Role of Ciliary Protein Intraflagellar Transport Protein 88 in the Regulation of Cartilage Thickness and Osteoarthritis Development in Mice

Mechanical and biologic cues drive cellular signaling in cartilage development, health, and disease. Primary cilia proteins, which are implicated in the transduction of biologic and physiochemical signals, control cartilage formation during skeletal development. This study was undertaken to assess the influence of the ciliary protein intraflagellar transport protein 88 (IFT88) on postnatal cartilage from mice with conditional knockout of the Ift88 gene (Ift88‐KO).


INTRODUCTION
Articular cartilage absorbs and transmits mechanical loads generated by muscle contraction and weight-bearing during physical activity. Cartilage can be broadly divided into a noncalcified zone and a calcified zone adjacent to bone. Total cartilage thickness is allometrically scaled to the size of the organism, ensuring that chondrocytes experience similar force irrespective of the mass of the individual animal (1). Articular cartilage is extremely mechanosensitive, with chondrocytes closely monitoring and remodeling the extracellular matrix in response to physiologic loading (2,3). Physiologic mechanics are critical for cartilage development and homeostasis (4), since their loss has been demonstrated to lead to cartilage thinning (atrophy) (5,6).
Pathologic, supraphysiologic mechanical loading leads to the development of osteoarthritis (OA), in which degradation of the cartilage occurs, resulting in a loss of integrity of the articular surface (6). Cellular responses to mechanical force include the release of fibroblast growth factor receptor type 2 and transforming growth factor β (TGFβ) from the matrix upon application of a mechanical load, as well as activation of a hedgehog (Hh) ligand, Indian hedgehog (7)(8)(9)(10). A number of other pathways are implicated in cellular mechanotransduction, including connexin and ion channel opening and integrin activation. It is not yet understood how chondrocytes in cartilage integrate these cues as the cartilage matures through adolescence and as adaptation occurs to prepare for lifelong loading.
As in most cells in the body, articular chondrocytes express a single immotile primary cilium (11), a microtubule-based organelle reliant on intraflagellar transport (IFT) proteins, including IFT88. Components of the Hh pathway localize to the cilium, supporting bidirectional modulation of signaling (12). Cilia have been directly linked with other growth factor-signaling pathways such as TGFβ signaling, and have also been indirectly linked with a large, growing list of signaling pathways (13), many of which are pertinent to cartilage health. The primary cilium has also been implicated as a "mechanosensory" structure (14). Findings from in vitro experiments in chondrocytes have implied that ciliary protein IFT88 is associated with both compressioninduced production of extracellular matrix proteins (15) and impaired clearance of aggrecanases, resulting in exacerbated aggrecan degradation (16).
Developmental mutations in ciliary genes, including Ift88, result in impaired embryonic patterning as a result of disrupted Hh signaling (17). Cartilage-specific deletion of Ift88 results in disrupted long bone and articular cartilage formation (18). Cartilagespecific deletion of Ift80 in mice in the first 2 weeks of postnatal life was found to result in thickening of the articular cartilage (19). However, despite its putative influence over a range of processes that regulate health and disease in cartilage, we have very limited direct knowledge regarding the postdevelopmental influence of ciliary IFT. We hypothesized that in a mouse model, ciliary protein IFT88 plays a crucial role in mediating cartilage homeostasis in adult animals.

MATERIALS AND METHODS
Animals. Ift88 fl/fl mice were obtained from The Jackson Laboratory (stock no. 022409) and were maintained as the control line. In parallel, offspring were crossed with aggrecanCre ERT2 mice, to generate aggrecanCre ERT2 ;Ift88 fl/fl mice (i.e., Ift88-knockout [Ift88-KO]), a mouse line that was originally generated at the Kennedy Institute of Rheumatology (20). The tdTomato reporter mouse line B6.Cg-Gt(ROSA)26Sor tm14(CAG-tdTomato)Hze /J was used to confirm tissue-specific Cre recombination (stock no. 007914; The Jackson Laboratory). For all experiments, apart from destabilization of the medial meniscus (DMM) and wheel exercise (in which only male mice were used), both sexes were analyzed. No effect of sex was observed.
DMM. Mice were given tamoxifen 2 weeks prior to surgery. Male Ift88 fl/fl control mice and aggrecanCre ERT2 ;Ift88 fl/fl mice underwent DMM surgery at 10 weeks of age, as previously described (21), or underwent a capsulotomy as sham surgery. Thereafter, the mice were anesthetized using a previously described method (22) and then killed at 8 or 12 weeks after DMM (at 8 weeks, n = 14 Ift88 fl/fl mice and n = 12 aggrecan-Cre ERT2 ;Ift88 fl/fl mice; at 12 weeks, n = 15 Ift88 fl/fl mice and n = 15 aggrecanCre ERT2 ;Ift88 fl/fl mice).
Histology. Knee joints from the mice were harvested into 10% neutral buffered formalin (CellPath). The joints were then decalcified (in EDTA), paraffin embedded, and cut coronally through the entire depth of the joint. Sections (4 μm in thickness) were stained with Safranin O at 80-μm intervals.
Osteoarthritis Research Society International (OARSI) scoring. Safranin O-stained joint sections were assessed using a scoring system for severity of cartilage damage, as described previously (21). OARSI scores were assessed by 2 observers (CRC and IP) who were blinded with regard to the mouse group. The summed score method was used (the sum of the 3 highest scores per section, per joint, with a minimum of 9 sections per joint).
Osteophyte quantification. Osteophyte size and maturity, as well as loss of medial tibial cartilage proteoglycan staining, were assessed using a 0-3 scale, as previously described (23).
Cartilage thickness measurements. Thickness measurements were obtained from the mean of 3 measurements from both the medial and lateral tibial plateaus (total of 6 measurements), from 3 consecutive sections in the middle of the joint (total of 18 measurements). The same method was used to measure noncalcified cartilage thickness from the same location from which the total cartilage thickness measurement was obtained. Calcified cartilage thickness was calculated by subtracting the noncalcified cartilage thickness from the total cartilage thickness. All images were independently scored twice; measurements were obtained using ImageJ software (Supplementary Figure 1A Subchondral bone measurements. Subchondral bone measurements were obtained using a previously described method (24). Briefly, using ImageJ software, 5 measurements were obtained perpendicular to the articular cartilage plateau, from the chondro-osseous junction to the top of the bone marrow or to the growth plate if there was no bone marrow. Thickness was measured in ≥3 sections of each plateau of each knee.
Micro-computed tomography (micro-CT) analysis of bone volume/total volume (BV/TV). Mouse knee joints were scanned using a SkyScan 1172 Micro-CT apparatus in 70% ethanol (10 μm/pixel). A 3-dimensional analysis was used to examine parameters including tissue and bone volume, tissue and bone surface area, trabecular thickness, separation, and volume, and trabecular pattern factor, measured using CTAn (Bruker).
Immunohistochemistry (IHC). Unstained fixed, decalcified coronal sections of the knee joints were deparaffinized, rehydrated, and quenched in 0.3M glycine and treated with proteinase K for 10 minutes. Tissue sections were treated with chondroitinase (0.1 units) for 30 minutes at 37 C, blocked in 5% goat serum and 10% bovine serum albumin in phosphate buffered saline, and permeabilized with 0.2% Triton X-100 for 15 minutes. The cartilage was then incubated overnight at 4 C with one of the following primary antibodies: anti-type X collagen in a 1:50 ratio (product no. ab58632; Abcam), anti-NITEGE in a 1:50 ratio (product no. PA1-1746; Thermo Scientific), IgG control in a 1:50 ratio, or no primary antibody. Sections were washed and incubated with Alexa Fluor 555 secondary antibodies for 30 minutes. Cartilage samples were incubated with the nuclear stain DAPI (in a 1:5,000 ratio), before being mounted in Prolong Gold.
RNA extraction. Articular cartilage was microdissected from the femoral and tibial plateaus using a scalpel (each data point from 2 animals) and was harvested directly into RNAlater (Invitrogen). Samples were transferred to RLT buffer (Qiagen) and were diced before pulverization using a PowerGen 125 Polytron instrument (Fisher Scientific) (3 times for 20 seconds each). RNA in mouse knee cartilage was isolated using an RNeasy micro kit (Qiagen).
Quantitative polymerase chain reaction (qPCR). Using RNA isolated from microdissected knee cartilage, complementary DNA was synthesized (Applied Biosystems) and realtime qPCR was performed using a ViiA 7 real-time PCR instrument on 384 custom-made TaqMan microfluidic cards (product no. 4342253; ThermoFisher). Threshold cycle values were normalized to the mean values for Gapdh and Hprt. Linear regression analyses were used to assess correlations with Ift88 expression, and subsequently raw P values and corrected P (P corr ) values (Bonferroni adjusted) were calculated using GraphPad Prism software version 9.
RNAScope analysis. Knee joints were harvested into icecold 4% paraformaldehyde and incubated in a refrigerated environment for 24 hours before being transferred into ice-cold 10%, 20%, and 30% sucrose, each for 24 hours. RNAScope probes Mm-Ift88-C1 (product no. 420211-C1) and Mm-Gli1-C2 (product no. 311001-C2) were used in combination with Opal 570 reagent pack signals (product no. FP14880001KT) and 690 reagent pack signals (product no. FP1497001KT) (all from ACD inc.), with the values normalized to those of both positive and negative controls. For further details on slide preparation and more information regarding the imaging protocols applied when using the RNAScope multiplex fluorescent reagent kit version 2 (product no. 323100), see Supplementary Methods (available on the Arthritis & Rheumatology website at http://onlinelibrary.wiley.com/ doi/10.1002/art.41894/abstract).

RESULTS
Thinner medial articular cartilage in adolescent mice following Ift88 deletion in chondrocytes. To confirm inducible tissue-specific Cre recombination, the aggrecanCre ERT2 mouse line was crossed with a tdTomato reporter mouse line, and tamoxifen was administered at 4 and 8 weeks of age. Two weeks following tamoxifen treatment, activation of Cre recombinase was observed in superficial hip articular cartilage chondrocytes (Supplementary Figure 2A, available on the Arthritis & Rheumatology website at http://onlinelibrary.wiley.com/ doi/10.1002/art.41894/abstract), as well as in knee articular cartilage chondrocytes and menisci (n = 3) ( Figure 1A).
In mice, throughout adolescence and adulthood, articular cartilage was thicker on the medial tibial plateau than on the lateral plateau (P < 0.0001) ( Figure 1B). Ift88 was deleted at 4 or 6 weeks of age, and analysis was conducted 2 weeks later ( Figure 1C). Deletion of Ift88 in mice at age 6 weeks resulted in reduced thickness of the medial and lateral knee cartilage, which was significantly different compared to that in the control group ( Figure 1D).
Thinner calcified articular cartilage resulting from Ift88 deletion. In a previous mouse study, reduced thickness of the articular cartilage was associated with acceleration of programmed hypertrophy upon postnatal activation of Hh signaling in mice (25). To assess the effects of Ift88 deletion on cartilage in maturing mice, we measured the thickness of noncalcified and calcified cartilage (  proportion of calcified articular cartilage increased from 25% at 4 weeks of age to 60% by 22 weeks of age ( Figure 2B) ;Ift88 fl/fl mice), though these reductions were only statistically significant on the medial plateau ( Figure 2C). Separating measurements by lateral and medial plateaus revealed reductions in calcified cartilage that were observable across both plateaus and time points, though these reductions were only significant on the medial plateau at 8 weeks of age ( Figure 2D).
Histologic measurements of the subchondral bone (Supplementary Figure 1B, available on the Arthritis & Rheumatology website at http://onlinelibrary.wiley.com/doi/10.1002/art. 41894/abstract) revealed that bone thickness approximately doubled between 6 and 8 weeks of age in control mice ( Figure 2E). However, deletion of Ift88 did not result in enhanced bone thickening. Additionally, there was no tartrate-resistant acid phosphatase staining (Supplementary Figure 1C) in the articular cartilage or the bone beneath, suggesting very little osteoclast activity in the calcified cartilage or subchondral bone at these time points.
Association of reductions in calcified cartilage thickness with articular cartilage atrophy following Ift88 deletion in mice at 8 weeks of age. Progressive calcification of articular cartilage continued when mice were between 8 and 22 weeks of age ( Figure 2B). Tamoxifen was administered at 8 weeks of age, and joints were collected at 10, 22, and 34 weeks of age ( Figure 3A). Deletion of Ift88 resulted in a 10-20% reduction in the thickness of tibial cartilage at every time point (Figures 3B and C). Based on the mean values across both plateaus, thinning was not cumulative over time and occurred without obvious surface fibrillation, thus representative of atrophy rather than degeneration (3,5,26,27). At 22 and 34 weeks of age, thinning of the medial plateau of Ift88-KO mice was significantly greater than The effect of Ift88 deletion on the thickness of the noncalcified cartilage was not statistically significant on either the medial or the lateral plateaus ( Figure 3E) Ift88 fl/fl mice) ( Figure 3E).
Prior evidence has shown that cartilage thinning and accelerated hypertrophy occur in association with reduced numbers of chondrocytes in articular cartilage following reactivation with Hh signaling (25). However, no significant reductions in cell number or density were observed in 10-week-old aggrecanCre ERT2 ; Ift88 fl/fl mice (Supplementary Figure 3B, Figure 3C).
Predisposition of joints to spontaneous OA following atrophy of the medial tibial cartilage. At 34 weeks of age, aggrecanCre ERT2 ;Ift88 fl/fl mice developed spontaneous cartilage damage in the medial compartment and this was found to be associated with osteophyte formation, indicating the development of OA ( Figure 4A). OARSI scoring revealed a significant increase in cartilage damage scores when mice were 34 weeks of age ( Figure 4B), but no significant increase when mice were 22 weeks of age (Supplementary Figure 4A, available on the Arthritis & Rheumatology website at http://onlinelibrary.wiley.com/doi/10.1002/art.41894/ abstract). The highest OARSI scores were found on the medial plateau ( Figure 4C). There was no difference in osteophyte scores, except in cartilage displaying severe damage ( Figure 4D Figure 4D).
Exacerbation of OA severity after surgical joint destabilization and deletion of Ift88. Mice were given tamoxifen at 8 weeks of age, and surgical DMM was performed  Figure 5F).
Lack of association between articular cartilage atrophy and markers of matrix catabolism or hypertrophy. To assess whether Ift88 deletion increased the catabolism of aggrecan, knee joints from control mice and aggrecanCre ERT2 ; Ift88 fl/fl mice were probed for protease-generated neoepitopes as described previously (22), using an antibody raised against the synthetic epitope NITEGE, and with an IgG antibody used as an isotype control (Supplementary Figure 6A Figure 6A), and no differences in NITEGE signaling were observed between the joints from control mice and joints from aggrecan-Cre ERT2 ;Ift88 fl/fl mice ( Figure 6A). Moreover, no differences in ColX expression, a marker of Hh-driven chondrocyte hypertrophy (28), were observed between Ift88-KO mice and control mice, as determined by IHC ( Figure 6B; results compared against the IgG control are shown in Supplementary Figure 6B).
Results of bulk RNA analyses implicating Tcf7l2 in cartilage atrophy. Two weeks after tamoxifen treatment, RNA was isolated from microdissected articular cartilage obtained from the knee joints of 10-week-old Ift88-KO and control mice (Supplementary Figure 6C, available on the Arthritis & Rheumatology website at http://onlinelibrary.wiley.com/doi/10.1002/ art.41894/abstract). A total of 44 candidate genes, which were all either associated with chondrocyte biology and/or implicated in primary ciliary signaling, were measured using qPCR. Due to the mosaic Cre activity ( Figure 1A) and sample pooling, we conducted a correlation analysis to assess correlations between Ift88 expression and expression of the candidate   Figure 6D), which was consistent with the IHC findings ( Figures 6A and B). Tcf7l2 (29) was the only gene found to be correlated with Ift88 expression at a level that was statistically significant after Bonferroni correction (P = 0.0006, P corr = 0.026, r 2 = 0.8811). Levels of Ctgf, Tgfbr3, Gli2, and the regulator of cartilage calcification Enpp1 (30) were positively correlated with Ift88 expression before Bonferroni correction (P = 0.002, P = 0.0107, P = 0.0037, and P = 0.009, respectively) ( Figure 6C and Supplementary Figure 6C Table 1). Effect of wheel exercise on cartilage atrophy and associated increases in Hh signaling in situ. RNAScope analysis was performed to investigate the expression of Gli1, a marker of Hh pathway activity, in the same cartilage sections in which reduced Ift88 expression was observed. A mean total of 24.12% of nuclei were positive for Gli1 in control mouse articular cartilage in comparison to a total of 45.46% of Gli1-positive nuclei in aggrecanCre ERT2 ;Ift88 fl/fl mouse articular cartilage ( Figure 6E). This increase in the percentage of Gli1-positive cells, indicative of increased Hh signaling, was statistically significant (n = 4 each; P < 0.0001) (for raw cell counts, see Supplementary Figure 6E, available on the Arthritis & Rheumatology website at http://onlinelibrary.wiley.com/doi/10.1002/art.41894/abstract). Analyzing Gli1 in terms of signal per cell also revealed higher expression of Gli1 in aggrecanCre ERT2 ;Ift88 fl/fl mice (2.88 mean dots per cell per animal) compared to control mice (1.49 mean dots per cell per animal), representing a statistically significant difference (n = 4 each; P < 0.01) ( Figure 6E). Mice were given free access to wheel exercise for 2 weeks immediately following the tamoxifen injection administered at 8 weeks of age (over the same time period that atrophy develops). In these experiments, no measurable difference in articular cartilage thickness was observed when comparing control mice to aggrecanCre ERT2 ;Ift88 fl/fl mice ( Figure 6F and Supplementary Figure 6F, available on the Arthritis & Rheumatology website at http://onlinelibrary.wiley.com/doi/10.1002/art.41894/ abstract). Wheel exercise did not affect the number of Gli1positive cells or Gli1 expression per cell in control mice, but, according to findings from both analyses, wheel exercise reduced Gli1 expression in aggrecanCre ERT2 ;Ift88 fl/fl mice, effectively restoring Hh signaling and cartilage thickness to the levels observed in control mice.

DISCUSSION
In the current study, we explored the influence of ciliary protein IFT88 in postnatal mouse articular cartilage in vivo, by depleting its expression in chondrocytes at different stages of postnatal skeletal maturation. When Ift88 was deleted in the mice at 4, 6, and 8 weeks of age, we observed rapid cartilage thinning, largely in the calcified cartilage within the medial joint compartment. Thinning was determined to be indicative of atrophy rather than degeneration, since there was no disruption of the articular surface. However, as mice aged, cartilage atrophy was associated with increased severity of spontaneous OA and DMMinduced OA.
We speculate that the Ift88-dependent effects in the thicker medial compartment of the mouse joint are mechanically driven (6) in a manner analogous to the bone "mechanostat" proposed by Frost in 1987 (31). In essence, this model ensures that chondrocytes mechanoadapt the extracellular matrix so as to experience force within a narrow window (1). Since cartilage thickness can be restored in Ift88-conditional KO mice with wheel exercise, this suggests that ciliary protein IFT88 may influence, but is not solely responsible for, cartilage mechanoadaptation in vivo. Similar modes of action have been proposed in the context of epithelial response to renal flow (14,32). Our findings contrast with the observation that in mice, deletion of Ift88 during development leads to increased cartilage thickness (17), possibly indicating a changing influence of Ift88 with skeletal maturation. Since atrophy in Ift88-conditional KO mice is largely restricted to calcified cartilage, we speculate that this represents a failure of cartilage hypertrophy during maturation, in a load-dependent manner. Cartilage thinning was not associated with increased matrix catabolism, enhanced subchondral bone thickness, osteoclast activity, or density changes (BV/TV) in the epiphysis, but we cannot exclude the possibility that calcified cartilage is transitioning to the bone in these mice. Calcification of mouse cartilage has recently been linked to Enpp1, a pyrophosphatase believed to inhibit calcification through Hh signaling (30). Findings from our molecular analyses indicated that Enpp1 positively correlates with Ift88 expression, indicating a reduction in the levels of an inhibitor of calcification in Ift88-conditional KO mice. This could result in accelerated ossification analogous to that seen in the growth plates of ciliary protein mutant mice and in articular cartilage upon postnatal activation of Hh (25).
Prior studies investigating congenital mutations (33) or constitutive deletions of Ift88 (17,18,34) demonstrated that Ift88 plays a role in mouse limb and joint development. Mouse models targeting other ciliary components, Kif3a, Bardet-Biedl syndrome proteins, and Ift80, also implicate ciliary machinery in musculoskeletal development (19,28). This influence over skeletal development is also exemplified by the human skeletal ciliopathies (35,36). The most important molecular pathway associated with ciliopathy is Hh, although other pathways have also been described (10,(37)(38)(39)(40)(41). Hh signaling largely switches off in adulthood (42) but is reactivated in OA (8). We investigated the molecular basis of Ift88-dependent cartilage atrophy by identifying the correlation between Ift88 expression (reflecting efficiency of deletion) and 44 molecules previously demonstrated to be associated with ciliary signaling and cartilage biology. In addition, we explored Hh signaling by directly visualizing Gli1 expression in murine cartilage using RNAScope. The gene that correlated most strongly with ciliary signaling was transcription factor Tcf7l2, previously shown to influence and interact with Hh and β-catenin signaling pathways in cartilage (37). Other genes whose levels correlated with Ift88 included Gli2, Ctgf, and Enpp1, although these correlations were only statistically significant before Bonferroni correction.
While a gene expression analysis of microdissected mouse cartilage did not show a correlation between Ift88 and classic Hh pathway molecules (Gli1, Ptch1), an individual cell analysis using RNAScope revealed increased levels of Gli1 expression in aggrecanCre ERT2 ;Ift88 fl/fl mouse chondrocytes, suggesting a reciprocal relationship between Hh signaling and Ift88. Therefore, we propose that loss of Ift88 disrupts ciliary-mediated repression of Hh signaling, resulting in net increases in Gli1 expression. This is consistent with findings from previous studies in constitutive Ift88 deletion (18) and endochondromas (43). In this model, our observation that cartilage atrophy was rescued and basal Gli1 expression levels were normalized in aggrecanCre ERT2 ;Ift88 fl/fl mice following wheel exercise provides critical evidence of a link with mechanical loading. Our data imply that in postnatal mouse articular cartilage, ciliary protein IFT88 safeguards the progressive mechanoadaptation of adolescent mouse cartilage, supporting the creation and maturation of fit-for-purpose adult mouse articular cartilage by ensuring appropriate levels of Hh signaling.
As previously described (44), Ift88 was deleted in mouse chondrocytes using induction of Cre recombinase expression on the aggrecan promotor, which ensured sufficient expression in mice from adolescence through adulthood. Despite the use of this method, we observed only a 40% reduction in Ift88-positive chondrocytes in the tibial articular cartilage of aggrecanCre ERT2 ;Ift88 fl/fl mice. Observations in the tdTomato reporter line of mice indicated that Ift88 deletion occurred in only a small proportion of chondrocytes, but this was not exclusive to any knee compartment, and therefore it is unlikely that this finding could be attributed to differences between the medial and lateral sides of the knee joint. We also recognize that, due to the challenges associated with imaging cilia in cartilage, we have not yet been able to address the question of whether primary cilia would be negatively impacted as a result of Ift88 deletion. Thus, we cannot conclude that molecular changes are a direct consequence of the loss of cilia. To date, we have not yet conducted experiments inducing Ift88 deletion in mice older than age 8 weeks to evaluate whether this gene is as influential later in adulthood, or assessed how behavior of the aggrecanCre ERT2 ; Ift88 fl/fl mice could be altered.
In summary, these data demonstrate that IFT88 is highly influential in adolescent mouse and adult mouse articular cartilage as a positive regulator of cartilage thickness, guiding cartilage calcification during maturation and safeguarding physiologic Hh signaling in adult mouse cartilage in response to mechanical cues.