Matrix Disruptions, Growth, and Degradation of Cartilage with Impaired Sulfation*

Background: Mutations in chondrocyte sulfate transporter cause dwarfism and joint degradation. Results: The resulting chondroitin undersulfation and collagen disorientation disrupt the protective articular surface and correlate with the chondroitin synthesis rate across epiphyseal cartilage. Conclusion: Matrix synthesis accelerates in enlarging chondrocytes, increasing their susceptibility to matrix-related mutations. Significance: Findings elucidate mechanisms of sulfation disorders, orientation of matrix collagen, and bone elongation. Diastrophic dysplasia (DTD) is an incurable recessive chondrodysplasia caused by mutations in the SLC26A2 transporter responsible for sulfate uptake by chondrocytes. The mutations cause undersulfation of glycosaminoglycans in cartilage. Studies of dtd mice with a knock-in Slc26a2 mutation showed an unusual progression of the disorder: net undersulfation is mild and normalizing with age, but the articular cartilage degrades with age and bones develop abnormally. To understand underlying mechanisms, we studied newborn dtd mice. We developed, verified and used high-definition infrared hyperspectral imaging of cartilage sections at physiological conditions, to quantify collagen and its orientation, noncollagenous proteins, and chondroitin chains, and their sulfation with 6-μm spatial resolution and without labeling. We found that chondroitin sulfation across the proximal femur cartilage varied dramatically in dtd, but not in the wild type. Corresponding undersulfation of dtd was mild in most regions, but strong in narrow articular and growth plate regions crucial for bone development. This undersulfation correlated with the chondroitin synthesis rate measured via radioactive sulfate incorporation, explaining the sulfation normalization with age. Collagen orientation was reduced, and the reduction correlated with chondroitin undersulfation. Such disorientation involved the layer of collagen covering the articular surface and protecting cartilage from degradation. Malformation of this layer may contribute to the degradation progression with age and to collagen and proteoglycan depletion from the articular region, which we observed in mice already at birth. The results provide clues to in vivo sulfation, DTD treatment, and cartilage growth.

mouse at birth when mechanical loads and their secondary effects on degradation of undersulfated cartilage should be negligible. We examined all regions between the articular surface and the mineralizing front of the epiphyseal cartilage. At birth, these regions are continuous hyaline cartilage. Yet, they are involved in different functions of articular cartilage development, preparation for subchondral bone formation, and longitudinal bone growth, all of which are abnormal in DTD (18).
As in many disorders compatible with life, early dtd abnormalities are mild, but may vary across functionally different regions separated by a few tens of micrometers. Such spatial resolution is not easily achievable by existing quantitative techniques, whereas mild defects are difficult to track by conventional histology.
To track mild defects, we advanced infrared (IR) hyperspectral imaging, which detects chemical groups via their characteristic peaks in the IR absorption spectra recorded with micrometer spatial resolution (see Refs. 19 -23 for previous applications to cartilage). Samples for this technique are usually dehydrated to eliminate strong absorption of IR light by water. However, the dehydration distorts the tissue and macromolecular structure and introduces multiple nonphysiological interactions between collapsed macromolecules, smearing spectral peaks (24) and reducing the number of resolved molecular components. We used a specially designed chamber with thermomechanical stabilization of spectral distortions in sections maintained in physiological-like solution (25,26). Combining this setup with spectral analysis based on purified and characterized model compounds, we devised a high-definition approach, which increased spectral reproducibility and the number of resolved molecular components, resulting in quantitative, label-free, "histological" images of multiple components from a single recording.
Using this new method and autoradiographic imaging of chondroitin synthesis, we found several extracellular abnormalities across functionally different cartilage regions. The findings provided insights into DTD, in vivo sulfation, orientation of matrix collagen, and bone elongation.

MATERIALS AND METHODS
This section outlines nonstandard methods. Supplemental Materials describes further details and standard techniques. Materials were bought from Sigma unless stated otherwise.

Animals
Newborn (0.5 day old) dtd mice with knocked-in homozygous A386V substitution in the Slc26a2 gene (8) and their wild type (wt) and heterozygous littermates were used. The mice were inbred progeny of F1 generation of C57Bl/6J and 129/SV backgrounds, which provided full penetration of the dtd phenotype.

Autoradiography of [ 35 S]Sulfate Incorporation into Explants
Whole femurs were incubated in a medium with 33 Ci/ml of [ 35 S]sulfate and total inorganic sulfate concentration of 0.8 mM (similar to that in mouse serum) for 2 or 4 h at 37°C and 5% CO 2 . The explants were fixed in 2% formaldehyde solution and cryosectioned to 12 m in the mid-coronal plane. The cryosec-tions deposited onto glass slides were exposed on phosphoimaging screens, and autoradiographic and autofluorescence images of the sections were recorded with an FLA-5000 scanner (Fujifilm) with 25/50-m nominal/actual spatial resolution. Visible transmission images of the sections with 0.5-m resolution were also recorded with a microscope (Fig. 1). The radiographic, fluorescence, and transmission images were registered to within ϳ13 m (confirmed via overlaid positions of thin radioactive debris on the slides). Cell positions at the transmission images were marked manually, and cell density and radiographic intensity profiles along the lines in Fig. 1 were evaluated using ImageJ software (NIH) (Fig. 1). The radiographic profiles were deconvoluted (to evaluate intensity near cartilage boundaries) and normalized by cell density.

High-definition Infrared Hyperspectral Imaging
Infrared hyperspectral imaging of cartilage composition in sections was performed in the "high definition" (HDIR) mode, in which accuracy and peak resolution, key factors in spectral analyses, were increased by advanced procedures for sample handling, spectra recording and analyses, and purification and characterization of model compounds. Advantages of our and previous analyses are compared in supplemental Materials.

Sample Handling
Proximal halves of the femurs were fixed mildly (1% formaldehyde in PBS for 4 -8 h at 22 Ϯ 2°C), and 12-m midcoronal cryosections were cut under a hard-tissue protocol. Sections were photographed to determine position of mineralizing cartilage and, then, demineralized in 0.5 M EDTA, pH 8.0, for 5 min. A section was placed between two 2-mm BaF 2 windows (avoiding section stretching) and mounted in a thermomechanically stabilized, flow-through chamber (25) and kept at 20 Ϯ 0.1°C. Sections were equilibrated with an excess of 2 mM CaCl 2 , 150 mM NaCl, pH 7.5 Ϯ 0.1 solution (CNC) flushed into the chamber. Model compounds solutions in CNC at pH 7-8 were flushed into a 4 -6-m gap between assembled windows or deposited as a 0.5-l droplet before assembly, and kept at 20 Ϯ 0.1°C. CNC contained calcium to maintain BaF 2 windows passivation and prevent adsorption and decomposition of sulfated glycosaminoglycans by barium. CNC had no buffer to avoid its interfering spectral contributions. Constancy of pH of samples in CNC was confirmed.

Spectra Recording
IR absorption spectra of the samples were recorded with a Hyperion 3000 microscope attached to a Vertex 70 Fourier transform infrared spectrometer equipped with ϫ15, 0.4 NA objective and condenser, low-pass 3850 cm Ϫ1 filter, holographic ZnSe polarizer inserted before the sample, and focal plane 128 ϫ 128 array of detectors (Bruker Optics). The spectra were collected with 4 cm Ϫ1 spectral resolution in the 820 -3950 cm Ϫ1 range by averaging 1000 scans accumulated for 60 min. Several hardware precautions were made to increase spectra quality.

Model Compounds
Our spectral analysis was based on model compounds representing major components of cartilage extracellular matrix.
Model compounds were carefully selected, purified, and prepared at aggregation state and conditions resembling those of the cartilage sections.
Glycosaminoglycans-Model chondroitins, unsulfated (C0S), -4-sulfate (C4S), and -6-sulfate (C6S), were analyzed in detail, as these GAGs dominated in our samples. Several commercial chondroitins were analyzed by ion exchange high performance liquid chromatography (HPLC) (8) and three compounds were selected based on their purity (90 -95%) and enrichment with a particular sulfated form (83-94%) (supplemental Table S1). Purity from low molecular weight compounds and residual buffer salts with interfering spectral contributions was confirmed by dialysis. Spectra of the selected chondroitins at ϳ2% concentrations were normalized by the IR light path length and decomposed onto individual C0S, C4S, and C6S components based on their HPLC determined concentrations. The individual spectra (Fig. 2B) were used to predict spectra of chondroitin compounds with lower enrichment (63-80%). The predicted and measured spectra agreed to within their purity (80 -90%). Spectra of 0.5 and 2% solutions of hyaluronate from rooster comb and dermatan sulfate from porcine mucosa ( Fig. 2B) were recorded "as is" because their content was small in our samples ("Results").
Collagen-Because of the limited quantity, purity, and calibration of mouse cartilage collagens, we used several collagen sources: pepsin-treated bovine type II (Elastin Products), a mixture of acid-soluble mouse type II and type I (Sigma), and mouse and rat type I from tail tendons. We removed impurities, degradation products, and proteoglycans from these samples by collagen precipitation induced by slow in vitro fibrilogenesis (26) (supplemental Materials). Purity (99%) and collagen types of these preparations were confirmed by electrophoresis, CNBr peptides, chondroitinase ABC digestion, and HDIR. The bovine type II collagen reconstituted into fibers ( Fig. 2A) was found to be a suitable model for isotropic spectra of mouse type II collagen dominating in cartilage (supplemental Fig. S2 and Table S2). HDIR absorptivity and concentration of the model type II collagen were calibrated via HDIR and circular dichroism spectra of solutions of pepsin-treated rat type I collagen monomers (supplemental Fig. S2). This calibration was confirmed by HDIR measurement of the water volume fraction in the fibers. Spectra of the selected models represented the triple helical motif of most collagen types because this motif is conserved and constitutes almost entire collagen molecules.
Noncollagenous Core Proteins (NCP)-To model NCP, we purified from smaller molecules by dialysis proteoglycans decorin, biglycan, and aggrecan (PG) abundant in cartilage and a "garden variety" protein bovine serum albumin (BSA, 99% pure) ( Fig. 2A and supplemental Fig. S3). We used several proteins because total mass of NCP in newborn cartilage matrix is not dominated by any particular protein: In agreement with data for bovine nasal cartilage (27), our HDIR analysis indicated that the measured mass ratio NCP/GAG ϳ 0.5 was higher than NCP/GAG ϳ 0.1 expected for aggrecan, which is the most abundant NCP and the main carrier of GAG chains (Fig. 4, A and C). After removing GAG contributions from the PGs spectra, we found that spectral characteristics of the PG cores and BSA were sufficiently similar and evaluated model NCP char-acteristics (supplemental Table S2). Concentrations of the cores and BSA were measured via integral intensities of the amide II band and BSA absorbance at 279 nm.

Spectral Analyses
Prior to spectral analysis, model compound and cartilage spectra were corrected for contributions from liquid water and residual atmospheric H 2 O and CO 2 (supplemental Fig. S1).
Spectral Decomposition-We evaluated concentrations (c j ) of major extracellular matrix components by decomposing the spectra onto individual components. We used Beer's law, cartilage section thickness (h), and baseline-corrected integral intensities (A i ) of selected peaks in unpolarized spectra of the sample.
This decomposition accounted for overlap of peaks of individual components via the elements ⑀ ji Ϫ1 of the inverse of square matrix of absorptivities ⑀ im whose elements ⑀ im ϭ A im /(c m l m ) are the integral intensities A im of the ith peak of mth model compound normalized by the concentration of the compound c m and path length l m .
To increase accuracy, the peaks and their integration limits were selected so that the intensities A im of a peak i were maximized for one compound and minimized for the others ( Fig. 2 and supplemental Table S2).
A set of four peaks marked in Fig. 3A was used in the main text analysis. The other peaks were composed into independent sets and used for verification of the method.
Collagen Orientation-Collagen orientation was evaluated from dependence of IR absorption on the orientation of the absorbing chemical group with respect to IR light polarization. The dichroic difference spectra (A ʈ Ϫ A Ќ ) (Fig. 3C) recorded with IR light polarization parallel (ʈ) and perpendicular (Ќ) to a selected direction were recorded and collagen orientation order parameters P and p were evaluated.
These parameters are related to angles between the collagen molecular axis and its projection onto the section plane, , and between the projection and a selected direction in the plane, .
Angle brackets ͗…͘ denote averaging over the effective illuminated volume (ϳ6 ϫ 6 ϫ 12 m 3 ) determined by the actual spatial resolution and the sample thickness l; c c is collagen concentration in this volume measured by the decomposition of depolarized spectra: (A ʈ ϩ A Ќ )/2. We used baseline-corrected integral intensity (A ʈ Ϫ A Ќ ) of the three bands in the amide III region (1135-1350 cm Ϫ1 ) whose vibration dipoles are nearly parallel to the collagen helix axis (Fig. 3B). The A ʈ Ϫ A Ќ spectra are dominated by collagen because chemical groups of the other components have quasi-random orientations (Fig. 3C). The sample intensity was normalized by the corresponding intensity (A ʈ Ϫ A Ќ ) 0 of a stretched, washed mouse tendon fiber and its collagen concentration c 0 and thickness l 0 . The concen-tration was determined from isotropic spectrum (A ʈ ϩ 2A Ќ ) 0 /3. The fiber was a suitable model because it is composed of ϳ95% type I collagen whose spectra are similar to the spectra of collagen from all cartilage regions (Fig. 3C and supplemental Table  S2).

Matrix Sampling, Definition of Cartilage Zones, and Statistical Analysis
Extracellular matrix areas (2-10 pixels) were selected based on visible images and HDIR hyperspectral images distinguishing matrix from cells (supplemental Fig. S4). The matrix areas were sampled in ϳ100-m wide regions along the lines shown in Fig. 1 with a density of ϳ100 areas per mm 2 . At the articular surface, 20 -30 areas (2 ϫ 2 pixels) were sampled on the inferomedial side every 8 m. Accidentally damaged regions were excluded. HDIR and autoradiographic characteristics were averaged at selected positions of cartilage zones for each section.
The continuous epiphyseal cartilage of 0.5-day-old mice was divided into articular and growth plate regions based on their resemblance to articular cartilage and growth plate, which become separated by the secondary ossification center after 7-14 days. Each region was divided into zones defined morphologically (Fig. 1). The articular region was divided into the articular (A) zone, where cells were flattened, and the subarticular (S) zone, which began where most cells became round (about 140 m below the articular surface). The boundary between the S and columnar (C) zones of the growth plate region was where most cells became flat and organized into stacks of two or more cells. The enlargement (E) zone began where the thickness (diameter along the bone axis) of most cells became 1.5 times larger than the thickness of the thinnest columnar cells. The hypertrophic (H) zone began where the thickness of most cells was 1 ⁄ 3 of the average thickness of cells where cell enlargement ceased and cartilage mineralization started. The mineralizing cartilage (M) zone began where septa of the mineralizing matrix appeared. The growth plate region ended at bone marrow where chondrocytes could no longer be distinguished. Within such definition, prehypertrophic zone is expected between the beginnings of the E and H zones. For data plotting, the growth plate zones of dtd were shifted so that the H zone beginning of dtd coincided with that of wt.
Mouse-to-mouse variability of the averaged spectroscopic and radiographic parameters at selected positions was analyzed by unequal variance, two-tail t-test. One section per mouse was included in the test because variation of the measured parameters between adjacent sections did not exceed the variation within a given region of a section.

HDIR Verification
Absence of several sampling, spectroscopic, and analytical artifacts was verified as detailed in supplemental Materials. In particular, we measured the matrix composition and its uncertainties using independent sets of spectral peaks, and compared the composition before and after enzymatic removal of GAG chains.
We found high correlation coefficients, R 2 ϭ 0.7-0.9, and small deviations, Յ10%, between the independent measurements of the same compounds (supplemental Figs. S5 and S6). Weak cross-correlation of different compounds, R 2 ϭ 0.04 -0.2, indicated that these compounds were resolved (supplemental Fig. S5). Reasonable quantification of the major matrix components was further suggested by reconstruction of matrix spectra from the model compounds spectra. The reconstruction based on decomposition of the main set of peaks reproduced the other peaks and fine structure of the peaks (Fig. 3A). We also estimated that interactions between major components within the matrix had a small effect on the spectra at our physiological-like conditions.

RESULTS
We examined sections of proximal femur cartilage of 0.5day-old mice near anatomical lines shown in Fig. 1. We recorded infrared absorption spectra using a spectrometer attached to a microscope with a 128 ϫ 128 array of detectors. Each recording collected spectra from a 355 ϫ 355 m 2 area with 2.8-m nominal and 6 -10-m actual spatial resolution. The spectra exhibited peaks typical of extracellular matrix (Figs. 2, A and B, and 3B) or cells (not shown). Using these peaks, the matrix spectra were decomposed onto individual components dominating in cartilage matrix: GAGs, NCP, and collagen.
The resulting representative distributions of densities of each component in the articular and growth plate regions marked by squares in Fig. 1 are plotted as gray-scale images in Fig. 4. We also extracted data from small matrix areas near the anatomical lines (Figs. 1 and 4) and graphed them versus anatomical distances (Fig. 4). The graphed profiles were similar in every analyzed section of 6 wt and 6 dtd mice.
Such analysis showed that the components' distributions in wt and dtd mice were different. These differences were inferred from the main set of spectral peaks marked in Fig. 3A. Similar differences were derived from independent sets of other peaks shown in Figs. 2, A and B, and 3B, suggesting that the method's uncertainties were small (see supplemental Figs. S5 and S6 and Methods for HDIR verifications).
Glycosaminoglycans-The dtd mutant exhibited depletion of GAGs relative to wt, which occurred in the articular region only, 1.35-1.65 Ϯ 0.35-fold, p ϭ 0.01-0.0001, respectively, at 25-240 m below the articular surface. (Mean Ϯ S.E. for n ϭ 6 wt and n ϭ 6 dtd mice are reported, unless stated otherwise). The depletion normalized toward the columnar (C) zone of the growth plate region where it became insignificant, 1.1 Ϯ 0.1fold or less, p ϭ 0.2-0.7 (see supplemental Methods and Fig. 1 for morphological definition of zones). Note that it is unclear if the decrease in densities of matrix components (Fig. 4, A, C, and D) in the hypertrophic (H) and mineralizing cartilage zones was real because thickness of the matrix septa became comparable  Table S2). B, spectra of chondroitins with C0S, C4S, and C6S forms of sulfation were derived by decomposition of spectra of highly enriched chondroitins. The available dermatan sulfate sample (DS) appeared contaminated by ϳ30% of GAGs other than HA or chondroitins (cf. supplemental Table S1). HA and C0S spectra were shifted down for clarity. Bold horizontal segments mark integration ranges of peaks used to analyze matrix composition. FIGURE 3. HDIR absorption spectra of cartilage extracellular matrix and a collagen fiber. A, matrix spectra from 5 ϫ 5 m 2 area 80 m below the articular surface in wt and dtd mice are compared with spectra reconstructed from the model chondroitins, collagen type II, and BSA spectra based on decomposition of 4 spectral peaks marked by horizontal segments: collagen prolines 1340 cm Ϫ1 band, which is weak in NCPs due to low proline content; the CH 2 bending of NCP of the 1401 cm Ϫ1 band, which is shifted to 1405 cm Ϫ1 and is weaker in collagen due to the high glycine content; sugar ring bands between 960 -1188 cm Ϫ1 , where the intensity of protein cores is weak; and GAG sulfate vibrations at 1235 cm Ϫ1 , where collagen also has a sizable contribution. B, polarized HDIR spectra of wt articular extracellular matrix and washed, stretched tendon fiber. The fiber spectra were normalized to have collagen content equivalent to that of the matrix. IR light polarization was parallel ʈ and perpendicular Ќ to the articular surface or to the fiber axis. C, dichroic difference ʈ Ϫ Ќ of the spectra shown in B. The fiber spectra at 1550 -1700 cm Ϫ1 were distorted due to high absorbance.
with the HDIR spatial resolution, resulting in marked "blurring" of HDIR images of the septa.
To confirm that GAG chains in 0.5-day-old mouse cartilage matrix were dominated by chondroitins, we exhaustively digested fixed wt and dtd sections in the HDIR chamber with testicular hyaluronidase, which cleaves and releases hyaluronate, chondroitin, and hybrid chondroitin/dermatan chains. The resulting intensity loss in IR absorption spectra (supplemental Fig. S6B) sug- . Densities of GAGs, noncollagenous core proteins, and collagen and sulfate/GAG ratio in proximal femur cartilage of 0.5-day-old wt and dtd mice measured by HDIR hyperspectral imaging. Images, hyperspectral images of articular and growth plate regions marked in Fig. 1. Pixel brightness is proportional to the densities and ratio according to vertical gray scale bars at y axes of graphs under the images. Pixels at cells appear darker in D or brighter in C than pixels at matrix due to, respectively, lower collagen or higher noncollagenous protein content at cells. The brightness of the cell pixels is quantitatively inaccurate. Graphs, the densities and ratio at the extracellular matrix areas near the anatomical lines (shown in Images and in Fig. 1) versus distance from the bone marrow and the articular surface. Shaded areas are confined by curves whose y values are mean Ϯ S.D. at each distance for a single section (cf. supplemental gested dominance of C4S and C0S. The estimated fraction of C6S was small (several percent), in agreement with a previous analysis of mice and rats (28) and HPLC analysis of whole 0.5-day-old femur heads, 3.1 Ϯ 0.2% in wt and 3.3 Ϯ 0.2% in dtd, n ϭ 4 (calculated from our previous data (8)). Low intensity of the characteristic peaks of hyaluronate and dermatan indicated their small content (less than 10 and 5% of GAGs, respectively (supplemental Fig.  S6, A and B)). This is consistent with previous biochemical analyses on neonatal mouse and bovine cartilages that can be summarized in GAG proportions of about 88% chondroitins, 7% hyaluronate, 0.4% dermatan, 5% keratan, and 1% heparan (9, 29 -31).
Glycosaminoglycan Sulfation-Taking into account the dominance of chondroitins, we evaluated the sulfate/chondroitin disaccharide ratio (cs) in the matrix using the main set of peaks including bands of C4S and C6S sulfate groups at about 1235 cm Ϫ1 and sugar rings at 960 -1188 cm Ϫ1 (Fig. 2B).
The apparent sulfation ratio cs varied across the proximal femur cartilage weakly in wt, but strongly in dtd (Fig. 4B). The strongest undersulfation occurred at the articular region 80 m below the articular surface and at the H zone of dtd. The dtd sulfation was normalizing toward resting-like zone (expected before the columnar zone) and articular surface, and toward the columnar zone.
Specifically These results were consistent with independent measurements of sulfation by spectral decomposition involving the 928 cm Ϫ1 peak of C4S (supplemental Fig. S5), by analyzing the spectra of GAG chains released by hyaluronidase (supplemental Fig.  S6D), and by HPLC disaccharide analysis of whole femur head, cs dtd ϭ 68.2 Ϯ 0.8 and cs wt ϭ 90.4 Ϯ 0.3, n ϭ 4 (8). The cs's determined by released chains spectra and HPLC were 5-10% higher than in Fig. 4B apparently due to small contributions from hyaluronate and oligosaccharides to the sugar ring bands used to measure cs in Fig. 4B. This error was smaller than the observed dtd defects and did not affect our conclusions.
The HDIR imaging (Fig. 4, A and B) was also consistent with reduced toluidine blue staining of sulfated GAGs in dtd (supplemental Fig. S8). The staining, however, could not distinguish between reduced GAG density and sulfation. In the articular region, the reduction was due to a decrease in both sulfation and GAG density. In the enlargement and hypertrophic zones, the staining remained intense despite undersulfation. Staining of human DTD samples at high and low pH values used to distinguish between unsulfated and sulfated GAGs did not allow unambiguous interpretation either (32).
Noncollagenous Core Proteins-Density of protein cores of all NCPs in the matrix of dtd was lower than that of wt, which occurred at the articular region only (Fig. 4C): wt/dtd density ratio was 1.5 Ϯ 0.2, p ϭ 0.01, and 1.3 Ϯ 0.1, p ϭ 0.02, at 25 and 80 m below the articular surface, respectively, and (0.9 -1.1) Ϯ 0.1, p ϭ 0.2-0.9, at deeper articular layers and the growth plate region.
Collagen Content-Likewise, density of all triple helical collagens in the matrix of dtd was lower than that of wt at the articular region only (Fig. 4D): the wt/dtd ratio was 1.6 Ϯ 0.3, p ϭ 0.02, and 1.4 Ϯ 0.2, p ϭ 0.07, at 25 and 80 m below the articular surface, respectively, and (0.9 -1.1) Ϯ 0.15, p ϭ 0.2-0.9, at deeper articular regions and at the growth plate region.
Collagen Orientation-We quantified collagen orientation across the sections using HDIR hyperspectral imaging with IR light polarized in two orthogonal directions with respect to the articular surface and the growth plate plane. A resulting collagen orientation order parameter P (Equation 2) is plotted as images with two colors marking areas where collagen is preferentially oriented parallel or perpendicular to the selected directions (Fig. 5).
The images in Fig. 5 indicate that the amount of preferentially aligned collagen varies across different cartilage regions. As collagen density also varied (Fig. 4D), we plotted order parameter p ϭ P/c c normalized for the collagen density c c for 15-50 m 2 areas of matrix (lower left panel of Fig. 5). In wt, this parameter varied between ͉p͉ ϭ 0.0 and 0.25. This range would correspond to 0 -25% of oriented molecules in a system where a fraction (1 Ϫ ͉p͉) of molecules is randomly oriented and the FIGURE 5. Collagen orientation order parameters of matrix of proximal femur cartilage from 0.5-day-old wt and dtd mice measured by polarized HDIR hyperspectral imaging. Images, parameter (P) uncorrected for collagen concentration. Cyan and magenta pixels mark areas with preferred collagen alignment indicated by orthogonal arrows of the corresponding color. The cyan-black-magenta bar corresponds pixel brightness to P values with 1 unit equal to 1 Ϯ 0.1 pg/picoliter. Bottom, parameter (p) corrected for collagen concentration sampled at extracellular matrix near the lines (shown in these images and in Fig. 1) versus distance from the bone marrow and articular surface. Shaded areas represent mean Ϯ S.D. for a single section (cf. supplemental Fig. S7E). Letters and ticks mark the cartilage zones shown in Fig. 1. Right, parameter (P) and its coefficient of variation CV at the articular surface. * and ** denote p values Յ 0.05 and 0.005, respectively. other fraction ͉p͉ is aligned ideally in the section plane and at zero angle to a preferred direction (33). Because aligned molecules have nonideal distribution of alignment angles, the actual oriented fraction is larger than 0 -25%.
In dtd, distribution of collagen orientation was qualitatively similar, but absolute values of the orientation order parameters were reduced in the articular region (Fig. 5). The strongest reduction was ϳ80 m below the articular surface, p wt /p dtd ϭ 1.5 Ϯ 0.3, p ϭ 0.046 (5 wt and 5 dtd mice). It normalized toward deeper and shallower regions, p wt /p dtd ϭ (1.1-1.25) Ϯ (0.2-0.3), p ϭ 0.2-0.6, with the collagen disorientation parameter, p wt -p dtd , being higher at the 80 m depth than at 25 and 240 m, p ϭ 0.05 and p ϭ 0.001, respectively.
Substantial disorientation also occurred in the thin layer of collagen at the articular surface where the order parameter P was P wt /P dtd ϭ 1.8 Ϯ 0.3 times lower in dtd (p ϭ 0.05, Fig. 5). Furthermore, the layer was diffuse in dtd, but sharp and thin in wt in every analyzed section (Fig. 5). Because the layer thickness in wt, ϳ5 m, was comparable with the HDIR spatial resolution, ϳ6 m, the measured P wt values were underestimated (the image of the layer in Fig. 5 was blurred). Thus, the actual P wt /P dtd ratio should be even higher.
In addition, the surface collagen layer in dtd was highly heterogeneous along the articular surface: coefficient of variation (CV ϭ S.D./mean) of its parameter P was higher than in wt (CV dtd /CV wt ϭ 2.6 Ϯ 0.4, p ϭ 0.002). The dtd variation was so high, CV dtd ϭ 0.38 Ϯ 0.04, that the p values were two or more times lower than average at ϳ10% of the surface in dtd. The corresponding surface fraction would be ϳ0.04% in wt (assuming normal distribution), but such poorly organized areas were not found in any wt section.
At the growth plate regions, the order parameters had high sample-to-sample variability which obscured differences between wt and dtd, if any (Fig. 5).
Chondroitin Synthesis Rate-To relate GAG composition and synthesis, we analyzed [ 35 S]sulfate incorporation into femur explants incubated shortly after excision from 0.5-dayold mice. In the resulting autoradiographs of explants cryosections, intensity distribution was similar for 4- (Fig. 1) and 2-h (not shown) incubations, suggesting that these profiles were not limited by [ 35 S]sulfate diffusion (as expected from ϳ20 min diffusion time estimated from sulfate diffusion coefficient in cartilage (34) and the explant geometry). Because ϳ90% of sulfate groups of this mouse cartilage belongs to chondroitin sulfate, and, because ϳ5% variability of chondroitin sulfation levels in wt was smaller than severalfold variability of incorporation levels, the radiographs reflected variability of chondroitin synthesis rates in wt.
From the [ 35 S]sulfate incorporation images, we extracted profiles for the same regions sampled in the HDIR experiment (Fig. 1). The 35 S intensities were normalized by areal cell density to correct for its ϳ10-fold variability. We pooled together 3 wt and 3 heterozygous littermates because wt and heterozygous mice had indistinguishable radiographic profiles (supplemental Fig. S9C), chondroitin sulfation, sulfate uptake, and morphologic and phenotypic characteristics (8). The resulting profile and its variability were plotted in Fig. 1  (bottom). The pooled data were also plotted in Fig. 6 after cor-recting for the fact that the areal cell density underestimates volumetric density of smaller cells. We verified that data processing uncertainties, ϳ15%, were smaller than the following observed effects.
The proximal femur cartilage showed two peaks in the incorporation profiles: near the boundary between articular and subarticular zones and near the beginning of the H zone ( Figs. 1 and  6). Specifically, the incorporation rate per cell at the subarticular peak was 3.8 Ϯ 0.3 and 1.27 Ϯ 0.06 times higher than at shallower (25 m) and deeper (240 m) distances from the articular surface, respectively (p ϭ 0.0004 and p ϭ 0.008, onesample, two-tail t-test). The rate at the H zone peak was 2.1 Ϯ 0.1 and 4.3 Ϯ 1.7 higher than at the middle of the columnar and beginning of mineralizing cartilage zones, p ϭ 0.0004 and p ϭ 0.05, respectively.
In a dtd mouse, the incorporation profiles in proximal and distal femoral cartilages had shapes similar to its wt littermates (Figs. 1, 6, supplemental Fig. S10). The incorporation intensities in dtd were 1.5-2 times lower than in wt. The corresponding difference in chondroitin synthesis rate should be smaller because the sulfate/chondroitin ratio in dtd is 1.1-1.9 lower than in wt and because a non-negligible fraction of incorporated sulfate in dtd originates from intracellular catabolism of sulfur-containing compounds (35,36). The semi-quantitative similarity of the incorporation profiles indicates that uptake of extracellular sulfate was still the major source of intracellular sulfate in dtd. The similarity also indicates that the chondroitin synthesis rates in wt and dtd had similar profiles and absolute values differing less than 2-fold. Although we had only one dtd mouse in the two litters used for radiolabeling, its phenotype and cartilage morphology were typical of all dtd mice. Because the sulfate incorporation measurement underestimates the chondroitin synthesis rate in dtd, it was not pursued further.
Note that similar incorporation profiles were observed in other epiphyseal cartilages. We found that the distal femurs had a wider peak at the H zone due to wider distal zones (supplemental Fig. S10). Qualitatively similar profiles were found for distal tibia explants from chick embryos and for in vivo incorporation into humeri of 4-and 7-day-old rats (37)(38)(39)(40). It has been unclear, however, if those profiles were distorted by limited [ 35 S]sulfate diffusion during very short (5 min) incubation of the explants and by variable cell density or by terminal cell hypertrophy initiating secondary ossification center in the articular region of rat humeri at 7 days.
Correlation between Chondroitin Synthesis and Matrix Disruptions-The radiographic and sulfation profiles are plotted together in Fig. 6. This plot indicates that undersulfation of chondroitin correlated with its synthesis rate in the articular region of dtd. Such correlation seems to hold also in the growth plate region, although undersulfation increased through this region while the rate increased and then declined. Because cells and their matrix progress from columnar to the mineralizing zone, progression in matrix composition reflects accumulation of chondroitin chains made through these zones rather than composition of chains made at each location. Synthesis in the middle of the H zone of chains whose sulfation is higher than that of chains made in the beginning of the H zone but lower than in the columnar zone can explain the observed decrease in the net sulfation across the H zone.
Disorientation of collagen in the articular region of dtd also correlated with the chondroitin synthesis rate and chondroitin undersulfation (Fig. 6). In the growth plate region, high variability of the measured orientation parameters obscured such correlation, if any.

DISCUSSION
In chondrocytes, the SLC26A2 sulfate/chloride antiporter is the main source of inorganic sulfate. Most of this sulfate is spent to sulfate chondroitin, which is then secreted to constitute ϳ90% of cartilage extracellular GAGs. The homozygous A386V mutation knocked into the antiporter reduces the sulfate transport rate into chondrocytes from dtd mouse (8). This reduction causes chondroitin undersulfation and yet a barely detectable clinical phenotype at birth (8).
Undersulfation Varies Across Cartilage and Correlates with Chondroitin Synthesis Rate-We found that the undersulfation was mild in most regions of epiphyseal cartilage, but strong in (sub)articular and (pre)hypertrophic zones of the growth plate region (Fig. 4B). The undersulfated regions gave small contributions to the net sulfation of the whole tissue because they were narrow and had thin matrix and lower GAG content (Fig.  4A, and supplemental Figs. S4 and S8). Yet, these regions are crucial for cartilage and bone development and maintenance, suggesting that the undersulfation variability may be a clue to DTD bone malformation and joint degradation.
In growing cartilage of newborn mice, matrix composition should be dominated by matrix synthesis rather than degradation. The rate of chondroitin synthesis in cartilage explants was also highly variable across the epiphyseal cartilage in both wt and dtd (Fig. 1). Furthermore, the synthesis rate and the undersulfation of chondroitin correlated (Fig. 6).
Variation of Matrix Synthesis Rate Is Related to Cartilage Growth-Profile of the chondroitin synthesis rate showed two peaks (Fig. 1), which seem related to cell/matrix morphology of the growing cartilage. In the upper (proximal to the articular surface) part of the columnar zone, the stacked cells are thin and their proliferation is accompanied by production of a thin layer of matrix. In the lower part of the growth plate, the cells stop proliferating, but enlarge their size and augment their matrix volume per cell up to 2.7 times by the end of the hypertrophy (despite making matrix septa thinner) (41). Because density of GAGs, collagen, and NCPs within the matrix changes little between columnar and hypertrophic zones, the matrix augmentation requires synthesis of matrix components such as chondroitin ( Fig. 1). At the end of cell enlargement, matrix augmentation and synthesis slow down, resulting in the peak in the profile of chondroitin synthesis rate (Fig. 1). We expect the main carrier of chondroitin chains aggrecan and perhaps other components to follow a similar synthesis rate profile, as observed for collagen in chick embryo explants (37).
Cartilage matrix provides mechanical integrity to growth plate that must withstand mechanical loads in growing bone. About 75% of the axial bone growth occurs at the chondrocyte enlargement zone (41,42). Therefore, the elevated matrix synthesis in enlarging chondrocytes is needed, at least, to support bone integrity during growth.
Whether the elevated chondroitin synthesis is also a driving force for chondrocyte enlargement is unclear. Of three forces hypothesized so far, static hydration pressure of chondroitin sulfate proteoglycans (which swell matrix around chondrocytes) does not seem dominating: chondroitin sulfate density and, hence, hydration pressure at the columnar zone are similar to those at the enlargement and hypertrophic zones (Fig. 4B) but the columnar chondrocytes do not enlarge. However, we cannot completely exclude a second force due to dynamic hydration pressure from sulfated proteoglycans secreted and transiently concentrated near cells because our measurements cannot resolve pericellular matrix. If this force contributes, the nearly 2-fold reduction in chondroitin sulfation, believed to be the main cause of the pressure, is insufficient to appreciably reduce enlargement in the dtd growth plate region.
The third hypothesized force driving cell enlargement is exerted from within chondrocytes by synthesizing intracellular components and uptaking water (43). This force does not seem to have substantial contribution from the elevated chondrotin sulfate synthesis because sulfation occurs in Golgi whose volume is small and does not augment (43). However, elevated synthesis of matrix proteins may contribute indirectly because it is accompanied by ϳ2.5-fold volumetric augmentation of the synthetic apparatus, endoplasmic reticulum, whose volume fraction per cell is substantial, ϳ15% (43).
Profile of chondroitin synthesis rate across the articular cartilage also seems related to growing morphology. In the articu-lar zone up to ϳ80 m below the articular surface, chondrocytes are small and need to synthesize little chondroitin sulfate because its density within the matrix is low and the matrix is thin (Fig. 4B, and supplemental Figs. S4 and S8). In the subarticular zone, cells enlarge, proliferate, and augment their matrix volume and GAG density in the matrix (Fig. 4B and supplemental Figs. S4 and S8), which requires elevated rates of matrix synthesis. The synthesis rate drops toward a subzone preceding the columnar zone. This subzone often called "resting" does not show a dramatic increase in matrix thickness, GAG density, and cell size, resulting in a lower synthesis rate and in a peak in the rate profile. Thus, such chondrocyte enlargement concomitant with elevated chondrotin synthesis at the subarticular zone may contribute to growth of articular cartilage as well.
Rapid Chondroitin Synthesis Creates Sulfate Deficit-The chondroitin sulfation inversely correlated with the chondroitin synthesis rate in dtd but not in wt (Fig. 6). We interpret such dependences via balance of sulfate supply, sulfate consumption by chondrotin sulfate synthesis/secretion, and enzymatic activity of intermediate steps of the sulfation pathway. The activity determines a threshold below which sulfation efficiency becomes dependent on its substrate, intracellular inorganic sulfate. In wt, the efficiency (sulfate/chondroitin ratio, cs) remains close to 100% and the sulfate remains above the threshold at any physiological rates of sulfate consumption.
In dtd, the reduced sulfation indicates that the intracellular sulfate reaches the threshold. The undersulfation would be trivially determined by the law of mass conservation, cs Յ V S /V C , if the chondrotin synthesis rate V C is much faster than rate of intracellular sulfate generation V S . This would happen when chondroitin is strongly undersulfated, cs Ͻ Ͻ 1. When the undersulfation is mild, as in dtd and nonlethal patients, it can depend on the enzymatic activity as well.
The in vivo correlation suggests that the chondrotin synthesis rate is the primary factor determining the intracellular sulfate deficit and the undersulfation in dtd cartilage. Activity of other steps of the sulfation and sulfate production pathway seems to have secondary, yet non-negligible compensating effects on the undersulfation. This is indicated by breakdown of the correlation upon transition from growth plate to articular region (Fig. 6): The columnar zone had a higher synthesis rate, but better sulfation than the resting-like and peri-articular subzones.
The undersulfation-synthesis rate correlation can explain the normalizing sulfation in dtd mice with age (8). As growth and chondroitin synthesis slow down with age (44), sulfation of newly synthesized chains should increase. The correlation, however, does not explain rapid progression of skeletal deformities and articular cartilage degradation with age despite the normalizing sulfation (8).
Reduced Collagen Orientation May Cause Cartilage Degradation-We found other matrix abnormalities potentially related to the articular degradation. First, densities of the major components, GAGs, collagen, and NCP, were reduced at the articular region only, indicating pathological processes specific to this region (Fig. 4, A, C, and D).
Second, collagen was partly disorientated in the articular region in dtd (Fig. 5), which could contribute to the proteogly-can depletion and the degradation (Fig. 4). The disorientation affected a thin layer of well oriented, densely packed collagen fibers normally present at the articular surface. This layer protects the articular cartilage from frictional damage by articulation of the opposite bone and, perhaps, from degrading enzymes whose concentration in synovial fluid may be higher than in cartilage. Its malformation may also lead to diffusional loss of proteoglycans and growth factors from the articular region and penetration of extraneous factors from synovial fluid, compromising functions of the whole joint.
Origin of Collagen Disorientation-Disruption of the surface layer occurs at early stages of joint degeneration in another genetic pathology (45) and in arthritis. Despite different causes, arthritic and DTD humans and mice exhibit common signs of chondrocyte and matrix degeneration including proteoglycan depletion from the articular region and accumulation of thick collagen fibrils (32, 46 -50), indicating potentially common intermediate mechanisms of degradation. In degenerating arthritic and DTD cartilages, accumulating collagen fibers often appear irregularly oriented (32, 46 -50), which should reduce collagen orientation order parameters. The lack of accumulation-disorientation correlation (Figs. 4D and 5) suggests that the accumulation is not the primary cause of the disorientation in dtd.
Collagen disorientation correlated with undersulfation and, likely, with the inorganic sulfate deficit (Fig. 5), indicating that sulfated species may be involved in processes of collagen orientation in growing articular cartilage in dtd. Because the disorientation appears in dtd cartilage whose sulfation and organization are perturbed by impaired sulfate transport mildly, the same sulfation-dependent processes may be involved in collagen orientation in wt cartilage as well.
Which Undersulfated Species Cause DTD?-Although this and previous works indicate a relationship of cartilage dysfunctions to chondroitin undersulfation, it is unclear if undersulfation of chondroitin is the direct cause of DTD. Chondrocytes of dtd and other mice with sulfation pathway defects accompanied by chondroitin undersulfation show reduced Indian Hedgehog (IHH) signaling and reduced proliferation (seen in the growth plate region (51)(52)(53), in the dtd articular region (supplemental Fig. S8) and in dtd cell culture (8)). Proliferation in cartilage is partly regulated by long range IHH and parathyroid hormone-related protein signalings generated by the prehypertrophic and articular zones (54,55), which experience the strongest undersulfation in dtd. In brachimorphic mouse with impaired synthesis of the organic sulfate donor, the IHH reduction has been ascribed to IHH binding to C4S whose content is reduced 1.3-fold (51). However, in a mouse chondrodysplasia caused by C4S-sulfotransferase knock-out with 20-fold C4S reduction, IHH signaling is unaffected, but dysfunctions have been ascribed to other factors (9).
Apart from chondroitin, other sulfated species are present in cartilage in minor amounts, but known to affect signaling and matrix. For example, cholesterol sulfate regulates the synthesis pathway of cholesterol (56 -58), a key regulator of IHH signaling at the protein level. Also, heparan sulfate is a co-factor of several signaling receptors and growth factors (59 -61); heparan sulfate undersulfation has been hypothesized to disorganize collagen (62) because suppression of sulfation or synthesis of a heparan sulfate proteoglycan in a cancer cell line or zebrafish embryos disorganizes fibronectin (62,63). A connection between undersulfation of minor species and of chondroitin has been hypothesized because all species accept the sulfonate group from 3Ј-phosphoadenosine 5Ј-phosphosulfate (PAPS) synthesized from inorganic sulfate. Whether this connection is significant depends on multiple enzymatic and transport steps whose kinetic relationships in vivo are not understood.
The observed dependence of undersulfation of chondroitin on its synthesis rate indicates that the PAPS concentration in dtd drops to the Michaelis constant for PAPS of chondroitin-4-sulfotransferase, K m ϳ0.3 M (64). Such concentration is lower than the K m of sulfotransferases to heparan, 0.4 -100 M, and cholesterol, 2-7 M (65-69). Therefore, such species may be undersulfated, and the domain structure of heparan sulfate implicated in signaling may be disrupted (70) in the regions with substantial chondroitin undersulfation. As these regions are narrow, undersulfation of minor species may be difficult to detect by existing analyses requiring samples larger than a mouse limb. Analytical advances and manipulation of genes may further clarify mechanisms of DTD and its high phenotype variability even between siblings (13).
Implications for Cartilage Growth-At present, we obtained evidence that sulfate deficit causes disorientation of cartilage collagen in the studied mouse background, suggesting that processes involved in collagen orientation may depend on sulfated species.
We observed that the chondroitin synthesis rate per cell varies ϳ10-fold across growing epiphyseal cartilage and peaks at the subarticular and pre-hypertrophic regions. The latter peak supports matrix and cell enlargement near the pre-hypertrophic zone and, therefore, may be needed for robust bone growth driven by the enlargement.
Implications for Chondrodysplasias-We observed correlations of undersulfation and collagen disorientation with chondroitin synthesis rate across cartilage with mildly impaired sulfation.
This example suggests that rapid matrix synthesis in prehypertrophic and subarticular regions can make these narrow, yet vital regions particularly susceptible to mutations whose adverse effects increase with the synthesis rate of matrix molecules. This may apply, for example, to chondrodysplasias whose mutations in aggrecan or other matrix molecules can cause endoplasmic reticulum stress. The stress and the resulting defects increase with the number of mutant alleles and, hence, with the synthesis rate (71)(72)(73).
For DTD, the observed correlations imply that improving sulfation during development may improve formation of bone, articular cartilage, and the periarticular collagen layer and thereby reduce predisposition to joint degradation at older age when the rate diminishes. These correlations may help to determine the kinetic regime of the in vivo sulfation pathway, develop its theoretical model, and predict pathway steps that can be targeted for improving sulfation. Targeting treatments are difficult to deliver to deep avascular cartilage regions. Observed in this work proximity of the most affected regions to synovial fluid and bone marrow provides a hope for systemic delivery from blood circulation.