|Home | About | Journals | Submit | Contact Us | Français|
To investigate the fibril architecture of the collage matrix in the superficial zone of articular cartilage non-destructively by microscopic MRI (μMRI) T2 anisotropy.
Six specimens of canine humeral cartilage were rotated in such a way that the normal axis of the articular surface of the cartilage specimen remained stationary and perpendicular to the static magnetic field, over a range of 180° and at a step of 15°. At each rotation angle, a quantitative T2 image was constructed at 13μm pixel resolution.
A set of complex and depth-dependent patterns was found in the μMRI T2 anisotropy along the depth of the tissue. In the superficial zone, the T2 anisotropy is clearly periodic, which demonstrates that the distribution of the collagen fibrils in the superficial zone is not random. In the transitional zone, the periodicity of the T2 anisotropy approximately doubles with respect to that in the superficial zone. In the initial part of the radial zone, the T2 anisotropy is also periodic but inverse to that in the superficial zone. In the deep part of the radial zone, the T2 anisotropy becomes increasingly weaker and eventually disappears.
There exists a certain degree of collagen anisotropy in all zones of articular cartilage. The anisotropic imaging data can be interpreted with the aid of a collagen architecture model.
As one of the major molecular constituents in articular cartilage, the fine fibrils of collagen have some unique architectural features in its structure [1-3]. From the articular surface of the tissue towards the cartilage-bone interface, the collagen fibrils are initially oriented perpendicular to the normal axis of the articular surface near the tissue surface, become somewhat obliquely oriented afterwards, and are in parallel with the normal axis of the articular surface for most of the deep tissue. These architectural features form the criteria in histology to conceptually subdivide the thickness (depth) of articular cartilage into three consecutive zones, namely the superficial, transitional, and radial zones (SZ, TZ, RZ) respectively [4, 5].
Among the histological zones in articular cartilage, the structural integrities of SZ are of particular importance to the entire cartilage tissue as a load-bearing material in a synovial joint [6-10]. Since the SZ tissue has a thickness about tens to one hundred microns [11, 12], transmission and scanning electron microscopy have been used to image the thin sections of the tissue or the surface of a tissue block [11, 13-21]. In recent years, atomic force microscopy [22-24], confocal microscopy  and X-ray diffraction  have been used to image the surface layer of an intact tissue block, which minimizes some potential artifacts resulting from the specimen preparation. Although one might be tempted to consider the orientation of the interwoven collagen fibrils in SZ to be somewhat randomly distributed at any 2D plane that is in parallel with the surface [11, 16], the existence of the split-line patterns on the joint surface [7, 14, 27, 28] implies that the fibril orientation in SZ might contain some form of macroscopic organization.
Because of the architecture of its collagen matrix, articular cartilage is known to have strong anisotropy when it is examined in a magnetic field  or under polarized light (visible  and infrared ), which offers a powerful way to monitor the ultrastructure of the tissue - the disruption of which is an early sign in the tissue degradation leading to osteoarthritis and other joint diseases. Among the quantifiable parameters in MRI, the anisotropy of T2 relaxation is of particular importance [29, 31]. This is because T2 anisotropy has been shown to be sensitive to the fibril structure and orientation in cartilage and is responsible for the MRI visualization of cartilage laminae (the so-called magic angle effect in MRI of cartilage) [31, 32]. In recent years, T2 anisotropy has been used to divide the MRI-visible features of cartilage into MRI-zones non-destructively and quantitatively, which have been proved to be statistically equivalent to the histological zones based on the collagen orientation by quantitative polarized light microscopy (PLM) [12, 33]. Consequently, when the spatial resolution of an MRI experiment is fine-tuned to tens of microns per pixel, although still not sufficient to visualize the individual fibrils, one can study the averaged physical and biological properties of articular cartilage in any sub-tissue histology zones, enabling the detection of early osteoarthritis non-destructively .
The MRI procedure that can be used to derive T2 anisotropy requires the rotation of the specimen in the magnetic field B0. In most, if not all, MRI studies of cartilage using T2 anisotropy, the rotation is in such a way that the normal axis of the articular surface is rotated with respect to the B0 direction (Fig 1a). This rotation, coupled with a vertical imaging slice, can produce a 2D image that contains all histology zones of the tissue. The schematic drawing in Fig 1a, however, contains one over-simplification for the collagen structure in SZ of articular cartilage. Instead of the parallel lines, as shown in the drawing, the actual collagen fibrils in SZ are distributing themselves in a 2D plane that runs in parallel with the articular surface of the tissue block. Consequently, this type of sample rotation cannot accurately determine the influence of the collagen fibrils in SZ to T2 anisotropy, a phenomenon that has been termed as the `ambiguity of the surface collagen orientation' in the MRI literature .
This work aimed to be the first quantitative study in microscopic MRI (μMRI) that characterizes the superficial fibrils in articular cartilage in its near native conditions (intact tissue blocks in aqueous solution) by rotating the specimen in such a way that the normal axis of the tissue surface remains stationary (Fig 1b). To resolve the superficial zone of articular cartilage, which is about 60-100μm thick in this type of canine cartilage , the transverse resolution in the μMRI experiments was fine tuned to 13μm per pixel. To unambiguously study the fibril structure using T2 anisotropy, the specimens were rotated from 0° to 180° at a step of 15°. Quantitative T2 images were obtained at each rotation and only the angular dependent features in the cartilage results were interpreted, based on fitting the experimental data to a schematic model.
Six articular cartilage specimens were harvested from the central load-bearing area of two canine humeral heads and kept at -20 °C until the μMRI experiments. These dogs were skeletally mature and healthy, sacrificed for an unrelated experimental study. Each specimen, which consisted of the full thickness of the articular cartilage still attached to the underlying bone, had the dimensions of about 1.75×2×10 mm for the rotation experiment as in Fig 1a (termed as the axial rotation) and trimmed afterwards to about 1.75×2×2 mm for the rotation experiment as in Fig 1b (termed as the planar rotation). The specimens were bathed in physiological saline with 1% protease inhibitor (Sigma, Missouri) and sealed in precision glass tubes with an internal diameter of 2.34 mm.
A homemade 3mm solenoid coil was used for the μMRI experiments, which contains a rotation device to orient the specimen . Two different types of rotations were used in the T2 anisotropy experiments. First, the samples were rotated so that the normal axis of the specimen (which is in parallel with the long axis of the fibrils in RZ) was at 0° and 55° (the so-called magic angle) with respect to the external magnetic field B0 (Fig 1a). This axial rotation has been used extensively in our lab for over 15 years to probe the T2 anisotropy in articular cartilage [29, 31, 35]. Second, the same samples after trimming were rotated so that the normal axis of the specimen remained stationary but a 2D tissue plane that was in parallel with the articular surface was rotated for 180° at a step of 15° (θ in Fig 1b). A 1.0mm or 0.5mm thick imaging slice was selected at the middle of the specimen for the two rotation experiments shown in Fig 1a and 1b, respectively. The direction of the B0 was horizontal on the humeral surface, in parallel with `a line' between the greater and lesser tubercles of a humeral joint. To ensure that the T2 anisotropy at different orientations were obtained from the same location of the specimen, the orientation of the imaging slice was set to follow the rotation of the specimen in the planar rotation. Consequently, the region of the specimen being slice-selected in the μMRI experiments remained stationary in the 2D images. The in-plane pixel size, which was across the depth of the cartilage tissue, was 13 μm in both types of rotation.
Microscopic MRI experiments were conducted at room temperature on a Bruker AVANCE II 300 NMR spectrometer equipped with a 7-Tesla/89-mm magnet and micro-imaging accessory (Bruker Instrument, Billerica, MA). At each sample rotation, a quantitative T2 imaging experiment was performed using a CPMG magnetization-prepared T2 imaging sequence . The echo spacing in the CPMG T2-weighting segment was 1ms. The five echo delays were 2, 6, 20, 50, 110 ms respectively for the five T2-weighted images in the planar experiments. The five echo delays were 2, 4, 10, 30, 60 ms and 2, 14, 36, 60, 120 ms for the five T2-weighted images at 0° and 55° respectively in the axial experiments. From these T2-weighted intensity images, the T2 relaxation map in cartilage was calculated by a single exponential fitting of the data on a pixel-by-pixel basis, which assumes that there is only one T2 component in cartilage . The imaging parameters were: the imaging echo time 7.2 ms, the field of view 0.32 cm × 0.32 cm, the imaging matrix size 128 × 256 (256 was in the readout direction which was the same as the normal axis of cartilage surface), the spectral bandwidth 50 kHz (corresponding to a readout sampling dwell time of 20 μs). 0.8 ms and 0.507 ms hermite shape pulses were used as excitation and refocusing pulse in imaging segment, respectively. The repetition time (TR) of the imaging experiment was 2 seconds.
To interpret the T2 anisotropy at different depths in cartilage during the planar rotation, a model of collagen distribution was developed, as shown in Fig 2. To be consistent with the practice in the literature [37-39], the anisotropy of the T2 relaxation rate R2 (R2 = 1/T2) was used, which follows the square of the dipolar interaction, R2 (1-3cos2θ)2,
where θ is the angle between the B0 direction and the direction of the dipole moment. Consequently, the following equation was used for any given distribution function f(θ), where A and B are two `amplitude' constants representing the isotropic distribution and the anisotropy distribution of the collagen matrix in the calculation respectively. The modeling was implemented with the MatLab codes (MathWorks, Natick, MA).
Fig 3 shows the T2 results from a representative cartilage-bone plug where the rotation method is axial (cf Fig 1a), i.e., the normal axis of the tissue block was set at 0° and 55° respectively. The T2 images and profiles in Fig 3 are highly consistent with our previous published results[12, 31]. As discussed earlier in the Introduction, this type of rotation is capable of probing the collagen fibrils in RZ unequivocally while leaving some ambiguity for the collagen fibrils in the 2D SZ[29,31].
Fig 4 shows the T2 results from a representative cartilage-bone plug where the rotation method is planar (cf Fig 1b), i.e., the normal axis of the tissue block remained stationary. Since the orientation of the imaging slice followed the rotation of the tissue block, the specimen appeared unchanged in all images shown in Fig 4a. Fig 4b shows the T2 profiles from the 1D array of T2 data at the centers of all specimen images. While the T2 profiles for the deep cartilage remain essentially identical, the variation of T2 in the first half of the tissue clearly demonstrates the effect of specimen rotation, not only in SZ, but also in TZ. These trends are illustrated in Fig 4c, which show the biggest difference in T2 (ΔT2 among all angles during the planar rotation as a function of tissue depth. An interesting feature in Fig 4c is its two peaks, one at the 0μm (the articular surface) and the other at 160μm, approximately the interface between TZ and RZ. The minimum between the two peaks occurs at about 100μm, which is approximately the center of TZ based on our previous PLM study, using the nearly identical specimens . Comparing with the well-organized fibril bundles in the deep part of RZ, which has the smallest ΔT2, the local minimum in TZ is still significant.
By plotting the T2 values as a function of the specimen rotation angle at each pixel location (every 13μm) along the tissue depth, we have examined the characteristics of the T2 anisotropy in cartilage due to the planar rotation, which are shown in Fig 5. Several distinct features can be identified. Firstly, the T2 anisotropy for the SZ fibrils (Fig 5a) is clearly periodic, which demonstrates that the distribution of the collagen fibrils in the 2D plane of SZ is not random (for the size of our image voxel). Secondly, the T2 anisotropy for the TZ fibrils (Fig 5b) seems to be approximately doubled its periodicity with respect to that of the SZ fibrils. Thirdly, the T2 anisotropy for the initial part of the RZ fibrils has a periodicity that seems inverse to that in SZ (eg, T2 at 195μm in Fig 5c vs T2 at 0μm in Fig 5a). Fourthly, the T2 anisotropy in RZ become weaker as one moves towards the deeper part of the tissue, which agrees with the observation in PLM experiments that the collagen fibrils in the deep zone become more organized towards the cartilage-bone interface . Finally, T2 is nearly isotropic for about one half of the thickness of the tissue, the lower half of the articular cartilage above the bone. This final feature attests two facts: (1) our experiments were highly reproducible since each pixel in the figure came from a different and independent T2 experiment, and (2) most of the fibrils in the deep part of RZ are indeed parallel with the normal axis of the tissue block, which has been illustrated earlier by the mapping of collagen angles in PLM .
To interpret the periodicity of the collagen fibrils during the planar rotation, a model of the collagen distribution in the planar rotation was developed (Fig 2). The two amplitude parameters in this model (a1 and a2) have arbitrary units, reflecting the `amount' of the fibrils that play an active role in the T2 anisotropy. Since each fitting (at each tissue depth) was carried out independently, we did not seek a quantitative correlation among all the amplitude terms along the tissue depth. The trends of the fitting parameters in the fibril orientation along the tissue depth are shown in Fig 6, with four representative fittings and their graphical interpretations as shown in Fig 7.
Several features can be identified: (a) One bundle of fibrils is sufficient to fit the T2 anisotropy in SZ and RZ (i.e., a2 = 0), while two bundles are needed for the fitting of the TZ fibrils. (b) For the SZ fibrils (Fig 7a), the overall orientation of the fibril bundle (Φ) is about 10° - 15° relative to B0, which means that this bundle of fibrils in SZ has an averaged orientation that is approximately transverse on the humeral head (i.e., in parallel with B0). (c) For the surface part of SZ (Fig 7a), the `span' of this fibril bundle (δ1) is approximately 75° (varying slightly from sample to sample in our group of six specimens), which is the origin of the T2 anisotropy in SZ. (d) When one moves into the deep part of SZ, and continues into TZ, the span of this fibril bundle (δ1) keeps decreasing, and the need for the second fibril bundle becomes apparent in the fitting (Fig 7b, Fig 7c). (e) Around the center of TZ, the best fitting of the experimental data requires the exchange of two bundles' amplitudes from 2:1 (Fig 7b) to 1:2 (Fig 7c). (Graphically, the switching of the bundle amplitudes changes the heights of the two peaks from `low-high' in Fig 7b to `high-low' in Fig 7c.) (f) For most of RZ, the T2 anisotropy has a 90° phase difference in orientation (Φ) with respect to that in SZ (Fig 7a vs Fig 7d). (g) When one gets into the deep part of RZ, T2 value reduces significantly and T2 anisotropy eventually disappears (Fig 5c). This fact confirms that the deep part of RZ indeed contains very organized fibrils, in parallel with the rotation axis.
The organizational structure of the collagen matrix in articular cartilage causes the non-zero averaging of the dipolar interaction among the protons in the tissue, which results in the motional anisotropy in the dynamics of water protons, represented by the T2 anisotropy in proton MRI [5, 29, 35]. When the normal axis of the cartilage surface is at 55° with respect to B0 (the magic angle), the minimization of the dipolar interaction restores the T2 relaxation to approximately isotropic (for the RZ tissue), which results in the vanishing of the laminar appearance of articular cartilage in MRI . In addition, since articular cartilage has multiple histological zones across its thin depth (thickness), the T2 anisotropy in articular cartilage is strongly depth-dependent , which mandates the use of high resolution in cartilage imaging. Furthermore, the variation of the fibril orientation in different histological zones implies that any simple rotation alone (either axial or planar, as shown in Fig 1) cannot probe the complete characteristics of the collagen fibril architecture in articular cartilage.
Since the fibril integrity of the surface structure of cartilage is of great importance to the load-bearing ability of the tissue , SZ has been called the `skin' of the tissue . Consequently, one of the early degenerative changes in the development of cartilage diseases is the surface fibrillation or the disruption of the superficial fibrils, which can currently be observed by clinical arthroscopy. This microscopic imaging study concerns the non-invasive monitoring of the averaged architectural properties of the collagen matrix in SZ of cartilage on a scale of tens of microns, a length scale that could be too small for an arthroscopic visualization.
A unique feature of articular cartilage is the so-called split-line pattern over a joint surface, which can be produced on some joint surfaces following pricking of the tissue using a sharp pin [7, 14, 16, 27, 28, 40]. Although the precise mechanism that contributes to the split lines on the joint surface has long been debated in literature [7, 14, 16, 27], the formation of the split-line pattern must be related in some way to the anisotropic properties of the surface fibril architecture, not necessarily just the superficial layer of the tissue but also the tissue beneath the superficial layer. This is because there is no abrupt boundary that separates any two histological zones in cartilage, and the definition of individual zones in histology merely represents a conceptual `discretization' of the continuous functions (the 3D fibril orientational structure) . The gradual decrease in T2 anisotropy from the surface tissue to the deep tissue in this study (Figs 3 and and4)4) clearly supports this understanding of fibril orientation; there exists a certain degree of collagen alignment and organization in all zones of articular cartilage [5, 16, 40, 41].
In this work, the placement of the tissue block is in such a way that the B0 direction is approximately transverse to the humeral head. Among all six specimens, the orientation of the overall fibril bundle (Φ) in the superficial zone was nearly identical, at about 10° - 15° respect to B0. (In our laboratory during the last fifteen years, we have never used a circular punch, but a table saw to harvest the rectangular tissue blocks from the animal joints. Special attention has always been paid to the orientation of the individual specimen on the original joint surface.) Hence the surface fibrils at the central load-bearing area of canine humeral heads must have an averaged anisotropy that is approximately in parallel with B0. For this type of canine humeral heads, we had carried out multiple tests to determine the split-line patterns on the intact surfaces using nearly identical joints in the past. However, we have never obtained a split-line pattern as clear as those shown in the literature [22, 42]. If one would consider this lack of clear split-line patterns as an indication of a random distribution of collagen fibrils on the surface, this current study clearly indicates that this is not true.
The architecture of collagen fibrils in articular cartilage has been described by several schematic models [4, 14, 16-18, 20, 21, 40, 43-46], the first being the arcade model of Benninghoff . Although these models differ in fine details, the overall orientation of the fibrils is consistent: the collagen fibrils arise in a radial manner from the subchondral bone, become oblique in TZ, and finally reach SZ in a transverse manner.
In any imaging experiment, unless one has the ability to visualize each individual fibril (as in electron microscopy), an image voxel (a small cube) contains numerous collagen fibrils and other molecules. It is the averaged property of the collagen matrix within the voxel that influences the anisotropy of T2 relaxation in MRI during any rotation experiment . For anyrotation experiment, one can separate this averaged property of collagens into two components: an axial component that is in parallel with the rotation axis and a planar component that is distributed on the 2D rotation plane (Fig 2). The anisotropic influence to the T2 relaxation is entirely due to an anisotropic distribution of the planar component. Any axial component, as well as any isotropically distributed planar component, has no effect on the T2 anisotropy during a rotation experiment.
During the axial rotation experiments (Fig 1a), we have in the past confirmed the general trends of the fibril orientation in both μMRI and PLM experiments [12, 29]. During a planar rotation experiment (Fig 1b), since the deep fibrils in RZ have little planar components, the planar rotation of an image voxel in RZ causes little T2 anisotropy. In contrast, since the surface fibrils have the most planar components, a planar rotation of any surface voxel can cause a significant T2 anisotropy. Approaching TZ from the cartilage surface, the planar component in any voxel decreases in `quantity' while the axial component increases. The contrary is true when approaching TZ from the deep tissue: the radial component in any voxel becomes smaller and the planar component becomes larger. In the middle of TZ (between Fig 7b and Fig 7c), the axial and planar components switch their dominance, which results in the exchange of the ratio of the fibril amplitudes in the data fitting from 2:1 to 1:2 in Fig 7.
Note that there is a 90° difference in the overall orientation angle Φ of the fibril bundles between SZ and RZ in Fig 7. This 90° difference in Φ is not the 90° difference between the long axis of the collagen fibrils in SZ and RZ. This 90° difference in Φ is the difference in the `projection' of the fibril architecture on the rotation plane between these two zones. For SZ, this architecture contains mainly the long axis of the fibrils; for the upper part of RZ, this architecture mainly contains the `anisotropic projection'of the bending radial fibrils on this plane, similar to looking at a bending arrow when it is pointed toward the viewer. Interestingly, Mollenhauser et al  also made a nearly identical observation in a small-angle X-ray diffraction experiment - there is an approximate 90° difference in the averaged fibril orientation in the parallel planes between SZ and RZ in articular cartilage from human talar dome.
In summary, this μMRI study uses the dynamic anisotropy of water molecules in articular cartilage to probe the architecture of the collagen matrix mainly in SZ and TZ of cartilage non-invasively. A complex set of distinct and depth-dependent patterns were found along the depth of the tissue at every 13μm pixel resolution. This work demonstrates that the fibril organization at SZ of cartilage is not random and that the orientational transition of the fibril structure in articular cartilage is not sudden along its depth. A collagen architecture model was formulated to interpret the experimental data. With future improvements in the resolution of clinical MRI , this work suggests a potential mechanism that might be used to detect the fibrillation on the surface of articular cartilage in vivo, which signals the clinical initiation of the tissue degradation and joint diseases.
Y Xia is grateful to Drs. C Les and H Sabbah (Henry Ford Hospital, Detroit) for providing the canine joints, and to the National Institutes of Health for the R01 grants (AR 45172, AR52353). The assistance in specimen harvesting by Mr. Farid Badar is greatly appreciated. The authors thank Miss Carol Searight (Dept of Physics, Oakland University) for editorial comments.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.