Introduction

Intervertebral disc (IVD) is a complex cartilaginous structure consisting of gel-like material in the center (the nucleus pulposus), surrounded by concentric lamellar layers, i.e., the annulus fibrosus (AF). This structure, responsible for the biomechanical behavior of IVD, is synthesized and maintained by the cells present in the extracellular matrix of the disc.20 Thus, it is crucial for the cells within the tissue to have adequate nutrient supply in order to maintain the health of the disc.12,17,35,48 The IVD is the largest avascular structure of the human body and the supply of nutrients to disc cells is provided by the blood vessels in the surrounding tissues, through cartilage endplate route and periannular route.3,24,36,49,50 The distance of its cells from the nearest blood vessel can be up to 8 mm,7,45 and diffusion is the major mechanism for the transport of nutrients (e.g., glucose and oxygen) in IVD.48

Poor nutritional supply is believed to be one of the mechanisms for disc degeneration.8,25,26,35,48,51 However, relatively little is known on the nutrition transport and levels within the IVD, even though many studies have been done experimentally and theoretically. For examples, investigators have used quantitative mathematical models and finite element methods to study the mechanisms and pathways of nutrient transport into the disc.18,40,42,53,54 In these numerical studies, however, the diffusion of solutes in AF was assumed to be isotropic.

It is well known that the morphology of AF is characterized by concentric lamellar layers. Adjacent lamellae are composed by collagen fiber bundles inclined, with respect to the vertical axis of the spine, with angles varying from ±62° to ±45°.13 This peculiar structure is responsible for an anisotropic behavior of AF observed in experiments, such as mechanical stiffness,16,19 hydraulic permeability,22 water diffusion,15,27,36 and electrical conductivity.29

Thus, it is reasonable to hypothesize that the diffusion of solutes within the AF is anisotropic. As a first order approximation, in the present study, the diffusion tensor in AF is assumed to be orthotropic, with three principal directions along the axial, circumferential, and radial directions of disc. The objective of this study was to experimentally determine the anisotropic diffusion tensor (D) of a relatively small solute (fluorescein, molecular weight: 332 Da) in AF using a fluorescent recovery after photobleaching (FRAP) technique.

FRAP has been extensively used to determine diffusion of macromolecules within tissues.5,6,14,31,33,39,43 The basic principle of FRAP is based on photobleaching the fluorescence of molecular probes and analysis of the recovery of the bleached area.2,30,38

In performing a FRAP test in the case of two-dimensional (2D) diffusion (e.g., translational mobility of molecules in membranes34 or polymeric films47), a common approach for calculating diffusion coefficient of fluorescent probes by video-FRAP technique is based on the transformation of the 2D mass transport field equation into the Fourier space by performing a Spatial Fourier Analysis (SFA).5,6,39,41,46 With respect to other methods for isotropic2 or anisotropic32,44 2D diffusion cases, the advantage of using a video-FRAP SFA approach is that the analysis is independent from the initial conditions of the bleached spot of the sample. On the other hand, when FRAP is performed in the case of 3D diffusion, fluorescent probe transport occurs in the plane of bleaching as well as in the direction orthogonal to this plane (z-direction). While several analytical and numerical models for 3D FRAP tests with Confocal Laser Scanning Microscope (CLSM) have been proposed for the case of isotropic diffusion,9–11 an extension of the SFA approach to the 3D case has not been developed yet, due to the difficulty in collecting 3D images of fluorescence recovery in real time.

In this work we propose an extension of the use of 2D SFA to the 3D diffusion case by combining the Fourier transformation approach with a multilayer bleaching (MLB) testing protocol. The basic principle of MLB consists of sequentially bleaching the sample over several layers parallel to the plane in which the fluorescence recovery is observed. Theoretically, if MLB is performed over the whole thickness of the sample, fluorescence recovery within the plane of interest could be two-dimensional. In practice only a limited number of sample layers can be sequentially bleached before fluorescence recovery occurs in the earlier bleached planes, therefore the diffusion is three-dimensional and the use of a 2D SFA approach leads to an overestimation of the diffusion coefficient. However, by optimizing the number and the size of the bleached layers, it is possible to minimize the error in the estimation of the diffusion tensor (see Appendix).

In the following, a new approach for determining 3D anisotropic diffusion tensor using video-SFA and MLB protocol will be described. It will be illustrated that, from three independent FRAP tests performed in the three principal planes of the IVD, it is possible to extract the three principal components of the anisotropic diffusion tensor of AF. The results are compared to those in the literature and correlated to the anisotropic structure of the tissue.

Theoretical Background

When MLB protocol is applied in FRAP test, diffusion of fluorescent probes in a bulk sample could be assumed to be confined within the focal plane of the microscope objective, see Appendix. Under the assumptions that diffusion is 2D and orthotropic, the field equation for solute concentration (c) within a medium can be obtained based on the conservation of mass and Fick’s law:

$$ \frac{{\partial c(x,y,t)}} {{\partial t}} = D_{{xx}} \frac{{\partial ^{2} c(x,y,t)}} {{\partial x^{2} }} + D_{{yy}} \frac{{\partial ^{2} c(x,y,t)}} {{\partial y^{2} }} + 2D_{{xy}} \frac{{\partial ^{2} c(x,y,t)}} {{\partial x\partial y}}, $$
(1)

where (x,y) are the coordinates of the imaging system within the focal plan, D xx , D xy , and D yy are components of diffusion coefficient tensor (D) in the (x,y) coordinate system. In arriving at Eq. (1), the diffusion tensor is assumed to be constant with respect to time (t) and position (x,y) within the focal plane (∼460 × 460 μm2, see Materials and Methods below). At the boundaries the concentration c(x,y,t) is constant. For an arbitrary initial condition Eq. (1) can be solved using 2D SFA:41,46

$$ C(u,v,t) = C(u,v,0)\exp {\left[ { - 4\pi ^{2} {\left( {u^{2} + v^{2} } \right)}D(\xi )t} \right]}. $$
(2)

In Eq. (2), C(u,v,t) is the Fourier transform of c(x,y,t), defined as:

$$ C(u,v,t) = {\int\limits_{ - \infty }^\infty {{\int\limits_{ - \infty }^\infty {c(x,y,t)e^{{ - i2\pi (ux + vy)}} } }dxdy,} } $$
(3)

and

$$ D{\left( \xi \right)} = D_{{xx}} \frac{{u^{2} }} {{u^{2} + v^{2} }} + D_{{yy}} \frac{{v^{2} }} {{u^{2} + v^{2} }} + D_{{xy}} \frac{{2uv}} {{u^{2} + v^{2} }} = D_{{xx}} \cos ^{2} \xi + D_{{yy}} \sin ^{2} \xi + 2D_{{xy}} \sin \xi \cos \xi , $$
(4)

where the angle ξ is defined by the wave vector specified by u and v, and the positive u-axis (Fig. 1):

$$ \xi = \tan ^{{ - 1}} \frac{v} {u}. $$
(5)
Figure 1
figure 1

Transformation of a video-FRAP image into the frequency space using SFA

Full size image

For a given interval of time [t 1, t 2], from Eq. (2), the following relationship is obtained:

$$ R = REAL{\left[ {\frac{{C(u,v,t_{2} )}} {{C(u,v,t_{1} )}}} \right]} = \exp {\left[ { - 4\pi ^{2} {\left( {u^{2} + v^{2} } \right)}D(\xi ){\left( {t_{2} - t_{1} } \right)}} \right]}. $$
(6)

By curve-fitting of experimental data with Eq. (6), one can obtain D(ξ) which depends on the choice of the frequencies u and v (or ξ), see Eq. (4). For this orthotropic diffusion case, it can be shown that the mean value of D(ξ) [averaged over the entire frequency space (u,v)] is given as follows:

$$ D_{{av}} = \pi ^{{ - 1}} {\int\limits_0^\pi {D(\xi )} }d\xi . $$
(7)

Equation (7) indicates that only the values of D(ξ) of the ‘ring’ [i.e., an arch of circumference with u 2 + v 2 = constant in the (u,v) space, see Fig. 1] are needed in order to evaluate D av, which is similar to the isotropic case.46 By use of Eq. (4), the solution of the above Eq. (7) yields:

$$ D_{{av}} = \frac{{D_{{xx}} + D_{{yy}} }} {2}. $$
(8)

Since the sum (D xx  + D yy ) is the trace of the diffusion tensor, it follows that the value of D av is equal to the averaged value of two diffusion coefficients in the principal directions (within the focal plane with surface normal in the z-direction). By measuring D av in the three principal planes, one can determine the three diffusion coefficients in the principal directions. Note that for an isotropic case where D xx  = D yy  = D, Eq. (8) recovers that D av  = D. Here D is a constant (independent of ξ). However, it is a common practice to average D over a range of [0,π/2] in order to reduce the noise and to improve the accuracy.46 For an anisotropic case, it is possible to obtain the value of D xx or D yy individually by choosing special frequency couples along the axes of the Fourier space, namely (u,0) and (0,v),46 see Eq. (4). In the present study, the averaged diffusion coefficient is obtained by Eq. (7) and individual components of diffusion tensor are obtained by use of Eq. (8), see Materials and Methods.

Materials and Methods

Specimens Preparation

Nine discs (S2-3 and S3-4) were harvested from five bovine coccygeal spines obtained from a local supermarket, wrapped in plastic (in order to prevent dehydration) and stored at −20 °C. Within a week, the specimens were partially thawed at room temperature and AF blocks were excised from the anterior and posterior regions of each disc in its three principal directions: axial, circumferential, and radial, respectively, see Fig. 2. From each AF block four cylindrical specimens of 5 mm diameter and 540 μm thickness were obtained using a microtome (SM24000, Leica Microsystems GmbH, Wetzlar, Germany) with a freezing stage (Model BFS-30, Physitemp, Clifton, NJ).

Figure 2
figure 2

Schematic of specimen preparation. The size and orientation of test specimens are shown

Full size image

Cylindrical specimens were equilibrated in a fluorescein (332 Da, λ ex 490 nm; λ em 514 nm, Fluka-Sigma-Aldrich®, St. Louis, MO, USA) water solution (0.1 mol/m3). In order to prevent tissue swelling, the AF specimens were confined within two sinterized stainless steel platens (20 μm porosity, Mott Corp., Farmington, CT) and an impermeable spacer (see Fig. 3). Measurements of the height of the cylindrical specimens were performed by a custom-made current sensing micrometer21 before and after the samples reached equilibrium in the fluorescein solution. No variation in thickness of the sample was detected (data not shown), indicating that no swelling of the sample occurred during its equilibration in the fluorescein solution.

Figure 3
figure 3

Setup for equilibrating specimens in a fluorescein water solution (0.1 mol/m3)

Full size image

Diffusivity Measurement

Fluorescein diffusion coefficients within the three principal planes of AF were determined by means of FRAP tests conducted at room temperature (22 °C). AF specimens were photobleached using an argon laser (488 nm wave length) of a CLSM (LSM 510 Zeiss, Jena, Germany). A Plan-Neofluar 20×/0.50 WD 2.0 objective (Zeiss, Jena, Germany) was used in these experiments: for each test 200 frames (128 × 128 pixels, 460.7 × 460.7 μm2) were collected, five of them before bleaching the sample. For each region (anterior or posterior) and direction (axial, circumferential, or radial) 48 tests were performed. The collected data (images) were analyzed by custom-made software performing an algorithm of SFA and curve-fitting the experimental results with Eq. (6),4,5,46 to yield D(ξ). Note that D(ξ) is a function of the u–v frequency couples. The diffusion coefficient D av was obtained by averaging D(ξ) over Ring 3 and Ring 4 by Eq. (7), similar to the isotropic case proposed in the literature.46

MLB Protocol

For each FRAP test four different layers of AF samples were sequentially bleached. The distance between the bottom glass slide (see Fig. 7) and the focal plane of the microscope objective, where fluorescence recovery was observed, was 7 μm and the diameter of the bleach spot was 28.75 μm. The other bleach spots were produced in layers at 17, 27, and 32 μm from the bottom of the sample and their diameters were 43.12, 50.31, and 71.88 μm, respectively. The bleached spots were produced from top to bottom. Measurements of fluorescence intensity within the sample indicated that, after bleaching the four sample layers, the final shape of the bleached region is a cylinder of approximately 28 μm diameter and 47 μm height. Numerical simulations demonstrated that in these conditions the highest relative error (in the case D⊥/D = 2) committed in the determination of D av using 2D SFA is estimated to be approximately 18% (see Appendix for details).

Determination Diffusion Tensor Components

Let D axi , D cir , and D rad stand for the averaged diffusion coefficients measured in the IVD principal planes with surface normal along the axial, circumferential, and radial directions of the disc, respectively; D z , D θ , and D r for the principal components of the diffusion tensor in the axial, circumferential, and radial directions, respectively. Applying Eq. (8) to the three principal IVD planes, it follows that:

$$ \begin{aligned}{} & D_{\theta } + D_{r} = 2D_{{axi}} \\ & D_{z} + D_{r} = 2D_{{cir}} \\ & D_{\theta } + D_{z} = 2D_{{rad}}. \\ \end{aligned} $$
(9)

The above set of linear equations allows the extraction of the principal components of the diffusion tensor (D z , D θ , and D r ) from the measured, averaged diffusion coefficients within the three principal planes of the IVD.

Scanning Electron Microscopy (SEM)

SEM of two bovine coccygeal IVDs was performed in order to correlate the collagen fiber structure of AF to the diffusion coefficient data. Axial and radial sections of anterior and posterior regions of AF were fixed with a 2% gluteraldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) in phosphate buffer solution (Sigma®, St. Louis, MO, USA), dehydrated in a graded series of ethanol (20, 40, 60, 80, 100%) and then dried by immersion in hexamethyldisilizane (Electron Microscopy Sciences, Hatfield, PA, USA).23 Finally, the samples were sputter coated with Pd (Sputter Coater 108auto, Cressington, Watford, UK). High-resolution SEM images were taken on an Environmental Scanning Electron Microscope (XL30 ESEM-FEG, FEI Company, Hillsborg, OR, USA) in high-vacuum mode.

Results

The values of averaged diffusion coefficient in axial, circumferential, and radial planes of AF were different (Table 1). A two-way analysis of variance (ANOVA) was performed to investigate the effects of the region and the direction on the measured, averaged diffusion coefficients in AF using Microsoft® Office Excel 2003 (Microsoft Corp., Seattle, WA). The two factors considered in the analysis were the disc region (two levels: anterior and posterior) and the direction (three levels: axial, circumferential, and radial).

Table 1 Measured diffusion coefficients of fluorescein within each of the principal planes in the anterior and posterior regions of AF
Full size table

Both disc region and principal direction were shown to affect the fluorescein diffusion coefficient significantly (p < 0.05). It was found that the mean values of the diffusion coefficient in anterior region were significantly different from those in the posterior region (p < 0.05). These findings indicated that D is inhomogeneous within AF. In both posterior and anterior regions, the diffusion coefficient within the radial plane was significantly higher than that within the circumferential or axial plane. In both anterior and posterior regions, axial and circumferential diffusion coefficients were not significantly different.

The principal components of the diffusion tensor (D) were extracted from the experimental data (Table 1) according to Eq. (9), see Fig. 4. In both regions, the mean value of the radial diffusion coefficient (D r ) was the lowest with a value of D Ant r  = 8.14 × 10−7 cm2 s−1 in the anterior region and D Pos r  = 7.93 × 10−7 cm2 s−1 in the posterior region. In comparison, the mean values of the circumferential and axial diffusion coefficients were similar: along the θ-direction D Ant θ  = 1.22 × 10−6 cm2 s−1 for the anterior region and D Pos θ  = 1.05 × 10−6 cm2 s−1 for the posterior region; along the z-direction D Ant z  = 1.26 × 10−6 cm2 s−1 and D Pos z  = 1.05 × 10−6 cm2 s−1 for the anterior and posterior regions, respectively, see Fig. 4.

Figure 4
figure 4

Anisotropic diffusion coefficients of fluorescein in axial, circumferential, and radial directions of AF

Full size image

Discussion

The main objective of this study was to investigate the anisotropic diffusion of solute in AF using the FRAP technique. The results showed that the diffusion coefficients of the fluorescein dye in AF are different along its principal directions (axial, circumferential, and radial). In particular, the results showed that the diffusion coefficient in the radial direction is about 66–75% the value of axial or circumferential direction for specimens harvested from anterior and posterior regions, respectively, see Fig. 4.

This study represents the first measurement of anisotropic diffusion of a relatively small solute in AF using a video-FRAP technique. A new FRAP testing protocol (i.e., multilayer bleaching) was developed for bulk samples to achieve an approximately 2D diffusion condition (see Appendix). Numerical simulations showed that it was possible to combine multilayer bleaching and the 2D SFA algorithm for determining anisotropic diffusion coefficients in AF (less than 18% error, Appendix). To further validate our method, our results were compared with the anisotropic diffusion coefficients of glucose in axial and radial directions of bovine coccygeal AF, measured by a direct diffusion experiment (manuscript in preparation). It was found that our results were consistent with those from the direct diffusion experiment in which the value of diffusion coefficient of glucose in the radial direction was about 66% of the value in the axial direction (manuscript in preparation). Moreover, the finding of anisotropic diffusion behavior of fluorescein in the present study is also consistent with a study on anisotropic ion diffusivity in bovine coccygeal AF in the literature.29 The values of ion (Na+ and Cl−) diffusivities are reported to be not significantly different between the axial and circumferential directions. In the radial direction, however, the ion diffusivity is, again, about 2/3 of the values in the axial or circumferential direction.29

Our findings can be related to the tissue structure of the AF in terms of its collagen fiber organization. Figures 5a–5d show SEM images of radial and axial sections (or planes) in both posterior and anterior regions of AF. In contrast with the radial sections, axial sections show orderly distributed cavities (Figs. 5c, 5d). Higher magnifications of the anterior radial section show the cavities to be microtubules (∼5 μm diameter) parallel to the direction of the collagen bundle fibers, see Figs. 6a and 6b. In case microtubules represent a preferential pathway for mass transport, this would explain the direction-dependent diffusive behavior of solutes in AF: while in the axial direction solutes could preferentially move through the microtubular pathway, in the radial direction this would not be possible since microtubules are not contiguous. Note that a similar structure (i.e., microtubules) has been reported in the literature for IVD of rat tails.28

Figure 5
figure 5

SEM images of AF sections: (a, b) anterior and posterior radial sections; (c, d) anterior and posterior axial sections

Full size image
Figure 6
figure 6

SEM images of anterior axial section of AF. (a) Microtubules are shown; (b) Magnification of a microtubule in picture (a)

Full size image

In this study, the diffusion tensor is assumed to be orthotropic with three principal (axial, circumferential, and radial) directions. Indeed, at collagen fiber level, the principal directions should be parallel to the orientation of the collagen bundles within the lamellae34 which varies from ±62° to ±45° respect to the axial direction.13 However, collagen fiber orientations within two adjacent lamellae are alternating; therefore, the diffusivity values in the axial and circumferential directions should be interpreted as averaged values over two adjacent lamellae.

Conclusions

A new method has been developed for determining anisotropic diffusion of solute in bulk tissue using FRAP. This method has been successfully applied, for the first time, to the study on the anisotropic diffusion of solute in IVD. It has been found that the diffusion of solute in bovine coccygeal AF is inhomogeneous (varying from anterior to posterior regions) and anisotropic, with radial diffusion coefficient being the lowest among three major directions (axial, circumferential, and radial). This finding is consistent with the findings on anisotropic diffusion of glucose (manuscript in preparation) and ions in bovine coccygeal IVD.29

The anisotropic diffusion behavior in AF is due to the organization of the collagen fibers in tissue. The microtubules parallel to the collagen fiber bundles may be a preferential pathway for solutes diffusing in the axial and circumferential directions.

Our results provide new understandings on the relationship between solute diffusion coefficient and the structure of AF. The methods and findings reported in the present study could help elucidate the pathways of nutritional supply to the human IVD (the largest avascular structure in the human body),1,40,42,52 which is crucial for understanding its degeneration.8,25,26,35,48