Diabetes impairs fracture healing through disruption of cilia formation in osteoblasts

Abstract

Diabetes-associated fracture risk and impaired fracture healing represents a serious health threat. It is well known that type 1 diabetes mellitus (T1DM) impairs fracture healing due to its effect on osteoblasts and their progenitor cells. Previous studies have showed that primary cilia and intraflagellar transport protein 80 (IFT80) are critical for bone formation. However, whether TIDM impairs fracture healing due to influencing ciliary gene expression and cilia formation is unknown. Here, we investigated the effect of T1DM on primary cilia in a streptozotocin induced diabetes mouse model and examined the impact of cilia on fracture healing in osteoblasts by deletion of IFT80 in osteoblast linage using osterix (OSX)-cre (OSX^cretTAIFT80^f/f). The results showed that diabetes inhibited ciliary gene expression and primary cilia formation to an extent that was similar to normoglycemic mice with IFT80 deletion. Moreover, diabetic mice and normoglycemic mice with cilia loss in osteoblasts (OSX^cretTAIFT80^f/f) both exhibited delayed fracture healing with significantly reduced bone density and mechanical strength as well as with reduced expression of osteoblast markers, decreased angiogenesis and proliferation of bone lining cells at the fracture sites. In vitro studies showed that advanced glycation end products (AGEs) downregulated IFT80 expression in osteoblast progenitors. Moreover, AGEs and IFT80 deletion significantly reduced cilia number and length which inhibited differentiation of primary osteoblast precursors. Thus, this study for the first time report that primary cilia are essential for bone regeneration during fracture healing and loss of cilia caused by diabetes in osteoblasts resulted in defective diabetic fracture healing.

1. Introduction

In United States more than 1.25 million people suffer from type 1 diabetes mellitus (T1DM), which is increasing yearly by 2–4% [1–3]. Approximately 7.9 million people in the United States suffer from fractures annually and their treatment imposes a financial burden of $21 billion per year [4,5]. T1DM causes an increased risk of fracture and complications such as delayed union and non-union fracture healing [6]. T1DM-associated inflammation and hyperglycemia adversely affect bone, contributing to higher rates of osteopenia and impaired fracture healing [7,8]. Understanding the underlying mechanism responsible for defective bone repair in diabetic fracture healing is imperative for developing new and effective therapeutic strategies.

Primary cilium is a hair-like cellular organelle projecting from the cell membrane. It functions as a sensor for chemical and mechanical stimuli to regulate the development and function of many organs including bone [9]. Intraflagellar transport (IFT) complexes (A and B) are essential for cilia assembly and function. IFT serves as a bidirectionary transport system along the ciliary axoneme to shuttle proteins from the basal body to the tip (anterograde transport) mediated by the IFT-B complex and from the tip of cilia to the cell body (retrograde transport) mediated by the IFT-A complex [10]. Mutations in IFT proteins cause cilia loss and/or malfunction and are associated with various bone disorders such as Ellis–van Creveld syndrome, Sensenbrenner syndrome, and short-rib polydactyly syndrome [11–13]. Our laboratory has shown that IFT80 (one of the components of IFT-B complex) plays a key role in bone development. Knockdown of IFT80 in bone marrow derived stromal cells impairs primary cilia formation and osteogenesis [14]. IFT80 deletion and loss of cilia in osteoblast precursor cells causes significant growth retardation and osteopenia with reduced osteogenesis [15]. Beside the indispensable role of primary cilia in bone development, the significance of primary cilia in acquired human diseases such as diabetes, hypertension and obesity has been recently shown [16,17]. Increased in glomerular filtration rate in diabetic kidney leads to sustained exposure of proximal tubular cells to fluid shear stress which cause cilia disappearance [18]. Hyperglycemia induced by streptozotocin (STZ) with loss of cilia -via conditional deletion of IFT88 using tamoxifen inducible systemic Cre with actin promoter- increases inflammation and accelerates cyst formation in the kidney [19]. Primary cilia also play a critical role in the function of β-cells in the pancreas [20–22]. Basal body and ciliary defects in murine pancreatic islets impair glucose stimulated insulin secretion. Ciliary localization of insulin receptor in insulin simulated β-cells is required for activation of downstream targets of insulin signaling pathway. In pancreatic islet of diabetic Goto-Kakizaki rats model ciliary genes expression are dysregulated and ciliated β-cells are dramatically decreased [20]. These findings demonstrate that primary cilia play critical roles in diabetes. However, to our knowledge, it is largely unknown the effect of cilia in diabetic fracture healing. Since T1DM increases the risk of fractures and impaired fracture healing more than type 2 diabetes mellitus (T2DM) [23,24] and the adverse effect of T1DM on osteoblasts and their progenitor cells is well documented [7], we aimed to explore the role of primary cilia in the regulation of T1DM fracture healing.

In this study, we induced femoral fractures in STZ-induced T1DM mice to investigate the impact of diabetes on cilia formation and compared the fracture healing among diabetic fractured mice, the fractured mice with IFT80 ablation in osteoblast lineage (OSX^cretTAIFT80^f/f) and matched control mice. Our findings demonstrate for the first time that T1DM impairs cilia formation and thereby disturbs fracture healing, providing new insights as to how diabetes interferes with osteogenesis in fracture repair.

2. Material and methods

2.1. Animal model

All studies and procedures performed on mice were approved by the University of Pennsylvania Institutional Animal Care and Use Committee. A floxed IFT80^f/f mouse line with flanking loxP cites on exon 6 was used as previously described [15]. To delete IFT80 in the osteoblast lineage cells, IFT80^f/f mice were crossed with Osx^cretTA transgenic mice, which targets osteoblast precursor cells [25]. Osx^cretTA mediated recombination contains a reverse tetracycline–dependent transactivator (tTA) which the cre expression is prevented with administration of tetracycline (or its more stable derivative, doxycycline) and initiated when its administration is stopped [25]. Standard breeding schemes (crossing Osx^cretTAIFT80^f/f mice with IFT80^f/f mice) were used to produce the experimental mice of Osx^cretTAIFT80^f/f. Osx^cretTA mice served as control. Mice genotypes were confirmed by PCR analysis. The IFT80 genotyping primers were IFT80F (5′-TGTGAGGCCAGCCCGAGTTA-3′) and IFT80R (5′-GCCTGAGCTACAGAGAGACCCCACG-3′). The Cre transgene was detected using two primers: CreF (5′-CCTGGAAAATGCTTCTGTCCGTTTGCC-3′) and CreR (5′-GGCGCGGCAACACCATTTTT-3′).

Drinking water containing 2 mg/ml doxycycline was given to pregnant female mice and their offspring to block Cre expression during skeletal development. Doxycycline was then withdrawn to induce Cre recombinase expressed in osteoblasts just prior to fracture (12–14-week old). The Osx^cretTA control mice were also bred and raised with doxycycline containing drinking water matching the experimental mice. T1DM was induced in Osx^cretTA and Osx^cretTAIFT80^f/f mice at 9 weeks of age by intraperitoneal injections of streptozotocin (STZ, 40 mg/kg, Sigma-Aldrich, St. Louis, MO) dissolved in sodium citrate buffer (pH = 4.5) once per day for 5 consecutive days [26,27]. Control mice received only vehicle (sodium citrate buffer). Mice were considered to be diabetic when blood glucose levels exceeded 220 mg/dl for two consecutive measurements. Femur fractures were produced when mice had been hyperglycemic for at least 3 weeks.

2.2. Closed femoral fracture model and timepoints

A closed femoral midshaft fracture model with intramedullary nail fixation was created in male and female mice as described previously [28–30]. Briefly, an incision was made lateral to the knee to expose femoral head. A 27-gauge needle was used to access the femoral bone marrow channel and inserted a 0.01 stainless steel pin for stabilization. After suturing the incision, a three-point blunt guillotine apparatus was used to create a uniform transverse fracture in the middle of the femur with internal pin. The fracture healing process including the primary bone formation and bone remodeling phases, was respectively assessed on Day (D) 21 and D35 following fracture surgery [31].

2.3. Micro-computed tomography

Micro-computed tomography (microCT) was performed to evaluate bone microarchitecture of the fractured femur. All fractured femurs were scanned by a μCT 35 (Scanco Medical AG, Bassersdorf, Switzerland) with a 10-μm nominal voxel size at the Micro-CT Imaging Core Facility, McKay Orthopaedic Research Laboratory, Perelman School of Medicine, University of Pennsylvania. Scan settings were 70 kVp, 114 mA, a 200-ms integration time. The scan area was manually adjusted on 2D CT images to include only the callus zone and the original cortical bone tissue and the medullary canal were excluded from evaluation according to previous reports [32–34]. A fixed threshold of 333 mg hydroxyapatite (HA)/cm was then applied for all the samples to evaluate the bone tissue in the callus. After segmentation, the ratio of the bone volume to the total volume (BV/TV), connectivity density (Conn-Dens) and bone mineral density (BMD) were analyzed (200 slices). 3D images were also reconstructed at the fracture line extending at least 3 mm to the proximal and distal.

2.4. Histological and histomorphometric analyses

Fractured femurs at D21 following the fracture surgery were fixed in 4% paraformaldehyde overnight and then decalcified for 6–8 weeks with 10% EDTA (pH 7.4) at 4 °C. Following completion of decalcifying process, the samples were soaked in 20% sucrose overnight and then embedded in optimum cutting temperature (O.C.T.). Longitudinal sections (8 μm thick) were cut from the mid-portion of the callus and collected on charged glass slides (Globe Scientific Inc.) for histology and immunofluorescent measurements. Safranin O staining was performed to visualize bone and the quantification of bone area was performed with Image J software. The callus total volume (TV) was defined as the volume enclosed by all tissues in the callus zone, and the callus bone volume (BV) was defined as the volume of the callus containing only mineralized tissue. The percentage of BV/TV fraction was calculated as the ratio of bone volume to total volume multiply by 100.

To visualize cilia on osteoblasts, examine osteoblast proliferation and identify blood vessels, immunofluorescence staining was performed using primary acetylated α-tubulin antibody (Sigma-Aldrich, St. Louis, MO, USA; 1:500), Ki67 antibody (Abcam, Cambridge, MA, USA; 1:300, ab16667) and CD31 antibody (eBioscience 14–0311–81,1:100), respectively. After washing with PBS, slides were permeabilized with 0.05% Triton X-100 and then incubated with the primary antibodies overnight at 4 °C. Slides were then incubated with Alexa Fluor 647-conjugated anti-mouse (1:1000, A-21236, Invitrogen), Alexa Fluor 594-conjugated anti-rabbit (1:1000, 111–585–144, Jackson ImmunoResearch) or Alexa Fluor 594-conjugated anti-rat (1:1000, A-11007, Invitrogen) secondary antibodies for 1 h at room temperature. Slides were washed and mounted with VectaShield containing 4,6-diamidino-2-phenylindole (DAPI) (Vector Laboratories, Burlingame, CA, USA). Images were captured and processed with Leica DMI6000 inverted epifluorescence microscope. To quantify ciliated cell percentage, twelve fields of callus were randomly selected from each slide and Z-stacked pictures were captured. Six mice were analyzed in each group. The percentage of ciliated osteoblasts for each slide was obtained by calculating the number of ciliated bone lining cells divided by the number of DAPI-positive cells around the bone particles (approximately 1000 bone lining cells per group was counted). The same procedure was used for Ki67 staining analysis. To assess angiogenesis, image J software was used to separate the positively stained CD31 areas from background and measure the relative positive CD31 area in the fracture femur callus. Results were compared to matched control antibody (1:500, Thermo Scientific 31,903) for acetylated α-tubulin antibody, match control antibody (1:300, Thermo Scientific, 02–6102) for Ki67 antibody and match control antibody (1:100, GenScript, A02004) for CD31 antibody.

2.5. Four-point bending assay

Femur fracture calluses at D35 mice following fracture surgery were isolated by removing the soft tissue thoroughly, evaluated by MicroCT and then subjected to the 4-point bending by standard procedures at the Biomechanics Core in Penn Center for Musculoskeletal Disorders (PCMD) using electromechanical testing machine (Instron 5542, Instron Inc., Norwood, MA). The lower supports were 5.4 mm apart, and the upper loading pin was in the center of the lower supports. The bending load (F) was applied with two tips of a bracket. The distance (d) between each support and the bracket tip was kept constant for all samples. The femur fracture calluses were loaded in the longitudinal direction using a 50 N load cell (Instron Inc., Norwood, MA) until the callus failed.

The load was recorded via the Instron system against sample deflection up to a maximum force of 50 N at a crosshead displacement of 0.03 mm s⁻¹. The slope (Δy/Δx) of the linear region of the load-displacement curve was used to measure stiffness (N/mm). Elastic modulus was calculated by the following formula: KL³/(48 I_min), which K = stiffness, L = span length, I_min = minimum value of moment of inertia. The I_min value which corresponds to the orientation of the callus that provides the least resistance to bending was calculated from MicroCT images [35].

2.6. Quantitative RT-PCR and Western blot

Western blots were performed as described previously [29]. Briefly, the fracture calluses at D21 post fracture were cut from the femurs right after removal of adjacent soft tissue. Fracture calluses were snap-frozen, pulverized with a cold mortar and pestle and homogenized in radioimmunoprecipitation assay (RIPA) buffer (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with protease and phosphatase inhibitors (Thermo Fisher Scientific, Waltham, MA, USA). Lysates were centrifuged, and the protein concentration in the supernatant of each sample was determined using a BCA protein assay (Pierce, Rochford, IL). The protein samples (20μg/lane) where subjected to electrophoresis on 12% SDS–PAGE gels were transferred to nitrocellulose membranes. After incubation in 5% milk–Tris-buffered saline Tween-20, the membranes were incubated with primary rabbit anti-IFT80 antibody (1:400, PAB15842, Abnova) overnight at 4 °C and then incubated with horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG antibody (1:10,000, A-11034, Novex, Carlsbad, CA) at room temperature for 1h. Visualization was performed with ChemiDoc Touch (Bio-Rad Laboratories, Hercules, CA, USA) imaging system. Beta-actin (1:2000, A00191, Genscript) was used as a loading control. Protein band intensities were measured using ImageJ software, normalized by band intensity of beta actin.

For quantitative RT-PCR, total RNA was isolated from each snap frozen and finely crushed callus at D14 post fracture using Trizol Reagent (Life Technologies, Inc., Grand Island, NY, USA). cDNA was prepared from 1 μg total RNA using the PrimeScript™ RT reagent kit (RR37B, Takara Bio). Quantitative RT-PCR was performed with SYBR Green PCR master Mix (Invitrogen) and mouse-specific primers (Bio-Rad Laboratories, Hercules, CA, USA). Relative gene expression was calculated using the delta-delta comparative threshold cycle algorithm (2^−ΔΔCT) method. To calculate the fold changes in the target gene, the relative gene expression of each group was referenced to the control group (NG Osx^cretTA). Primer sequences are listed in Table 1.

Table 1.

List of primers sequence for RT-qPCR.

Name	Gene (ID)	Forward primer sequence	Reverse primer sequence	Reference
IFT80	IFT80 (68259)	AAGGAACCAAAGCATCAAGAATTAG	AGATGTCATCAGGCAGCTTGAC	[15]
IFT20	IFT20 (55978)	GGATGCTGGTGCTTCTGGACTC	GCTCTGCGGCCCTGACGACTGT	[73]
IFT88	IFT88 (21821)	TGGTCAGCCCGCTCCTCCTC	ACCCGTGTCATTCTCCAACTCCTC	[73]
KIF-3A	KIF-3A (16568)	CCAGCCTCGCCCCCAAACC	CCGGCACCTAACCACCACCTTC	[73]
Runx2	Runx2 (12393)	ACAACCACAGAACCACAAG	TCTCGGTGGCTGGTAGTGA	[15]
OSX	SP7 (170574)	AGCGACCACTTGAGCAAACAT	GCGGCTGATTGGCTTCTTCT	[15]
ALP	ALPL (11647)	ATCTTTGGTCTGGCTCCCATG	TTTCCCGTTCACCGTCCAC	[74]
OCN	BGLAP (12096)	GCAATAAGGTAGTGAACAGACTCC	GTTTGTAGGCGGTCTTCAAGC	[74]
Col-1	COL1A1 (12842)	GCAACAGTCGCTTCACCTACA	CAATGTCCAAGGGAGCCACAT	[74]
GAPDH	GAPDH (14433)	ACTTTGTCAAGCTCATTTCC	TGCAGCGAACTTTATTGATG	[74]

2.7. In vitro preparation of advanced glycation end products (AGEs)

AGE-modified bovine serum albumin (BSA) was prepared as described by Valcourt et al. [36]. Briefly, 50 mg/ml BSA was incubated with 0.6 M D-ribose in phosphate- buffered saline (PBS; pH 7.4) at 37 °C, 5% CO2 humidified incubator, under sterile conditions and protected from light for 1 week. The unincorporated ribose was removed by dialysis against PBS for 3 days. BSA without ribose was incubated in the same conditions and served as control to AGEs in the in vitro experiments.

2.7.1. Isolation of pre-osteoblasts and viral transduction of cre-recombinase

Pre-osteoblasts (POBs) were isolated from IFT80^f/f mouse calvaria bone (postnatal day 3–5) by serial digestion as we performed previously [37,38]. Briefly, calvariae were isolated, washed with cold PBS, and incubated in collagenase type II (2mg/ml, EMD, Darmstadt, Germany), Trypsin (0.25%, Corning, Manassas, VA) for 30min. Calvaria bone was then minced and subjected to collagenase type II (2mg/ml) and trypsin again for 10 min before being plated in complete α-modified Eagle medium (Fisher Scientific) (α-MEM) supplemented with 10% fetal bovine serum (Gibco), L-glutamine (2 mM/ml) (Life Technologies) and penicillin/streptomycin (100 U/ml) (Life Technologies). POBs were infected with Adenoviruses- Cre (Ad-CMV-Cre; Vector Biolabs, Malvern, PA, USA; #1405) or Adenoviruses-Null (Ad-Null; Vector Biolabs, Malvern, PA, USA; #1300) as described [38]. Cells treated with Ad-cre yielded an approximate 85% deletion of IFT80, which were labeled as IFT80^d/d. Ad-Null-treated cells were used as control and named as IFT80^f/f. Cells at passage 3 were used for experiments which were done in triplicate.

2.7.2. Cytotoxicity analysis of AGEs

The cytotoxic effect of AGEs on POBs was examined with the Alamar blue assay (Bio-Rad Laboratories, Hercules, CA, USA). POBs were isolated from mouse calvaria bone and seeded in 96-well plates at a density of 3 × 103 cells/well for 1 day. Cells were then incubated with different doses of AGEs (50,150 and 250 μg/ml) or control BSA (150 μg/ml) for 24 h (h), 48 h and 72 h. Then, the Alamar blue reagent was added to each well and gently mixed. After 4 h incubation at 37 °C, 5% CO₂ humidified incubator, the absorbance at 570 nm and 600 nm was measured in microplate reader.

2.7.3. Osteogenic differentiation

For osteogenic differentiation, POBs were seeded at density of 1000,000 cells/ml and reached 100% confluency, and then treated with osteogenic medium in α-MEM (Gibco) containing 10% FBS, 10mM β-glycerophosphate (Sigma, St Louis, MO), 50μg/ml ascorbic acid (Sigma) and 10⁻⁸M dexamethasone (Sigma) supplemented with AGEs (150 μg/ml) and control BSA (150 μg/ml) for 21 days. To measure bone nodule formation, extracellular matrix calcium deposits were stained with Alizarin Red solution as described previously [15]. For quantification, stained cells were destained with 10% cetylpyridinium chloride in 10mM sodium phosphate (pH 7.0) and optical density was measured at the wavelength of 570 nm.

2.7.4. Quantitative real-time polymerase chain reaction

POBs were seeded at density of 1000,000 cells/ml to reach 100% confluency and then induced with osteogenic media supplemented with AGEs (150 μg/ml) or BSA (150 μg/ml) for 5 days. Total RNA was isolated from the cells with Trizol Reagent (Life Technologies, Inc., Grand Island, NY, USA) as mentioned above for assessing IFT80 and osteoblast markers expression levels. After calculating relative gene expression using 2^−ΔΔCT method, the relative gene expression of each group was referenced to the control group (BSA IFT80 ^f/f) to calculate the fold changes in the target genes.

2.7.5. Immunofluorescence in vitro

POBs were seeded at the density of 8 × 10⁵ cells/ml to reach 70% confluency. After pretreated with AGEs and control BSA for 5 days, the POBs were incubated in a serum free media overnight with AGEs or BSA for cell ciliation. Then, cells were washed with PBS and fixed with 4% paraformaldehyde. Fixed cells were permeabilized with 0.05% Triton X-100 and then incubated with acetylated α-tubulin antibody (1:500, T6793, Sigma) overnight at 4 °C followed by alexa Fluor 647-conjugated anti-mouse (1:1000, A-21235, Invitrogen) antibody as a secondary antibody for detecting cilia. Results were compared to matched control antibody (1:500, Thermo Scientific 31,903). Cilia numbers and length in the above cells were measured from the images captured with inverted epifluorescence microscope under ×40 lens. Twenty-five fields per coverslip (three coverslips per group) were randomly selected and Z-stacked pictures were captured. The percentage of ciliated cells number over total cells and cilia length were analyzed in each field using free hand tool in Leica microscope software.

2.8. Statistical analysis

Statistical analysis was performed using Prism software (GraphPad version 8). When only two groups were analyzed, statistical significance was determined using Studenťs t-test. For multiple groups analysis, two-way ANOVA was used and when significance difference was observed Tukey's multiple-comparison post hoc test was carried out to determine pairwise significance. In case there was an interaction between diabetes and genotyping, we fixed one factor (diabetes) to compare different groups. The interaction information is provided in figure legends. When the interaction did not exist, we only compared the main effect (NG Osx^cretTA, NG Osx^cretTAIFT80^f/f, Dia Osx^cretTA and Dia Osx^cretTAIFT80^f/f). All data are presented as the mean ± standard deviation (SD). P < 0.05 was considered to be statistically significant.

3. Results

3.1. Diabetes downregulates ciliary gene expression in fracture calluses

To examine the effect of diabetes on the expression of ciliary genes, we induced diabetes in C57BL/6 J mice via multiple injections of low doses of STZ while the control mice received only vehicle. Femoral fracture model was then produced at 12 weeks of age (Fig. 1A). RT-qPCR was performed for analyzing the expression levels of IFT80, IFT88, IFT20, and KIF3A in the fracture callus of diabetic (Dia) and normoglycemic (NG) mice. The results show that IFT88, IFT20 and KIF3A levels were reduced by 50%–63% (P < 0.05) and IFT80 was decreased by 70% (P < 0.01) in diabetic mice compared to normoglycemic mice (Fig. 1B). IFT20 and IFT80 protein levels were also significantly reduced in diabetic fracture callus by 65% and 75% compared to normoglycemic mice (Fig. 1C & D). To further examine whether the negative impact of diabetes on osseous healing is cilia dependent, we generated an osteoblast specific cilia loss model by conditional deletion of IFT80 (Osx^cretTAIFT80^f/f) using Osx-cre transgenic line [15]. Parallel experiments utilized diabetic mice without or with conditional IFT80 deletion (Fig. 1E) (see Supplementary Fig. 1 for blood glucose levels). Genotyping of the mice was confirmed via PCR using IFT80 and Cre primers (Fig. 1F). Water containing doxycycline was administered to pregnant female mice and their offspring to block Cre recombinase activity until inducing fracture model to ensure no effect of IFT80 deletion on bone development. Doxycycline was withdrawn immediately after fracture to induce Cre recombinase expression. RT-qPCR analysis confirmed that there was no IFT80 deletion before inducing fracture (Fig. 1G) and IFT80 was deleted in the fracture callus after withdrawing doxycycline (Fig. 1H).

Fig. 1.

Diabetes disrupts ciliogenic markers in fracture callus. (A) Schematic figure for experimental study on investigating the effect of diabetes on ciliary genes expression (B) Real time RT-qPCR analysis of ciliary genes expression (IFT80, IFT20, IFT88 and KIF3A) in fracture callus of normoglycemic (control) and diabetic mice at D14 post fracture (n = 3). The expression levels of the genes were normalized to GAPDH. (B) Western blot of IFT80 and IFT20 protein levels in fracture callus of normoglycemic (control) and diabetic mice at D21 post fracture (n = 3). (C) Quantification of protein levels of IFT80 and IFT20 were obtained by imageJ. The protein levels were normalized to beta-actin. The intensity of the bands was averaged from 3 different mice per group. (D) Experimental design and the effect of doxycycline on Cre recombinase activity. (E) Genotyping of OSX^cretTAIFT80^f/f and control littermate mice. IFT80 with flanking loxP sites (469 bp) with Cre recombinase (281 bp) is detected in the groups with floxed IFT80, whereas the wild type band (247 bp) is found only in control mice with Cre recombinase (281 bp). (F) Real time RT-qPCR analysis of IFT80 mRNA levels in control and experimental femurs immediately before withdrawing doxycycline to ensure blocking of Cre recombinase expression before performing fracture model (n = 3). There was no IFT80 deletion in OSX^cretTAIFT80^f/f femur before withdrawing doxycycline. (G) Real time RT-qPCR analysis of IFT80 expression of control and experimental fracture callus 2 weeks after withdrawing doxycycline to ensure IFT80 deletion in OSX^cretTAIFT80^f/f callus. Significance was determined by Student's t-test. * P < 0.05; Δ P < 0.01; # P < 0.001. NG: Normoglycemic; Dia: Diabetic.

3.2. IFT80 deletion in osteoblasts is comparable to diabetes in inhibiting new bone formation in the fracture callus

Axial and 3D reconstruction of microCT images from fracture calluses on D21 demonstrated less bone mass with porous woven bone in the experimental groups (NG Osx^cretTAIFT80^f/f, Dia Osx^cretTA and Dia Osx^cretTAIFT80^f/f) compared to normoglycemic controls (NG Osx^cretTA) (Fig. 2A). Quantitative microCT analysis demonstrated that the BV/TV value in diabetic control (Dia Osx^cretTA) and normoglycemic (NG) experimental (Osx^cretTAIFT80^f/f) mice was decreased by 60% and 64%, respectively, compared to NG Osx^cretTA mice (Fig. 2B). It is striking that IFT80 deletion in diabetic mice did not have a further impact on fracture healing compared to the impact of diabetes alone (P > 0.05). Similarly, the Conn-Dens and BMD significantly decreased in Dia Osx^cretTA mice as well as NG Osx^cretTAIFT80^f/f and Dia Osx^cretTAIFT80^f/f mice compared to NG control mice with no statistical difference between the three experimental groups (P > 0.05) (Fig. 2B).

Fig. 2.

IFT80 deletion and diabetes impair the callus bone architecture. (A) Representative images of 3D reconstruction and sections of μCT scans of the fracture site at D21 post fracture of NG OSX^cretTA (control), OSX^cretTAIFT80^f/f, Dia OSX^cretTA and Dia OSX^cretTAIFT80^f/f mice (experimental groups). (B) Quantitative measurements of the percentage of bone volume to total bone volume (BV/TV), connectivity density (Conn-Dens) and bone mineral density (BMD) at fracture site on D21 (n = 8–9 mice per group). (C, D) Safranin O staining of longitudinal sections of the fracture site from NG OSX^cretTA (control), OSX^cretTAIFT80^f/f, Dia OSX^cretTA and Dia OSX^cretTAIFT80^f/f mice (experimental groups) at D21 post fracture and quantification of bone area normalized to callus area (BV/TV) (n = 6 mice per group). (E) Real time RT-qPCR analysis of markers of osteogenesis, alkaline phosphatase (ALP) and osteocalcin (OCN), extracted from the fracture calluses of four groups (NG OSX^cretTA, NG OSX^cretTAIFT80^f/f, Dia OSX^cretTA and Dia OSX^cretTAIFT80^f/f) at D14 post fracture (n = 3 mice per group). Significance was determined by two-way ANOVA followed by Tukey post hoc test. Two-way ANOVA determined a significant interaction between diabetes and genotyping regarding MicroCT (P = 0.006, P = 0.038 and P = 0.035 for BV/TV, Conn-Dense and BMD, respectively) and histological analysis (P = 0.001), but No interaction was noticed between diabetes and genotyping for osteogenesis markers (P > 0.05). * P < 0.05; Δ P < 0.01; # P < 0.001. No statistically significant difference was found among experimental groups at D21 regarding MicroCT, histology and osteogenesis markers level analysis. NG: normoglycemic, Dia: Diabetic. Scale bar: 1 mm.

Specimens from D21 post fracture were further analyzed histologically to assess the effect of IFT80 deletion and diabetes on bone formation. Consistent with micro-CT results, BV/TV in Dia Osx^cretTA mice and Dia Osx^cretTAIFT80^f/f mice significantly was reduced by 44% and 49% compared to that in NG Osx^cretTA group (P < 0.05). Diabetic mice had a reduction in bone formation that was comparable to the decrease in NG Osx^cretTAIFT80^f/f (P > 0.05) and the deletion of IFT80 in diabetic mice had no further affect (P > 0.05) (Fig. 2C & D). To further investigate the influence of diabetes and IFT80 deletion on osteogenic differentiation in the fracture callus, mRNA levels of ALP and OCN were measured in callus specimens. Expression of ALP and OCN were down-regulated to a similar extent in the three experimental groups (NG Osx^cretTAIFT80^f/f, Dia Osx^cretTA and Dia Osx^cretTAIFT80^f/f) (P > 0.05) (Fig. 2F).

3.3. Ablation of IFT80 in osteoblast lineage shows the same phenotype as diabetes with reduced bone regeneration and mechanical strength during fracture healing

To further investigate bone formation in the fracture at a later stage, the fractured femurs were examined at D35 post fracture by Micro-CT analysis. The results showed that IFT80 deletion in osteoblasts in the NG mice (NG Osx^cretTAIFT80^f/f) caused a 58% reduction in BV/TV compared to NG control mice (P < 0.001). Additionally, diabetic control (Dia Osx^cretTA) and diabetic mice with IFT80 deletion (Dia Osx^cretTAIFT80^f/f) mice had 60% and 64% reduction in BV/TV compared to the NG control (NG Osx^cretTA) mice, respectively (P < 0.001) (Fig. 3A). The impact of IFT80 deletion in diabetic mice on bone formation was similar to the effect of diabetes alone (P > 0.05) (Fig. 3A & B). Similar results were observed for Conn-Dens and BMD (Fig. 3B).

Fig. 3.

IFT80 deletion and diabetes impair bone quality and mechanical strength of fracture callus. (A, B) MicroCT scan, 3D reconstruction and quantitative analysis of the fracture site of NG OSX^cretTA (control), NG OSX^cretTAIFT80^f/f, Dia OSX^cretTA and Dia OSX^cretTAIFT80^f/f mice (experimental groups) at D35 post fracture (n = 8 mice per group). (C) Modulus and stiffness of the femur at D35 post fracture from NG OSX^cretTA (control), NG OSX^cretTAIFT80^f/f, Dia OSX^cretTA and Dia OSX^cretTAIFT80^f/f mice (n = 7 per mice group). Significance was determined by two-way ANOVA followed by Tukey post hoc test. Two-way ANOVA showed a significant interaction between diabetes and genotyping regarding MicroCT (P < 0.001 for BV/TV and BMD and P = 0.003 for Conn-Dense) and mechanical analysis (P = 0.02 for stiffness and P = 0.011 for modulus). * P < 0.05; Δ P < 0.01; # P < 0.001. No statistically significant difference was found among three experimental groups regarding D35 MicroCT and mechanical analysis. NG: normoglycemic, Dia: Diabetic. Scale bar: 1 mm.

To test the mechanical strength of the femur fracture callus at D35 post fracture, four-point bending analysis was performed. In all cases failureoccurredinthemiddleof the callus (Supplementary Fig. 2). The callus stiffness in the NG Osx^cretTAIFT80^f/f, diabetes and Dia Osx^cretTAIFT80^f/f groups were significantly decreased by 33%, 50% and 49% compared to that in the control NG Osx^cretTA mice, respectively (Fig. 3C). Modulus was significantly decreased by IFT80 deletion (56%, P < 0.01) and diabetes (66% in both Dia Osx^cretTA and Dia Osx^cretTAIFT80^f/f groups) compared with that in NG Osx^cretTA control mice (P < 0.001) (Fig. 3C).

3.4. Ablation of IFT80 in osteoblast lineage and diabetes impairs cilia formation and proliferation of bone lining cells and inhibits angiogenesis in fracture callus

To assess the impact of diabetes and IFT80 deletion on cilia formation, immunofluorescence assays were performed with antibody specific for acetylated tubulin. Primary cilia were detected in 51% of bone lining cells in the NG Osx^cretTA control group, while only 14% of bone lining cells had cilia in the NG Osx^cretTAIFT80^f/f (P < 0.001), 9% in Dia Osx^cretTA and 6% in the Dia Osx^cretTAIFT80^f/f groups (P < 0.001) (Fig. 4A & B). The diabetic mice had a similar reduction in cilia in bone lining cells as diabetic mice with IFT80 deletion (P > 0.05). To further evaluate the IFT80 protein levels in fracture calluses at D21, Western blot analysis was carried out. The results show that the IFT80 protein levels were reduced by 60% in the NG Osx^cretTAIFT80^f/f group, 75% in the Dia Osx^cretTA mice and and 82% in Dia Osx^cretTAIFT80^f/f mice compared with IFT80 protein levels in the control NG Osx^cretTA group (P < 0.001) (Fig. 4C & D).

Fig. 4.

IFT80 deletion in fracture callus and diabetes decreased the cilia formation and impaired osteoblasts proliferation in fracture callus. (A, B) Z-stacked 3D deconvolution immunofluorescent images with antibody specific for acetylated α-tubulin (red) to visualize cilia and quantitative analysis of the percentciliated bone lining cells in the fracture calluses of NG OSX^cretTA, NG OSX^cretTAIFT80^f/f, Dia OSX^cretTA and Dia OSX^cretTAIFT80^f/f mice at D21 post fracture (n = 6 mice per group). Nuclei were identified by DAPI counterstained. The white dotted line indicates the bone lining cells and the black dotted bone indicates the surface of newly formed bone in fracture callus. Magnification: 40×. Scale bar: 10 μm. (C, D) Western blot for IFT80 protein levels in fracture callus of the four groups mice at D21 post fracture. IFT80 protein level was normalized to beta-actin. Quantification of protein levels of IFT80 was obtained by imageJ. The intensity of the bands was averaged from 3 different mice per group. (E, F) Immunofluorescent images and analysis of Ki67-positive bone lining cells with representative bright field images in sections from fracture calluses of NG OSX^cretTA (control), NG OSX^cretTAIFT80^f/f, Dia OSX^cretTA and Dia OSX^cretTAIFT80^f/f mice (experimental groups) at D21 post fracture (n = 6 mice per group). Nuclei were stained with DAPI. The white dotted line marks the bone lining cells and the black dotted bone indicates the surface of newly formed bone in fracture calluses. Magnification: 40×. Scale bar: 10 μm. (E, F) Immunofluorescent images with CD31 antibody and quantification of CD31-positive area in sections from fracture calluses of NG OSX^cretTA (control), NG OSX^cretTAIFT80^f/f, Dia OSX^cretTA and Dia OSX^cretTAIFT80^f/f mice (experimental groups) at D21 post fracture (n = 6 mice per group). Nuclei were stained with DAPI. Magnification: 20×. Scale bar: 50 μm. Significance was determined by two-way ANOVA followed by Tukey post hoc test. In all the analysis, a significant interaction between diabetes and genotyping was found (P < 0.001, P = 0.043 for cilia and Ki67 staining, respectively and P < 0.001 for CD31 staining and IFT80 protein level). * P < 0.05; Δ P < 0.01; # P < 0.001. No statistically significant difference was found among the three experimental groups regarding Ki67 and anti-acetylated α-tubulin immunofluorescence staining. No statistically significant difference was found in IFT80 protein level between Dia OSX^cretTA and Dia OSX^cretTAIFT80^f/f mice. Scale bar: 10 μm.

To better understand the impact of diabetes and IFT80 deletion on fracture healing we examined osteoblast proliferation in D21 fracture callus sections by immunofluorescence with antibody specific for Ki67 (Fig. 4E). In the normal control group 49% of the bone lining cells were Ki67+ while the number was decreased to 30%, 32% and 27% in NG Osx^cretTAIFT80^f/f, Dia Osx^cretTA and Dia Osx^cretTAIFT80^f/f groups, respectively (P < 0.01) (Fig. 4F).

Since blood vessels formation is an essential component of fracture healing and bone formation [39], we investigated the effect of IFT80 deletion and diabetes on vascularization in fracture callus at D21 by immunofluorescence assay using CD31 antibody. The result showed that CD31+ areas in IFT80 deleted group and diabetes groups (Dia Osx^cretTA group and Dia Osx^cretTAIFT80^f/f group) were respectively decreased to 55%, 70% and 73% compared with that in NG Osx^cretTA control mice (P < 0.001) (Fig. 4G & H). No signal was detected in callus sections stained with matched control antibodies (Supplementary Fig. 3A–C).

3.5. IFT80 deletion and AGEs impair ciliogenesis during osteoblast differentiation in vitro

To investigate mechanistically how diabetes affects osteoblasts, we investigated the role of AGEs in suppressing expression of ciliary genes and cilia formation. AGEs were produced as previously described [36]. The formation of AGEs was confirmed by pentosidine production and induction of the receptor for AGEs (RAGE) in POBs (Supplementary Fig. 4). POBs treated with a high dose of AGEs (250 μg/ml) had reduced viability; however, lower doses of AGEs (50 and 150 μg/ml) did not reduce viability (Supplementary Fig. 5A). Based on the cell viability result and previous literatures [40,41], the concentration of 150 μg/ml of AGEs was chosen to mimic the diabetic condition in-vitro.

To investigate the effect IFT80 with or without AGE stimulation on osteoblast differentiation, POBs were isolated from IFT80^f/f mice, infected with either Ad-Cre or Ad-null viruses and incubated with either 150 μg/ml AGEs or 150 μg/ml BSA (control). IFT80 deletion or AGEs treatment substantially reduced formation of a mineralized matrix as demonstrated by reduced Alizarin Red accumulation (Fig. 5A) and inhibited osteoblast differentiation as shown by reduced ALP, OSX and type 1 collagen (Col-1) expression (Fig. 5B). Consistent with in-vivo findings (Fig. 2E), IFT80 deletion combined with incubation with AGEs had no further effect on osteoblast differentiation compared to cells incubated with AGEs alone (AGEs IFT80^f/f).

Fig. 5.

IFT80 deletion and AGEs impair osteoblasts differentiation and ciliogenesis in-vitro. (A) Alizarin Red staining and quantitative analysis of IFT80^f/f and IFT80^d/d POBs treated with 150 μg/ml unmodified control BSA and 150 μg/ml AGEs at 21 days after incubation in osteogenic media. (B) Real time RT-qPCR analysis of markers of osteogenesis, ALP, osterix (OSX) and collagen type 1 (Col-1) in POB cultures from IFT80^f/f and IFT80^d/d mice in osteogenic media with 150 μg/ml unmodified control BSA or 150 μg/ml AGEs for 5 days. (C, D) Immunofluorescence of IFT80^f/f and IFT80^d/d in POB cultures with antibody specific for acetylated α-tubulin (red) to visualize cilia and quantitative analysis of the percent ciliated cells and cilia length. (E) Real time RT-qPCR analysis for IFT80 expression level in IFT80^f/f and IFT80^d/d POBs treated with 150 μg/ml unmodified BSA or 150 μg/ml AGEs for 5 days. In vitro experiments were carried out in triplicate. Magnification: 40×. Significance was determined by two-way ANOVA followed by Tukey post hoc test. Significant interaction was identified between AGEs and cells' genotyping in all the analysis (P < 0.001 for Alizarin Red, OSX marker, ciliated POBs and cilia length. P = 0.004, P = 0.003 and P = 0.001 for Col-1, ALP and IFT80 markers, respectively). * P < 0.05; Δ P < 0.01; # P < 0.001. No statistically significant difference was found among three experimental groups (BSA IFT80^d/d, AGEs IFT80^f/f, AGEs IFT80^d/d) in regard to mineralization, osteogenic markers expression level and anti-acetylated α-tubulin immunofluorescence staining. Scale bar: 10 μm.

To determine whether AGEs reduced osteoblastic cilia formation, immunofluorescence staining assay was carried out for analyzing cilia number and length in POBs. The result showed that 78% of the POBs had cilia in the control group, unmodified BSA, while cilia numbers were significantly reduced to 33%, 38% and 30% in the groups of IFT80 deletion, AGEs and IFT80 deletion plus AGE incubation, respectively (P < 0.001). Moreover, cilia length in BSA control group was 2.4 μm which was significantly reduced by either IFT80 deletion, exposure to AGEs or IFT80 deletion cells plus AGEs (0.93, 1.1 and 0.88 μm, respectively) (P < 0.001) (Fig. 5C & D). No signal was detected in POBs stained with the match control antibody (Supplementary Fig. 3D). To further assess the cilia formation in response to the different doses of AGEs, we perform the dose response study. Low dose of AGEs (50 μg/ml) did not affect cilia formation in POBs whereas higher dose of AGEs (150 μg/ml and 250 μg/ml) significantly reduced cilia formation in POBs (Supplementary Fig. 5B & C, P < 0.001). Consistently, POBs incubated with AGEs demonstrated a dramatic reduction in ciliary IFT80 expression comparable to Cre-recombinase mediated gene deletion (Fig. 5E).

4. Discussion

The mechanisms by which diabetes affects osteoblasts in fracture healing have drawn considerable interest due to the impact of this disease on the healing process [8,42,43]. The effect of diabetes on primary cilia of several organs such as kidney and pancreas has been recently demonstrated [18–20]. Diabetes induced hyperglycemia in the absence of cilia accelerates cystogenesis in the kidney and causes renal damages [19]. Diabetes decreases the number of ciliated beta cells along with dysregulated ciliary genes expression [20]. Previous studies have shown that impaired glucose tolerance and diabetes are common in patients with syndromic ciliopathies such as Bardet-Biedl and Alström Syndromes [44–46]. However, the effect of diabetes on primary cilia during fracture healing has not been well defined. This study for the first time uncovered the role of primary cilia in fracture healing and a novel mechanism by which diabetes interferes with fracture healing through a dramatic downregulation of cilia formation in osteoblasts, demonstrating that primary cilia play an essential role in facilitating bone formation during the healing process and the restoration of bone mechanical strength in diabetic fracture healing.

Recent studies have shown that loss of primary cilia increase diabetes susceptibility. Additionally, in diabetic animals, ciliated β-cells are significantly reduced in the pancreas [20–22], demonstrating the important role of primary cilia in regulation of diabetic β-cells. Consistently, in this study, we found that diabetes suppressed cilia formation in osteoblasts during fracture healing. Interestingly, the expression levels of ciliary genes (IFT80, IFT88, IFT20 and KIF3A) in the diabetic fracture callus were significantly downregulated compared to those in the NG control mice. These results provide a reasonable explanation for diabetic mediated suppression in the formation of primary cilia, given deletion of ciliary genes such as IFT88, IFT80 in osteoblasts disrupts cilia formation and function and impairs bone development [14,15,47,48]. Supportively, the deletion of IFT80 in osteoblasts disrupted cilia formation and caused a similar reduction in bone formation and mechanical strength in the fracture calluses compared to the diabetic mice. Furthermore, osteoblast-deletion of IFT80 in diabetic mice did not further suppress ciliogenesis compared to diabetic mice, reflecting the dramatic and profound effect of diabetes on cilia formation. Noteworthily, Osterix-cre mediated IFT80 deletion occurs in osteochondroprogenitors [49,50]. Our previous study also revealed the essential role of IFT80 and primary cilia in chondrocytes during fracture healing. Deletion of IFT80 in chondrocytes dramatically decreased primary cilia formation in callus chondrocytes resulting in defective fracture healing with significant reduction in the cartilaginous callus formation [29], which was further supported by previous findings that smaller cartilaginous callus is formed in diabetic mice [51–53]. This study mainly focuses on fracture healing at D21 and D35 (bone formation and remodeling stage) which further demonstrates that diabetes-reduced ciliogenesis in osteoblasts is a key event needed for fracture healing.

During diabetic fracture healing process, angiogenesis dramatically reduces which has been shown to be associated with fracture healing complications [53,54]. Reduced angiogenesis disrupts not only the supply of nutrients and growth factors, but also the recruitment of precursor cells to the fracture site and supportive soft tissue [39]. Application of angiogenesis factors in diabetic murine models promotes vascularization at early and late stages of bone healing process and improves the healing process [55–57]. In this study, we also found a significant reduction in the blood vessel formation of diabetic fracture callus. Notably, we found that IFT80 and primary cilia in osteoblasts are essential for blood vessel formation in fracture callus as osteoblastlineage IFT80 deletion markedly decreased angiogenesis in fracture callus. Supportively, our previous study also confirmed that IFT80 and primary cilia in chondrocytes play a critical role in angiogenesis during fracture healing [29]. Diabetes and ablation of IFT80 in osteoblasts also significantly reduced osteoblast proliferation and expression of osteogenic markers during fracture healing in-vivo. These results are consistent with the concept that cilia are needed for both osteoblastic proliferation and differentiation, which are also impaired by diabetes. The impact of diabetes on cilia provides insight on how diabetes may negatively affect osteoblast proliferation and differentiation as previous studies have reported [58–60].

STZ-induced T1DM model is a well-established diabetic model and has many similarities with T1DM complications in humans such as chronic hyperglycemia, impaired bone quality and reduced bone strength [54,61,62]. However, the toxic effect of STZ is not only restricted to pancreatic β-cells but also may adversely affect other organs including renal, liver and cardiac. To reduce the nonspecific toxicity, STZ was administered in multiple injections of low doses to cause repeated low grade β-cells damage [63–65]. The link between the accumulation of AGEs in STZ-induced diabetic mice and bone related complication has been extensively shown [60,66,67]. AGEs are complex and heterogenous compounds formed via a nonenzymatic reaction between reducing sugars and amine residues on proteins, lipids, or nucleic acids. AGEs formation is strongly accelerated in diabetes due to the longterm hyperglycemia [68]. Hansen et al. found CML (N-carboxy-ethyl-lysine, one of the most abundant components of AGEs) concentration in the blood serum of STZ-induced diabetic mice is between 100 and 150 ng/ml and in muscle lysate between 150 and 200 ng/ml [67]. Another study showed a significant increase in collagen glycation in bone of STZ mice determined by assessing the fluorescence emission at 440 nm [66]. Although no studies showed the AGEs concentration in bone tissue in SZT-induced diabetic model, we used 150 μg/ml concertation since it is one of the most common used of AGEs in in-vitro studies [40,41] and also based on our optimization results, 150 μg/ml was the highest dose of AGEs that was not toxic to osteoblasts and it inhibits cilia formation. Consistent to the in-vivo findings, our in-vitro results also demonstrate that AGEs reduced primary cilia formation and cilia length in osteoblasts and downregulated IFT80 expression levels. Both AGEs and IFT80 knockdown in osteoblasts reduced osteogenesis and negatively impacted osteoblast differentiation and formation of a mineralized matrix. These results are also consistent with previously reported studies on the effect of diabetes on these parameters [36–39].

Here we found for the first time that primary cilia and IFT80 protein in osteoblasts are essential for fracture healing process, and diabetes negatively affects primary cilia formation and function in osteoblasts thereby leading to defective fracture healing. Our ongoing study is targeting on rescuing ciliogenesis in diabetic fracture healing to more firmly establish the link between diabetes mediated cilia loss and impaired bone healing and providing a promising therapeutic strategy for enhancing the healing process.