Primary Cilia Direct Murine Articular Cartilage Tidemark Patterning Through Hedgehog Signaling and Ambulatory Load

Abstract

Articular cartilage (AC) is essential for body movement but is highly susceptible to degenerative diseases and has poor self-repair capacity. To improve current subpar regenerative treatments, developmental mechanisms of AC should be clarified and, specifically, how its postnatal multizone organization is acquired. Primary cilia are cell surface organelles crucial for mammalian tissue morphogenesis. Although their importance for chondrocyte function is appreciated, their specific roles in postnatal AC morphogenesis remain unclear. To explore these mechanisms, we used a murine conditional loss-of-function approach (Ift88-flox) targeting joint-lineage progenitors (Gdf5Cre) and monitored postnatal knee AC development. Joint formation and growth up to juvenile stages were largely unaffected. However, mature AC (aged 2 months) exhibited disorganized extracellular matrix, decreased aggrecan and collagen II due to reduced gene expression (not increased catabolism), and marked reduction of AC modulus by 30%–50%. In addition, and unexpectedly, we discovered that tidemark patterning was severely disrupted, as was hedgehog signaling, and exhibited specificity based on regional load-bearing functions of AC. Interestingly, Prg4 expression was markedly increased in highly loaded sites in mutants. Together, our data provide evidence that primary cilia orchestrate postnatal AC morphogenesis including tidemark topography, zonal matrix composition, and ambulation load responses.

Introduction

Body movement requires uniquely shaped synovial joints to maximize range and type of motion based on anatomical location and load-bearing needs.⁽¹⁾ Much has been learned about the structure and function of adult joint tissues—articular cartilage (AC), meniscus, and intrajoint ligaments—including the biomechanical and lubricating roles each tissue plays in sustaining motion and resilience.⁽²⁾ Although it is clear that the majority of morphogenesis and terminal organization of mature AC occur during postnatal life in coordination with growth of the skeleton,^(3–6) the underlying cellular mechanisms remain poorly understood. A full understanding of this morphogenetic process would be critical to inform on how joint developmental defects/diseases are established and, as importantly, how defective and damaged joint tissues could be repaired and regenerated.

In adults, the fully mature AC displays a stereotypic zonal organization of cells and extracellular matrix (ECM) and contains a top non-mineralized portion and a bottom mineralized portion. The nonmineralized cartilage includes: (i) a superficial zone of flat cells juxtaposed with the synovial space and parallel to the surface, producing ECM components such as lubricin/Prg4; (ii) an intermediate zone of round chondrocytes amidst aggrecan and randomly aligned collagen II fibrils; and (iii) a thick, deep zone of polarized chondrocytes oriented into columns and containing abundant ECM rich in aggrecan and collagen II fibrils aligned perpendicular to the surface. The mineralized cartilage comprises a bottom zone with large hypertrophic-like chondrocytes that conjoins AC to the underlying subchondral bone.⁽⁷⁾ The tidemark represents the mineralization front demarcating the boundary between nonmineralized and mineralized cartilage. Although little is known of tidemark developmental origin, significant evidence exists that it advances toward the cartilage surface with age^(8–12) or with alterations in mechanical load,^(13–15) reducing nonmineralized cartilage thickness. In addition, the transitioning of aquatic axolotl salamanders to terrestrial life results in the formation of a tidemark,^(16,17) suggesting that ambulation influences its initiation and positioning. Despite these observations, no cellular mechanisms on tidemark patterning or disease-state repositioning have been clarified.

Postnatal AC morphogenesis into zones is complex because it needs to eventually establish a highly structured, anisotropic tissue critical for joint mechanical function.^(7,18,19) Recent studies by our group and others have led to a better understanding of basic tenets of postnatal morphogenesis of AC in mouse knee joints.^(4–6) At birth, nascent AC is disorganized and contains a pool of highly proliferative progenitors that give rise to all the cells of adult AC. By 2–3 weeks of age, AC is matrix-rich, but still lacks a clear zonal organization. By 8 weeks, the stereotypical zonal organization is established. Importantly, we found that cell proliferation decreases rapidly soon after birth, whereas the tissue continues to thicken through matrix production/accumulation and increases in average cell size.⁽⁴⁾ Thus, there are three stages to postnatal AC development that have become clear: proliferation (neonatal), growth (adolescent), and maturation (adult). Lineage tracing in our recent study strongly suggests that chondrocytes establish zonal organization and vertical columns during the maturation phase,⁽⁴⁾ but the mechanisms by which this occurs have not been elucidated.

Primary cilia are mechano-transducing and morphogen-transducing cell surface organelles present on nearly every cell in mammals and are crucial for tissue morphogenesis. Cytoskeletal regulation, cell polarity, and a number of signaling pathways (eg, Hedgehog, Wnt, and transforming growth factor β [TGFβ]) are governed, at least in part, by the primary cilium.^(20,21) The chondrocyte primary cilium is an important regulator of cartilage function^(22–24) and lengthening has been correlated with human osteoarthritis (OA).⁽²⁵⁾ In mice, conditional loss of primary cilium function in chondrocytes (including in the growth plate) results in AC deficits that often lead to early onset OA.^(26–31) It is thus clear that primary cilia are important to AC health, but the specific mechanisms by which they function during postnatal morphogenesis remain unclearly defined. In addition, the concurrent ablation of cilia in growth plate carried out in previous studies could have had secondary and severe influences on AC, thus requiring new studies selectively interrogating primary cilia function in AC. To identify mechanisms of postnatal AC development in mice and how cilia may function in this context, we developed a joint-specific (Gdf5Cre) disruption of primary cilia (Ift88-flox) and delineated AC development up to 8 weeks of age.

Materials and Methods

Animals

All studies involving mice were reviewed by our institutional IACUC at The Children’s Hospital of Philadelphia. All animals were handled, treated, and cared for according to the approved protocols in conventional group housing and provided standard diet (Lab Diet 5001; LabDiet, St. Louis, MO, USA). IFT88-flox (Stock no. 022409), Rosa-TdTomato (Stock no. 007914), and Gli1-LacZ (Stock no. 008211) mice were obtained from the Jackson Laboratory (Bar Harbor, ME, USA). Gdf5Cre mice (kindly provided by Dr. D. Kingsley [Dept. of Developmental Biology, Stanford University School of Medicine]) were described previously⁽³²⁾ and line B was used in the present study. Wild-type, sex-matched littermate animals (either Ift88^F/F or Ift88^F/+ depending on litter genotyping) were used for controls. Males and females were used equally. Cage activity was not monitored and is a limitation of the study. In approximately one-third of Gdf5Cre animals, Cre recombines randomly early in embryonic development resulting in a mosaic recombination throughout the adult animal. These animals were not used in the study, were monitored by genotyping, and were neonatal lethal consistent with global loss of IFT88 function.⁽³³⁾

Atomic force microscopy–based nanoindentation

Atomic force microscopy (AFM) nanoindentation was applied to freshly dissected femoral condyle and tibial plateau cartilage from mice at 8 weeks of age (n = 5 animals), following the established procedure.^(34,35) Immediately after euthanasia, cartilage was dissected free of tendon and ligament tissues and glued onto AFM sample discs by a cyanoacrylate adhesive gel (Loctite 409; Henkel Corp., Rocky Hill, CT, USA). Throughout the procedure, tissues were immersed in PBS with protease inhibitors (Pierce 88266; Fisher Scientific, Rockford, IL, USA) to minimize postmortem degradation. Nanoindentation was performed on the surfaces of cartilage on a Dimension Icon AFM (Bruker Nano, Santa Barbara, CA, USA) using a colloidal AFM probe. The probe was custom-made by attaching a polystyrene microsphere (radius [R] ≈ 12.5 μm; PolySciences, Warrington, PA, USA) to a tipless cantilever (nominal force constant [k] ≈ 5.4 N/m, cantilever C, HQ:NSC35/tipless/Cr–Au; NanoAndMore, Watsonville, CA, USA) using liquid epoxy glue (M-bond 610; SPI Supplies, West Chester, PA, USA). For each sample, at least 10 random, different indentation locations were tested on each selected region of interest up to ~1μN force at 10 μm/s AFM z-piezo displacement rate. The effective indentation modulus (E_ind) was calculated by fitting the entire loading portion of each indentation force delpth (F–D) curve to the Hertz model, where R is the tip radius and ν is the Poisson’s ratio (0.1 for cartilage).⁽³⁶⁾

β-galactosidase staining

To visualize β-galactosidase (β-gal, LacZ) in Gli1-LacZ animals, β-gal staining was performed by wholemount (n = 4). Following euthanasia, joints were disarticulated and cleaned to expose the articular cartilage and fixed for 1 hour at 4°C in 1% paraformaldehyde (PFA). Tissues were washed three times for 5 minutes in LacZ wash buffer (2mM MgCl2, 0.01% Na Deoxycholate, and 0.02% NP-40 in 1× PBS) followed by staining at 37°C with rocking in the dark for either 3 hours (3-week-old samples) or 16 hours (8-week-old samples). To stain, X-gal (Millipore, Billerica, MA, USA; 71077-3) was freshly diluted 1:40 in LacZ base staining solution (5mM K₃Fe(CN)₆, 5mM K₄Fe(CN)₆ in LacZ wash buffer). All staining was performed with samples from Gli1--LacZ–negative animals to control for background staining. Following staining, samples were fixed in 4% PFA and imaged by wholemount using a Nikon dissecting microscope (Nikon, Tokyo, Japan) equipped with camera and SPOT imaging software (Diagnostic Instruments, Sterling Heights, MI, USA). For sectioning, samples were decalcified in 20% EDTA and processed for paraffin embedding. Six-micrometer-thick (6-μm-thick) sections were counterstained with eosin.

Tissue processing

Tissues processed for frozen sectioning were injected with 4% PFA and fixed for 2 days at 4°C, followed by decalcification in 20% EDTA for 6–10 days (depending on age), washes in 1× PBS, equilibration in 30% sucrose for 16–36 hours (depending on age), and embedding in optimal cutting temperature (OCT) compound. Sections were either taken at 18 μm for normal slide collection or at 10 μm on Kawamoto’s cryofilm⁽³⁷⁾ to preserve tissue morphology. Tissues processed for paraffin sectioning (basic histology and immunofluorescence) were injected with 10% neutral buffered formalin (NBF) and fixed for 2 days at 4°C, followed by decalcification in 20% EDTA for 6–10 days (depending on age), washes in 1× PBS and water, and dehydration in 70% ethanol overnight before automated processing into paraffin overnight. Tissues processed for RNAscope (Advanced Cell Diagnostics, Inc., San Jose, CA, USA) were decalcified using Morse’s solution as described⁽³⁸⁾ and processed into paraffin overnight. All sections were cut at 6-μm-thick and slides were prepared with mutant and littermate control sections on a single slide for each probe (n = 5 per time point).

Histology/immunofluorescence and imaging

Safranin O staining was performed using standard protocols and imaged using a Nikon Eclipse Ci-L upright microscope with attached DS-Fi3 camera (Nikon). Immunofluorescence staining was carried out on frozen or paraffin sections depending on antibody optimization. For staining on frozen sections, 1× PBS + 0.1% Tween was used for all washes and incubations. Sections were blocked with 5% donkey serum and incubated overnight at 4°C with primary antibody and for 2 hours at room temperature with a Cy3-conjugated secondary antibody (1:500; Jackson Immunoresearch, West Grove, PA, USA). For staining on paraffin sections, antigen retrieval was performed as indicated in Table 1. For non-amplified staining, sections were blocked with 5% donkey serum (and Mouse-on-Mouse block as necessary; Vector Laboratories, Inc., Burlingame, CA, USA) and incubated overnight at 4C with primary antibody and for 2 hours at room temperature with a Cy3- or AF488-conjugated secondary antibody (1:500; Jackson Immunoresearch). For amplified staining, sections were treated with 3.6% H₂O₂ (made from 30% stock in water) and permeabilized with 1× PBS + 0.1% Triton for 20 minutes. 1× PBS + 0.1% Tween was used for all following washes and incubations. Sections were blocked with 5% donkey serum (and Mouse-on-Mouse block as necessary; Vector Laboratories). Sections were incubated at room temperature for 90 minutes with primary antibody, for 1 hour with biotin-conjugated secondary antibodies (1:500; Jackson Immunoresearch), for 1 hour with Vectastain Elite ABC amplification kit (Vector Laboratories) and developed for 10 minutes with fluorescein-conjugated or Cy3-conjugated Tyramide amplification system (Perkin Elmer, Shelton, CT, USA). Amplified staining for Aggrecan was performed without detergent (1× PBS only). All sections were counterstained with DAPI (1 mg/mL stock per manufacturer guidelines; Sigma-Aldrich, St. Louis, MO, USA; D9542) diluted 1:10,000 in wash buffer for 10 minutes and coverslipped using Prolong Gold (Invitrogen, Waltham, MA, USA; P36930). A Leica DMi8CEL Advanced inverted SP8 confocal was used to collect fluorescent and DIC images (Leica Camera, Wetzlar, Germany).

Table 1.

Primary Antibody Information and Immunofluorescence Protocols

Antibody	Manufacturer	Catalog #	Protocol	Amplified	Antigen retrieval	Dilution
MMP13	Abcam (Cambridge, UK)	Ab39012	Frozen	No	-	1:100
Perlecan	Life Technologies (Carlsbad, CA, USA)	MA1-06821	Paraffin	No	(ii)	1:50
Col2a1	MilliporeSigma (Burington, MA, USA)	MAB8887	Paraffin	No	(ii)	1:50
ColX	MilliporeSigma	234196	Paraffin	No	(ii)	1:50
Acetylated tubulin	MilliporeSigma	T7451	Paraffin	No	(ii)	1:200
Sox9	Novus Biologicals (Littleton, CO, USA)	AF3075	Paraffin	Yes	(ii)	1:2000
Aggrecan	Dev Studies Hyb Bank (Iowa City, IA, USA)	12/21/1-C-6	Paraffin	Yes	(ii) + (iii)	1:25
pSmad2	Cell Signaling (Danvers, MA, USA)	3108S	Paraffin	Yes	(ii)	1:1500
pSmad159	Cell Signaling	13820	Paraffin	Yes	(i)	1:800

Antigen retrieval methods

1. 1 mg/mL trypsin (Sigma-Aldrich; T7168) for 20 minutes at 37°C.

2. 10 μg/mL Proteinase K for 20 minutes at 37°C.

3. 10mM dithiothreitol (DTT) for 2 hours at 37°C (reduction) followed by 40mM iodoacetamide for 2 hours at room temperature (alkylation) followed by 5 mg/mL Hyaluronidase (in 10% serum) for 1 hour at 37°C.

RNAscope

RNAscope in situ hybridization (n = 5 per time point) was carried out using RNAscope2.5 HD Detection reagent-RED (Advanced Cell Diagnostics) to visualize the spatiotemporal expression of genes listed in Table 2. Briefly, sections were pretreated with a custom reagent and the probe was hybridized for 2 hours at 40°C in a custom oven. Signal was amplified with multiple reagents as per manufacturer’s protocols and final signal was detected and visualized by reaction with Fast Red substrate for 10–20 minutes (dependent on age) at room temperature. Companion sections were hybridized with positive (Advanced Cell Diagnostics; Cat. No. 313911) or negative control probes (Advanced Cell Diagnostics; Cat. No. 310043) to assure signaling specificity (data not shown). Sections imaged with brightfield were counterstained with hematoxylin, dried, and sealed with Permount. Sections imaged with confocal microscopy were counterstained with Hoescht and coverslipped with Prolong Gold. A Nikon Eclipse Ci-L upright microscope with attached DS-Fi3 camera was used to capture brightfield images. A Leica DMi8CEL Advanced inverted SP8 confocal was used to collect fluorescent and differential interference contrast (DIC) images.

Table 2.

RNAscope Probes

Gene name	Probe name	Catalog number
Adamts4	Mm-Adamts4	497161
Adamts5	Mm-Adamts5	427621
Col2a1	Mm-Col2a1	407221
Acan	Mm-Acan	439101
Prg4	Mm-Prg4	437661
Gli1	Mm-Gli1	311001
Ptch1	Mm-Ptch1	402811
Ihh	Mm-Ihh	413091

Quantification from tissue sections

All quantification was carried out using ImageJ software (National Institutes of Health [NIH], Bethesda, MD, USA; https://imagej.nih.gov/ij/) with FIJI plugins. All sections were chosen from the central region of the medial knee that includes that concentrated load-bearing region of the tibial plateau. For cilia incidence, acetylated tubulin immunofluorescence DAPI signal was quantified manually using the cell counter plugin in FIJI. For AC thickness measurements (n = 5 controls and four mutants at 3 weeks; n = 6 controls and 7 mutants at 8 weeks). Safranin O-stained sections were used. For each section (three to six per animal), the thickest region of AC was identified, and eight lines were drawn manually and measured. An average and standard deviation (SD) was calculated from all lines in all sections per region of interest. Percent Sox9-positive (n = 5 at 3 weeks, n = 6 at 8 weeks), pSmad2-positive (n = 7 at 8 weeks), and pSmad159-positive (n = 7 at 8 weeks) cells was calculated from immunofluorescent-stained slides counterstained with DAPI and additionally imaged with DIC to outline nonmineralized and mineralized cartilage zones (two sections per animal per stain). The total number of cells in each region was calculated by thresholding the DAPI signal in ImageJ and analyzing particles to count each nucleus. The same thresholding technique was applied to count Sox9, pSmad2, and pSmad159-positive cells and divided by the total (DAPI) number of cells per region. Cell density at 3 weeks and 8 weeks was calculated from DAPI staining and DIC imaging (from Sox9, pSmad2 and pSmad159-stained sections). AC composition at 8 weeks was calculated from DIC imaging (from Sox9, pSmad2 and pSmad159-stained sections). To quantify Collagen II immunofluorescence, total AC area for was manually outlined on each section (two sections per animal) and the mean gray value was measured separately for femoral condyle and tibial plateau. RNAscope for Prg4 was quantified by thresholding the signal from brightfield images manually and applying the same threshold value to all images. The thresholded area was quantified over total AC area (drawn manually) in all sections (one section per animal, n = 5). Gli1-LacZ in tissue sections at 8 weeks was quantified manually using the cell counter plugin in FIJI (n = 6). The approximate location of the tidemark in control animals was drawn, as described in the Results section, using knowledge from DIC images of previous quantification, and applied to all control and mutant tissue sections. The femoral condyle tidemark was drawn four cell layers from the surface and the tibial plateau tidemark was drawn seven cell layers from the surface.

Statistical methods

Sample sizes were justified based on prior experimentation using the described techniques. A relative standard deviation of 15%–20% is found in cartilage biomechanics for mice. Based on a 30% estimated difference between groups, a statistical power (1 – β) or 80% at a significance level of α = 5% is obtained with a sample size of n = 6 biological repeats per group. For histologic assessments, we also aimed for n = 6 for most measurements but used n = 4 for quantification analyses of 3-week samples where no visual differences were detected for multiple stains and n = 5 for RNAscope where the procedure limits the number of slides per experiment. Data were analyzed using Prism 9 (Graphpad Software, Inc., San Diego, CA, USA). For all analyses, two groups were compared to each other (control versus mutant) and we used unpaired t tests with Welch correction where p < 0.05 was considered statistically significant. All data are shown as box plots with interquartile range and minimum and maximum.

Results

Gdf5Cre is joint specific, and conditional deletion of Ift88 results in loss of primary cilia

To target primary cilia specifically in synovial joints, we used Gdf5Cre in combination with an Ift88 conditional allele. Gdf5 is a unique marker for the embryonic mesenchymal progenitor cells of synovial joints called the interzone. Gdf5Cre faithfully marks these cells in the embryo and lineage-traces to all of the mature tissues of synovial joints^(4,39) (Fig. 1Aa–c) but not growth plate chondrocytes (Fig. 1Ad). Thus, we combined Gdf5Cre with an Ift88 conditional allele whose product (intraflagellar transport protein 88, IFT88) is important in anterograde transport in cilia and maintenance of cilium structure. We visualized primary cilia using an antibody against acetylated tubulin. They were evident as expected in growth plate chondrocytes in control and mutant animals (Fig. 1B). They were also present and clear in control AC (Fig. 1Ca,b) but much reduced in the AC of mutant animals (Fig. 1Cc,d). The incidence of primary cilia was quantified in control and mutant AC (Fig. 1D), confirming that our model was reliable and the loss of IFT88 resulted in a significant reduction of primary cilia specifically in articular chondrocytes.

Fig. 1.

Gdf5Cre is joint specific and conditional deletion of Ift88 results in loss of primary cilia. (A) Gdf5Cre/+; Rosa-tdTomato lineage cells are present in adult joints (Aa) and all zones of AC as shown with DIC imaging to visualize the tidemark (Ab,c) but are absent in growth plate (Ad) (n = 3). (B,C) Primary cilia are visualized by immunofluorescence staining for acetylated tubulin. They are present in control and mutant growth plate chondrocytes (B) and in control AC (Ca,b), and visibly reduced in mutant AC (Cc,d). (D) The incidence of cilia in AC is quantified as a percentage of AC nuclei (n = 5). Statistical analysis was performed by unpaired t tests with Welch correction. Scale bar = 100 μm (Ab–d), 10 μm (B), 50 μm (C). All images displayed with FC on top and TP on bottom unless otherwise noted. FC = femoral condyle; sb = subchondral bone; TP = tibial plateau.

Joint-specific deletion of Ift88 leads to disrupted AC tidemark patterning and reduced AC stiffness

Next, we carried out histological analysis and performed mechanical testing to assess the consequences of primary cilia loss on postnatal AC morphogenesis. For all images provided, the femoral condyle is on top and the tibial plateau is on the bottom, unless otherwise noted. We used Safranin-O staining to visualize basic histology and proteoglycan content in the ECM (Fig. 2A). At birth, nascent AC of mutants was indistinguishable from that in controls (Fig. 2Aa,b, brackets). Up to 3 weeks, immature AC developed similarly in all animals (Fig. 2Ac,d) and higher magnification images revealed no discernible differences in cell morphology or patterning (Fig. 2Ae–h). At 8 weeks, however, we noted irregular accumulation of interterritorial ECM (arrows) and varying chondrocyte size (asterisks) in the femoral condyle (Fig. 2Ai–l) and a consistent reduction of Safranin-O in the tibial plateau (Fig. 2i,j,m,n). Thus, Ift88 deletion primarily affected the maturation phase of postnatal AC morphogenesis. Growth plates were unaffected at all stages consistent with no recombination from Gdf5Cre (Fig. S1). To assess possible functional consequences of these morphologic changes in mature AC, we used AFM nanoindentation to measure tissue stiffness. We measured two regions on the medial tibial plateau (the edge covered by meniscus and the center load-bearing region) and one region on the femoral condyle. In all regions of interest, the indentation modulus (ie, resistance to deformation) was significantly decreased in mutants (Fig. 2B). Thus, Ift88 deletion in AC results in impaired load-bearing function during tissue maturation.

Fig. 2.

Joint-specific deletion of Ift88 leads to disrupted AC tidemark patterning and reduced AC stiffness. (A) Safranin-O staining at birth through 8 weeks. Nascent AC at birth is unchanged in mutant joints (Aa,b, brackets). AC is also unchanged at 3 weeks. Low magnification (Ac,d) and high magnification images of femoral condyle (Ae,f) or tibial plateau (Ag,h) are comparable. At 8 weeks, control AC (Ai,k,m) displays strong safranin O staining throughout the femoral condyle (Ak) and tibial plateau (Am). In mutant AC (Aj,l,n), the femoral condyle displays aberrant ECM deposition (Al, arrows) and increased cell size (Al, asterisks), while the tibial plateau displays reduced Safranin-O staining (An). (B) Biomechanical function was measured by AFM-snanoindentation from three regions. Schematic shows approximate location of the two regions of tibial plateau AC. Indentation modulus was significantly decreased in all locations. (C) DIC imaging of the tidemark (dashed yellow line) reveals ordered patterning in control AC (Ca). In mutant AC, the tidemark of the femoral condyle is disordered and its distance from the surface is increased in the tibial plateau (Cb, bracket). (D) Nonmineralized (non-mAC) composition is shown as a percent area from multiple tissue sections and is significantly increased in mutant AC. (E,F) Measurements of maximum AC thickness and cell density. In the femoral condyle (E), maximum AC thickness of total cartilage and non-mAC is significantly increased. Cell density is significantly decreased throughout mutant AC. In the tibial plateau (F), maximum AC thickness of total cartilage is unchanged, however non-mAC is significantly increased and mAC is significantly decreased. Cell density is unchanged. Statistical analyses were performed by unpaired t tests with Welch correction. Scale bar = 100 μm. All images displayed with FC on top and TP on bottom unless otherwise noted. AC = articular cartilage; FC = femoral condyle; SB = subchondral bone; mAC = mineralized AC; non-mAC = nonmineralized AC; TP = tibial plateau.

Next, we utilized a combination of Safranin-O–stained sections and sections imaged with DIC to visualize the tidemark and quantify AC composition of mineralized (mAC) and nonmineralized regions (non-mAC), maximum AC thickness, and cell density. These analyses revealed several notable changes in mutants. First, in healthy AC the tidemark (Fig. 2Ca, dashed yellow line) closely followed the contour of the femoral condyle surface and was two to three cell layers below the surface. Instead, it was relatively flat in the central load-bearing region of the tibial plateau and was located about five to six cell layers from the surface (Fig. 2Ca, bracket). In mutants, the tidemark was more difficult to visualize, but followed an irregular path in the femoral condyle and was flat, but noticeably further from the surface, in the tibial plateau (Fig. 2Cb, bracket). Quantifying AC composition as a percent area (Fig. S2A), we found that nonmineralized articular cartilage constituted approximately 31% of the femoral condyle and 51% of the tibial plateau AC in control animals. In mutants, this fraction increased to 36% and 70% in the femoral condyle and tibial plateau, respectively (Fig. 2D) and was accompanied by respective decreases in mineralized AC (Fig. S2B). We measured maximum AC thickness and cell density to further understand changes in zonal composition. In the femoral condyle, we found that the transition of mutant AC to subchondral bone tended to be more irregular and significantly increased maximal non-mAC thickness (Fig. 2E, top). Cell density was in general decreased in femoral condyle AC (Fig. 2E, bottom). In the tibial plateau, although there was no overall change in AC thickness, the maximal thickness of non-mAC increased whereas that of mAC decreased (Fig. 2F, top). Cell density remained unchanged in the tibial plateau (Fig. 2F, bottom). Cell density and AC thickness were both unchanged at 3 weeks (Fig. S2C).

Primary cilia regulate chondrocyte fate and matrix deposition into mineralized and nonmineralized zones

Next, we performed immunofluorescence staining to examine several key ECM and chondrocyte-specific markers. Collagen II is a major structural component of cartilage ECM, responsible for tissue tensile strength.⁽⁴⁰⁾ In control animals, it was uniformly dispersed in AC at 3 and 8 weeks (Fig. 3Aa,b). In mutant animals, we noted no changes at 3 weeks (Fig. 3Ac,B, top) but a reduction at 8 weeks (Fig. 3Ad,B, bottom). Aggrecan is the major proteoglycan in cartilage ECM that yields compressive load-bearing and energy dissipation properties. In healthy tissue, it was uniformly deposited over the entire thickness of developing AC at 3 weeks (Fig. 3Ca) and was enriched in nonmineralized cartilage at 8 weeks (Fig. 3Cb). A clear boundary (ie, tidemark) could be visualized where the bottom aggrecan-low mineralized cartilage met the top aggrecan-rich nonmineralized cartilage (Fig. 3Cb, arrows). In Ift88 mutants, aggrecan deposition was comparable to controls in immature cartilage at 3 weeks (Fig. 3Cc), but at 8 weeks the mutant tissue displayed notable disorganization, lack of tidemark, and variable aggrecan deposition (high, arrows; low, arrowheads) (Fig. 3Cd). Sox9 is considered a master transcriptional regulator of chondrocyte fate and can directly regulate the expression of aggrecan and collagen II.⁽⁴¹⁾ In control immature AC, Sox9 was present uniformly in articular chondrocytes (Fig. 3Da) consistent with the uniform distribution of collagen II and aggrecan at this 3-week stage. In mature AC, Sox9-positive cells were largely restricted to the nonmineralized cartilage above the tidemark (Fig. 3Db, dashed line), as visualized by DIC imaging (Fig. S3a–c), and positive cell distribution was consistent with previous reports of Sox9 function in AC.^(42,43) Mutant tissues at 3 weeks showed comparable Sox9 localization to controls (Fig. 3Dc,E, top). At 8 weeks Sox9-positive chondrocytes continued to be broadly present in nonmineralized cartilage (Fig. 3Dd and Fig. S3d–f), but positive cells could also be seen beneath the tidemark in femoral condyle (Fig. 3D, arrows; Fig. 3E, bottom). Taken together, these data show that Ift88 mutants display alterations in the patterning of nonmineralized chondrocytes and in the distribution of ECM components in mature AC, but not in immature AC.

Fig. 3.

Primary cilia regulate chondrocyte fate and matrix deposition in mineralized and nonmineralized zones. (A,B) Collagen II immunostaining is broadly distributed in AC and comparable at 3 weeks (Aa–c) and reduced at 8 weeks in mutants (Ad–f). The mean fluorescence intensity is quantified at 3 weeks (top) and 8 weeks (bottom) in B. (C) Aggrecan immunostaining is broadly distributed and comparable at 3 weeks (Ca,c). It is enriched in nonmineralized 8-week AC of controls (Cb, white arrows point to approximate tidemark) and is disorganized in 8-week AC of mutants (Cd), with irregular areas of high staining (green arrows) and low staining (green arrowheads). (D,E) Sox9 is broadly distributed and comparable in AC at 3 weeks (Da,c). It is also broadly distributed in nonmineralized AC of controls (Db) and mutants (Dd) at 8 weeks (dashed line represents the tidemark), but mutant femoral condyle AC also displays ectopic Sox9-positive cells beyond the tidemark (Dd, white arrows). The percentage of Sox9-positive cells is quantified of all cells at 3 weeks (top) and in zones at 8 weeks (bottom) in E. (F–H) Perlecan immunostaining detects the pericellular chondrocyte matrix. At 3 weeks (F), Perlecan is broadly distributed in AC and comparable in controls and mutants. At 8 weeks (G), Perlecan is enriched nearest the surface in both controls (Ga) and mutants (Gd). High magnification with DIC shows most perlecan staining in nonmineralized AC above the tidemark (Gb–c, e–f, white arrows). Ectopic perlecan-positive cells are present below the mutant femoral condyle tidemark (e, asterisks), quantified in H. Statistical analyses were performed by unpaired t tests with Welch correction. Scale bar = 100 μm unless otherwise noted. All images displayed with FC on top and TP on bottom unless otherwise noted. FC = femoral condyle; TP = tibial plateau; mAC = mineralized articular cartilage; non-mAC = nonmineralized articular cartilage.

Pericellular matrix (PCM) creates a thin microenvironment surrounding chondrocytes within the broader ECM and is considered a crucial transducer of biomechanical signals.⁽⁴⁴⁾ The PCM is rich in certain matrix macromolecules including collagen VI and perlecan; together with the enclosed chondrocytes, the PCM is referred to as a chondron.^(45–47) To analyze the PCM, we performed immunofluorescence staining for perlecan. In immature AC at 3 weeks, perlecan was broadly distributed in both control and mutant tissue (Fig. 3F). In mature AC at 8 weeks, perlecan was localized to chondrocytes nearest to the surface in controls (Fig. 3Ga) and high magnification images overlaid with DIC imaging revealed restriction to the nonmineralized cartilage above the tidemark (Fig. 3Gb,c, arrows). Perlecan deposition was significantly disrupted in mutant AC at 8 weeks (Fig. 3Gd). In the femoral condyle, chondrocytes with a perlecan-rich PCM could be visualized far below the tidemark (Fig. 3Ge, asterisks; Fig. 3H), but in tibial plateau AC there was an expansion of perlecan-positive chondrocytes extending deeper into the concentrated load-bearing region above the tidemark (Fig. 3Gf). Together, these data provide strong evidence that deletion of Ift88 in AC disrupts tidemark patterning and chondrocyte fate/microenvironment during postnatal maturation.

Ift88 mutants do not show indications of AC matrix degradation or precocious hypertrophy

Previous in vitro studies have shown that primary cilia, and intraflagellar transport (IFT) proteins, regulate ECM proteases in chondrocytes.⁽⁴⁸⁾ To test whether this applied to our in vivo model, we performed immunofluorescence staining for MMP13 (collagenase) and in situ hybridization for Adamts4 and Adamts5 (aggrecanases) at 8 weeks. Matrix metalloproteinase 13 (MMP13) was restricted to the base of AC at the transition to subchondral bone in controls (Fig. 4Aa) and was unchanged in mutants (Fig. 4Ab). We used RNAscope (FastRed detection) to perform in situ hybridization and took advantage of FastRed fluorescence for imaging low-expressing genes. In AC, we found no expression of either Adamts4 or Adamts5 in controls (Fig. 4Ba,Ca) or mutants (Fig. 4Bb,Cb). To ensure these data were reliable, we compared to other tissues in and around the joint. Adamts4 (Fig. S4A) was expressed in primary spongiosa (PS, arrowheads) beneath the growth plate and in periosteum (PO, arrows), and Adamts5 (Fig. S4B) was highly expressed by inner synovial fibroblasts in both controls and mutants. These data indicate that the changes observed in collagen II and aggrecan deposition in mutant AC (Fig. 2A–C) were likely not caused by matrix degradation. Though protease gene expression and enzymatic activity does not always correspond with each other, the lack of detectable protease gene expression supports our conclusion. Collagen X is a marker of hypertrophic and mineralizing chondrocytes and often accompanies precocious matrix degradation. We found collagen X similarly localized in control and mutant cells juxtaposed to the developing secondary ossification center at 3 weeks (Fig. 4Da, b) and 8 weeks (Fig. 4Dc,d), consistent with a lack of precocious hypertrophy closer to the AC surface.

Fig. 4.

Ift88 mutants do not show indications of AC matrix degradation or precocious hypertrophy. (A) Immunofluorescence for MMP13 is restricted to the subchondral bone at 8 weeks and is unchanged in controls (Aa) and mutants (Ab). (B,C) Adamts4 and Adamts5 are not detectable by RNAscope in control (Ba, Ca) or mutant (Bb, Cb) AC at 8 weeks. (D) Immunofluorescence for CollagenX is restricted to chondrocytes near the subchondral bone and is comparable in controls and mutants at 3 weeks (Da,b) and 8 weeks (Dc,d). Scale bar = 100 μm. All images displayed with FC on top and TP on bottom. FC = femoral condyle; TP = tibial plateau.

Altered ECM in mature mutant AC correlates with abnormal gene expression

To assess transcriptional changes of Col2a1, Acan, and Prg4, we utilized RNAscope (FastRed detection) and imaged these high-expressing genes with brightfield. At 3 weeks, Col2a1 and Acan were broadly expressed in immature AC of both controls (Fig. 5Aa,Ba) and mutants (Fig. 5Ab,Bb), consistent with our findings from immunofluorescent staining at this stage described in Fig. 3A–C. In mature control AC at 8 weeks, Col2a1 (Fig. 5Ac,d) and Acan (Fig. 5Bc,d) transcripts were restricted to layers of AC corresponding with the domain of nonmineralized cartilage, two to three layers deep in the femoral condyle and five to six layers deep in the tibial plateau. In mutant AC, both Col2a1 (Fig. 5Ae,f) and Acan (Fig. 5Be,f) expression appeared much more disorganized, particularly in the femoral condyle (arrows), consistent with deviant Sox9-positive cells outside of the nonmineralized cartilage (Fig. 3D,E). High-magnification images (Fig. 5Af, Bf) highlighted the clear disorganization of gene expression in all mutant cartilage, consistent with dysregulated tidemark patterning and ECM deposition (Figs. 2 and 3) and reduced expression per cell of Col2a1 and Acan in the concentrated load-bearing region of the tibial plateau compared to this region in controls. Prg4 (lubricin) is important for protecting AC from OA.^(49–53) Prg4 was highly expressed and uniformly distributed in cell layers closest to the AC surface in control animals at 3 weeks (Fig. 5Ca) and at 8 weeks (Fig. 5Cc–e), extending slightly farther in the tibial plateau (Fig. 5Ce, bracket). In mutant femoral condyle cartilage, we found a significant reduction of Prg4 at both 3 weeks (Fig. 5Cb) and at 8 weeks (Fig. 5Cd,f; Fig. 5D top). In mutant tibial plateau, there was no change in Prg4 at 3 weeks (Fig. 5Cb) and a striking increase in the concentrated load-bearing region at 8 weeks (Fig. 5Cd,f, bracket; Fig. 5D, bottom). These data suggest that reduction/disorganization of collagen II and aggrecan deposition in mature cartilage were due to changes at the transcriptional level and that, surprisingly, Prg4 expression was altered in both immature and mature cartilage of mutants.

Fig. 5.

Altered ECM in mature mutant AC correlates with abnormal gene expression. (A,B) RNAscope for Col2a1 (A) and Acan (B) is comparable in controls and mutants at 3 weeks (Aa,b, Ba,b). In control AC at 8 weeks, low magnification images (Ac, Bc) and high magnification images (Ad, Bd) show organization and high levels of expression closest to the surface. In mutant AC, low magnification images (Ae, Be) and high magnification images (Af, Bf) show generally disorganized expression that is displayed deeper into the tissue than in controls (arrows) and decreased particularly in the tibial plateau. (C,D) RNAscope for Prg4 is shown with quantification in D. In the femoral condyle, Prg4 is decreased at 3 weeks (Ca,b, top). At 8 weeks, low magnification images (Cc,d, top) and high magnification images (Ce,f, top) show generally decreased but patchy expression (arrows). In the tibial plateau, Prg4 expression is unchanged at 3 weeks (Ca,b, bottom). At 8 weeks, low-magnification images (Cc,d, bottom) and high-magnification images (Ce,f, bottom) show a significant increase of Prg4 that extends further from the surface (bracket). Quantification of Prg4 in D is shown as a percentage area of AC per section after staining area was measured by thresholding the red signal in ImageJ. Statistical analyses performed by unpaired t tests with Welch correction. M (with arrow) denotes regions of mAC in images. Scale bar = 100 μm. All images displayed with FC on top and TP on bottom unless otherwise noted. FC = femoral condyle; TP = tibial plateau; mAC = mineralized articular cartilage; non-mAC = nonmineralized articular cartilage.

Primary cilia impart load/location-dependent functions for TGFβ and bone morphogenetic protein signaling

Primary cilia are considered a signaling nexus that exert regulatory influences on most major signaling pathways.⁽⁵⁴⁾ TGFβ and bone morphogenetic protein (BMP) regulate chondrocyte fate and ECM anabolism by distinct mechanisms via downstream SMAD-dependent mechanisms.⁽⁵⁵⁾ Although TGFβ signaling largely maintains nonmineralized chondrocyte function and prevents hypertrophy,^(56–64) BMP signaling can influence both nonmineralized and mineralized functions dependent on context.^(65–70) To assess the involvement of these pathways, we used immunofluorescence for pSmad2 (TGFβ) or pSmad159 (BMP). In immature AC at 3 weeks, pSmad2-positive cells were broadly distributed in all layers (Fig. 6Aa), whereas pSmad159-positive cells were largely present in the deepest layers of cartilage (Fig. 6Ba). These patterns were unchanged in mutant cartilage (Fig. 6Ab,Bb). In mature control AC at 8 weeks, pSmad2 was abundantly present (Fig. 6Ca) and localized to non-mineralized AC above the tidemark (Fig. 6Cb,c, dashed line). In mutant femoral condyle AC, pSmad2 was similarly localized to non-mineralized cartilage despite tidemark topographical irregularities (Fig. 6Cd,e,D). However, in mutant tibial plateau AC, areas of pSmad2-negative cells could be visualized (Fig. 6Cf, arrow) and were quantifiably reduced (Fig. 6D). We performed parallel experiments for pSmad159 and found similar localization in nonmineralized cartilage in control animals (Fig. 6Ea–c) and in mutant femoral condyle (Fig. 6Ed,e), but a significant, quantifiable and specific loss only in mutant tibial plateau (Fig. 6Ef,F). These data suggest that disruption of IFT88 leads to alterations in BMP signaling and, to a lesser degree, TGFβ signaling in the concentrated load-bearing region of the tibial plateau simultaneously with ECM compositional alterations to collagenII and aggrecan described in Fig. 3 and Fig. 5.

Fig. 6.

Primary cilia impart load/location-dependent functions for TGFβ and BMP signaling. (A,B) Immunofluorescence for pSmad2 (TGFβ) and pSmad159 (BMP) at 3 weeks. pSmad2-positive chondrocytes are distributed throughout AC in both controls (Aa) and mutants (Ab). pSmad159-positive chondrocytes at 3 weeks are localized in middle and bottom layers of AC in both controls (Ba) and mutants (Bb). (C–F) Immunofluorescence and quantification for pSmad2 (TGFβ) and pSmad159 (BMP) at 8 weeks. DIC imaging is used to visualize the tidemark (dashed white line). (C) Low magnification (Ca,d) and high magnification of femoral condyle (Cb,e) and tibial plateau (Cc,f) shows that pSmad2 is mostly localized to non-mAC (above the tidemark). Positive cells are unchanged in the femoral condyle (Ce) and displays abnormal negative regions in the tibial plateau (Cf, arrow). Quantification (D) shows a significant reduction in the non-mAC of the tibial plateau. (E) Low magnification (Ea,d) and high magnification of femoral condyle (Eb,e) and tibial plateau (Ec,f) shows that pSmad159 is mostly localized to non-mAC (above the tidemark). Positive cells are unchanged in the femoral condyle (Ee) and displays a significant reduction in the tibial plateau (Ef). Quantification confirms this in F. Statistical analyses were performed by unpaired t tests with Welch correction. M (with arrow) denotes regions of mAC in images. Scale bar = 100 μm. All images displayed with FC on top and TP on bottom unless otherwise noted. FC = femoral condyle; TP = tibial plateau; mAC = mineralized articular cartilage; non-mAC = nonmineralized articular cartilage.

Patterning of Hedgehog-responsive cells is coupled to tidemark patterning

Hedgehog signaling is a regulator of chondrocyte function and is dependent on the primary cilium for proper transduction through processing of glioma-associated oncogene (Gli) transcriptional effectors (Gli1, Gli2, and Gli3) for pathway activation or repression.⁽²¹⁾ In the absence of hedgehog ligands (Sonic, Desert, or Indian), the transmembrane receptor Patched1 (Ptch1) inhibits the transmembrane protein Smoothened (Smo) to sequester Gli factors in the cytoplasm. When hedgehog ligand is present, it binds to Ptch1 and, in turn, the inhibition on Smo is relieved and Gli proteins are processed into active forms translocating to the nucleus for transcriptional activation. Thus, loss of primary cilia generally results in loss of hedgehog signaling. During embryonic development, hedgehog signaling is tightly linked to tissue patterning through established gradients of ligand distribution. A notable example of this is neural tube formation where cells establish specific identities dependent on their exposure, and response, to hedgehog ligands⁽⁷¹⁾ and finetuned by a number of co-activators/antagonists.^(72–74) Zonal chondrocyte patterning of the growth plate is also dependent on hedgehog exposure in a signaling loop that requires Indian Hedgehog (Ihh)-expressing prehypertrophic chondrocytes and Parathyroid related protein (Pthrp)-expressing reserve zone chondrocytes.^(75,76) Hedgehog signaling patterns have not been well established in AC.

We sought to elucidate a possible role for hedgehog signaling in our AC model. To visualize active signaling, we used the Gli1--LacZ reporter mouse in combination with whole mount staining and imaging. Background staining was monitored using Gli1--LacZ–negative tissues (Fig. S5). We obtained several key findings. First, hedgehog signaling was highly active in immature AC of control animals at 3 weeks (Fig. 7Aa,Ba), and tissue sections revealed differences in Gli1-LacZ patterning of the femoral condyle and concentrated load-bearing region of the tibial plateau, localized to the first three to four cells layers of the femoral condyle (Fig. 7Ac) and to the bottom three to four cells layers in the tibial plateau (Fig. 7Bc). Gli1-LacZ persisted in mature AC at 8 weeks (Fig. 7Ca,Da) with similar cell layer patterning of immature cartilage and near the approximate location of the tidemark (Fig. 7Cc,Dc, dashed line). In immature cartilage of mutants, Gli1--LacZ was strongly downregulated in both the femoral condyle (Fig. 7Ab,d) and the tibial plateau (Fig. 7Bb,d). In mature mutant femoral condyle AC, Gli1-LacZ repression was maintained from 3 weeks (Fig. 7Cb), but exhibited randomly distributed positive cells (Fig. 7Cd). In the concentrated load-bearing region of mutant AC tibial plateau, Gli1-LacZ was strongly upregulated (Fig. 7Db, arrows) including at the surface (Fig. 7Dd), a pattern not seen in controls. The zonal properties of Gli1-LacZ in mature AC was quantified by approximating the tidemark location of control tissues based on knowledge from DIC imaging (Fig. 7Cc,d,Dc,d, dashed line). Staining with eosin and brightfield imaging to visualize LacZ did not allow actual tidemark visualization. Quantification confirmed a significant reduction of Gli1--LacZ in nonmineralized femoral condyle AC of mutants and a sharp increase in the tibial plateau (Fig. 7E, left). Interestingly, Gli1-LacZ was increased in mineralized AC of both regions, suggesting extension beyond the normal tidemark location (Fig. 7E, right). Taken together, the data show significant hedgehog signal disruption before a structural AC phenotype is apparent, largely consistent with known roles for primary cilia in hedgehog signaling. However, our data also shows hyperactive Gli1-LacZ activity in the absence of primary cilia but in conjunction with high mechanical load during ambulation, suggesting load-dependent alterations.

Fig. 7.

Primary cilia impart age-dependent and anatomic location-dependent functions for Hedgehog signaling. Gli1-LacZ (active hedgehog signaling) was developed by whole mount β-gal staining. Yellow bars on whole mount images indicate the approximate location of tissue section images. (A,B) At 3 weeks, Gli1-LacZ staining is strong in control femoral condyle (Aa) and tibial plateau (Ba). Staining is reduced in mutant femoral condyle (Ab) and tibial plateau (Bb). Tissue sections at 3 weeks from controls reveal Gli1-LacZ-expressing cells localized near the surface in the femoral condyle (Ac) and three to four cell layers from the surface in the tibial plateau (Bc). In mutant tissue sections, Gli1-LacZ is patchy in the femoral condyle (Ab) and nearly absent in the tibial plateau (Bd). (C–E) Dashed line denotes approximate location of the tidemark in control tissues. At 8 weeks, Gli1-LacZ staining is strong in control femoral condyle (Ca) and tibial plateau (Da). Staining is reduced in mutant femoral condyle (Cb) and increased in the central region of the tibial plateau (Db, arrows). Tissue sections at 8 weeks from controls reveal Gli1-LacZ–expressing cells localized near the surface in the femoral condyle (Cc) and three to four cell layers from the surface in the tibial plateau (Dc). In mutant tissue sections, Gli1-LacZ is patchy in the femoral condyle (Cb) and dramatically increased in the tibial plateau up to the surface (Dd). The percent of Gli1-LacZ postitive cells was quantified by zone using an approximation of location that is known in control tissues. Mutants display decreased positive cells in non-mAC of the femoral condyle and increases in all other regions. Statistical analyses performed by unpaired t-tests with Welch correction. M (with arrow) denotes regions of mAC in images. Scale bar = 100 μm. FC = femoral condyle; TP = tibial plateau; mAC = mineralized articular cartilage; non-mAC = nonmineralized articular cartilage.

Using RNAscope, we assessed two major hedgehog response genes (Gli1 and Ptch1) as well as Ihh in immature and mature AC and compared these patterns to the response when Ift88 was disrupted. Once again, we used confocal microscopy to image the FastRed detection and overlaid these results with DIC to visualize cell morphology, the tidemark (Fig. 8, dashed blue line), and the subchondral bone (Fig. 8, sb, solid white line). First, we established patterning in immature AC at 3 weeks. In control AC, Gli1 and Ptch1 were expressed in overlapping domains surrounding the future tidemark located three to four cell layers below the surface in femoral condyle (Fig. 8Aa,b) and four to six cell layers in tibial plateau (Fig. 8Ba,b). Importantly, these data were consistent with those from our Gli-LacZ studies (Fig. 7A,B). In both regions only cells juxtaposed to the underlying subchondral bone expressed Ihh (Fig. 8Ac,Bc). Two things were particularly notable about these patterns in immature AC: (i) the location of the future tidemark correlated strongly with Gli1 and Ptch1 expression; and (ii) despite Ihh-expressing cells being located always nearest the subchondral bone, the responding cells were closer to the surface in condylar cartilage than in tibial plateau. In mutant AC at this stage, Gli1 was significantly reduced in both femoral condyle and tibial plateau AC (Fig. 8Ad,Bd), consistent with Gli1-LacZ studies (Fig. 7C,D). Ptch1 expression generally followed the same profile, but with some notable groups of positive cells in the femoral condyle (Fig. 8Ae, arrows; Fig. 8Be). Importantly, Ihh expression was unchanged compared to controls (Fig. 8Af,Bf). Thus, downregulation of signaling in this context is due to lack of response rather than lack of ligand expression.

Fig. 8.

Patterning of hedgehog responsive cells correlates with tidemark patterning. RNAscope detected by FastRed is visualized by confocal microscopy and overlaid with DIC imaging to visualize cell morphology and tidemark patterning. The blue dashed line in (A,B) represents the approximate level of the future tidemark in control AC and in (C,D) it follows the tidemark visualized by DIC. (A) Femoral condyle AC at 3 weeks. Control AC expresses Gli1 in multiple cell layers near the AC surface (Aa), Ptch1 in one to two cell layers around the presumptive tidemark (Ab) and Ihh in chondrocytes adjacent to the subchondral bone (Ac). Mutant AC displays decreased but patchy Gli1 (Ad, arrows) and Ptch1 (Ae, arrows) expression and normal Ihh in chondrocytes adjacent to the subchondral bone (Af). (B) Tibial plateau AC at 3 weeks. Control AC expresses Gli1 (Ba) and Ptch1 (Bb) in similar layers surrounding the presumptive tidemark and Ihh in chondrocytes adjacent to the subchondral bone (Bc). Mutant AC displays decreased Gli1 (Bd) and Ptch1 (Be) expression and normal Ihh in chondrocytes adjacent to the subchondral bone (Bf). (C) Femoral condyle AC at 8 weeks. Control AC expresses Gli1 in multiple cell layers near the AC surface (Ca), Ptch1 in one to two cell layers around the tidemark (Cb) and Ihh in chondrocytes just beneath the tidemark (Cc). Mutant AC displays decreased expression of Gli1 (Cd), decreased and patchy expression of Ptch1 that follows the irregular tidemark (Ce, arrows) and ectopic Ihh in nonmineralized chondrocytes (Cf, arrows). (D) Tibial plateau AC at 8 weeks. Control AC expresses Gli1 (Da) and Ptch1 (Db) in similar layers surrounding the tidemark and Ihh in chondrocytes just beneath the tidemark (Dc). Mutant AC displays increased and expanded expression of Gli1 (Dd) and Ptch1 (De) but normal Ihh just beneath the tidemark (Df). Scale bar = 100 μm.

We performed identical analyses in mature AC at 8 weeks of age. Femoral condyle AC of controls displayed Gli1 expression in nonmineralized chondrocytes including surface zone cells (Fig. 8Ca), Ptch1 expression in cells at the level of the tidemark (Fig. 8Cb), and Ihh expression in mineralized chondrocytes beneath the tidemark (Fig. 8Cc). Tibial plateau AC displayed overlapping Gli1 and Ptch1 in a zone two to three cell layers thick surrounding the tidemark (Fig. 8Da,b) and Ihh in mineralized chondrocytes beneath the tidemark (Fig. 8Dc). Femoral condyle cartilage of mutants displayed generally reduced Gli1 above the tidemark (Fig. 8Cd), consistent with Gli1-LacZ studies. Interestingly, although Ptch1 was also significantly reduced compared to control cartilage, the random distribution of persistent Ptch1-expressing chondrocytes correlated with the irregularly shaped tidemark (Fig. 8Ce, arrows). Ihh also persisted but was ectopically expressing in some nonmineralized chondrocytes (Fig. 8Cf, arrows). In tibial plateau cartilage, Gli1 and Ptch1 expression was highly increased in the concentrated load-bearing region (Fig. 8Dd,e) consistent with Gli1-LacZ, but, unlike the condylar cartilage, Ihh remained exclusively expressed by nonmineralized chondrocytes beneath the tidemark (Fig. 8Df). Taken together, these results provide evidence that tidemark patterning is closely tied to hedgehog ligand gradients and local ambulation load, and that primary cilia integrate cellular responses to appropriately pattern the location of the tidemark during postnatal development.

Discussion

Our study of joint-specific loss of primary cilia function provides new and probing insights into AC zonal patterning and morphogenesis during postnatal growth and development. We find that the loss of IFT88 function (and primary cilia) has little consequence on embryonic joint formation, in line with previous in vivo studies.^(26,28–31) However, we have now uncovered the that primary cilia exert major roles in AC structural and functional maturation between adolescence (3 weeks) and adulthood (8 weeks). During this postnatal period loss of primary cilia alters chondrocyte behavior and fate in both mineralized or nonmineralized cartilage, with consequent disruption of ECM component production and overall organization. However, we observed no clear evidence of cartilage matrix destruction in mutants, a result consistent with recent work by Coveney and colleagues,⁽²⁷⁾ who conditionally ablated Ift88 in adult articular cartilage and growth plate using AggrecanCreER mice. Interestingly, that study showed that Ift88 ablation at 6 or 8 weeks of age resulted in articular tissue thinning, but no such effect was noted when ablation was induced at 4 weeks.⁽²⁷⁾ We also did not observe tissue thinning in adults when Ift88 is ablated starting in embryonic joint progenitors. Similarly, we observed no obvious phenotype in adult synovium despite its cells being targeted by Gdf5Cre. Together, the data lead to the notion that conditional IFT88 deficiency in embryonic joint progenitors and early postnatal life may at least in part be compensated to maintain certain articular tissue traits such as overall thickness and synovial fibroblast function. However, other important traits cannot be compensated for, and mutants do display articular tissue defects as we clearly document in the present study.

One of our most important and surprising findings is that genetic deletion of Ift88 alters tidemark patterning and topography in a manner that mirrors alterations in hedgehog signaling. These analyses allow us to establish, for the first time, the interrelationships between spatial patterning of Ihh and cellular responsiveness in developing and mature AC. Our transcriptional analysis shows that even before the tidemark emerges, hedgehog-responding cells topographically align with the future location of the tidemark eventually seen in mature AC. This is further supported by our joint-specific loss of primary cilia model that disrupts cellular responsiveness to hedgehog signaling and alters tidemark topography. Cells that chance to avoid such loss and continue to be hedgehog-responsive—such as those in femoral condyle—remain located where the tidemark is patterned (Fig. 9, femoral condyle). The disruptions to hedgehog signaling lead to irregular zonal patterning, particularly in much of the femoral condyle where hedgehog signaling remains low throughout postnatal growth and likely contributes to ectopic Sox9-positive and Perlecan-positive cells that become mislocated in the mineralized zone (Fig. 9, non-mACs in mineralized zone). In the tibial plateau, the tidemark is relatively well patterned closer to the subchondral bone likely due to a switch from low hedgehog response to high hedgehog response in the timeframe where the tidemark is formed (Fig. 9, tibial plateau). In future work, the mechanisms linking hedgehog signaling to tidemark patterning should be tested directly. Is the tidemark repositioning dependent on a changing hedgehog response? What is the response in tissue integrity? Tidemark duplication and advancement toward the cartilage surface are highly associated with OA disease progression, but it is not understood whether this movement enhances disease progression or whether it is correlative.^(8–12) In addition, overactivation of hedgehog signaling in mice leads to early onset of OA,⁽⁷⁷⁾ possibly through inhibition of ectonucleotide pyrophosphatase/phosphodiesterase 1 (Enpp1), a suppressor of pathological calcification.⁽⁷⁸⁾ In sum, future studies delineating how the tidemark forms during postnatal development can inform why it is altered in disease, offering additional mechanistic insight into disease progression.

Fig. 9.

Summary schematic of findings. Patterning of control AC (left) and joint-specific (Gdf5Cre) loss of primary cilia (Ift88-flox) mutant AC (right). In the femoral condyle (top), the tidemark is located three to four cell layers from the surface in both controls and mutants but includes additional irregular patterning in mutants. In the tibial plateau (bottom), the tidemark is located five to six cell layers from the surface in controls and seven to eight cell layers from the surface in mutants. Alterations to mutant nonmineralized AC is shown in the boxed panels on the far right. Hh = hedgehog; mAC = mineralized articular cartilage; non-mAC, nonmineralized articular cartilage.

On a broad level, our study underscores the differential topographical character of AC in normal joints and the differential phenotype due to the loss of primary cilia during postnatal growth. Each joint must be uniquely fitted to the load-bearing and movement requirements at each anatomic location, and our study provides evidence that tidemark patterning is involved in this evolution and adaptation. Specifically, we highlight differential tidemark patterning in opposing AC in the same joint during normal tissue postnatal growth that bear unique ambulatory loads (Fig. 9, control). In the femoral condyle, ambulatory load is dispersed throughout cartilage as the entire surface slides along the underlying meniscus and tibial plateau with joint flexion and body movement. In the tibial plateau instead, ambulatory load is highly concentrated within a central location that is not covered by the meniscus. Tissue and tidemark patterning reflects such differing load-bearing requirements. In the femoral condyle, nonmineralized cartilage makes up ~30% of the tissue in control mice, with the tidemark located two to three cell layers beneath the surface. In the tibial plateau, nonmineralized cartilage makes up a significantly larger percentage of the tissue (~52%), with the tidemark located five to six cell layers beneath the surface. It is likely that the differing load-bearing topography in femur and tibia is also at play in response to the loss of primary cilia in articular cartilage and contributes to differential tissue responses. In this context, we have identified three major distinctions in the response of femoral condyle and tibial plateau to loss of primary cilia (highlighted in Fig. 9): (i) cellular responsiveness to hedgehog; (ii) cellular responsiveness to BMP and TGFβ; and (iii) expression of Prg4.

Our study shows that joint-specific loss of primary cilia leads to decreased hedgehog responsiveness in the femoral condyle and increased responsiveness in the tibial plateau in mature tissues. Most tissues follow the dogma that loss of the primary cilia leads to loss of hedgehog signaling. However, in the musculoskeletal system, Gli3 repressor plays a large role in tissue function and is also regulated by the primary cilium. Several studies have provided evidence that the loss of primary cilia, in addition to preventing Gli1 and Gli2 processing to active forms, affects the processing of Gli3 to its repressor form, resulting in abrogation of its repression on Gli1.^{(26,33,79,80)} Our data are consistent with these studies and, in particular, with a recent study in skeletal muscle suggesting that Gli1 overactivation also has a mechanical load component.⁽⁸⁰⁾ However, in the femoral condyle, hedgehog signaling in mutants is decreased compared to controls with only patchy expression where ambulation load is distributed across the tissue. Of note, the loss of ciliation in the femoral condyle is not as strong as in the tibial plateau and the remaining hedgehog response could be the result of some ciliated cells still able to respond. In future work testing hedgehog signaling more directly, it will be interesting to see if these differential loading responses of femur and tibia persist over time.

We explored TGFβ and BMP signaling in order to gain mechanistic insights into changes in ECM production in mature AC. Our data indicate that both signaling pathways remain unchanged at 3 weeks in both mutant femoral condyle and tibial plateau, consistent with no apparent changes in Col2a1 and Acan gene expression. However, at 8 weeks, BMP or TGFβ responsiveness is significantly decreased in the load-bearing region of the tibial plateau. Knowledge of the role(s) for BMP signaling in postnatal AC is limited, but two studies of conditional ablation of either Bmpr1a or Bmp2 in embryonic joints (Gdf5Cre) reported no embryonic phenotype and a progressive loss of ECM during postnatal stages, without hypertrophy or loss of Sox9.^(32,81) These findings are strikingly similar to those we report here. Hedgehog signaling is also a known negative regulator of BMP during embryonic somitogenesis by enhancing the expression of BMP antagonists Noggin and Gremlin.^(82,83) Perhaps, a similar mechanism is at play here. The loss of TGFβ responsiveness is less clear but might include negative regulation by pathways not assessed in this study or an altered adaptation to mechanical load without primary cilia. In the femoral condyle at 8 weeks, we detected no changes to BMP or TGFβ signaling in nonmineralized cartilage, although the tissue was disorganized in correlation with a disorganized tidemark. Changes to Col2a1 or Acan gene expression were also not as prominent in the femoral condyle. It is notable that tissue stiffness (by AFM nanoindentation) was equally decreased in femoral condyle and tibial plateau. Thus, despite a lack of signaling changes, the ECM is still not normal. Notable decreases in cell density in the femoral condyle could contribute to this altered ECM organization; however, we believe that irregular patterning of nonmineralized and mineralized zones also plays a significant role. The additional decrease of TGFβ and BMP signaling in mutant tibial plateau might explain the more significant loss of collagen II compared to mutant femoral condyle (p = 0.03), with more clearly visible reductions in Safranin-O and aggrecan immunostaining. To test this theory directly, future work could use fluorescence-guided AFM nanomechanical mapping⁽⁸⁴⁾ to delineate local micromechanical properties that could differentially contribute to ECM composition and properties in femoral condyle versus tibial plateau.

Prg4 (lubricin) is an important regulator of AC homeostasis⁽⁸⁵⁾ and prevents tissue and matrix degradation that would lead to OA.^(49,53) Prg4 can be regulated by multiple pathways including TGFβ,⁽⁴⁹⁾ Creb5,⁽⁸⁶⁾ Yap/Taz,⁽⁸⁷⁾ and mechanically activated Ca²⁺ signaling via Cox2.^(88,89) The primary cilia also function to transduce mechano-osmotic signals via the mechanically activated cation channels such asTrpv4,^(90–92) Polycystins1 and 2,⁽⁹³⁾ and Piezo1/2⁽⁹⁴⁾ that are present through the cell surface, but also on the primary cilium.^(95–98) The sensitivity of Prg4 expression to mechanical load may explain the differential expression of Prg4 in femoral condyle versus tibial plateau. Consistently decreased expression in the femoral condyle suggests that full transcriptional activation of Prg4 requires the function of the primary cilium where ambulation load is dispersed throughout the tissue. However, in the tibial plateau, where ambulation load is concentrated to the tissue center, Prg4 is initially unchanged at 3 weeks and then increases markedly by 8 weeks, possibly due to the presence of cation channels residing outside the primary cilia that respond to this more concentrated load. The presence of Trvp4 channels on the primary cilium could also be important for the broader irregular ECM in our Ift88 mutants due to known functions for Trpv4 in ECM remodeling (reviewed by Ji and McCullough⁽⁹⁹⁾), an active area of research in cartilage.

Although our mouse study is limited in its direct translation to human cartilage development and disease, previous work has shown that primary cilia are lengthened during human osteoarthritis.⁽²⁵⁾ Whether this accelerates disease progression or is a failed tissue response to subvert disease progression is not understood due, in large part, to a lack of understanding of how primary cilia function in articular cartilage postnatal development and growth. Better understanding of these facets of AC biology could be exploited to create more effective strategies for cartilage regeneration and repair. Hence, our study is a significant step ahead in providing important and novel insights into zonal and tidemark patterning, but also highlights that there are many remaining questions left to resolve regarding articular cartilage development.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.