Automatic Three-Dimensional Neuron Reconstruction Software. J. Vis. Exp. (211), e66493, doi:10.3791/66493 (2024).