Analysis and Imaging of Osteocytes

Abstract

Osteocytes are the bone cells that are thought to respond to mechanical strains and fluid flow shear stress (FFSS) by activating various biological pathways in a process known as mechanotransduction. Confocal image-derived models of osteocyte networks are a valuable tool for conducting CFD analysis to evaluate shear stresses on the osteocyte membrane, which cannot be determined by direct measurement. Computational modeling using these high-resolution images of the microstructural architecture of bone was used to numerically simulate the mechanical loading exerted on bone and understand the load-induced stimulation of osteocytes. This study elaborates on the methods to develop 3D single osteocyte models using confocal microscope images of the LCN to perform CFD analysis utilizing various computational modeling software. Prior to confocal microscopy, the mouse bones are sectioned and stained with Fluorescein isothiocyanate (FITC) dye to label the LCN. At 100x resolution, Z-stack images are collected using a confocal microscope and imported into MIMICS software (3D image-based processing software) to construct a surface model of the LCN and osteocyte-dendritic processes. These surfaces are then subtracted using a Boolean operation in 3-Matic software (3D data optimization software) to model the lacunar fluidic space around the osteocyte cell body and canalicular space around the dendrites containing lacunocanalicular fluid. 3D volumetric fluid geometry is imported into ANSYS software (simulation software) for CFD analysis. ANSYS CFX (CFD software) is used to apply physiological loading on the bone as fluid pressure, and the wall shear stresses on the osteocytes and dendritic processes are determined. The morphology of the LCN affects the shear stress values sensed by the osteocyte cell membrane and cell processes. Therefore, the details of how confocal image-based models are developed can be valuable in understanding osteocyte mechanosensation and can lay the groundwork for future studies in this area.

SUMMARY:

This study outlines the method to visualize and develop three-dimensional (3D) models of osteocytes within the lacunar-canalicular network (LCN) for computational fluid dynamics (CFD) analysis. The generated models using this method help to understand osteocyte mechanosensation in healthy or diseased bones.

INTRODUCTION:

Osteocytes are postulated to regulate bone mass in response to physical exercise¹. Membrane deformation of osteocytes and their dendritic processes due to mechanical loading, subjects them to FFSS, which is detected by the osteocytes and triggers intracellular signaling^2-4. Bone microstructure goes through deterioration or alterations in its lacunar-canalicular morphology due to aging or bone diseases such as osteoporosis and diabetes and in conditions such as perlecan deficiency that cause impaired mechano-responsiveness of osteocytes^5,6. These changes in bone architecture cause osteocytes to experience different levels of FFSS and strains^7,8. Importantly, FFSS experienced by osteocytes in response to mechanical loading is difficult to quantify in vivo because they are embedded in the calcified bone matrix.

Confocal image-based modeling is a powerful technique to overcome the limitations of studying inaccessible osteocytes in their natural environment by replicating computer models of the LCN^9,10. Processing and modeling the interconnected network of LCN in 3D has been challenging. There are several imaging techniques, such as Transmission electron microscopy (TEM), scanning electron microscopy (SEM), serial block face sectioning, and serial focused ion beam scanning electron microscopy (FIB/SEM)^2,11,12. A valuable technique was developed to visualize bone^13-15 and generate 3D osteocyte models via confocal laser scanning microscopy (CLSM). CLSM was chosen here for computational modeling rather than other imaging techniques due to its ability to image all of the lacuna volume and most of the canaliculi in 3D^16,17. The LCN geometry can be generated using CLSM for osteocyte FEA to predict bone strains. However, fluid analysis to predict FFSS experienced by osteocytes is more complicated as it requires modeling of the cell membrane of the osteocyte and its dendrites within the LCN to enable modeling of the narrow lacunar-canalicular space, in which the interstitial fluid moves around¹⁸.

In this protocol, fluorescein isothiocyanate (FITC) dye is applied to undecalcified thick bone sections before confocal microscopy to label the LCN inside the bone, and osteocyte-dendritic membranes are modeled based on imaging data from the LCN. The lacunar-canalicular space is simulated using computational modeling, and physiological loading due to physical activity is modeled using a CFD approach. The osteocytes are subjected to a fluid pressure gradient in the CFD software to analyze the fluid profile inside the LCN and measure FFSS on the osteocyte and dendritic membranes. Furthermore, a finite element analysis (FEA) approach can measure osteocyte strains or stresses by applying compressive mechanical loading.

A geometry modification technique was also developed to modify the microstructures derived from images of young, healthy bone in order to simulate the altered lacunar-canalicular morphology in aged animals or ones with bone disease. Alterations of the bone microstructure included reducing the number of canaliculi with aging, reducing the lacunar-canalicular space area to model what happens in perlecan deficiency and increasing it to model aging effects, and reducing the canalicular and dendritic wall area to model diabetic bone^5,6. The geometry modification technique allows us to compare FFSS experienced by osteocytes in bone with different microstructures, such as young versus aged or bones in healthy versus diseased animals.

Overall, confocal image-based modeling is a valuable tool for simulating the morphology of osteocytes in healthy bone as well as aging/disease-associated changes in osteocyte morphology. Also, osteocyte morphological parameters, such as surface area and volume of the lacunar-canalicular space, can be measured and compared in various bones to predict cellular responses to mechanical strain.

PROTOCOL:

Animal experiments were carried out with the approval of the Institutional Animal Care and Use Committee at the University of Missouri, Kansas City (UMKC), and conformed to relevant federal guidelines.

1. Bone preparation process

1.1. Collect femurs from 4-month-old and 22-month-old female C57BL6 mice and fix them in cold 4% paraformaldehyde in PBS for 24 h at 4 °C with gentle rocking, then rinse them in PBS and store them in 70% ethanol before embedding.

NOTE: Fixative volume should be about 20x the tissue volume

1.2. Embed bones quickly in a fast polymerizing Acrylic (Table of Materials) following the manufacturer's instructions.

Table of Materials

Name of Material/ Equipment	Company	Catalog Number	Comments/Description
50ml centrifuge tube
Single Platform Laboratory Shaker,Model 55S	Reliable scientific INC
Sampl-Kwick Fast Cure Acrylic Kit	Buehler	20-3560
Leitz 1600 inner hole diamond saw	Leica
600 grit sandpaper	Buehler	30-5118-600-100
800 Grit sandpaper	Buehler	30-5170-800-100
1,200 Grit sandpaper	Buehler	30-5170-012-100
Fluorescein Isothiocyanate (FITC)	Sigma-Aldrich	F7250
permount mount medium	Fisher scientific	SP15-500
Leica TCS Sp5 II confocal microscope	Leica Microsystems
Immersion Oil for Microscopes	Leica Microsystems	195371-10-9
ImageJ software
MIMICS Innovation Suite® software	Materialise
3-Matic software	Materialise
ANSYS software	ANSYS

NOTE: It is important to use a fast polymerizing resin for this step (~10 min). The purpose is to support the bone tissue during sectioning using the diamond saw, but without the resin penetrating into the LCN, which would block the FITC stain from penetrating in.

1.3. Cut thick 300 μm transverse slices from a standardized site above the third trochanter using a diamond saw and store them at 4 °C in 70% ethanol prior to FITC staining.

1.4. Polish the sections using 600, 800, and then 1200 grit sandpaper to a final thickness of ~90–100 μm.

NOTE: Attaining proper thickness was ensured using a digital caliper.

1.5. Rinse the sections in 70%, 95%, and 100% ethanol for 5 min each.

1.6. Stain them in 1% FITC in 100% ethanol for 4 h at room temperature (RT) in the dark with moderate shaking.

1.7. Wash the sections in 100% ethanol for 30 min with gentle shaking in the dark. Then, air-dry them overnight in the dark.

1.8. To mount, place the section into a drop of mounting media on a glass microscope slide. Use forceps to position the section as flat as possible against the slide, avoiding generating air bubbles and using the surrounding resin to manipulate the sample. Mount a coverslip on the specimen.

2. Confocal microscopy

2.1. Use a confocal microscope for imaging of FITC-stained bone slices.

2.2. Use a 100x 1.44NA oil objective with a digital zoom of 1.7 and a step size of 0.126 μm to collect detailed Z-stacks of 400 Z-planes at 1024 x 1024 pixels, 0.089 μm pixel resolution.

2.3. Use a 488 nm laser for excitation, with an emission collection window of 496–596 nm. Collect the image stacks using compensation settings to correct for signal loss with increasing imaging depth.

2.4. Increase the accuracy and resolution of the images by using image collection techniques such as oversampling and increased line averaging. Also, collect the images with a magnification of 5x, 20x, and 100x resolution of transverse sections of femur sections as shown in Figure 1.

Figure 1: Single Z-plane confocal images of a transverse section of a FITC-stained femur taken with 5x, 20x, and 100x oil objectives.

The Lacunar-canalicular network is visible in the image taken with the 100x objective.

NOTE: The low-resolution (5x) image in Figure 1 shows the complete cross-section area of the femur, with three regions selected for 100x imaging fields.

2.5. Use the 100x Z-stacks for the computer modeling of osteocytes.

3. Computer modeling

3.1. Import the collected 100x images, in tif format, to ImageJ software to build a sequence of images of the LCN in the Z-direction.

3.2. Import the Z-stacks into the 3D image-based processing software to construct a mask of the LCN after defining the image orientation.

3.3. Threshold the original image from the young and aged mice between 30,012–45,677 Hounsfield units and 15,000–46,701 Hounsfield units, respectively, to closely resemble the LCN. Adjust the threshold in the section menu to change pixel intensity limits for inclusion in a mask.

3.4. Crop one lacuna with its canaliculi as the region of interest (ROI) from the stack using the Crop Mask operation. Define the ROI so that it encircles the lacuna in the center of the cube, and all its connected canaliculi extend to the sides of the cube. Encase the lacuna in an imaginary bigger cube with side lengths of 21 μm, 14 μm, and 19 μm.

3.5. Since the model is created out of multiple parts, perform a region grow operation to select the connected pixel regions, remove noise and despeckle to generate a uniform LCN.

3.6. Convert the lacunar-canalicular mask into an object using the Calculated Part operation in the 3D image-based processing software.

3.7. Build the osteocyte and dendritic membranes by reducing the LCN volume using the Smoothing operation. Perform this operation several times to achieve a lacunar and canalicular space thickness of 0.75 μm and 0.08 μm, respectively^9,18.

3.8. Export the objects (STL format) as the last step in the 3D image-based processing software.

3.9. Import two layers of the LCN and osteocyte-dendritic membranes into the 3D data optimization software to generate a volume mesh.

3.10. Utilize the Fix Wizard tools in software to identify mesh problems in each part. Check the mesh quality in the diagnostic section of Fix Wizard after each operation.

3.11. Remove the inverted normal parts, intersecting triangles, and bad contours using the Auto-Fix operation in Fix Wizard.

3.12. Replace the overlapping triangles manually by defining new ones or automatically via the Fill Hole Normal operation.

3.13. Improve the mesh quality using operations including filtering sharp triangles, small edges, and small shells.

3.14. After improving the mesh quality, combine two surfaces of the LCN and osteocyte-dendritic membranes into one surface (lacunar-canalicular fluidic space) that belongs to both parts using a non-manifold assembly.

3.15. Create a volumetric model of the lacunar-canalicular space using the Remesh operation and then export it as an STL file. Adjust the object scale to be in micrometers in the export section.

4. Geometry modification technique in the 3D image-based processing software and 3D data optimization software

NOTE: The geometry modification technique is used to model changes in osteocyte morphology, such as canalicular density and diameter and lacunar-canalicular thickness owing to aging or bone disease.

4.1. Choose the young osteocyte as the base model and modify it to build other distinct osteocyte models by applying morphological alterations.

4.2. Generate an osteocyte model with different canalicular densities from the base model by changing the image thresholding in the 3D image-based processing software.

4.2.1. Select a lower threshold to reduce the light intensity of the image and obtain a lacuna with fewer canaliculi. The advantage of the thresholding technique is that the lacuna shape and size remain the same, and only the influence of canalicular density is studied. Figure 2 shows the simulated aged model generated from the young osteocyte using the geometry modification technique.

Figure 2: Confocal image-based models of a young (Model 1) and an aged (Model 2) osteocyte generated from the FITC-stained femur of 4-month-old and 22-month-old C57BL6 mouse.

Model 1 image has been modified from²³. The magnified image shows the lacunar-canalicular (LC) thickness for young osteocyte and the same for the aged osteocyte.

4.3. Develop osteocyte models with different lacunar-canalicular space thicknesses or dendrite/canalicular diameters in the 3D image-based processing software and 3D data optimization software. Build larger or smaller osteocyte models by Wrapping or Smoothing operations, respectively. Figure 3 shows six osteocyte models with altered geometry developed from the young osteocyte.

Figure 3: Six osteocyte models developed from Model 1 using the geometry modification technique.

Canalicular density was reduced to generate Model 3 (simulated middle-aged) and Model 4 (simulated aged), while dendrite diameter and LC thickness were the same as Model 1. This image has been modified from ²³. Model 5 and Model 6 were developed with the same canalicular density and LC thickness as Model 1, whereas their dendrite diameter was two and four times greater than Model 1. Model 7 and Model 8 had two times and four times greater lacunar-canalicular thickness, respectively, while canalicular density and dendrite diameter remained the same as in Model 1.

5. CFD analysis

NOTE: After generating the volumetric osteocyte models, several steps, including geometry, mesh, and setup, are conducted in the CFX module of the simulation software.

5.1. Create a Fluid Flow (CFX) in the simulation software to prepare the models for the CFD analysis.

5.2. Import the developed confocal image-based geometries into the geometry section of CFX, known as ANSYS SpaceClaim (3D modeling tool). Set the unit dimensions to nanometers in the setting.

5.3. Geometry appears as two facets of the LCN and osteocyte-dendritic processes. Click on facet on the top menu and remove the geometrical errors such as intersections, sharp or overconnected edges, and vertices, openings, or holes for each facet.

5.4. Click on subtract on the facet menu to reduce the smaller facet, osteocyte-dendritic processes, from the larger facet, the LCN, to achieve a single body of lacunar-canalicular space. Then, righ-click on the generated facet and convert it from facets to a solid domain without merging faces. Figure 4 depicts the cross-sectional area of the young osteocyte model, which represents the lacunar-canalicular space.

Figure 4: 3D view confocal image-based young osteocyte model.

3D view of the cross-sectional area of the confocal image-based young osteocyte model depicting the modeling of the lacunar-canalicular space, which is used for fluid analysis around the osteocyte and dendrites.

5.5. Click on mesh and select linear tetrahedral elements using an element size of 0.06 μm. Refine the mesh with a mesh convergence study to have enough elements in the tiny dendritic system to ensure the results are independent of mesh size.

5.6. Select surface and choose the canaliculi on the top side of the imaginary cube as fluid inlets and select the canaliculi on the other five faces as fluid outlets using box select.

5.7. Export the mesh (fluent file format) as it loads faster in the setup in the next step.

5.8. Create another Fluid Flow (CFX) in the simulation software and import the fluent mesh into the setup section of CFX. Define two boundary conditions of inlets and outlets for the faces preselected as inlets/outlets using the insert boundary option.

5.9. To imitate physiological conditions, exert a fluid inlet pressure of 300 Pa and 0 Pa on the inlets and outlets, respectively^19,20. Treat the remaining surfaces as walls with a no-slip condition in that fluid has zero velocity at the interface of walls. Fluid flows from the inlets around the dendrites and osteocyte cell body and exits from the other canaliculi assigned as outlets.

5.10. Treat the interstitial laminar fluid as water ⁹, chosen from the material library. Set the heat transfer, combustion, and thermal radiation sections to None, as no heat transfer is defined in the problem. Select the Turbulence mode as the fluid characteristic in the LCN, which is a laminar fluid⁹.

5.11. Run the software using Double Precision and Direct Start as the submission type. Monitor mass and momentum until residuals drop and become constant. After solution convergence, measure FFSS data using the CFD-post section of the CFD software.

6. CFD post processing

6.1. To depict FFSS experienced by osteocytes and their dendrites, insert a new contour in the results section of the CFD software. Create an FFSS contour by choosing the wall shear on the osteocyte-dendritic membranes as the variable at the domain.

NOTE: To better show high FFSS on dendritic membranes, the range of FFSS is set to User Specified to modify min/max values of FFSS.

6.2. Insert a velocity streamlines contour inside the lacunar-canalicular domain starting from the inlets. Set the sampling to equally spaced and select the number of points as 2500. An animation section in the CFD software accurately displays in 3D how the fluid particles flow inside the lacunar-canalicular spaces using the velocity streamline graph.

6.3. Use the Function Calculator tool in the CFD software for analyzing the magnitude of FFSS or velocity based on geometrical parameters, especially where there are various osteocyte models (i.e., young vs. aged). Measure the volume and surface area of the lacunar-canalicular space as geometrical parameters along with maximum, minimum, or average FFSS values.

REPRESENTATIVE RESULTS:

This protocol describes how to develop confocal-derived osteocyte models to investigate the amount of fluid flow shear stress an osteocyte and its dendritic processes are subjected to due to mechanical loading. An aged and a young C57BL6 mouse were selected to build young and aged confocal image-based osteocyte models. Six other simulated osteocyte models were generated from the same young osteocyte model using the geometry modification technique to study the alteration of LCN morphology due to aging or bone disease. Geometry-modified parameters included the lacunar-canalicular thickness, canalicular density, and diameter. Modified osteocyte models were developed by changing one parameter at a time while holding the other parameters the same. Therefore, the effect of each of the morphological parameters on FFSS is examined in isolation. The results revealed that the amount of FFSS experienced by an osteocyte was different in these osteocyte models, showing the dependency of FFSS on LCN morphology. FFSS values were predicted on various parts of the osteocyte model, such as the osteocyte cell body or dendritic membranes. Dendrites in all models experienced greater FFSS values than the osteocyte cell body membrane.

Figure 5 shows the FFSS contours for young vs. aged osteocyte models generated directly from confocal FITC image stacks and six other osteocyte models developed from Model 1 using the geometry modification technique. The average FFSS on osteocyte and dendritic membranes in the young osteocyte model (Model 1) was 0.42 Pa, more significant than the average FFSS of 0.13 Pa in the aged osteocyte model (Model 2). The obtained data show an increase in average FFSS in osteocytes with larger lacunar-canalicular space (Model 7 and Model 8) and a decrease in FFSS in osteocytes with lower canalicular density (Model 3 and Model 4), while FFSS did not significantly change with changing dendrite diameter in Model 5 and Model 6. Model 2 and Model 4 had the lowest FFSS values of 0.19 Pa and 0.13 Pa, respectively. The magnitude of FFSS in Model 8 is greater than in the other osteocyte models. Not only are most of the dendrites in red color in Model 8 but also, the osteocyte cell body experiences higher FFSS than in other models.

Figure 5: FFSS distribution for all osteocyte models shows the dependence of FFSS magnitude on osteocyte morphology.

Red regions show the highest FFSS values, especially on the dendritic membranes rather than the osteocyte cell body. This image has been modified from ²³.

FFSS experienced by osteocytes was also presented based on the lacunar-canalicular space volume and surface area, which were obtained by post-processing in ANSYS CFX to find out the correlation between FFSS and the altered structure of the LCN. It was also possible to measure the surface area of each model exposed to FFSS greater than 0.8 Pa, which is thought to be the level of FFSS that causes bone formation^21,22. The osteocyte modeling showed a correlation between the average FFSS and the lacunar-canalicular space volume or surface area for all osteocyte models, as shown in Table 1. The aged osteocyte (Model 2) and simulated middle-aged osteocyte (Model 4) with the lowest canalicular density had the lowest lacunar-canalicular surface area of 1372.7 μm² and 1122.6 μm². Model 2 and Model 4 also had the lowest lacunar-canalicular space volume of 229.4 μm³ and 203.5 μm³, respectively, which can be correlated to their low amounts of FFSS. Model 6, with the largest dendrite diameter, had the largest lacunar-canalicular surface area. Model 8 had the largest lacunar-canalicular space volume and the greatest FFSS values.

Table 1: Comparison of osteocyte models.

Comparison of average FFSS, lacunar-canalicular space volume, and surface area in eight distinct osteocyte models, including two confocal image-based models and six simulated models generated from Model 1 using the geometry modification technique.

	Models	Lacunar-canalicular surface area (μm2)	Lacunar-canalicular space volume (μm3)	Ave FFSS (Pa)
Confocal image-based osteocyte models	Model 1	2652.9	261	0.42
	Young (4-month-old)
	Model 2	1372.7	229.4	0.19
	Aged (22-month-old)
Simulated osteocyte models generated using geometry modification technique	Model 3	1530.4	242.6	0.27
	(Simulated middle-aged)
	Model 4	1122.6	203.5	0.13
	(Simulated aged)
	Model 5	3724.9	302.9	0.49
	(2 x dendrite diameter)
	Model 6	4791	337	0.57
	(4 x dendrite diameter)
	Model 7	3116.3	440.6	1.17
	(2 x LC thickness)
	Model 8	3542.2	658.2	1.89
	(4 x LC thickness)

DISCUSSION:

This protocol outlines a confocal imaging technique for visualization and computational modeling of the osteocytes. Before confocal imaging, the bone preparation process for sectioning and staining bone samples is performed. Confocal images of 100x magnification are imported into various software to develop computer models of osteocytes and the lacunar-canalicular space. A CFD analysis is conducted lastly on the confocal image-based models to model FFSS surrounding the osteocytes and dendritic membranes due to physical activity, which is technically challenging in vivo due to the inaccessibility of osteocytes in bone matrix. The provided protocol is also helpful in determining other stress or strains using a finite element or fluid-structure interaction (FSI) approach in multiphysics modeling.

Confocal imaging has been used to visualize the LCN^7,9,24-25 owing to the high-resolution images that can be achieved using FITC staining of the bone. Although the diameter of the canaliculi is close to the resolution limit of the microscope, reliable confocal images of the canaliculi can be obtained, which can capture most of the canaliculi. Due to the smaller diameter of the osteocyte dendrites, it is technically more challenging to accurately capture images of the dendrites. A helpful technique was therefore developed to model the osteocyte and dendrites from confocal images of the LCN by reducing their volume using a smoothing operation in the software. Some other published studies used a similar approach to offset the LCN to generate the osteocyte-dendritic membrane⁹. The other benefit of the smoothing technique is in improving some imaging errors in CLSM, such as the elongation of canalicular diameter that occurs in the z-axis due to the lower imaging resolution in the Z-axis, which is a feature of optical imaging. Deconvolution is also considered an alternative method to decrease the optical stretching in the Z-direction since it improves the signal-to-noise in the original data set. The lacunar-canalicular space is modeled from LCN and osteocyte-dendritic membranes to analyze the fluid profile inside the LCN.

Notably, a new technique was introduced with respect to existing methods to develop geometry-modified osteocyte models without sacrificing another mouse. The geometry modification technique also helps to simulate the alteration of bone microstructure due to aging or bone diseases such as osteoporosis, diabetes, and conditions such as perlecan deficiency, etc. In this technique, a confocal image-based osteocyte model of a young mouse was used to model the morphology of aged or other bone-diseased osteocytes by modifying parameters such as canalicular density, lacunar-canalicular space, and dendrite diameter). This approach was validated by the similarity of the two aged osteocyte models, one from confocal image-based modeling of an actual osteocyte from a 22-month-old mouse and the other from geometry modification modeling of an osteocyte originally imaged from a 4-month-old mouse by reducing its canalicular density. These two aged osteocyte models showed a similar FFSS magnitude, the lowest among all osteocyte models. The geometry modification technique could have many applications in different biology fields as it can predict comparable results to osteocyte models generated using real osteocytes from an animal. Some of the limitations of this work include losing some of the detail of surface roughness due to the smoothing technique and the fact that the canalicular diameters are at the limit of resolution of the confocal microscopes, which could potentially lead to some inaccuracy for canaliculi with smaller diameters, and also fluid flow idealization instead of modeling using oscillatory flow.

To demonstrate the technique's functionality, osteocyte FFSS values were measured for distinct osteocyte models generated from a young and an aged mouse or the geometry modification technique. One of the noteworthy findings using these computer models was that the volume of lacunar-canalicular space is directly proportional to the average FFSS experienced by the osteocytes. This new finding which was not investigated before, may have various applications in determining the osteocyte mechano-responsiveness as it can correlate the osteocyte network morphology to detected FFSS by the osteocytes and their processes.