CRELD2 Is a Novel LRP1 Chaperone That Regulates Noncanonical WNT Signaling in Skeletal Development

Cysteine‐rich with epidermal growth factor (EGF)‐like domains 2 (CRELD2) is an endoplasmic reticulum (ER)‐resident chaperone highly activated under ER stress in conditions such as chondrodysplasias; however, its role in healthy skeletal development is unknown. We show for the first time that cartilage‐specific deletion of Creld2 results in disrupted endochondral ossification and short limbed dwarfism, whereas deletion of Creld2 in bone results in osteopenia, with a low bone density and altered trabecular architecture. Our study provides the first evidence that CRELD2 promotes the differentiation and maturation of skeletal cells by modulating noncanonical WNT4 signaling regulated by p38 MAPK. Furthermore, we show that CRELD2 is a novel chaperone for the receptor low‐density lipoprotein receptor‐related protein 1 (LRP1), promoting its transport to the cell surface, and that LRP1 directly regulates WNT4 expression in chondrocytes through TGF‐β1 signaling. Therefore, our data provide a novel link between an ER‐resident chaperone and the essential WNT signaling pathways active during skeletal differentiation that could be applicable in other WNT‐responsive tissues. © 2020 American Society for Bone and Mineral Research. © 2020 The Authors. Journal of Bone and Mineral Research published by American Society for Bone and Mineral Research..


Introduction
T he endoplasmic reticulum (ER) is the largest organelle within the cell and predominantly functions as the site of protein synthesis and folding. Despite a wide range of ER-resident folding enzymes and chaperones, proteins can misfold and accumulate within the ER lumen, for instance due to high secretory load or the presence of pathological mutations. This triggers an evolutionarily conserved process called the unfolded protein response (UPR).
The UPR is initiated by the ER-resident chaperone BiP dissociating from its three transmembrane receptors, protein kinase RNAlike ER kinase (PERK), inositol-requiring enzyme 1 (IRE-1), and activating transcription factor 6 (ATF-6), and binding to exposed hydrophobic residues on the unfolded protein. (1) The UPR initially acts with the aim of decreasing protein translation, increasing refolding capacity, and degrading the misfolded protein; however, if the stress is not returned to physiological levels, the UPR can trigger cell death by inducing C/EBP homologous protein (CHOP)dependent/−independent signaling cascades. (2,3) Endochondral ossification is the process responsible for the development and postnatal growth of long bones and requires a cartilage precursor formed by chondrocytes. The cartilage growth plate, organized into three distinct chondrocyte differentiation zones-the resting, proliferative, and hypertrophic zones-drives long bone growth through a tightly regulated process involving proliferation, hypertrophy, mineralization, and apoptosis. Bone then forms on the cartilage scaffold and is continually remodeled throughout life through the combined actions of bone-forming osteoblasts and bone-resorbing osteoclasts. Interestingly, the UPR has also been implicated in the differentiation of these highly secretory skeletal cells. For example, ATF4 is a transcription factor operating directly downstream of PERK; Atf4-null mice (4) have a skeletal dysplasia phenotype due to reduced chondrocyte differentiation and Perk-null mice (5) exhibit osteopenia due to impaired osteoblast maturation and activity.
The CRELD (cysteine rich with epidermal growth factor [EGF]like domains) family of proteins comprises CRELD1 and CRELD2 (6) and these are ER-localized (7) multidomain proteins composed of EGF-like and calcium-binding EGF-like repeats and a unique and highly conserved tryptophan-aspartic acid (WE) (8) domain. During embryogenesis, CRELD1 is expressed in various soft tissues and has been shown to play a crucial role in heart development with CRELD1 mutations associated with atrioventricular septal defects. (9) CRELD2 is expressed in the developing skeleton as well as several soft tissues and to date five human isoforms, (8) comprising varying numbers of EGF-like domains, have been identified. In contrast, only one CRELD2 isoform, homologous to the CRELD2α human isoform, has been identified in mice. The precise function of CRELD2 is largely unknown; however, CRELD2β has been shown to bind to and be involved in the folding of the nicotinic acetylcholine receptor α4 and β2 subunits. (10) CRELD2 was first identified as an ER stress-inducible gene after treatment of neuro2α cells with thapsigargin (7) and importantly, the promoter of CRELD2 contains an AFT6 response element. (7) Unlike CRELD1, which is anchored by two C-terminal type II transmembrane domains, (6) CRELD2 has been shown to be secreted during ER-stress (11,12) suggesting a functional divergence between the two family members. The four C-terminal amino acids (REDL), comprising an imperfect C-terminal KDEL ER-retention sequence, play an important role in the secretion of CRELD2 that is further enhanced by BiP (11) and MANF (13) ; however, its exact role outside of the cell is currently unknown. (7) Several studies have implicated CRELD2 in the pathobiology of ER-stress-mediated conditions such as kidney disease (14) and aortic aneurysm in Marfan syndrome. (15) Creld2 is also upregulated and secreted in mouse models of multiple epiphyseal dysplasia (16) and metaphyseal chondrodysplasia type Schmid, (17) skeletal dysplasias characterized by the retention of mutant proteins in the ER of chondrocytes that results in prolonged ER stress. Substrate-trapping experiments have shown that CRELD2 acts as a putative protein disulphide isomerase (PDI) (12) for substrates such as matrilin-3, laminin 5 β3, type VI collagen, and thrombospondin 1. Moreover, CRELD2 is upregulated in BMP9 stimulated osteogenic differentiation of mesenchymal stem cells (MSCs) through SMAD1/6/5 binding directly to its promoter, suggesting a novel role for CRELD2 in bone development. (18) Despite these indications, the role of CRELD2 in skeletal development and homeostasis is not clearly understood. We therefore generated and studied in depth cartilage-specific and bonespecific Creld2 knockout mouse models to investigate the role of CRELD2 in skeletal development and homeostasis.

Materials and Methods
Generation of Creld2 conditional knockout mice A Creld2 gene trap (Creld2 tm1b(EUCOMM)Hmgu ) targeting vector containing a large selection cassette was obtained from the European Conditional Mouse Mutagenesis Programme (EUCOMM), electroporated into C57BL6 blastocysts, and used to generate chimeras that were bred onto the C57BL6 background (Supplemental Fig. S1A). The introduction of this large targeting vector did not abolish CRELD2 expression as identified by Western blotting of liver tissue (Supplemental Fig. S1B). Therefore, mice heterozygous for the targeted allele were crossed with Actin-flp mice generating offspring with the conditional Creld2 allele, which was identified by genomic DNA PCR using the primers 5 0 -GGT GGA CAC CTG TCA ATG GT-3 0 and 5'-CAG CAT GAG AAA GGG AGC TTA-3 0 (Supplemental Fig. S1C). These mice were then bred with mice expressing the Cre transgene under type II collagen (Col2a1) promoter to generate Creld2 cartilage-specific knockout mice (Creld2 CartΔex3-5 ) and under osteocalcin (Bglap) promoter to generate bone-specific (Creld2 BoneΔex3-5 ) knockout mice by Cre recombination, resulting in the deletion of exons 3-5 of Creld2. Mice with the conditional Creld2 allele were also crossed with global deleter Cre mice, and Western blotting of liver tissue was used to confirm that CRELD2 expression was abolished by Cre recombination (Supplemental Fig. S1D). The presence of the Cre transgene in the conditional lines was identified by genomic DNA PCR using the primers 5'-TGC CAC CAG CCA GCT ATC AAC T-3 0 and 5'-AGC CAC CAG CTT GCA TGA TCT C-3 0 (Supplemental Fig. S1E). RT-PCR was performed to confirm the deletion of Creld2 in chondrocytes and osteoblasts (Supplemental Fig. S1F). All experiments were carried out in compliance with the Animals Scientific Procedures Act 1986 according to Directive 2010/63/EU of the European Parliament, under project license PPL60/4525. X-ray and micro-computed tomography (μCT) X-ray images were obtained using the MX-20 Cabinet X-ray System (Faxitron, Tucson, AZ, USA). Bone measurements were taken at 3, 6, and 9 weeks of age using the image processing program FIJI (ImageJ). (19) To analyze bone structure, undecalcified bone samples fixed in 4% paraformaldehyde (PFA) were subjected to μCT analysis using the SkyScan 1174v2 (Bruker, Kontich, Belgium) μCT device. Multiple 2D image projections were acquired at 50 kV and 800 mA with a resolution of 6.74 nm. Image reconstruction was carried out using the SkyScan NRecon program. Common standard 3D bone morphometric parameters were determined for trabecular and cortical bone using the CTAn program. (20) Trabecular bone architecture was calculated from 350 serial slides of 6 microns thick, 650 μm from the growth plate. Cortical bone architecture was calculated from 100 serial slides of 6 microns thick, 3.5 mm from the growth plate.

Histological and immunohistochemical staining of cartilage
Cartilage samples were fixed for histology and immunohistochemistry and processed as outlined previously. (21) Formalinfixed sections were stained with hematoxylin and eosin (H&E) and toluidine blue. For immunohistochemistry (IHC), antigen retrieval was performed by incubation in 0.2% bovine hyaluronidase, 0.2% Triton-X, and 20 μg/mL proteinase K. Sections were blocked in serum for 1 hour, incubated in primary antibody overnight at 4 C, and probed with the appropriate secondary antibody for 1 hour at room temperature. Sections were mounted in Fluoroshield mounting media with DAPI (Abcam, Cambridge, UK). The following primary antibodies were used: acetylatedtubulin (ab24610), collagen I (ab34710), collagen II (ab34712) (all from Abcam), CRELD2 (sc86110, Santa Cruz Biotechnology, Dallas, TX, USA), matrilin-3 (AF3357, R&D Systems, Minneapolis, MN, USA), aggrecan (Professor Tim Hardingham, The University of Manchester), and collagen X (Professor Karl Kadler, The University of Manchester). Negative controls are shown in Supplemental Fig. S2F. To analyze chondrocyte proliferation, 3-week-old mice were injected subcutaneously with BrdU labeling reagent according to weight (0.1 mg per 10 g). Two hours after injection, mice were euthanized and limbs harvested. IHC was then performed to detect BrdU-positive cells using an anti-BrdU antibody (ab6326, Abcam). BrdU-positive cells in the proliferative zone of the growth plate were counted using the watershed algorithm in the FIJI program and expressed as a percentage of the total number of osteoblasts or the total number of chondrocytes within the proliferative zone. Analysis was performed on three matched slides per mouse from three mice per genotype.

TUNEL analysis
Cell death was detected by fragmented DNA and quantified using the DeadEND Fluorometric TUNEL System (Promega, Madison, WI, USA) according to manufacturer's instructions. Permeabilization was performed by boiling samples in citric buffer pH 6 as Proteinase K treatment can generate false positives. (22) TUNEL-positive cells were counted using the watershed algorithm in FIJI and were expressed as a percentage of the total number of osteoblasts or the total number of chondrocytes within each zone of the growth plate. Analysis was performed on three matched slides per mouse from three mice per genotype.

Transmission electron microscopy (TEM)
Tibial cartilage from 1-week-old pups was dissected and fixed in 2% glutaraldehyde in 0.1 M sodium cacodylate buffer at 4 C. Samples were then processed, embedded, and sectioned by the Electron Microscopy Research Services Unit at Newcastle University. The protocol was as follows. Cartilage samples were then washed three times in 0.1 M sodium cacodylate and fixed a second time in 2% (w/v) osmium tetroxide in distilled water. Tissues were then rinsed twice in water, immersed in 2% uranyl acetate, and rinsed twice again in water before being washed with 25%, 50%, 75% acetone. Samples were then incubated in two changes of 100% acetone and embedded using the TAAB epoxy resin kit according to manufacturer's specification. To allow for resin to fully penetrate the tissues, samples were incubated in 100% resin for 2 days. Tibial cartilage samples were then placed in a mold and resin was polymerized overnight at 60 C. After embedding, samples were sectioned to ultrathin sections of 70 nm using a diamond knife on a Leica (Buffalo Grove, IL, USA) EM UC7 ultramicrotome, stretched with chloroform and mounted onto Pioloform-filmed copper grids. Prepared grids were then stained with 2% aqueous uranyl acetate and lead citrate solution. Prepared grids were imaged using the Phillips CM100 Transmission Electron Microscope.

Bone histomorphometry
Nine-week-old tibias were dissected and fixed in 4% formaldehyde for 24 hours and embedded in methylmethacrylate without decalcification. Sections (6 μm) were stained with Von Kossa (to analyze calcification), 1% toluidine blue (to analyze osteoblast number (23) ), and stained for TRAcP (to analyze osteoclast number and surface (23) ). To study the dynamic parameters of bone, mice were injected in the intraperitoneal cavity 1 week and 1 day before euthanization with 30 mg/kg alizarin complexone and 20 mg/kg calcein in a 2% sodium bicarbonate pH 7.4 solution, respectively. Trabecular bone histomorphometric measurements were performed on 6-μm-thick sections from fluorescent images acquired using a Zeiss (Thornwood, NY, USA) Axio imager 2 microscope. The analysis of the bone formation rate (BFR) (23) was carried out on matched sections from 5 mice per genotype and was performed on the trabecular bone in the proximal tibia, 100 μm below the growth plate.

Primary cell extraction and cell culture
MSCs were isolated from bone marrow aspirates, cultured and phenotyped as previously described. (24,25) Primary chondrocytes were isolated after digestion of 5-day-old costal and tibial cartilage with 2.5 mg/mL collagenase IA as outlined previously. (21) Primary chondrocytes were cultured for up to 7 days in complete DMEM/F-12 (supplemented with 10% FBS, 5% non-essential amino acids, 1 U/mL penicillin, 1 μg/mL streptomycin, and 50 μg/mL L-ascorbate-2-phosphate). Primary osteoblasts were isolated from 1-week-old neonatal calvarias by sequential enzymatic digestion of the bone matrix. (26) Cells released in the second and third digestion were pelleted by centrifugation and cultured in complete alpha MEM (supplemented with 10% FBS, 1 U/mL penicillin, and 1 μg/mL streptomycin). To generate mature osteoblasts, primary osteoblasts were cultured for 21 days in osteogenic media (complete alpha MEM containing 50 μg/mL L-ascorbate-2-phosphate, 10 mM β-glycerophosphate, and 0.1 μM dexamethasone). Osteoclast precursors were isolated from the long bones of 6-to 9-day-old mice. (27) The bone marrow mononuclear cell fraction was obtained by centrifugation of Histopaque-1077. Cells were plated in 12-well plates and cultured for 48 hours in complete alpha MEM containing 30 ng/mL M-CSF (R&D Systems) generating M-CSF-dependent macrophages. Osteoclastogenesis was performed by culturing macrophages for 1 week in complete alpha MEM supplemented with 30 ng/mL M-CSF and 100 ng/mL RANKL (R&D Systems). Osteoclastogenesis media was refreshed every 2 days. To analyze the effect of wild-type and Creld2 knockout primary osteoblast conditioned media, osteoclastogenesis was performed in conditioned media. The SW1353 chondrocyte-like and the SAOS2 osteoblast-like cell lines were maintained in complete DMEM/F-12 and complete alpha MEM, respectively. For WNT4 stimulation studies, recombinant WNT4 (R&D Systems) was added at a concentration of 100 ng/mL. For TGFβ1 stimulation studies, recombinant TGFβ1 (R&D Systems) was added at a concentration of 5 ng/mL.

Protein extraction and immunoblotting
Protein lysates were extracted as outlined previously. (21) Protein concentrations were obtained using the Pierce (Rockford, IL, USA) BCA kit. Samples were denatured by boiling and reduced by DTT where appropriate. For analysis of plasma membrane protein expression, plasma membranes were extracted using the plasma membrane protein extraction kit from Abcam. Western blotting was performed as outlined previously. (21) The following primary antibodies were used: p38 (ab170099), Phospho p38 (T180 + Y182) (ab195049), β-catenin (ab32572), osteoprotegerin (ab183910), LRP1 β-chain (ab92544), alpha 1 sodium potassium ATPase antibody (ab7671) (all from Abcam), GAPDH (Millipore, Burlington, MA, USA), and CRELD2 and RANKL (sc377079, Santa Cruz). Membranes were imaged on the LI-COR Odyssey CLx Imaging System. Densitometric analysis of the fluorescent protein bands was presented relative to protein loading or GAPDH levels. Analysis was performed on three samples per condition.

Enzyme-linked immunosorbent assay (ELISA)
To determine the concentration of secreted RANKL and OPG in osteoblast conditioned media, media was concentrated using the Vivaspin 15 MWCO centrifugal concentrators for 30 minutes at 15,000g. The mouse RANKL and mouse Osteoprotegerin Quantikine ELISA kits from R&D systems were then used to determine protein concentration in the concentrated conditioned media as outlined in the manufacturer's instructions.

RNA extraction and sequencing
RNA was extracted from whole knee joint cartilage of 5-day-old mice using the Promega Reliaprep RNA Tissue Miniprep System as outlined previously. (21) RNA was also extracted from cells using the Promega Reliaprep RNA Cell Miniprep System according to the manufacturer's protocol. The extracted RNA from cartilage knee joints and primary osteoblasts were sent to GATC Biotech for RNAseq analysis. Each of the three samples contained two individually extracted RNA samples, pooled into one sample at a concentration of 1 μg RNA with a RIN >8. The RNA was sequenced and analyzed by GATC Biotech. Briefly, Bowtie was used to align the RNA sequencing reads to the mouse reference genome. TopHat, Cufflinks, and Cuffmerge were used to identify exon-exon splice junctions, and identify, quantify, and annotate the processed RNA-sequencing alignment assembly. Cuffdiff was then used to determine differential expression levels of wild-type and knockout merged transcripts. Several RNAsequencing hits were confirmed by quantitative-PCR (qPCR).
Quantitative-PCR (qPCR) cDNA was synthesized from RNA using the GoScript Reverse Transcription System according to the manufacturer's protocol. RNA was removed after the incubation with 1 U RNase H for 20 minutes at 37 C. qPCR was performed using the Power SYBR Green method as outlined in the manufacturer's instructions on a Bio-Rad (Hercules, CA, USA) DNA engine Peltier Thermal Cycler. For primer sequences, see Supplemental Table S7. Samples were analyzed in duplicate on each plate and the level of cDNA was normalized to the level of 18S. qPCR results presented in Fig. 4 were obtained by TaqMan real-time PCR as described previously. (25) Sequences for TaqMan qPCR primers and corresponding Universal Probe Library (UPL) probes are outlined in Supplemental Table S8.

siRNA gene silencing
For siRNA transfection, 100 nM SMARTpool ON-TARGETplus siRNA of four specific siRNA duplexes were used to target CRELD2 (L-010856, Horizon, Dharmacon, Lafayette, CO, USA) and LRP1 (L-004721, Horizon, Dharmacon). This was transfected into cells using the Dharmafect transfection reagent 1 (Horizon, Dharmacon). Reduction of gene-specific mRNA levels was calculated by comparing levels on cells transfected with 100 nM siCONTROL non-targetting siRNA (D-001206-14). For chondrogenesis, MSCs were transfected once before the induction of chondrogenesis. For osteogenesis experiments, MSCs in monolayer were transfected every week with siRNA for the duration of the experiment.

MSC differentiation and validation
For chondrogenesis, MSCs were cultured as a pellet in chondrogenic medium (DMEM supplemented with 1 U/mL penicillin, 1 μg/mL streptomycin, 100 μg/mL sodium pyruvate, 100 nM dexamethasone, 50 μg/mL L-ascorbate-2-phosphate, 1X ITS-1 premix, 10 ng/mL TGFβ3). For histology, either 500,000 (large) or 100,000 (mini) MSCs were seeded into a pellet. For gene expression analysis, 150,000 MSCs were seeded into a pellet. Chondrogenic media was replaced every 2 to 3 days for 14 days. For osteogenesis, MSCs were seeded in monolayer and induced to differentiate in osteogenic media as outlined above. Osteogenic media was replaced every 2 to 3 days for 21 days. For WNT4 recovery studies, recombinant WNT4 (R&D Systems) was added at a concentration of 100 ng/mL and was added to the culture from day 4 (chondrogenesis) and day 7 (osteogenesis).

Analysis of osteoblast differentiation and activity in vitro
Alizarin Red S staining was performed on mature osteoblast monolayers to analyze mineralization. Briefly, monolayers were fixed in 4% PFA and stained with 40 mM Alizarin Red S for 30 minutes. Staining intensity was calculated relative to Alizarin Red S standards at a wavelength of 405 nm. Alkaline phosphatase activity was determined on mature osteoblast protein lysates using the alkaline phosphatase colorimetric assay kit from Abcam. Analysis was performed on three samples per condition.

Co-immunoprecipitation and mass spectrometry
A previously generated V5-tagged wild-type CRELD2 cDNA construct in pcDNA 3.1 (+) (12) was used to overexpress CRELD2 in the chondrocyte-like SW1353 and osteoblast-like SAOS2 cell lines. To assess the binding partners of CRELD2, protein-protein interactions were cross-linked by incubation in 1.0 mM DSP for 10 minutes at room temperature. The reaction was stopped upon the addition of Tris-HCl pH 7.5 added to a final concentration of 20 mM for 15 minutes at room temperature. Cellular proteins were harvested and clarified by centrifugation at 10,000g. Lysates were precleared with 5% (v/v) protein A Sepharose for 30 minutes at 4 C with gentle rotation. Protein lysate (500 μg) was added to the anti-V5 agarose beads and incubated overnight at 4 C under gentle rotation. CRELD2 interacting proteins were identified by LC/MS-MS performed by the Mass Spectrometry Facility at The University of Manchester.

Immunocytochemistry
To assess the intracellular localization of CRELD2, primary chondrocytes were fixed in 10% formalin for 10 minutes at room temperature. Plasma membranes were labeled with wheat germ agglutinin (WGA) AlexaFluor 594 conjugate. Cells were then permeabilized for 5 minutes in 0.5% Triton-X before immunolabeling. The following primary antibodies were used: CRELD2, GRP94 (sc1794, Santa Cruz), GLG1 (AF7879, Biotechne), and EEA1 (AF8047, R&D Systems). For labeling of membrane LRP1, living cells were incubated for 2 hours at 37 C in a primary antibody raised against LRP-1 clone 8G1 that detects the extracellular LRP1 α-chain (Abcam). The cells were washed three times in PBS and incubated in an anti-mouse secondary antibody labeled with AlexaFluor 488 for 20 minutes at 13 C. Cells were then fixed in 10% formalin for 10 minutes at room temperature. Immunocytochemistry slides were mounted in Fluoroshield mounting media with DAPI and images were acquired at ×630 magnification using a Nikon (Tokyo, Japan) A1R confocal microscope.

Statistics
Results are expressed as the mean AE the standard deviation (SD) of a minimum of three individual experiments/mice. Statistical significance was analyzed by Student's unpaired t test, and statistical significance was p < 0.05.

Results
The ablation of Creld2 impairs skeletal development To study the role of CRELD2 specifically in cartilage and bone development, we generated Creld2 conditional knockout mouse models (Supplemental Fig. S1).
Interestingly, cartilage-specific Creld2 knockout mice (Creld2-CartΔEx3-5 ) displayed a chondrodysplasia-like phenotype with disproportionate short stature due to impaired endochondral ossification (Fig. 1A). For example, the long bones of Creld2-CartΔEx3-5 mice were significantly shorter than controls and they also displayed reduced bone density with disrupted trabecular bone architecture (Fig. 1B). In contrast, intramembranous ossification of the cranial bones, measured by the inner canthal distance (ICD), was unaffected. Hematoxylin and eosin (H&E) staining of cartilage growth plate sections from 3-week-old Creld2 CartΔEx3-5 mice revealed a disrupted growth plate with numerous areas of hypocellularity and an overall reduction in the number of chondrocytes within individual growth plate zones (Fig. 1C).
Disruptions to primary cilia length and orientation result in growth plate defects and lead to cell misalignment (28) and abnormal cell morphology. It has been suggested that the columnar organization of chondrocytes within the proliferative zone may be controlled by primary cilia that normally align in an axis parallel to bone growth. (29) Interestingly, the proliferative zone of Creld2 CartΔEx3-5 growth plates lacked the typical columnar organization observed in controls, and primary cilia in the mutant mice were no longer aligned in an axis parallel to bone growth (29) (Fig. 1C and Supplemental Fig. S2B). In the hypertrophic zone, matrix mineralization creates a hypoxic environment depleted of nutrients and results in chondrocyte death. Whereas the late hypertrophic zone of growth plates in control mice were characterized by cellular debris, in Creld2 CartΔEx3-5 mice, chondrocytes were present within the newly ossified bone (Fig. 1C) and displayed an abnormal morphology that was more indicative of elliptical-looking proliferative chondrocytes (Supplemental Fig. S2A). Additionally, the height of the zone of type X collagen expression (a marker of chondrocyte hypertrophy) was significantly reduced in the Creld2-null growth plates, suggesting a decrease in numbers of mature hypertrophic chondrocytes (Fig. 1D). In contrast, the expression of other cartilage extracellular matrix proteins was unaffected (Supplemental Fig. S2C). Consistent with changes observed in other mouse models of human growth plate dysplasias, (16,30,31) moreover, the deletion of Creld2 in chondrocytes significantly reduced cell proliferation and increased apoptosis (Fig. 1E). Thus, we conclude that Creld2 ablation in chondrocytes impairs endochondral ossification by disrupting chondrocyte maturation and survival within the growth plate.
The ablation of Creld2 impairs bone homeostasis Interestingly, the bone-specific Creld2 knockout mice (Creld2 BoneΔEx3-5 ) demonstrated growth retardation ( Fig. 2A) and showed an osteopenic phenotype characterized by significantly reduced bone mass with altered trabecular and cortical architecture (Fig. 2B). Toluidine blue staining revealed significantly less mature bone-forming osteoblasts and osteocytes in Creld2 BoneΔEx3-5 mice compared with controls (Fig. 2C). Furthermore, deletion of Creld2 in mature osteoblasts resulted in increased osteoblast apoptosis in vitro (Fig. 2D) and a reduction in the rate of bone formation as demonstrated by dynamic histomorphometry (Fig. 2E). In addition, there was also a significant increase in the osteoclast number and the erosion surface in Creld2 BoneΔEx3-5 bones as identified by staining for the osteoclast marker tartrate-resistant acid phosphatase (TRAcP) (Supplemental Fig. S3A).
The cross-talk between osteoblasts and osteoclasts is central to maintaining bone homeostasis. Therefore, we next measured the expression of the osteoclastogenic cytokine receptor activator of nuclear factor kappa-Β ligand (RANKL) and its decoy receptor osteoprotegerin (OPG) that inhibits osteoclastogenesis. The RANKL/OPG axis was disrupted at both the RNA and protein level in osteoblasts from Creld2 BoneΔEx3-5 mice. RANKL expression at both the gene (Tnfsf11) and protein level was upregulated in Creld2 BoneΔEx3-5 osteoblasts; however, OPG expression at the gene (Tnfrsf11b) and protein level was unchanged (Supplemental Fig. S3B). Additionally, the levels of secreted RANKL were significantly increased in media collected from Creld2 BoneΔEx3-5 osteoblasts in vitro, whereas the secretion of OPG was increased (Supplemental Fig. S3C). Overall these data suggested that the upregulation of RANKL expression could result in the significant increase in osteoclast number in Creld2 BoneΔEx3-5 mice. To verify whether the impaired RANKL/ OPG ratio in Creld2 BoneΔEx3-5 osteoblasts stimulated osteoclastogenesis, we differentiated pre-osteoclasts in vitro in conditioned media from the osteoblasts of Creld2 BoneΔEx3-5 and control mice. The levels of several osteoclast markers (Acp5, Calcr, and Ctsk) were significantly upregulated in primary osteoclasts differentiated in Creld2 BoneΔEx3-5 conditioned media, indicating that disrupted RANKL/OPG axis promotes osteoclast formation (Supplemental Fig. S3D). Thus, we conclude that CRELD2 deficiency in mature osteoblasts impairs bone homeostasis by impairing osteoblast maturation, activity, and survival and by disrupting osteoblast/osteoclast cross-talk.
The deletion of Creld2 disrupts WNT signaling To further assess the effect of CRELD2 deletion on chondrogenesis and osteogenesis, we analyzed the global gene expression profiles of both Creld2 knockout chondrocytes and mature osteoblasts by RNA-sequencing. After the ablation of Creld2, 220 genes were differentially expressed in chondrocytes (Fig.  3A) compared with 1021 in osteoblasts (Fig. 3B). The enriched Gene Ontology (GO) term clusters for the differentially expressed genes in Creld2 CartΔEx3-5 chondrocytes (Supplemental Table  Table S1) and Creld2 BoneΔEx3-5 mature osteoblasts (Supplemental Table S2) comprised genes with ubiquitous cellular function and genes related to skeletal development and ossification. Of particular note, genes involved in cell cycle arrest, apoptosis, and embryonic skeletal system development were differentially expressed in Creld2 CartΔEx3-5 chondrocytes. The expression of several key genes was verified by qPCR (Fig. 3C). For example, Mmp13 encoding a matrix metalloproteinase and a marker of chondrocyte hypertrophy (32) was significantly downregulated in Creld2 CartΔEx3-5 chondrocytes. Additionally, the expression of Cyr61, encoding Cysteine-rich protein 61 that promotes chondrogenesis in vitro, (33) was also downregulated in Creld2 CartΔEx3-5 mice. Similar to the GO clusters from Creld2 CartΔEx3-5 chondrocytes, differentially expressed genes in Creld2 BoneΔEx3-5 mature osteoblasts clustered into genes    involved in the cell cycle, apoptosis, and ossification. qPCR was used to verify the expression of four key genes differentially expressed in Creld2 knockout osteoblasts (Fig. 3D). The expression of the immature osteoblast marker Spp1 was upregulated, whereas Bglap and Sost, both markers of mature osteoblasts and osteocytes, were downregulated. Overall these data further confirm that the deletion of Creld2 in chondrocytes and mature osteoblasts results in impaired differentiation and maturation.

CRELD2 modulates WNT signaling during chondrogenic and osteogenic differentiation
We further examined the role of CRELD2 in chondrocyte and osteoblast differentiation by silencing CRELD2 expression using siRNA during in vitro chondrogenic and osteogenic differentiation of human mesenchymal stem cells (hMSCs). Successful knockdown of CRELD2 was confirmed by Western blotting before MSC differentiation (Fig. 4A). CRELD2 knockdown (siCRELD2) resulted in smaller cartilage pellets compared with control (siCON) pellets, indicating that CRELD2 plays a prominent role in regulating early chondrogenesis (Fig. 4B, C). siCRELD2 pellets showed a disrupted ECM composition characterized by a reduction in toluidine blue (proteoglycan stain) and type II collagen staining (Fig. 4C). Additionally, expression levels of several chondrogenic markers (SOX9, IHH, COL2A1, ACAN, and COL10A1) were significantly reduced in siCRELD2 chondrocytes, indicating chondrocyte differentiation was impaired after the knockdown of CRELD2 (Fig. 4D).
Osteogenic differentiation of siCRELD2 hMSCs was also disrupted, as characterized by a significant reduction in matrix mineralization (Fig. 4E) and alkaline phosphatase (ALP) activity (Fig. 4F). Several well-established osteoblast marker genes were differentially expressed after the knockdown of CRELD2 (Fig. 4G). For example, COL1A1 and ALPL involved in bone matrix formation and mineralization; SPP1, BGLAP, and SOST, markers of immature and mature osteoblasts and osteocytes, respectively, were significantly reduced in siCRELD2 osteoblasts, indicating the knockdown of CRELD2 impaired osteoblast differentiation and osteogenesis.
It was of particular significance that both siCRELD2 hMSCderived chondrocytes and osteoblasts exhibited a significant reduction in WNT4 expression (Fig. 4D, G), consistent with its reduced expression levels in primary chondrocytes and osteoblasts from the Creld2 knockout mice (Fig. 3). These data suggest that CRELD2 plays an important role in hMSC chondrogenic and osteogenic differentiation potentially via modulating WNT4 signaling.

CRELD2 promotes the membrane expression of LRP1
Immunocytochemistry was used to analyze the subcellular localization of CRELD2 in primary chondrocytes. CRELD2 co-localized with the ER marker heat shock protein 90 kDa beta member 1 (also known as GRP94), the Golgi marker Golgi apparatus protein 1 (GLG1), and the exosomal marker Cluster of differentiation 81 (CD81), thereby suggesting a potential role for this chaperone in trafficking proteins from the ER through the secretory pathway (Supplemental Fig. S4).
To further elucidate the role of CRELD2 in chondrocyte and osteoblast biology, putative CRELD2 binding partners were identified by co-immunoprecipitation (Co-IP) performed using recombinant V5-tagged CRELD2 overexpressed in the SW1353 (chondrocyte-like) and the SAOS2 (osteoblast-like) cell lines as previously described. (12) Co-IP proteins were identified by LC-MS/MS performed on Co-IP protein pools and 66 potential CRELD2 binding partners in SW1353 cells (top 50 shown in Supplemental Table S4) and 99 potential CRELD2 binding partners in SAOS2 cells (top 50 shown in Supplemental Table S5) were identified. Interestingly, there were only 28 binding partners in common between SW1353 and SAOS2 cells (Supplemental Table S6). In both cell types, CRELD2 was found to bind to DNAJ proteins, peptidyl prolyl-isomerases (PPIs) and PDIs that function as chaperones involved in the folding and processing of nascent proteins, further confirming its ER localization. CRELD2 also bound to protein complexes involved in the processing of collagen fibrils, such as collagen prolyl-hydroxylases as well as several collagen alpha chains including collagen alpha-2 (1) suggesting a putative role for CRELD2 in collagen folding; however, there was no evidence of collagen retention in knockout cells. Furthermore, CRELD2 was also found to bind the receptor low-density lipoprotein receptor-related protein 1 (LRP1) and the mesoderm development LRP chaperone (MESD) in both cell lines, indicating that it may function as a chaperone in assisting the folding, maturation, and transport of LRP1 through the secretory pathway during skeletal development.
To study the role of CRELD2 in LRP1 membrane trafficking, CRELD2 expression was knocked down in SW1353 and SAOS2 cells using siRNA. Successful knockdown was confirmed by qPCR and Western blotting (Fig. 5A, B). It is important to note that the expression of LRP1 was not changed at either the RNA or protein level after the ablation of CRELD2 in siRNA-treated SW1353 and SAOS2 cells (Fig. 5A, B). However, live immunolabeling (Fig. 5C) and Western blotting (Fig. 5D) of the membrane fraction revealed that there was significantly less staining of LRP1 at the membrane in both SW1353 and SAOS2 knockdown cells. Thus, in chondrocytes and osteoblasts, CRELD2 potentially functions as a novel LRP1 chaperone through the secretory pathway and promotes LRP1 expression at the cell surface.

WNT4 expression is driven by LRP1 in chondrocytes but not in osteoblasts
To determine whether reduced LRP1 membrane expression results in impaired chondrocyte and osteoblast maturation, LRP1 expression was silenced during in vitro chondrogenic and osteogenic differentiation of hMSCs using siRNA. Successful knockdown of LRP1 was confirmed by Western blotting (Fig. 6A).
The knockdown of LRP1 during chondrogenesis resulted in significantly smaller siLRP1 cartilage pellets (Fig. 6B, C); however, this difference in size was not as pronounced as observed with siCRELD2 cartilage pellets. siLRP1 pellets displayed reduced matrix staining for type II collagen; however, unlike the altered proteoglycan composition observed in siCRELD2 pellets, matrix staining (Fig. 6C) and gene expression levels (Fig. 6D) for the major cartilage proteoglycan aggrecan were comparable between control and siLRP1 pellets, suggesting that CRELD2   may play additional roles during chondrogenesis besides chaperoning LRP1. The expression levels of several chondrogenic markers (IHH, COL2A1, and COL10A1) were significantly reduced in siLRP1 chondrocytes, confirming that chondrogenesis was disrupted after the knockdown of LRP1 expression (Fig. 6D).
Unlike the knockdown of CRELD2, the knockdown of LRP1 did not affect matrix mineralization (Fig. 6E) despite a significant decrease in ALP activity in siLRP1 osteoblasts (Fig. 6F). Many of the osteoblast markers were also unaffected after the knockdown of LRP1 (Fig. 6G); however, ALPL, involved in matrix mineralization, was downregulated and the expression of the immature osteoblast marker SPP1 was upregulated in siLRP1 osteoblasts, indicating that LRP1 potentially plays a role in early osteoblast differentiation.
WNT4 expression was significantly reduced in siLRP1 chondrocytes ( Fig. 6D) but was unchanged in siLRP1 osteoblasts (Fig. 6G). These data suggest that LRP1 drives WNT4 expression during chondrogenic differentiation, but the expression of WNT4 is LRP1-independent during osteogenesis.
WNT4 signals via p38 mitogen-activated protein kinase (MAPK) in chondrocytes and osteoblasts As previously shown in primary chondrocytes and during hMSC differentiation, the knockdown of CRELD2 in the chondrocytelike SW1353 cell line resulted in a significant reduction of WNT4 expression (Fig. 7A). As observed in siLRP1 chondrocytes, WNT4 expression was also reduced in siLRP1 SW1353 cells, providing further evidence that LRP1 drives WNT4 expression in chondrocytes (Fig. 7A).
Previous studies have shown that p38 MAPK regulates noncanonical WNT signaling. (35,36) Pharmacological inhibition of p38 signaling inhibits chondrogenesis, suggesting an obligatory role for p38 MAPK signaling in chondrocyte differentiation. (37,38) To further dissect the role of noncanonical WNT4 signaling during chondrogenesis, the p38 MAPK signaling cascade was analyzed in SW1353 cells by Western blotting. The knockdown of CRELD2 in SW1353 cells did not affect canonical WNT signaling as β-catenin levels were unchanged; however, it significantly impaired p38 MAPK signaling as shown by a significant reduction in the pp38/p38 ratio in CRELD2 knockdown cells (Fig. 7B). Moreover, the addition of recombinant WNT4 to knockdown cells restored p38 signaling 15 to 30 minutes post-treatment (Fig. 7C). Recent studies have shown that LRP1 modulates WNT5A expression in mouse embryonic fibroblasts by binding TGF-β1. (39) The expression of WNT4 in SW1353 cells was upregulated after TGF-β1 treatment, whereas in LRP1 knockdown SW1353 cells, WNT4 expression was significantly reduced with and without TGF-β1 treatment (Fig. 7D), indicating that LRP1 modulates WNT4 expression via TGF-β1 in chondrocytes.
As previously observed in primary osteoblasts and hMSCs, the knockdown of CRELD2 in the osteoblast-like SAOS2 cell line also resulted in reduced expression of WNT4 (Fig. 7E). The levels of WNT4 expression were significantly higher in siLRP1 SAOS2 cells (Fig. 7E), opposite to that observed in siLRP1 SW1353 cells. These data highlight that WNT4 expression is regulated differently in chondrocytes and osteoblasts and that WNT4 downregulation in siCRELD2 osteoblasts is likely to be independent of LRP1. Studies have shown that noncanonical p38 MAPK signaling is crucial for osteogenesis as p38 inhibitors blocked osteogenic differentiation. (40,41) Western blotting was therefore used to analyze p38 MAPK signaling in siCRELD2 SASO2 cells. As demonstrated in the SW1353 cells, canonical WNT/β-catenin signaling in SAOS2 cells was unaffected after the knockdown of CRELD2; however, p38 MAPK signaling was impaired since there was a significant reduction in the pp38/p38 ratio in knockdown cells (Fig. 7F). As already demonstrated in SW1353 cells, this reduction in p38 signaling was restored by the addition of recombinant WNT4 to siCRELD2 SAOS2 cells (Fig. 7G). Thus, CRELD2 is important for WNT4 signaling via noncanonical p38 MAPK that has previously been shown to promote chondrogenic and osteogenic differentiation.
Exogenous WNT4 rescues the impaired chondrogenic and osteogenic differentiation of knockdown CRELD2 hMSCs Because WNT4 promotes chondrocyte (3) and osteoblast differentiation (4) we next wished to determine the effects of exogenous WNT4 on in vitro chondrogenic and osteogenic differentiation of CRELD2 knockdown hMSCs.
Consistent with experiments shown in Fig. 4, chondrogenesis was disrupted after CRELD2 knockdown because siCRELD2 pellets were significantly smaller (Fig. 8A, B) and had a disrupted ECM composition with reduced staining for type II collagen and aggrecan (Fig. 8B). The addition of exogenous WNT4 rescued this abnormal cell phenotype; indeed siCRELD2 pellets treated with WNT4 were significantly larger than the siCRELD2 pellets and were an equivalent size to control pellets (Fig. 8A). The expression of ECM structural proteins was restored in siCRELD2 pellets treated with WNT4 (Fig. 8B) and in addition the expression levels of several chondrogenic markers, which were significantly reduced in siCRELD2 chondrocytes, were significantly increased after the addition of WNT4 (Fig. 8C).
We have demonstrated that the in vitro osteogenic differentiation of hMSCs was delayed after CRELD2 knockdown (Fig. 4E-G). For example, matrix mineralization and ALP activity were significantly reduced in siCRELD2 osteoblasts and this was also rescued by the addition of exogenous WNT4 (Fig. 9A, B). Moreover, the knockdown of CRELD2 resulted in downregulation of several osteoblast marker genes; however, their expression returned to control levels in siCRELD2 osteoblasts treated with WNT4 (except for the osteocyte marker gene SOST) (Fig. 9C).
These data indicate that the addition of WNT4 restores the chondrogenic and osteogenic differentiation of CRELD2 knockdown hMSCs and confirms that CRELD2 plays an important role in chondrocyte differentiation and osteoblast maturation by directly modulating WNT4 signaling.

Discussion
CRELD2 has been identified as an ER-stress-inducible gene expressed in several tissues (16,17) ; however, very little is known about its role in skeletal development and homeostasis. Here we describe the first in-depth study of Creld2 transgenic mice and identify a novel important role for CRELD2 in regulating chondrogenic and osteogenic differentiation.
Creld2 CartΔEx3-5 mice displayed a chondrodysplasia-like phenotype with abnormal cartilage growth plates, whereas Creld2 BoneΔEx3-5 mice had an osteopenic phenotype with low bone density and altered trabecular architecture. Transcriptomic analysis of the knockout tissues showed dysregulation of genes important for chondrocyte differentiation and proliferation and for osteoblast maturation. Moreover, in vitro chondrogenesis and osteogenesis of hMSCs was impaired after CRELD2 knockdown, indicating that deletion of Creld2 impairs chondrocyte and osteoblast differentiation and maturation.
WNT signaling plays a major role in controlling skeletal development and homeostasis, and, importantly, noncanonical p38 MAPK WNT signaling promotes skeletal cell differentiation (37,38,40,41) Interestingly, Wnt4, encoding a secreted noncanonical (β-catenin independent) WNT signaling protein that modulates chondrocyte maturation (34) and promotes osteogenic differentiation, (35) was downregulated in both cell types after the knockout of Creld2. We confirmed that knocking down CRELD2 expression reduces noncanonical p38 signaling in both chondrocyte-like (SW1353) and osteoblast-like (SAOS2) cells, and importantly, this reduction is restored upon the addition of exogenous WNT4. Furthermore, we showed that the addition of recombinant WNT4 was able to rescue the delayed chondrogenic and osteogenic differentiation of siCRELD2 hMSCs.
To understand how CRELD2 regulates WNT4 expression, putative CRELD2 binding partners were identified by co-immunoprecipitation. CRELD2 contains an imperfect KDEL ER-retention sequence (REDL (11) ) that is believed to play a role in ER retrieval and is implicated in toxin endocytosis. (42) We therefore postulate that the REDL motif functions to retrieve extracellular CRELD2 back into the ER. A search of the UniProt database for proteins identified only four other mammalian proteins with an REDL motif. Two of these proteins, Canopy 4 (CPNY4) and Mesoderm development candidate 2 (MESDC2), function as chaperones and enhance the folding, maturation, and cell surface expression of receptors such as toll-like receptors (43) and low-density lipoprotein receptors (LRP) (44)(45)(46) respectively. Additionally, studies have shown that an isoform of CRELD1, an ortholog of CRELD2, enhances the maturation and cell surface expression of acetylcholine receptors. (47) It is particularly noteworthy therefore that we have shown CRELD2 co-immunoprecipitates with LRP1, and CRELD2 knockdown cells display less LRP1 on the plasma membrane. We therefore suggest that CRELD2 functions as a novel LRP1 chaperone by promoting the expression of LRP1 on the cell surface.
Previous studies have shown that LRP1 represses hypertrophy, and LRP1 knockdown in mature chondrocytes results in an upregulation of the hypertrophic marker COL10A1. (48) Here we show for the first time that LRP1 knockdown in hMSCs disrupted chondrogenesis and a reduction in WNT4 expression, indicating that WNT4 expression lies downstream of LRP1 during chondrogenesis. Additionally, we show that LRP1 modulates WNT4 expression in chondrocytes by binding TGFβ-1. The defect caused by LRP1 knockdown during chondrogenesis was less severe than that resulting from CRELD2 knockdown as the reduction in chondrocyte pellet size was not as pronounced in siLRP1 pellets compared with siCRELD2 pellets, suggesting that CRELD2 may modulate chondrogenesis through other pathways. Additionally, CRELD2 also co-immunoprecipitated with the prochondrogenic proteins CYR61 (33) and TGFβ-1 (49) suggesting it may regulate a wide range of signaling processes during chondrogenesis. The knockdown of LRP1 did not affect osteogenesis, which is consistent with previous studies showing that osteoblast differentiation was unaffected in Lrp1 knockout mice. However, the mice displayed an osteoporotic phenotype due to increased osteoclastogenesis (50) further supporting the contention that LRP1 limits receptor activation of RANKL and osteoclastogenesis. (51) We found reduced membrane expression of LRP1 in siCRELD2 SAOS2 cells and showed that RANKL is upregulated after Creld2 knockout in osteoblasts. This suggests that CRELD2 chaperoning of LRP1 to the cell membrane acts as a mediator of the RANKL/OPG signaling axis and therefore of osteoblastosteoclast cross-talk. Moreover, WNT4 expression was not reduced in siLRP1 osteoblasts, suggesting that CRELD2 modulates osteoblast WNT4 expression and maturation via a different pathway, potentially through directly binding to TGFβ-1 and stimulating WNT4 expression. (52) To summarize, our results propose a novel and important role of CRELD2 in chondrogenesis and osteogenesis, identifying it as a novel modulator of WNT signaling and as an evolutionarily conserved chaperone of LRP1, promoting its expression at the cell surface (Supplemental Fig. S5). To date, no human disease causing mutations has been identified in CRELD2, but due to its role in skeletogenesis, it is intriguing to speculate that it might be a potential genetic locus for skeletal diseases such as osteogenesis imperfecta (brittle bone disease).

Disclosures
All authors declare that they have no conflict of interest.