Quantitative μMRI and PLM Study of Rabbit Humeral and Femoral Head Cartilage at Sub-10μm Resolutions

Abstract

This study aimed to establish the baseline characteristics in humeral and femoral cartilage in rabbits, using quantitative MRI relaxation times (T2, T1ρ, T1) at 9.75μm and 70-82μm pixel resolution, and quantitative PLM measures (retardation, angle) at 1.0μm and 4.0μm pixel resolution. Five intact (i.e., unopened) shoulder joints (the scapula and humeral heads) and three femoral heads of the hip joints from five healthy rabbits were imaged in MRI at 70-82μm resolution. Thirteen cartilage-bone specimens were harvested from these joints and imaged in μMRI at 9.75μm resolution. Subsequently, quantitative PLM study of these specimens enabled the examination of the fibril orientation and organization in both intact joints and individual specimens. Quantitative MRI relaxation data and PLM fibril structural data show distinct features in tissue properties at different depths of cartilage, different in individual histological zones. The thicknesses of the histological zones in μMRI and PLM were successfully obtained. This is the first correlated and quantitative MRI and PLM study of rabbit cartilage at sub-10μm resolutions, which benefits future investigation of osteoarthritis using the rabbit model.

Introduction

Articular cartilage is a layer of avascular and aneural tissue on the articulating ends of long bones in joints ^{1; 2}. Cartilage degradation as a result of aging or trauma is the hallmark in osteoarthritis (OA) and other arthritis, which affects millions of adults. At least two features of arthritic diseases have so far prevented a successful detection of early OA in the clinics. First, the development of OA can span over many years, beginning with increased synovial inflammation and an upregulation of degradative enzymes ^{3; 4}. Since the tissue is avascular and aneural, the early degradation remains silent to the patients. By the time pain and stiffness are felt, opportunities for successful interventions could have been missed. Second, articular cartilage is a thin layer of tissue that has unique depth-dependent properties across its thickness, which is commonly sub-divided into three sub-tissue zones (the superficial zone SZ, the transitional zone TZ, the radial zone RZ). This non-uniform structural variation across its thin thickness implies that any diagnostic technique would benefit from high resolution in imaging; otherwise a volume averaging in any imaging pixel ⁵ could mask any chance for early detection.

The use of animal models plays a vital role in OA research, since a degradation can be initiated by a particular event. This certainty in the timing and type of the disease initiation would enable careful studies of molecular and pathological changes during the subsequent degradation process. Although canines are likely the most studied OA model ^6–9, the use of canines is being discouraged in recent years due to societal concerns and cost effectiveness. Among other animal models ^8–11, the use of rabbits is becoming popular because of its close phylogenetical and genetic resemblance to humans, sufficient thickness for easier creation of local defects (better than small rodents), relative inexpensiveness, and less societal sensitivity ^12–21.

This study aims to establish the characteristic T2, T1ρ and T1 relaxation times in humeral and femoral cartilage in rabbits using quantitative microscopic MRI (μMRI) and the quantitative fibril orientation (angle) and fibril organization (retardation) using polarized light microscopy (PLM), both at the highest possible resolutions. Humeral head cartilage and femoral head cartilage come from the shoulder and hip joints respectively - both belonging to a ball-and-socket type of joint. The use of rabbit humeral cartilage capitalizes on our extensive experience in the studies of canine humeral cartilage by multidisciplinary imaging at microscopic resolutions, including μMRI ²², PLM ²³, Fourier-transform Infrared imaging (FTIRI) ²⁴, and biomechanics imaging ²⁵. In these studies, we have established the quantitative correlations ²⁶ among the physical, chemical, biological and biomechanical properties in cartilage. The use of femoral cartilage is motivated by the high prevalence of OA in hip joints ²⁷. The relaxation times in MRI are extremely sensitive to molecular motions in soft tissues and have been used extensively studied in both high-resolution MRI and clinical MRI for the detection of cartilage degradation ^28–30. Since μMRI and clinical MRI share similar scientific principles and engineering architecture, the findings in μMRI can directly benefit the detection and management of OA by clinical MRI, which is the standard diagnostic tool for soft-tissue damages in the clinics. The use of PLM provides the correlation of MRI data with the gold standard in histology, whose optical resolution can further verify the MRI findings. The establishment of these characteristic features of rabbit cartilage would provide a solid foundation for future utilization of the rabbit model in the OA investigation.

Materials and Methods

Sample preparation

Five healthy rabbits (12-15-weeks New Zealand White, male) were used in this study. The animals were sacrificed for an unrelated biomedical study (which was approved by the relevant Institutional Review Committee), frozen immediately at −80°C after necropsy, and thawed at 4°C before the specimen harvesting.

Five intact (i.e., unopened) shoulder joints (the scapula and humeral heads) and three femoral heads from the hip joints were harvested; each was placed in a glass tube and imaged by MRI at low resolutions (70-82μm pixel size). A cotton swab soaked with 154mM protease-inhibitor (PI) saline solution was placed in the tube to avoid tissue drying. After each intact joint/bone was imaged, a diamond saw was used to cut each humeral head into 5 slices (Fig 1 a, b) and each femoral head into 4 slices (Fig 1 c, d), where each slice was about 1.8mm thick.

Figure 1.

(a) A humeral head, where the top left and right corners are the greater and lesser tubercles respectively. After MRI of the intact shoulder joint, four cuts were made over each intact head, as shown in the photo. The two peripheral slabs were not used in specimen imaging because of its irregularity. The second and third slabs (each about 1.8mm thick) were selected for the quantitative μMRI and PLM. The rectangular shape marks the approximate location of the imaging slice in the high-resolution imaging. (b) The side view of a central slab cut into small specimen blocks, through the humeral head. (c) A femoral head, where a central slab is shown in (d).

Two cartilage-bone specimens were selected from each humeral head (Fig 1a), which resulted in n=10 for the humeral specimens. Similarly, one cartilage-bone specimen was harvested from each femoral head (Fig 1c), which resulted in n=3 for the femoral specimens. Each cartilage-bone specimen (each about 1.8×2×2.5 mm³) had the full thickness of articular cartilage still attached to the underlying bone. These specimens were bathed in the PI solution and imaged using the same MRI system at high resolution (9.75μm pixel size).

MRI experimental

The MRI system was a Bruker AVANCE II console and a 7Tesla/89mm magnet (Bruker, Germany). The intact joints and bones were first imaged in sagittal orientation using the standard multi-slice multi-echo sequence, where each joint/bone had 8 or 10 slices and each slice had 10 incremental echo time intensity images. This resulted in 80 or 100 intensity images for each joint/bone. Each slice had a thickness of 1mm with a gap of 0.5mm (humeral) or 0.8mm (femoral) between any two neighboring slices. The humeral joint imaging had an echo time (TE) of 6.64ms, a repetition time (TR) of 2.5s. The 2D acquisition matrix was 256×256 and the field-of-view (FOV) was 18×18mm², resulting in a 2D pixel size of 70.31μm. The femoral head imaging had TE of 7ms, TR of 2.20s. The 2D acquisition matrix was 256×256 and FOV was 21×21mm², resulting in a 2D pixel size of 82.03μm.

All cartilage specimens (10 humeral and 3 femoral) were imaged for quantitative T2, T1ρ and T1 at two orientations with respect to the external magnetic field B0, 0° and 55°, which allowed the tissue properties to be influenced by the max and min dipolar interaction ³¹. All quantitative imaging used TR of 2s, TE of 6.47ms, and the slice thickness of 0.8mm. The 2D matrix was 256×128, which was reconstructed into 256×256; the FOV was 2.5×2.5mm², resulting in a 2D pixel size for all cartilage specimens at 9.75μm.

To determine the depth-dependent T2 anisotropy across the full depth of rabbit cartilage, quantitative T2 imaging was repeated at a series of angular orientations with respect to B0 for two humeral specimens. The selected orientations form a series of discrete sampling points along the geometrical factor (3cos²θ–1)² that is proportional to the non-zero dipolar Hamiltonian.

Quantitative μMRI relaxation protocol

Quantitative T2, T1ρ and T1 imaging adapted the concept of magnetization preparation ^{31; 32}, which consisted of a head segment to prepare the magnetization and an imaging sequence where all timings were constant. The head segment for T2 was a CPMG sequence ^{31; 33}, where the echo times were 2, 8, 16, 38ms for both humeral and femoral specimens. The head segment for T1ρ was a spin-lock sequence at a strength of 2000Hz ³⁴, where the spin-locking pulse lengths were 2, 8, 16, 38ms for humeral and 2, 16, 40, 60ms for femoral. The head segment for T1 was an inversion-recovery sequence ³¹, where the inversion recovery times were 0, 0.8, 1.4, 3.5s for all specimens. Since only the timing of the head segment is altered during the experiments, the relaxation effect of the magnetization can be determined accurately.

Quantitative PLM protocol

After μMRI, 7 humeral cartilage-bone specimens and 3 femoral cartilage-bone specimens were fixated in 10% neutral buffered formalin overnight at 4 °C. A number of cartilage-bone slabs (5-6 mm thick) were cut at the center of the humeral head (at the location of the μMRI slices) across its width horizontally (hence containing both lateral and medial sides) and fixated. The fixated slabs and the cartilage specimens in μMRI were sent to a histology service (Yale Pathology Tissue Services, CT) to section into 6μm thin sections using the paraffin method. 30 sections from the humeral cartilage specimens, 6 sections from the femoral cartilage specimens, and 5 sections from the whole-width humeral head slabs were imaged using PLM.

The PLM system consists of a digital setup (Cambridge Research & Instrumentation, MA) that is mounted on a polarized light microscope (Leica, Germany). The output of the CCD camera was processed ³⁵ to generated two quantitative images: the optical retardation and the angular orientation. Both quantitative images describe different aspects of the collagen matrix in cartilage ²³. All cartilage specimens were imaged at the identical orientation under a 10x objective which yielded a pixel size of 1.0μm. All intact humeral slabs were imaged under a 2.5x objective which yielded a pixel size of 4.0μm.

Image and data analysis

On each T2 map of the intact humeral or femoral bone, a three-pixel-wide region of interest (ROI) was selected, which covered the entire thickness of cartilage and extended beyond the cartilage. The ROI was chosen at an identical location on the joint surface from which the small cartilage-bone specimens would be harvested for μMRI. From each ROI, one 1D profile along the cartilage depth was generated by averaging the three-column data. On each quantitative 2D relaxation image, twenty parallel neighboring columns of data were selected and averaged to obtain one 1D profile. (Note that 20 consecutive columns in μMRI at 9.75μm have approximately the same tissue width to 3 consecutive columns in the intact humeral cartilage resolution at 70.31μm.) On each quantitative PLM image, 195 parallel neighboring columns, which would have approximately the same width as the ROI thickness in μMRI, were averaged to produce one 1D profile.

Since the averaging from 2D ROI to 1D profiles in μMRI and PLM occurs perpendicular to the tissue depth, the spatial resolutions of the 1D profiles along the tissue depth are still 70.31μm for MRI of the intact humeral, 82.03μm for MRI of the femoral head, 9.75μm for μMRI of the cartilage-bone specimens, and 1μm for PLM of the intact joint/bone. All 2D images and 1D profiles, where each came from one independent imaging experiment, used the identical scaling and settings without any additional adjustment. The post-acquisition image and statistical analyses used the public-domain software ImageJ (NIH) and a commercial software KaleidaGraph (Reading, PA).

Results

MRI of the intact joint/bone and cartilage specimen at different resolutions

Fig 2a and 2c showed four intensity images for one intact humeral joint and one intact femoral head respectively, together with their quantitative T2 images. Note that the surface of the humeral cartilage was bordered by the synovial fluid (and potentially other tissues) since the shoulder joint was intact; in comparison, the surface of the femoral cartilage was bordered with air (which appeared black in the intensity images) since only the femoral head was imaged in the glass tube (due to the large size of the acetabulum of the hip). Fig 2b and 2d showed the intensity and T2 images for one humeral specimen and one femoral specimen respectively. The high-resolution images of humeral and femoral cartilage appear similar. A thin black line was more visible near the femoral surface (Fig 2d) than the humeral surface (Fig 2b), indicating a lower T2 in the surface of femoral cartilage.

Figure 2.

T2-weighted intensity images of an intact shoulder joint (a) and a humeral cartilage specimen (b) at different echo times, at the pixel resolution of 70.31μm and 9.75μm respectively. (c) and (d) show similar images from a femoral head, at the pixel resolutions of 82.03μm and 9.75μm respectively. The quantitative T2 images from each series of ten T2-weighted intensity images were shown on the most-right images. Since the direction of the magnetic field B₀ was vertically up in the figure, the orientation of the cartilage specimens was 0° to B₀ (Fig 2b, 2d), which was similar to the surface orientation of the tissue when the intact joint/bone was imaged in the magnet in sagittal view (Fig 2a, 2c). All T2 images were displayed with the same up/low limits (0-100ms).

Fig 3 compares the T2 profiles from the T2 images in Fig 2. Two features can be identified for humeral cartilage (Fig 3a). First, for the deeper part of humeral cartilage (about 75% of the total cartilage), T2 profiles at both resolutions match each other. Second, for the top 25% of the tissue, the low-resolution imaging missed the critical feature of the surface cartilage. The continuously increased T2 values towards the articular surface (the depth of 0) and beyond the articular surface (into the neighboring space) in Fig 3a clearly demonstrates the consequence of low-resolution MRI of cartilage, which averaged cartilage with the surrounding synovial fluid – two tissues that have very different T2 relaxation times. Similarly, the low-resolution MRI of femoral cartilage (Fig 3b) missed out the critical features of the surface cartilage (the top 25%), where cartilage signal was reduced because of the averaging with air near the surface. In contrast, μMRI can resolve the T2 characteristics of the surface tissue faithfully (Fig 3), which represent the superficial and transitional zones.

Figure 3.

The depth-dependent T2 profiles from the T2 images in Fig 2: (a) humeral cartilage at 70.31μm/pixel (open squares) and 9.75μm/pixel (solid dots) and (b) femoral cartilage at 82.03μm/pixel (open squares) and 9.75μm/pixel (solid dots), respectively. The depth scale is relative, with 0 marks the articular surface and 1 marks the cartilage-bone interface. The dashed lines fit through the T2 profiles in the low-resolution MRI. The vertical error bars represent the standard deviation of the data; the horizontal error bars in the low-resolution mark the pixel sizes. Note that in (b), μMRI T2 (0°) are shown till the tissue depth of 0.6, after which the image intensities become very low and the calculated T2s become unreliable.

Quantitative relaxation measurements at 9.75 μm resolution

Fig 4 shows the quantitative T2, T1ρ and T1 images and profiles of both humeral and femoral cartilage at 9.75μm resolution, where the same tissue block was imaged using the identical protocol at two different orientations in B0. These images and profiles reveal similarities and differences between humeral and femoral cartilage in rabbits, and between rabbit and canine humeral cartilage in the literature ^{23; 36}.

Figure 4:

The quantitative T2, T1ρ, and T1 images as well as their respective profiles at 9.75μm/pixel in humeral cartilage (a)-(c), and femoral cartilage (d)-(f). The solid dots are the profiles at the 55° specimen orientation, while the open squares are the profiles at the 0° specimen orientation with respect to the external magnetic field B₀.

Between humeral (Fig 4a–c) and femoral (Fig 4d–f) cartilages in rabbits, the relaxation images and profiles appear similar at 55° orientation, at which the dipolar interaction has the weakest influence to spin relaxation. This is likely due to the fact that both cartilages belong to the same type of ball-and-socket joint and hence broadly structured similarly. At 0° to B0 (which has the strongest dipolar influence to spin relaxation), the shape of T2 profile in femoral cartilage is more well defined and narrower than humeral cartilage, implying that the fibril structure needs higher resolution for further investigation.

Between the humeral cartilage in rabbit (Fig 4a–c) and canine in our previous studies using the same protocols ^{23; 36}, the depth-dependent characteristics of these relaxation times in rabbit cartilage match broadly the corresponding characteristics in canine cartilage, where the bell-shaped curve in the T2 profiles (Fig 4a, 4d) implies a classical three-zone structure in cartilage, which can be used to sub-divide cartilage thickness into multiple histological zones ^{23; 37}. One noticed that the T2 profile at the 55° was not as uniform in values across the tissue depth, which should have an anatomic origin at the fibril level in the tissue.

Using the established criteria ^{23; 38}, the zonal division of cartilage thickness was carried out on all μMRI images (Table 1). In both humeral and femoral cartilages, SZ is about 21μm thick, just over 2 pixels in μMRI at 9.75μm/pixel. This thin thickness of SZ in rabbit cartilage puts a high-resolution requirement on MRI and any other imaging tools in the study of rabbit cartilage. A peculiar feature in the humeral data is that the L blocks were consistently thicker than the C blocks, which suggests a topographical variation of cartilage thickness over the joint surface (L and C blocks were two adjacent tissue blocks, about 1.8mm apart on the joint surface). Table 2 summarizes the zonal averaged relaxation times, based on the zonal thickness data in Table 1. Clearly, the biggest zonal variations are T2 at 0°, while the zonal data in T1ρ and T1 have much less depth dependence and little orientational dependence.

Table 1.

Averaged zonal thicknesses based on the T2-0° profiles of (a) humeral cartilage (n = 10 specimens) and (b) femoral cartilage (n = 3 specimens). σ represents the standard error. % is the relative percent thickness of each zone based on the total thickness.

(a) Humeral cartilage
Specimen	Superficial Zone (μm)	Transitional Zone (μm)	Radial Zone (μm)	Total Tissue (μm)
1L	19.50	48.75	234.25	302.25
IC	19.50	48.75	224.25	292.50
2L	19.50	58.50	292.50	380.25
2C	29.25	48.75	234.25	302.25
3L	19.50	48.75	380.25	458.25
3C	19.50	48.75	360.75	429.00
4L	19.50	58.50	380.25	458.25
4C	19.50	48.75	302.25	370.50
5L	19.50	48.75	399.75	468.00
5C	29.25	58.50	370.50	458.25
Mean ± σ	21.45 ± 1.3	51.67 ± 1.6	317.90 ± 21.8	391.95 ± 22.5
%	6	13	81	100
(b) Femoral cartilage
Specimen	Superficial Zone (μm)	Transitional Zone (μm)	Radial Zone (μm)	Total Tissue (μm)
1C	19.50	29.25	253.50	302.25
2C	29.25	48.75	282.75	360.75
3C	29.25	39.00	292.50	360.75
Mean ± σ	26 ± 3.3	39 ± 5.6	276.25 ± 11.7	341.25 ± 19.5
%	8	11	81	100

Table 2.

Averaged orientation-dependent zonal T2, T1ρ and T1 relaxation times based on the zone division data in Table 1. (a) humeral cartilage (n = 10 specimens) and (b) femoral cartilage (n = 3 specimens). Data are in mean ± standard error. RZ-I and RZ-II are the top and bottom equal-halves of the radial zone.

(a) Humeral cartilage
Zones	T2 (ms)		T1ρ (ms)		T1 (s)
	0°	55°	0°	55°	0°	55°
SZ	20.46 ± 1.55	51.31 ± 3.87	86.32 ± 2.92	88.82 ± 3.97	1.25 ± 0.02	1.32 ± 0.02
TZ	37.96 ± 1.66	58.99 ± 2.75	92.49 ± 3.12	93.60 ± 3.52	1.21 ± 0.02	1.21 ± 0.01
RZ-I	12.04 ± 0.58	62.94 ± 2.91	90.84 ± 2.67	103.63 ± 2.67	1.26 ± 0.03	1.27 ± 0.02
RZ-II	7.24 ± 0.61	37.55 ± 1.93	53.97 ± 2.27	65.67 ± 1.85	0.95 ± 0.03	1.01 ± 0.02
(b) Femoral cartilage
Zones	T2 (ms)		T1ρ (ms)		T1 (s)
	0°	55°	0°	55°	0°	55°
SZ	17.16 ± 3.59	48.10 ± 7.47	68.34 ± 3.88	76.01 ± 5.77	1.18 ± 0.09	1.29 ± 0.08
TZ	33.30 ± 3.73	51.15 ± 1.82	76.30 ± 3.23	79.94 ± 4.72	1.21 ± 0.01	1.26 ± 0.04
RZ-I	10.90 ± 0.65	46.19 ± 3.60	82.32 ± 9.06	89.58 ± 1.06	1.23 ± 0.05	1.27 ± 0.03
RZ-II	4.30 ± 2.10	25.89 ± 2.50	35.63 ± 4.22	51.25 ± 2.19	0.75 ± 0.07	0.90 ± 0.06

Depth-dependent T2 anisotropy

To study the depth-dependence of T2 anisotropy, two humeral specimens were imaged at six different orientations with respect to B0. Fig 5a showed one set of six T2 profiles from six independent T2 experiments (from one L block), plotted together without additional scaling. These profiles clearly demonstrated that T2 anisotropy in rabbit humeral cartilage is strongly depth-dependent and orientational dependent, with the biggest anisotropy around 80μm to 160μm depths, representing the upper part of the radial zone. A close examination can be done by plotting T2 anisotropy at every tissue depth (every 9.75μm). Four such plots are shown in Fig 5b. It is clear that T2 anisotropy in rabbit humeral cartilage is prominent at 0μm and stays equally prominent at 9.75μm (not shown) – both pixels represent SZ cartilage. As the depth goes deeper, T2 anisotropy weakens and remains nearly identical within the depths of 29.25μm to 48.75μm, which signals TZ cartilage. However, T2 anisotropy remains noticeable even at its weakest (up to 40%), which reflects a residual fibril structure in TZ. From the depth of 68.25μm, T2 anisotropy becomes stronger again, reaching a max at 351μm (near the bottom of RZ). These pixel-by-pixel features in the depth-dependent T2 anisotropy reflect precisely the variations in the collagen organization across the entire thickness of cartilage ^39–41.

Figure 5:

(a) Depth-dependent T2 profiles of humeral cartilage at six orientations to B₀, which follow the geometric factor in the dipolar interaction, (3cos² θ-1)², that manipulates T2 relaxation in MRI. (b) The plots of T2 anisotropy (% change in T2 with respect to the 55° data, as a function of the specimen orientation) at four tissue depths (at a 9.75μm tissue depth step): 0μm represents the first layer of the superficial zone; 39.00μm is inside the transitional zone; 78μm marks the boundary between the transitional zone and radial zone; and 351μm locates deep in the radial zone. The solid line plots the (3cos² θ-1)² in the dipolar interaction that dominates the T2 characteristics in articular cartilage.

Quantitative PLM Results

The angle and retardation maps from the quantitative PLM represent the pixel-averaged fibril orientation and fibril organization of the collagen matrix in cartilage respectively. Fig 6 showed the retardation image of an intact humeral head. Several observations can be noted. First, there is a dark ‘line’ in the retardation image at ~50μm below the articular surface, which represents the location of the least organized fibers in tissue (the center of TZ). Second, this dark line gradually vanishes towards both peripherals, which suggests the collagen architecture in peripheral cartilage is different from the central region. Thirdly, humeral cartilage has strong topographical variation in its thickness, from ~250μm at the joint center to ~500μm at the peripherals. Although this thickness variation is significant, it is not as complex as in some other joints such as tibia (refer to Figure 1.10 in Chapter 1 of the book ²⁹). Finally, our cartilage specimens for μMRI and PLM were harvested from a slightly off-center location in humeral head, which explains the consistent thickness difference between our L specimens and our C specimens (Table 1).

Figure 6 :

The retardation images from an intact humeral head of the shoulder joint, imaged by polarized light microscopy with a 2.5x objective (4μm pixel resolution). The histological slice of the intact humeral head was imaged in four parts to cover the entire humeral head; the four quantitative retardation images were scaled using the identical limits (0 – 10nm); the images were pieced together in photoshop. The humeral head shows a strong curvature, flat in the central area and rounded towards both peripherals. (Some of the curvature could also be influenced by the histological process, after the underlying bone becomes soft in decalcification.) In the retardation images, large and small values indicate that the collagen fibrils are more or less ordered, which could include the information for the fibril diameters and fibril packing density.

Fig 7 showed the quantitative PLM results from the same humeral specimen from which the μMRI data were discussed (Fig 4). These PLM images from rabbit are very similar to the PLM images from canine cartilage ²³. Briefly, the depth-dependent retardation (Fig 7a) had its lowest values about 50μm below the articular surface, which indicates the center of TZ; the high retardation values at the articular surface indicate the fine fibril organization at the articular surface while the high retardation values in the deep cartilage represent the fibril structure of RZ (closely and more densely packed). In the angle image (Fig 7b), the blue and red colors in this color scale showed ~85° orientational difference between the SZ and RZ fibrils. The tide mark in this rabbit tissue was not quite visible as in some mature cartilage ²³. Femoral cartilage was also analyzed by PLM, which yielded similar results as in humeral cartilage. Fig 7c showed one set of quantitative PLM profiles (from Fig 7a and 7b), which were averaged over 6 tissue sections from one humeral specimen, together with the T2 profile from the same tissue block (Fig 4a). These profiles solidify the descriptions and conclusions for the angle and retardation images in the early part of this section.

Figure 7:

2D retardation (a) and angle (b) images in PLM, at 1μm pixel resolution, showing the quantitative results of one humeral cartilage specimen (the same specimen from which the μMRI data were shown in Fig. 3–5). In the angle images, a typical arch structure of the collagen matrix should have a nominal 90° orientational difference between the superficial and radial fibers, hence represented by the blue and red colors. The rectangular boxes in (a) and (b) show the location where the depth-dependent profiles were calculated and showed in (c). Also in (c) is the T2 profile from μMRI of the same specimen. Note that the oval shapes in the retardation image are the locations of the chondrocytes that used to present in cartilage. Since no protocol was used in histology to preserve the cells, the cellular components were washed out, which yielded black color in the retardation images (i.e., background values). In contrast, the same empty cellular spaces are often filled with random numbers in the angle images, which is due to the imaging algorithm that determines two quantitative numbers at each pixel location). These inaccurate assigned angle and retardation values in the cellular cavities contribute to and exaggerate the size of the error bars in the PLM profiles. The vertical dashed lines mark the locations of the three histological zones. AS and CBI in the tissue depth mark the articular surface and cartilage-bone interface respectively.

Micro MRI zones vs Histological zones correlation.

Using the established criteria ^{23; 38}, the zonal division of cartilage thickness was carried out on all PLM images from 30 histological sections (from 5 humeral specimens). PLM zones from the humeral specimens were compared with the corresponding μMRI zones from the same specimens (Table 3), together with the statistical results from the Student t-tests. Some disagreements can be noted for two thin zones (SZ and TZ) between μMRI and PLM, which is likely caused by the difference in the imaging resolution (9.75μm in μMRI, 1.0μm in PLM). In addition, the physics and engineering behind μMRI and PLM instrument are very different, which could also lead to the measurement biases in different imaging techniques. However, the correlations in the total tissue thickness and the thick radial zone are excellent (i.e., no statistical difference between μMRI and PLM zones were found). Similarly, the zonal division for femoral cartilage has also been carried out, where the μMRI and PLM results from the same femoral specimens were compared for their zonal thicknesses. In femoral cartilage, SZ was ~ 7-9 %, TZ 8-9 %, and radial zone was ~82% of the total cartilage (~341μm).

Table 3.

Correlations of the averaged zonal thickness in humeral cartilage between μMRI (n = 5 cartilage specimens) and PLM (n = 30 histological sections from 5 μMRI specimens at 6 sections per specimen). Data are in mean ± standard error. The t-probabilities were calculated using Student t-test analysis.

	μMRI Thickness (μm)	PLM Thickness (μm)	t-probability
SZ	21.45 ± 1.95	29.09 ± 2.36	0.04
TZ	52.65 ± 2.38	31.69± 1.24	<0.001
RZ	366.60 ± 16.82	379.68 ± 15.76	0.58
Total Tissue	440.70 ± 18.13	440.46 ± 16.82	0.95

Discussions

Among many MRI parameters that could be used to study cartilage degradation, T2, T1ρ and T1 relaxation times are currently the most commonly used parameters in clinical MRI of OA ²⁸. These parameters are sensitive to molecular motions and their interactions with the tissue surroundings. This study has successfully quantified T2, T1ρ and T1 in both humeral and femoral cartilage in rabbits at 9.75μm resolution. To the best of our knowledge, these μMRI results are the first quantitative characterization of the entire depth of humeral and femoral cartilage in rabbits at a sub-10μm resolution. The complementary use of quantitative PLM in this study further enhances the μMRI results. Given our experience in canine studies ^{31; 38; 42–44}, the depth-dependent features of T2, T1ρ, and T1 in rabbit cartilage were found to be similar to those in canine cartilage. This similarity implies a classical three-zone structure in rabbit cartilage, and provides possibilities of doing comparative analyses between canine studies and leporine studies.

Challenges in studying rabbit cartilage – Age, thickness, topographical variations

There are at least three challenges in using the rabbit model for OA research. First, the growth plates in rabbit joints do not close until 6-9 months, beyond that animals can be considered skeletally mature ^45–48. The animals in this study (also in many studies in the literature) were 3-4 months old, hence skeletally immature. However, the same studies ^{45; 48} also noted that the collagen matrix of articular cartilage in mature rabbits appeared to be “similar” to that at 12–14 weeks. In contrast to this similarity between immature and mature rabbits, some previous μMRI and PLM studies of canine humeral cartilage have found vastly different characteristics between young and mature canines ^{36; 49}. In any case, the maturity of rabbit tissues should be noted for any study that creates an injury or intervention in the tissue and waits for a time period for its development, as well as for any comparative analyses of the rabbit data in the OA literature.

Second, rabbit cartilage is thin, a few hundreds of micron in thickness in the largest joints. This thin thickness possesses challenges for several types of studies, including diagnostic imaging and surgery. In this imaging study, based on the birefringent information at the 1.0μm optical resolution, we determined the zonal thicknesses for the central part of humeral cartilage, where SZ and TZ were just over 30μm each. These findings demonstrate that high resolution (e.g., approximately 10μm) is required to analyze zonal differences in the rabbit cartilage. One subtle but important reason to use two resolutions to study rabbit cartilage in this MRI study comes from the imaging scaling in MRI of cartilage ⁵. In this study, for a nominal thickness of 300-400μm in rabbit cartilage, there were approximately 4-5 pixels in our low-resolution MRI of intact bone heads (70-82μm pixel size) and 35 pixels in our μMRI of tissue blocks (9.75μm pixel size). In contrast, although human knee and hip cartilage are much thicker, clinical MRI also has a bigger pixel size, which commonly resolves human cartilage in 3-5 pixels. Consequently, our low-resolution MRI of rabbit cartilage mimics almost exactly the tissue averaging situation in clinical MRI of human cartilage ⁵. Hence our experimental methodology and approach can be adapted to perform a large number of OA projects using the rabbit model.

Thirdly, most of the tissue characteristics in cartilage change topographically over a joint surface, likely due to the consequence of load-bearing and motional patterns for a particular joint and a particular animal. Even for a ball-and-socket joint such as the humeral head in rabbit (which has considerably simpler variations than a knee joint), the thickness of cartilage ranges from about 250μm at the center to twice the thickness at the peripherals (Fig 6). These topographical variations exist on any joint surface ^{36; 49; 50}. In this study, it is clear that a slightly off-centre harvesting of the cartilage-bone specimens from the humeral head has led to noticeable variations in the quantitative measurement of the tissue properties by both μMRI and PLM (Fig 4–7). In μMRI results, for example, we can attribute the non-uniformity of the T2 depth profile at the 55° (Fig 4) to be the results of this off-center specimens, where the collagen organization was not entirely perpendicular to each other between the superficial and radial fibrils, which was also confirmed by the PLM results (Fig 7c). These topographical variations can make the interpretation of the individual study difficult and the comparison among different studies in the literature challenging.

Similarities between Leporine and Canine cartilage

This study demonstrates that the depth-dependent characteristics of the T2, T1ρ, and T1 relaxation times in humeral and femoral cartilage in rabbits are similar to those in canine cartilage^{23, 28}. This similarity implies a classical three-zone structure in the rabbit cartilage, and possibilities of doing comparative analyses between canine studies and leporine studies.

Limitations of This Study

This study has several experimental limitations. First, five animals were used in this study, from which five intact shoulder joints and three femoral heads from the hip joints were harvested. These animals belong to a total of over 20 nearly identical rabbits that we obtained from the same source and studied in our lab in different projects. We can confirm that the μMRI and histology features from the tissues of other animals were highly consistent with the presentation in this study. Second, quantitative imaging can be influenced by the topographical variation of cartilage, i.e., the surface site where the specimen is harvested and studied. This variation could be significant for joints such as knee. In rabbit humeral and femoral heads, by contrast, our result indicates that the topographical variation has smaller influence on cartilage properties and hence the limited sampling sites in this study can represent most of the joint surface (apart from the peripherals).

Conclusion

This study has successfully established the baseline characteristics in the humeral and femoral cartilage in rabbits, using quantitative MRI relaxation times (T2, T1ρ, T1) at 9.75μm and 70-82μm pixel resolutions, and quantitative PLM measures (retardation, angle) at 1.0μm and 4.0μm pixel resolutions. To the best of our knowledge, these μMRI results are the first quantitative characterization of the entire depth of humeral and femoral cartilage in rabbits at a sub-10μm resolution. The quantitative PLM imaging in this study (at 1.0 and 4.0μm/pixel) further verify and enhance the high resolution MRI results, which provides the anatomic knowledge that can be correlated with that from the non-invasive MRI. Since the low-resolution MRI settings in this study mimic the tissue averaging situation in clinical MRI of human cartilage in the clinics, this study demonstrates an animal model and an instrumental platform from which various issues in human OA research can be investigated. In addition, since μMRI in this study has one of the highest MRI resolution in cartilage research, our results provide the baseline data for any future use of the rabbit model in bench-top investigation of OA. The quantitative multi-resolution imaging approach in this study demonstrates a translational pathway from pre-clinical bench-top studies to bedside early OA diagnosis in the clinics.