Microstructural Remodeling of Articular Cartilage Following Defect Repair by Osteochondral Autograft Transfer

Abstract

Objective

To assess collagen network alterations occurring with flow and other abnormalities of articular cartilage at medial femoral condyle (MFC) sites repaired with osteochondral autograft (OATS) after 6 and 12 months, using quantitative polarized light microscopy (qPLM) and other histopathological methods

Design

The collagen network structure of articular cartilage of OATS-repaired defects and non-operated contralateral control sites were compared by qPLM analysis of parallelism index (PI), orientation angle (α) relative to the local tissue axes, and retardance (Γ) as a function of depth. qPLM parameter maps were also compared to ICRS and Modified O’Driscoll grades, and cell and matrix sub-scores, for sections stained with H&E and Safranin-O, and for Collagen-I and II

Results

Relative to non-operated normal cartilage, OATS-repaired regions exhibited structural deterioration, with low PI and more horizontal α, and unique structural alteration in adjacent host cartilage: more aligned superficial zone, and reoriented deep zone lateral to the graft, and matrix disorganization in cartilage overhanging the graft. Shifts in α and PI from normal site-specific values were correlated with histochemical abnormalities and co-localized with changes in cell organization/orientation, cloning, or loss, indicative of cartilage flow, remodeling, and deterioration, respectively

Conclusions

qPLM reveals a number of unique localized alterations of the collagen network in both adjacent host and implanted cartilage in OATS-repaired defects, associated with abnormal chondrocyte organization. These alterations are consistent with mechanobiological processes and the direction and magnitude of cartilage strain.

INTRODUCTION

The ideal biomimetic treatment for articular cartilage defects would restore normal tissue structure and function. Current treatment options for full-thickness focal defects in human articular cartilage include microfracture, osteochondral auto- and allo-grafting, and autologous chondrocyte implantation (ACI), as well as emerging treatments such as matrix-associated chondrocyte implantation¹. While defect filling by regenerative methods is gradual, that by osteochondral autografts or allografts is immediate with the mature articular cartilage possessing structure, composition, and mechanical function that vary in the superficial (SZ), middle (MZ) and deep zones (DZ). Osteochondral autograft transplantation surgery (OATS) has site-specific success rates of 74–92% based on clinical scores 2 months-11 years post-operatively². However, practical limitations in the analysis of human repair make it difficult to assess mechanisms leading to success or failure.

After treatment of focal articular defects, the surrounding host tissue can exhibit “cartilage flow”, the permanent deformation of cartilaginous tissue directed toward a defect or unstable implant under the influence of mechanical loading³. Although integrative healing occurs within subchondral bone, full-thickness clefts tend to remain at the interface between host cartilage and repair tissue⁴, especially with osteochondral grafts¹. Cartilage flow at such interfaces is prominent in weight-bearing regions and less filled defects⁵. When cartilage defects are loaded in compression⁶ and shear^{7, 8}, the lack of lateral support at the defect site leads to increased strain amplitude in the remaining cartilage, particularly near the edge. If cartilage flow is driven by such mechanical deformation, structural reorientation in the host and implanted cartilage near the cleft would be expected.

Post-implantation, osteochondral graft cartilage may be maintained or else undergo deterioration, associated with altered cell and matrix properties. Retrievals of failed human grafts variably exhibit regional alterations, including fibrillation, loss of viable cells, and loss of glycosaminoglycans at the cartilage periphery and SZ, extending to the upper DZ, often with viable cells in deeper regions^{9, 10}. Animal models indicate that repair success is associated with maintenance of chondrocyte viability within the graft^{11, 12} and graft placement such that the articular surface is neither substantially proud nor sunk^13–15. Histochemical and immunohistochemical analyses of cartilage provide substantial information about the quality of cells and matrix molecules, but not the collagen network organization, especially across the breadth of defect repair and host cartilage.

The collagen network organization of articular cartilage can be assessed by polarized light microscopy (PLM), applied qualitatively or quantitatively (qPLM). The orientation of the anisotropic collagen network, tangential in the SZ and vertical in the DZ, is manifest by birefringence when cartilage sections are viewed by PLM with crossed-polarizers oriented at 45°, between the horizontal and vertical axes¹⁶. Such PLM analysis has been used to assess zonal organization qualitatively^17–19 and overall cartilage structure after repair and in osteoarthritic degeneration, with semi-quantitative scores^{20, 21}. With additional PLM measures, cartilage structural features can be quantified by the qPLM parameters, anisotropy (as parallelism index, PI), orientation (as α), and birefringence (as birefringence and retardance parameters, B and Γ, respectively, the latter scaled by sample thickness²²). Consistent with the proposed collagen network structure of articular cartilage²³, the DZ of normal mature articular cartilage is highly aligned (PI~0.8) and vertically-directed (α~85°), the MZ is less aligned (PI~0.47) and less vertical (α~60°), and the SZ has an intermediate alignment (PI~0.57) with the most horizontal orientation (α~22°)²⁴. In addition, qPLM parameters vary during cartilage repair, development, aging, and degeneration^24–27. In osteochondral autografts evaluated 4 days after transplantation from the medial to lateral femoral condyle of adult rabbits, SZ alterations included loss of birefringence, associated with immunohistochemical evidence of collagen damage²⁸. After ACI repair of the patellofemoral groove of immature pigs, PI was variable but higher at 3 than 12 months²⁷. Additional qPLM analysis of cartilage repair may reveal relationships between structural alterations in the collagen network, traditional histological and histochemical indices, and load-induced deformation.

The hypothesis of the present study was that in OATS-repair sites, the collagen network of cartilage, as assessed by qPLM, deviates from the normal zonal structure, in association with histopathological indices of poor repair and co-localizing with cartilage flow at graft interfaces and cartilage remodeling or deterioration in the bulk. To test this hypothesis, the aims of this study were to: (1) characterize non-operated condyle cartilage for zonal collagen network microstructure, as SZ, MZ, and DZ thickness from α patterns, and zonal qPLM parameters (PI, α, and Γ); (2) map and quantify the effect of osteochondral autograft repair on zonal qPLM parameters as a function of region (proximal interface, central implant, distal interface) and post-operative time-point (6 and 12 months); and (3) compare spatial variations in PI, α, and Γ with traditional histological and histochemical indices of cartilage structure.

METHODS

Experimental Design

Histological sections were obtained from the repaired and contralateral non-operated stifle joints of 3–4 y.o. Spanish goats from a recently-described in vivo study of repair of cartilage defects¹³, and also from non-operated stifle joints of other age-matched goats. In the repair study, OATS grafts in the form of osteochondral cylinders (diameter=3.5mm, height=3.0 mm) were taken from the lateral trochlea (LT) of the operated knee (stifle) and grafted into similarly-sized defects in the medial femoral condyle (MFC). Both non-operated LT and MFC sites from contralateral knees and from other age-matched animals served as controls to the grafted donor recipient site. In the present study, a total of 32 osteochondral samples were analyzed from 13 goats (Table 1): (1) 8 samples from OATS-repaired MFCs (one sample from one knee of each of four animals at 6 months, and another four animals at 12 months), and (2) 24 samples from non-operated control joints (two samples per knee, one from the MFC and the other from the LT, from each of 12 age-matched animals, 6 animals at 6 months and 6 at 12 months). Seven of the twelve non-operated knees were contralateral to OATS-repaired knees; the remaining five were from age-matched goats. Each sample was fixed with paraformaldehyde, decalcified with 10% formic acid, and processed by paraffin embedding and sectioning to a thickness of 5 µm. All sections were taken in the mid-sagittal plane (see electronic supplement). One unstained section from each site was analyzed by qPLM in the present study, and this section was adjacent to or nearby those that were analyzed previously by histochemistry (Hematoxylin and Eosin, Safranin-O) and immunohistochemistry (types I and II collagen)¹³. ICRS²⁹ 85 and Modified O’Driscoll (MOD)^{30, 31} histopathology scores had been determined previously¹³ for each sample.

Table 1.

Study groups, histopathological grades, and qPLM scores of non-operated and repair sites

Group	Treatment	Site	Post-op. Period [mo.]	n	ICRS Score	Mod O'Driscoll score	qPLM raw score	qPLM weighted score
1A	Non-op	MFC	6	6	17.8(175–18.2)	27.8(27.4–28.1 )	8.1(5.0–11.2)	1.5(0.6–2.4)
1B	Non-op	MFC	12	6	17.6(17.2–18.0)	27.7 (27.3–28.1)	12.5(3.1–21.9)	1.4(0.2–2.5)
2A	Non-op	LT	6	6	17.6(17.4–17.9)	26.2 (25.9–26.6)	10.5(2.4–18.7)	1.7(0.8–2.6)
2B	Non-op	LT	12	6	17.9(17.6–18.1)	26.5(26.1–27.2)	14.4(3.0–25.7)	6.2(−1.7–14.0)
3A	OATS	MFC	6	6	7.9(5.1–10.7)	12.0(9.0–15.0)	69(34–103)	14(5–23)
3B	OATS	MFC	12	6	9.0(9.0–15.0)	13.1(8.2–18.1)	74(48–101)	15(6–23)
Comparison					p-value	p-value	p-value	p-value
1v2					0.98	0.37	0.94	0.93
2v3					<0 001	<0001	<0 001	<0 001
1v3					<0.001	<0.001	<0 001	<0.001

^*results (p-values) of Tukey post-hoc tests following ANOVA effect of group on(ICRS, MOD qPLM) scores, p<0 001.

To assess cartilage microstructure, qPLM was performed on unstained sections on a region (~10mm×3mm) encompassing graft and adjacent (proximal and distal) tissues. (see electronic supplement). Hydrated unstained sections were imaged with a polarized light microscope (Polam - 213 TE, LOMO America, Northbrook, IL) and digital camera essentially as described previously^{22, 26}. PI, α, and Γ were computed from raw image data and mosaiced with custom software (Matlab R2010a, The Mathworks, Natick, MA)

Statistics

Statistical analysis was performed with Systat v10.2 (Chicago, IL). Data are presented as mean and either SD (charts) or 95% confidence intervals (estimation uncertainties, when of interest). ANOVAs were used to compare in various combinations (see below) of site (non-operated LT and MFC, OATS MFC), time (6 months, 12 months), zone (SZ, MZ, DZ), and sublocation of OATS grafts (proximal, P; central, C; and distal, D) and adjacent host cartilage (AHC) to the grafts, which were all considered repeated-measures, except time (since data at different times was from different animals). Dependent variables assessed by ANOVAs included qPLM parameters (Γ, PI, α) and scores, and, for non-operated groups, zonal thickness. Before ANOVAs and Student’s t-tests were performed, groups were tested for assumptions of equal variance (Levene) and normality of data and residuals (Kolmogorov-Smirnov). If assumptions were violated, data were log- or arcsine-transformed (thereupon satisfying ANOVA and Student’s t-test assumptions). Before repeated-measures ANOVAs were performed, the sphericity of factors (with three levels) was assessed using Mauchly’s test. The assumption of sphericity was confirmed for sublocation and zone (p>0.05), and the Huynh-Feldt correction did not significantly alter p-values. To minimize the chance of a type I error, tests were limited to variations of interest (one-way ANOVAs only when the factor was significant by multi-way ANOVA, and not evaluating zone effects for OATS repairs). Significance was p<0.05, except when Bonferroni correction was applied (for multiple post-hoc tests).

To address Aim 1, the structural characteristics of normal articular cartilage from the MFC and LT sites of the non-operated control knees were assessed (see electronic supplement). Articular cartilage zone boundaries and thicknesses were determined based on a fit of depth-averaged α to a logistic equation; the SZ-MZ and MZ-DZ boundaries were taken to be the depths at which the fit reached 10% and 90% of the transition between the lower and upper plateaus, respectively. The zonal averages of PI, α, and Γ were determined. The effects of site on the thickness of each zone were assessed with one-factor ANOVA. The effects of site and zone (repeated-measures), and time, on Γ, PI, and α in non-operated joints were assessed with three-factor ANOVAs. Since zones were distinctly different, the effects of site and time on thickness, Γ, PI, and α of each zone in non-operated joints were assessed with two-factor repeated-measures ANOVAs and Bonferroni post-hoc tests.

To address Aim 2, the structural characteristics of repaired defects and non-operated controls were compared qualitatively and quantitatively. Repair sites were analyzed for qPLM values in P, C, D and AHC sublocations (see electronic supplement)¹³. To illustrate differences in graft structure, the PI and α of OATS and non-operated groups were shown in polar plots, and were assessed visually from qPLM colormaps. To quantify the overall differences between OATS and non-operated qPLM, several overall qPLM scores were computed (see electronic supplement). The effects of sublocation and zone (repeated-measures) and time, on Γ, PI, and α in graft-implanted joints were assessed with 3-factor ANOVAs. Then, since the zones were distinctly different, the effects of sublocation on Γ, PI, and α were assessed with one-factor repeated-measures ANOVAs and Bonferroni post-hoc test; also, Non-Op MFC were compared to each sublocation using unpaired Student’s t-tests with Bonferoni correction. To compare composite qPLM scores, the effects of site were assessed at each time-point with one-factor ANOVA and post-hoc Dunnett test (including controls) or post-hoc Tukey test (grafts only).

To address Aim 3, images of serial sections, processed for H&E, Safranin-O, and collagen types I and II were aligned with qPLM colormaps and compared in matched graft regions and zones. First, qPLM structural features were identified, by zone and distance, <0.5 mm and >0.5 mm from the graft-host cleft, in repair and AHC. Then, cells, glycosaminoglycan, and collagen types were assessed in matching areas. To assess the monotonic correlation of qPLM scores and histopathology scores (ICRS, MOD), Spearman’s rank correlation (ρ) coefficient was computed. Correlation coefficients were calculated for graft and non-operated scores combined, as well as graft scores alone (the latter analyses avoiding the possible effects of correlated experimental and control data from an individual animal). Significance was determined by comparing ρ to critical values from the z-distribution³². Finally, qPLM PI and α data were represented as ellipses and overlaid on co-registered H&E serial sections. Overlays were presented for selected non-operated, good repair, and poor repair in areas including graft-host interfaces and extending ~1.5 mm proximally and distally (see electronic supplement).

RESULTS

Zonal Structure of Non-Operated Cartilage

The articular cartilage of non-operated joints appeared histologically normal (see supplementary results and Figure 1A,B,I,M), and based on the depth-variation in the qPLM parameter, α, the SZ, MZ, and DZ boundaries and thicknesses were readily defined (see supplementary results and Figure S1). Zonally-averaged qPLM parameter values in non-operated samples were analyzed (Figure S3, see electronic supplement). Overall, for the non-operated LT (6+12 months), zonal values (SZ, MZ, DZ) averaged PI=0.42 (0.34–0.49), 0.39 (0.32–0.45), and 0.78 (0.71–0.84), α=16° (6–10°), 44° (11–33°), and 83° (80–85°), and Γ=3.4 nm (2.9–3.9 nm), 3.2 nm (2.7–3.7 nm), and 7.0 nm (5.8–8.1 nm), respectively. At the MFC, PI, α, or Γ did not vary significantly between 6 and 12 months (Figure 2). Overall, for the non-operated MFC (6+12 months), zonal values (SZ, MZ, DZ) averaged PI=0.33 (0.28–0.39), 0.54 (0.47–0.60), and 0.82 (0.78–0.85), α=20° (14–26°), 47° (37–56°), and 83° (81–85°), and Γ=3.0 nm (2.5–3.4 nm), 3.6 nm (3.0–4.1 nm), and 7.0 nm (6.0–7.9 nm), respectively.

Figure 1. Safranin O histology and quantitative polarized light microscopy (qPLM) maps of osteochondral sections from non-operated knees.

A,B) Safranin O, (C,D) parallelism index (PI), (E,F) orientation (α) and (G,H) retardance (Γ) maps for representative samples of contralateral non-operated medial femoral condyle (MFC) (A,C,E,G,I–L) and lateral trochlea (LT) (B,D,F,H,M–P), with boxed areas in A–H shown in insets in I–L and M–P. ICRS and qPLM scores, and qPLM scores for superficial zone (SZ), middle zone (MZ) and deep zone (DZ) are indicated. ICRS subcategory A=surface, B=matrix, C=cell distribution, D=cell population viability, E=subchondral bone, and F=cartilage mineralization.

Figure 2. Effect of OATS grafts on qPLM parameters at 6 and 12 months.

PI (A–C), α (D–F), and Γ (G–I) of superficial (SZ, A,D,G), middle (MZ, B,E,H) and deep (DZ, C,F,I) zones of articular cartilage of the contralateral non-operated (Non-Op) or osteochondral autograft (OATS) study groups, at 6 and 12 months post-implantation. OATS repair regions were analyzed at the proximal (P) and distal (D) host (h), as well as the proximal interface (P), central region (C), and distal interface (D) of the graft (g). Data are mean ± SD (n=4 for OATS; n=6 for Non-Op). * indicates difference vs. time-matched Non-Op (p < 0.05).

Effect of OATS Repair on Cartilage Structure

Certain zonal qPLM parameters were affected by OATS repair and time after surgery (Figure 2). In OATS repair, the central implant DZ was affected, with PI at 12 months being 29% lower (PI=0.60 repair vs. 0.83 non-operated, p<0.01, Figure 2C) and α at 6 and 12 months being ~34% lower (α=54° repair vs. 83° non-operated at 6 months, p<0.001; α=56° repair vs. 84° non-operated at 12 months, p<0.01, Figure 2F), but Γ was not different (Figure 2I). At the distal interface, DZ, PI was lower than non-operated at 12 months (p<0.01, Figure 2C). At proximal and distal interfaces, DZ α was lower than time-matched control values at 6 and 12 months (p<0.01, Figure 2F). SZ PI at proximal and distal graft interfaces (which included both graft and host cartilage near the interface) were not significantly different; however, the SZ PI of the AHC (Figure 2A) was 41% higher than that of non-operated tissue (p<0.05).

The qPLM composite scores were substantially higher for OATS repair samples than non-operated samples (<0.001–0.01, see supplementary results), mirroring trends in histological scores (Table 1). Graft repairs with low, moderate and high composite qPLM scores showed distinctive qPLM features with position, particularly in the central graft and at graft-host interfaces. Even for low qPLM scores (good repairs), the central graft displayed some abnormalities, with PI and Γ lower than in non-operated in the upper DZ (Figure 3D,G,J). For moderate qPLM scores (moderate repairs), the central graft region of relatively low PI and Γ extended deeper (Figure 3E,H,K). For high qPLM scores (poor repairs), the central graft region exhibited relatively low PI and Γ through the full-thickness of cartilage (Figure 3F,I,L). At interfaces of all repairs, relative to non-operated controls, PI was higher in SZ, and α lower in the DZ (Figure 3D–I).

Figure 3. Co-localization of qPLM parameters with histological indices of graft deterioration.

(A–C) Safranin O, (D–F) PI, (G–I) α, and (J–L) Γ maps for representative samples of good OATS repair (A,D,H,J), moderate OATS repair (B,E,H,K), and poor OATS repair (C,F,I,L) at 6 months post-operatively. ICRS scores, qPLM overall scores, and qPLM zonal values are indicated. ICRS subcategory A=surface, B=matrix, C=cell distribution, D=cell population viability, E=subchondral bone, and F=cartilage mineralization. PI, α, and Γ averages in regions from proximal (P), central (C), and distal (D) graft superficial (SZ), middle (MZ) and deep zones (DZ) are tabulated.

Correlations between qPLM and Histopathology Scores

Composite qPLM and histopathological repair scores and subscores exhibited a number of rank-order correlations (Table 1, Figure 4). Ranked zone- and area-weighted qPLM composite scores were correlated with ICRS and MOD scores for repairs and non-operated controls (Figure 4, p<0.001); since results were similar for zone- and area-weighted composite scores, only the former are provided here. Correlation coefficients were negative, since larger composite qPLM scores and smaller histological scores indicated increasingly abnormal characteristics. There was a strong association between qPLM and ICRS scores (ρ=−0.74, Figure 4A), and qPLM and MOD scores (ρ=−0.71, Figure 4B). qPLM characteristics of DZ cartilage associated with ICRS cell distribution (ρ=−0.84, Figure 4C), MOD cell morphology (ρ=−0.90, Figure 4D), ICRS cartilage mineralization (ρ=−0.88, Figure 4E), and MOD structural characteristics (ρ=−0.84, Figure 4F). The rank-order correlations displayed in Figure 4 remained strong (ρ>0.74, p<0.05) even among OATS samples only, excluding non-operated groups (Table S2).

Figure 4. Correlation of qPLM scores with histopathological indices of osteochondral deterioration.

Weighted (open markers) and zone-weighted (filled markers) qPLM scores vs. (A) summed ICRS score and (B) summed Modified O’Driscoll (MOD) score, qPLM DZ subscore vs. (C) ICRS cell distribution subscore and (D) MOD cell morphology subscore, and qPLM DZ subscore vs. (E) ICRS cartilage mineralization subscore and (F) MOD structural characteristics subscore. Data are from n=8 OATS and n=12 contralateral non-operated samples. Values of Spearman’s correlation coefficient, ρ, are indicated. Data marker types, defined in the legend, indicate OATS and Non-Op samples for weighted and unweighted scores.

Direct overlays of local qPLM parameters, displayed as ellipses, onto histochemical section images clarified qPLM variations with repair region and cartilage zone, as well as the correlations between qPLM and histopathology scores (Figure 5). The graft-host cleft between the AHC and graft was oriented near-vertical for good repairs (Figure 5B), but off-vertical, with overhanging host cartilage exhibiting flow over the graft cartilage especially for poor repairs (Figure 5D). In the adjacent host cartilage overhanging the graft, PI was low, indicating a less anisotropic matrix. The AHC, >0.5 mm from the cleft, had zonal qPLM structure similar to non-operated cartilage. However, the AHC near (<0.5 mm from) the cleft was variably abnormal with α parallel to the articular surface along with a high PI. In addition, in both the deeper AHC and graft cartilage located near the cleft, PI was heterogeneous and α was oriented parallel to the cleft.

Figure 5. Overlay of ellipse representation of qPLM scores with H&E sections of non-operated, good OATS repair, and poor OATS repair samples.

Overlays are shown for (A) non-operated MFC, (B,C) a good OATS repair, and (D–F) a poor OATS repair. qPLM parameters PI, α, and Γ were averaged over local areas, both in the host (h) and graft (g) regions, separated by dotted line in repairs (B,D), and used to calculate ellipse aspect ratio, orientation, and area, respectively (see cartoon). Boxes areas in (B,D) are shown in detail in insets (C,E,F) to show chondrocyte clusters (arrows) near the host-graft cleft.

In addition, certain cellular features were associated with local qPLM abnormalities. In AHC >0.5–1.5 mm from the cleft, chondrocyte patterns and the long axes of cells and cell pairs were similar to non-operated. However, in AHC nearer the cleft, within ~0.5–1.5 mm (extending to more distant regions for poorer repairs, Figure 3H,I), DZ cell columns were shifted from vertical to oblique, corresponding to off-axis α. Within 0.5 mm of the cleft, in both the AHC and graft, cell clusters were apparent and aligned with long axes parallel to the cleft, especially horizontal for poor repairs (Figure 5E,F, arrows) more than good repairs (Figure 5C, arrows) in association with variable low PI and Γ. Areas of graft cartilage devoid of cells, extending variably from the articular surface and also from the cleft, exhibited low PI and Γ (Figure 5).

DISCUSSION

Analyses by qPLM of articular cartilage from a goat defect repair model revealed localized areas of structural maintenance and alteration that varied according to cartilage zone and in association with certain histopathological features. In non-operated knees, goat articular cartilage of the donor (LT) and recipient (MFC) sites exhibited characteristic zones with relatively constant SZ and MZ depths along with normal histological features. At 6 and 12 months, repair regions of MFC defects exhibited variable qPLM indices and histopathological scores. In the adjacent host SZ corresponding to areas of cartilage flow, PI was higher than normal. In the adjacent host cartilage DZ, α was shifted from vertical to oblique in association with reorientation of cell columns. In the central graft DZ, PI and α were reduced. Areas of decreased PI, near the graft-host interface and in regions of graft tissue, co-localized with altered cellularity. Thus, qPLM revealed a variety of zone-specific and regional alterations in the structure of both the graft and adjacent host articular cartilage after OATS repair of cartilage defects.

There are several advantages and disadvantages to the qPLM technique, implementation, and study design. To analyze the entire repair site including the 3.5 mm diameter graft and its environs (~10×3 mm²), the current study obtained qPLM data using an established technique²⁶, but from 4–6 adjacent, slightly overlapping fields-of-view that were subsequently stitched together. Factors that involve a tradeoff between image detail and data acquisition time are the microscope field-of-view (which depends on objective magnification) and image pixel resolution. In the present study, qPLM, with pixel resolution (6.6 µm)², required 60–90 minutes/sample. A qPLM system with liquid crystal variable retarders³³ reduces acquisition time, but would similarly require post-processing to calculate PI or contour-corrected α. Finally, limited sample numbers were available at each post-operative time-point. Nevertheless, this analysis revealed that qPLM parameters in cartilage zones varied substantially between graft and adjacent host regions, and also non-operated joints.

The interpretation of the qPLM results depends on several factors and study design choices. Interpretation of qPLM parameters in terms of collagen network structure is based on it being the major component of the tissue, and the minimal effect of glycosaminoglycan removal on birefringence³⁴. Since the area fraction of cells within the cartilage of adult goats is low (~5% from (1.6 µm)² resolution images, data not shown), 95% of the area-averaged qPLM parameter values reflect matrix properties. While PI and α reflect the local alignment and orientation, respectively, of network structure in the image/section plane, qPLM does not reveal details such as fibril distribution³⁵. Γ is more complex to interpret, since it depends upon shape, composition, concentration, alignment and orientation of birefringent molecules within each voxel^{22, 26, 36}. Other techniques to analyze collagen network structure provide complementary perspectives. Second harmonic generation microscopy can interrogate tissue surfaces rather than sections ³⁷ and electron microscopy provides more detailed information²¹.

The qPLM features of adjacent host cartilage near the graft-host interface reveal that cartilage flow involves local alteration of the collagen network (higher SZ PI and lower DZ α, tilted toward graft). These deviations from normal cartilage were accentuated with the extent of histopathological deterioration (Figures 3,5). Cartilage flow appears to be affected by the mechanical environment, with more flow in cartilage adjacent to defects³, into empty defects and at weight-bearing joint locations⁵. Such flow and the corresponding qPLM alterations are generally consistent with altered loading patterns and associated strain profiles in the cartilage adjacent to defects^{6, 15} and with the osteochondral grafts being sunk relative to adjacent cartilage¹⁵ and also relative to the cartilage contour of contralateral joints⁸. In host cartilage overhanging the graft, PI and Γ were lower than in non-overhanging host, suggesting disorganization of existing or new matrix in these more deformed regions. Similar structural features of higher SZ and lower DZ Γ compared to unloaded cartilage occur with normal (subinjurious) cyclic loading of intact adult bovine patellar³⁸, and static loading of equine articular cartilage³⁹. PLM images and histological staining of host articular cartilage adjacent to a microfracture defect repair in skeletally mature rabbits⁴⁰, and to empty defects in 5-month old sheep⁵, were consistent with flow as well as the novel qPLM structural alterations quantified in the present study.

The qPLM abnormalities identified in the bulk graft, in the absence of frank fibrillation, coincident with certain histological alterations (Figures 2,3,5), suggest either remodeling/deterioration of existing matrix or deposition of new disorganized matrix. Most alterations in qPLM parameters occurred from the SZ through upper DZ of the grafted cartilage. In these regions, chondrocyte death and collagen degradation epitopes have been reported²⁸. In moderate repairs, qPLM parameter reduction occurred in parallel with hypocellularity and diminished Safranin-O staining, suggesting that matrix damage and loss of glycosaminoglycan may precede tissue loss due to wear. Conversely, qPLM parameters and Safranin-O staining were better maintained around chondrocyte clusters in the central graft DZ than in acellular regions of graft tissue. The relationship between such clusters and those present in osteoarthritis is unclear, and may be associated with chondrocyte expression of type II versus X collagen⁴¹.

The overall qPLM scores, whether zone or area-weighted, quantify net differences between structurally well-organized non-operated cartilage and less-organized graft cartilage. These qPLM scores were differential, with the average non-operated structure assigned a value of 0, and were thus lowest (best) for non-operated controls, and substantially higher (worse) for the OATS groups (Table 1), consistent with higher Mankin scores for more degeneration⁴². The correlations between total qPLM scores and histological scores (Table S2, Figure 4) suggest a general relationship between altered cartilage structure (detected with qPLM) and altered cell organization, morphology, and cartilage mineralization (scored from Safranin-O and H&E-stained sections). Chondrocyte clustering⁴³ and tidemark advancement occur in osteoarthritis⁴⁴ and during cartilage repair⁴⁵, and may indicate a typical response to an aberrant environment. Thus, the overall qPLM scores of the present study, and the recently introduced semi-quantitative PLM score^{20, 21} provide complementary descriptions of the overall collagen network structure of articular cartilage, in accordance with ICRS recommendations for assessment of cartilage repair quality, adding to other previously validated methods¹⁷. Alternate weighting of zone- and region-specific qPLM scores may provide additional insight into the stages or components of articular cartilage deterioration.

This study provides detailed characterization of collagen network remodeling in the cartilage of repair sites after OATS implantation, as a function of cartilage zone and graft region, including adjacent host cartilage and grafted cartilage. Histological and qPLM patterns in osteochondral grafts, taken together, delineate structural alterations with cartilage flow in response to in vivo loading. In contrast, graft structural alterations near interfaces and in cartilage DZ were associated with cell and glycosaminoglycan loss. The qPLM technique may be used to assess the outcome of a variety of therapies for cartilage defects, not only graft transplantation but also repair and regeneration techniques in which histological sections can be obtained, microfracture has been analyzed by traditional PLM^{40, 46}, and ACI has been analyzed by qPLM^{27, 47}. The interpretation of qPLM depends upon the context of the hypothesized mechanisms of the putative therapy. However, for many treatment strategies, restoration of zonal architecture is desired and may be assessed at appropriate time-points. Application of qPLM may be useful to the study not only of osteochondral therapies, but also fracture callus and other situations of connective tissue healing and repair, as long as relevant microstructural features are captured in the section plane, and appropriate sample and instrument controls are used.

Supplementary Material

10. Figure S7. Site-matched group PI, α, and Γ from OATS graft-implanted goat medial femoral condyles, 6 and 12 months post-implantation.

PI (A,B), α (C,D), and Γ (E,F) at 6 months (A,C,E) and 12 months (B,D,E) post-operatively averaged over matched sites at the proximal interface (P1,P2), central implant (C1–C5) and distal interface (D1,D2) for superficial (SZ), middle (MZ) and deep (DZ) zones. Graph data markers represent OATS group mean ± SE (n=4 condyles). Horizontal lines represent non-operated control group zonal mean ± SE (n=6 condyles/Non-Op group).

11. Figure S8. Effect of implant site, timepoint and zone on PI and α values.

Effect on PI vs. α means (polar coordinates) of OATS at (A) 6 and (B) 12 months post-operatively by implant site (proximal interface, P; central, C; distal interface, D) and zone (superficial, SZ; middle, MZ; deep, DZ), plotted with zone and time-matched controls. Data markers represent average values for n=4 condyles/experimental group and n=6 condyles/non-operated group.

02. Figure S1. Determination of boundaries between cartilage zones from α depth profiles for non-operated MFC (A–C,G–I) and LT (D–F,J–L) sites.

(A,D) Representative depth-averaged α curve, showing zonal boundaries (round blue markers), and (B,E) depth-averaged PI curve of a single ROI. (C,F) Zonal boundary positions averaged from normal (non-operated) samples, for regions along an 11 mm span along the articular surface. From logistic fits to α ROI data, zonal (G,J) boundary values and (H,K) thickness, as well as (I,L) % of total thickness. Data shown as mean±SD, n = 12 Non-Op condyles.

05. Figure S2. Quantification of PI, α, and Γ from non-operated goat MFC.

(A,C,E) Colormaps (qPLM) with ROI boundaries (black dotted line) and (B,D,F) ROI quantification, subdivided by zone, for PI (A,B), α (C,D), and Γ (E,F). Graph data markers and bars represent individual ROI mean ± SD (across all depth levels within a zone); lines represent non-operated control group zonal mean ± SE (n = 6 condyles/Non-Op group).

06. Figure S3. Effect of joint site, time post implantation and zone on qPLM parameters of articular cartilage of non-operated goat knees.

(A) PI, (B) α, and (C) Γ for articular cartilage from the non-o erated (contralateral) knee, medial femoral condyle (MFC) and lateral trochlea (LT) at 6 and 12 months ost-im lantation, for su erficial (SZ), middle (MZ) and dee (DZ) zones. Data are mean ± SD (n=6/grou ). * indicates difference vs. time- or site-matched grou (< 0.05).

07. Figure S4. Quantification of PI, α, and Γ from a good OATS repair of goat MFC. (A,C,E).

Colormaps (qPLM) with ROI boundaries (black dotted line) and (B,D,F) ROI quantification, subdivided by zone, for PI (A,B), α (C,D), and Γ (E,F). Graph data markers and bars represent individual ROI mean ± SD (across all depth levels within a zone); lines represent non-operated control group zonal mean ± SE (n = 6 condyles/Non-Op group).

08. Figure S5. Quantification of PI, α, and Γ from moderate OATS repair of goat MFC.

(A,C,E) Colormaps (qPLM) with ROI boundaries (black dotted line) and (B,D,F) ROI quantification, subdivided by zone, for PI (A,B), α (C,D), and Γ (E,F). Graph data markers and bars represent individual ROI mean ± SD (across all depth levels within a zone); lines represent non-operated control group zonal mean ± SE (n = 6 condyles/Non-Op group).

09. Figure S6. Quantification of PI, α, and Γ of a poor OATS repair of goat medial femoral condyle.

(A,C,E) Colormaps (qPLM) with ROI boundaries (black dotted line) and (B,D,F) ROI quantification, subdivided by zone, for PI (A,B), α (C,D), and Γ (E,F). Graph data markers and bars represent individual ROI mean ± SD (across all depth levels within a zone); lines represent non-operated control group zonal mean ± SE (n = 6 condyles/Non-Op group).