Introduction

Chondrocytes are the drivers of elongation within mammalian growth plates and are responsible for endochondral ossification [1]. During this process chondrocytes are arranged in well-organized columns as the cells undergo linear differentiation and maturation. Cells in the reserve zone (resting zone cartilage, RC) maintain a hyaline-like extracellular matrix (ECM) consisting of large proteoglycan aggregate and type II collagen. Cells in the lower reserve zone undergo a series of divisions, which are responsible for long bone growth. Post-proliferative cells form the growth zone (GC), which includes the pre-hypertrophic zone and the upper hypertrophic zones, where the cells increase in size and remodel their ECM in preparation for mineralization [2].

Key players in the mineralization process are matrix vesicles (MVs) [3,4,5].They are 50–150 nm in diameter lipid bound vesicles produced by the chondrocytes and secreted by lateral budding into the ECM [6]. MVs produced by GC cells contain enzymes needed for ECM mineralization, including matrix processing enzymes such as matrix metalloproteinases (MMPs) and enzymes involved in calcium phosphate deposition such as alkaline phosphatase, along with factors and microRNA (miRNA) for chondrocyte regulation [7,8,9].

Vitamin D metabolites are involved in the regulation of this entire process and the specific metabolite involved is zone dependent. RC cells respond primarily to 24R,25-dihydroxyvitamin D3 [24R,25(OH)2D3] with an increase in alkaline phosphatase activity and the stimulation of cell proliferation while GC cells primarily respond to 1α,25-dihydroxyvitamin D3 [1α,25(OH)2D3] by decreasing cell proliferation and increasing alkaline phosphatase activity [10,11,12]. This zonal specificity holds true for MVs produced by these cells. 24R,25(OH)2D3 regulates the production of RC MVs, which exhibit increased alkaline phosphatase activity and a reduction in phospholipase A2 (PLA2) activity, as well as incorporation of neutral metalloproteinases. In contrast, 1α,25(OH)2D3 regulates the production of GC MVs, with an increase in both alkaline phosphatase and PLA2 activity, and incorporation of acid matrix metalloproteinases (MMPs) [13, 14].

Regulation of post-fetal growth plate development has been studied using rat costochondral cartilage cells due to the comparative ease of dissection and the fact that the growth plates remain open throughout life. Cells present in the costochondral cartilage at the interface of the cartilage and bone are arranged similarly to the long bone growth plate. However, the long bone growth plates become mineralized as the rats mature and isolation of chondrocytes and MVs from them necessitates the induction of rickets to obtain sufficient uncalcified tissue. Despite differences in mechanical loading, published studies show that the costochondral growth plate cartilage cells and chondrocytes isolated from the long bone growth plates respond in comparable ways to regulation by vitamin D metabolites and the composition of their matrix vesicles are comparable [15, 16].

1α,25(OH)2D3 has two known receptors that mediate its role in GC cells, acting through both genomic and rapid acting pathways. The genomic effect of 1α,25(OH)2D3 is realized through binding of 1α,25(OH)2D3 with the canonical vitamin D receptor (VDR), which forms a heterodimer within the nucleus and is able to bind to vitamin D response elements (VDRE) located in the promoter regions of target genes [17]. Binding to the VDR regulates expression of over 1000 target genes in human tissues and in chondrocytes has been shown to upregulate receptor activator of NF-κΒ ligand (RANKL) and vascular endothelial growth factor (VEGF) production [18, 19]. Little is known about a direct role for 1α,25(OH)2D3 in post-translational regulation.

Rapid acting effects of 1α,25(OH)2D3 in GC chondrocytes take place through membrane-associated VDR, as well as through protein disulfide isomerase, family A, type 3 (PDIA3), which is also known as 1,25 membrane-associated rapid response steroid-binding protein (1,25-MARRS). Membrane-associated VDR and PDIA3 are able to regulate chondrocytes through various signal transduction systems (e.g. PLA2, phosphatidyl inositol 3-kinase [PI3K], protein kinase A and C [PKA and PKC] pathways that increase cytosolic calcium levels and activate transcription factors nuclear factor kappa Β [NF-κΒ] and extracellular-signal regulated kinase [ERK] 1/2) that elicit both a rapid response and activation of transcription factors [12, 17, 20,21,22].

Cartilage is not vascularized, making local production of many regulatory factors important for controlling chondrocyte proliferation, differentiation, and maturation. RC and GC cells are able to use 25(OH)D3 to produce metabolically active 24R,25(OH)2D3 and 1α,25(OH)2D3 [23, 24].These metabolites are secreted by the chondrocytes into the ECM under cell-specific regulation, where they interact directly with the MV membrane. Whereas 24R,25(OH)2D3 stabilizes the MV membrane, 1α,25(OH)2D3 destabilizes it through the activation of PLA2, releasing the MV contents into the ECM [25].

1α,25(OH)2D3 acts on MVs via rapid membrane-associated mechanisms. Both VDR and PDIA3 are present on the MV membrane [26, 27]. In the ECM, 1α,25(OH)2D3 activates PLA2, hydrolyzing membrane phospholipids to produce lysophospholipids, and destabilizing the MV membrane [28]. One consequence is the release of stromelysin 1 (MMP-3), which activates latent transforming growth factor beta 1 (TGF-β1) that has been secreted by chondrocytes into the ECM [29,30,31].

We recently demonstrated that the costochondral RC and GC chondrocytes produce MVs that contain a population of miRNA that is distinct from the parent cell, using smallRNAseq [32]. Treatment of intact MVs with RNase failed to destroy the miRNA. In contrast, all RNAs were destroyed by RNase when the MVs were first treated with detergent, indicated that one function of the MVs was to maintain miRNA safety [9, 15, 32]. We assessed the regulatory role of select miRNAs that were present almost exclusively in the MVs v. their parent cells and showed that they modulated aspects of proliferation and differentiation [15].

1α,25(OH)2D3 has been shown to regulate miRNA production in other cell types [33]. The observation that 1α,25(OH)2D3 regulates MV production by GC cells, including packaging of enzymes and other bioactive factors, suggests that it might also mediate the packaging of miRNA. Similarly, locally produced vitamin D metabolites contribute to the release of MV contents in the ECM, suggesting that 1α,25(OH)2D3 is also involved in the availability of miRNA for chondrocyte regulation. The purpose of this study was to determine if 1α,25(OH)2D3 regulates the production and/or packaging of miRNA into GC MVs as part of its regulatory role in the growth plate. In addition to the export of miRNA, we examined the ability of 1α,25(OH)2D3 to release miRNA from GC MVs and the impact that 1α,25(OH)2D3 has on the binding/endocytosis of GC MVs by GC cells.

Materials and Methods

Chondrocyte Cultures

Costochondral cartilage was removed from 100 to 125 g male Sprague Dawley rats. RC and GC chondrocytes were isolated by enzymatic digestion and cultured as previously described by Boyan et al. [34, 35]. All animal procedures conducted during the course of this work followed a protocol approved by the Institutional Animal Care and Use Committee (IACUC) at Virginia Commonwealth University. Experiments were carried out in accordance with all relevant guidelines and regulations. In brief, animals were killed by CO2 asphyxiation followed by cervical dislocation. Rib cages were removed by sharp dissection and placed in Dulbecco’s modified Eagle’s medium (DMEM, Life Technologies, Carlsbad, CA) 1 g/L glucose with 150 U/mL penicillin and 150 μg/mL streptomycin. All tissue was removed from around the ribs and the RC and GC cartilage sections were dissected out discarding one or two transition slices. The cartilage slices were incubated in 0.25% trypsin–EDTA (Gibco, Gaithersburg, MD) for 1 h, washed and incubated in 0.2% collagenase type II (Worthington Biochemical, Lakewood, NJ) for 3 h on a shaker. Cells were filtered through a 40 µm nylon mesh strainer, collected by centrifugation (500 g for 10 min) and resuspended in DMEM FM (1 g/L glucose DMEM with 10% FBS, 50 U/mL penicillin, 50 µg/mL streptomycin, and 50 µg/mL ascorbic acid). Cells were plated at a density of 20,000 GC cells/cm2. Cells were incubated at 37˚C and 5% CO2 and the culture media were changed 24 h after plating and then every 48 h thereafter. Confluent cells were passaged using 0.25% trypsin and plated as above.

Fourth passage cells were used for all experiments due to the need to generate sufficient MVs to conduct the analysis of MV miRNA. Previous studies demonstrated that the fourth passage growth plate chondrocytes retain their differential phenotype with respect to ECM composition and zone-specific responses to growth factors and hormones [34]. Fourth passage GC cells were grown to confluence and fresh media were added. 24 h later the media were aspirated and replaced with fresh media containing 10–8 M 1α,25(OH)2D3 for 24 h at which time the MVs were isolated from the ECM, as described below. The single time point was selected based on previous investigations of the reaction of GC cells to 1α,25(OH)2D3 and the need for over 100 T175 flasks to obtain sufficient MV RNA for sequencing [36].

Materials

Unless otherwise indicated: 0.9% NaCl was 0.2 µm filtered; and ultracentrifugation was performed in 32.5 mL straight wall tubes (Beranek Laborgeräte, Nuβloch, Germany) using a 50.2 Ti fixed-angle rotor (Beckman Coulter, Indianapolis, IN). Exosome-depleted FBS was derived from heat inactivated FBS (Life Technologies, Carlsbad, CA) that was centrifuged to remove exosomes. 30 mL of FBS was spun at 238,668 g for 4 h at 4˚C, after which the top 10 mL of supernatant were carefully pipetted off and 0.2 µm filtered for use in MV isolation and the remaining 20 mL was discarded.

MV Isolation

Media were aspirated from the culture flask and discarded. The cell layer was washed with 1 × DPBS and the MVs were separated from the ECM by digesting the cell layer with 0.25% trypsin and following that with differential centrifugation as has been described previously [34].In brief: fourth passage GC cell layers in T-175 flasks were incubated in 10 mL 0.25% trypsin for a total of 10 min at 37 °C. At minute 4, the bottom of each flask was gently scrapped to ensure detachment of the cell monolayer and aid in initial breakdown of the ECM. After being scraped, the flasks were returned to 37 °C for the remainder of the 10 min. 10 mL 1 × DPBS with 10% exosome-depleted FBS was used to quench the trypsin. The MVs were isolated from the digested ECM/cell suspension by differential ultracentrifugation. The suspension was spun (500 g for 10 min at 4 °C) in a benchtop centrifuge to pellet the cells leaving MVs in the supernatant. The MV supernatant was poured off and stored in − 80 °C and the cell pellet resuspended in 10 mL 0.9% NaCl. Two 1 mL aliquots were made from the resuspended cell pellet and spun (500 g at 4 °C for 10 min) in a benchtop centrifuge. The supernatant was discarded and one pellet resuspended in 700 μL of Qiazol lysis reagent (Qiagen, Hilden, Germany) for RNA extraction as described below, and the other in 200 μL of NP-40 lysis buffer (Boston BioProducts, Ashland, MA) with 1% protease inhibitor cocktail (Sigma, St. Louis, MO) as CP isolate. The remaining 8 mL of cell pellet resuspension was spun down (500 g at 4 °C for 10 min) in a benchtop centrifuge and resuspended in 5 mL 0.25 M sucrose solution (0.25 M sucrose, 1 mM EDTA tetrasodium salt, 100 mM Tris) for plasma membrane isolation. All suspensions were stored at – 80 °C.

To obtain the MVs, the MV supernatant containing exosome-depleted FBS was thawed on ice and processed in an ultracentrifuge. Any cell debris was spun down (24,000 g for 20 min at 4 °C) and the resulting supernatant was transferred to new tubes and spun (100,000 g for 70 min at 4 °C) to pellet the MVs. The supernatant was discarded and the MV pellet was washed with 20 mL 0.9% NaCl at 4 °C and centrifuged again at 100,000 g for 70 min at 4 °C. This final supernatant was discarded and the MV pellet was resuspended in fresh 0.9% NaCl (100 μL for RNA, 200 μL for MV miRNA release, MV endocytosis, and western blot, 500 μL for alkaline phosphatase activity and NanoSight characterization). Exosome-depleted FBS was used as a control for the NanoSight characterization.

Plasma Membrane (PM) Isolation

In order to validate the isolated MV preparation, we compared its alkaline phosphatase activity to that of isolated plasma membranes based on the greater than twofold enrichment in MV enzyme specific activity compared to that of the PM [37]. The PM suspension was thawed on ice, poured into a glass homogenizer, 1 mL 0.25 M sucrose was added to the sample tube, vortexed and added to the homogenizer. The tube was then rinsed with an additional 2 mL 0.25 M sucrose that was poured into the homogenizer. The pestle was inserted into the homogenizer, rotated 360° and lifted up. This was repeated a total of five times. The sample was poured out of the homogenizer into an ultracentrifuge tube; pestle and homogenizer were each rinsed with 2 mL 0.25 M sucrose solution that was poured into same ultracentrifuge tube. The tube was spun (1480 g at 4 °C for 20 min) and the supernatant was discarded. The pellet was resuspended with 2 M sucrose solution and spun (20,000 g at 4 °C for 20 min) and the supernatant was poured off into a new ultracentrifuge tube. The tube was filled with ice cold ultrapure water and spun (40,000 g at 4 °C for 30 min). The supernatant was carefully poured off and the PM pellet resuspended in 1 mL of 0.9% NaCl.

Alkaline Phosphatase Specific Activity

Alkaline phosphatase specific activity was measured in MV, PM, and cell pellet (CP) isolates (in MP-40) using a colorimetric assay that specifies the amount of phosphatase used at 37˚C to hydrolyze 1 μmole of p-nitrophenyl phosphate to 1 μmole of p-nitrophenol [37]. Sample absorbance was quantified using a plate reader at 405 nm. Enzyme activity was normalized to sample protein content as determined by the Pierce BCA assay for CP and PM and the micro BCA assay for MV samples (ThermoFisher, Waltham, MA).

Matrix Vesicle Composition

Nanosight

In order to be detected as individual vesicles, MV samples were diluted 1:50 in 0.9% NaCl and loaded into a NanoSight NS300 running NTA 3.3 Dev Build 3.3.104 (NanoSight, Salisbury, UK). Three samples were run for each variable. Data were analyzed by NanoSight’s NTA software Malvern WR14 1XZ, UK. Error bars represent ± 1 standard error of the mean.

Western Blot

In order to assess purity of the MV preparations, marker enzymes were determined using western blot. Protein concentration was determined by BCA (CP & PM) and microBCA (MV). 0.7 μg protein was loaded into each well with 6 × Laemmli SDS sample buffer (Alfa Aesar, Ward Hill, MA) using Mini-PROTEAN TGX 4–15% precast gels (Bio-Rad Laboratories, Hercules, CA) and electrophoresed at 120 V and then transferred to a polyvinylidene fluoride (PVDF) membrane (Bio-Rad Laboratories). The membranes were blocked for one hour and then incubated overnight with primary antibodies at 4 °C with gentle shaking with respective antibodies: (Alix, 1:1000) (Sigma); Annexin V (1:1,000), B-actin (1:5000), and Na + /K + ATPase (1:1000) (Cell Signaling Technology, Danvers, MA); CD63 (1:5000) (Bio-Rad); and CD81 (1:5000) (ABclonal, Woburn, MA). Membranes were washed and incubated with species-specific horseradish peroxidase-labeled secondary antibody for one hour at room temperature and washed again. Membranes were developed with chemiluminescense detection system (Pierce, Waltham, MA) to detect bound antibodies.

RNA Analysis

Immediately after isolation, the MV pellets were resuspended in the 100 μL volume remaining in the centrifuge tube after decanting the supernatant and mixed with 700 μL Qiazol lysis reagent. Exosome-depleted FBS was processed as above and resuspended in 100 μL 0.9% NaCl for RNA extraction. All samples were stored at − 80 °C. After thawing, samples in Qiazol were processed using Qiagen’s miRNeasy mini kit following the standard protocol for purification of total RNA from animal cells. In brief, chloroform was added, the aqueous phase was removed and combined with ethanol, RNA was bound to a spin filter and washed with RWT and RPE buffers. MV and exosome-depleted FBS samples were eluted in 30 μL of nuclease free water per sample and CP samples in 50 μL.

RNA isolations were quantified with an RNA 6000 Pico Chip on a BioAnalyzer (Agilent, Santa Clara, CA). Samples were then pooled to maximize RNA distribution between the MV groups and submitted for smallRNAseq (ArrayStar, Rockville, MD). Samples included quality control and library construction, and were sequenced on an Illumina instrument (Illumina, San Diego, CA).

MicroRNA Bioinformatics

Bioinformatic Analysis of smallRNAseq

Raw read files were processed with the miARma-Seq tool that evaluated read quality with FASTQC, aligned the reads using Bowtie2 against the Rattus norvegicus (Rnor_6.0) genome, and generated a count of the microRNA from the bam files. Bioinformatic analysis was carried out using R version 4.0.3, normalization and differential expression was calculated with edgeR, Venn diagrams built using VennDiagram, heatmaps with pheatmap and the static parallel coordinate plots with bigPint [38,39,40].

MiRNA Enrichment Analysis and Annotation Analysis

MicroRNA from four different comparisons (cell vehicle—cell 1α,25(OH)2D3, MV vehicle—MV 1α,25(OH)2D3, cell vehicle—MV vehicle, and cell 1α,25(OH)2D3—MV 1α,25(OH)2D3) with absolute log twofold change > 1 and a p value < 0.05 were uploaded to the miRNA enrichment analysis and annotation 2.0 tool (miEAA 2.0) and run through the over-representation analysis (ORA) pipeline [41]. Resulting annotations were searched for musculoskeletal associated terms (Table 1). Results from the full ORA as well as the musculoskeletal focus was loaded into Cytoscape 3.8.2 to visualize the microRNA interactions.

Table 1 Differentially expressed miRNA (p value < 0.05 & absolute log twofold change > 1)
Full size table
MIENTURNET

An unguided search for regulated genes was carried out using MIENTURNET [42]. MicroRNA with log twofold change > 1 and a p value < 0.05 from the miRNA that were uniquely differentially expressed in the vehicle group (27) or the 1α,25(OH)2D3 group (20) were uploaded to MEINTURNET and a target enrichment analysis was performed. The resulting genes were filtered to require a minimum of 2 miRNA-target interactions and sorted by p-value. The genes were then manually searched for known biological roles.

MV miRNA Release

We previously showed that MV metalloproteinases are released when MVs are treated with either detergent or 1α,25(OH)2D3 and small RNAs are released when MVs are treated with detergent, making them susceptible to degradation by RNase [30, 32]. In order to determine if miRNAs are released following loss of MV membrane integrity due to 1α,25(OH)2D3 action, we treated MVs with detergent or 1α,25(OH)2D3, followed by treatment with RNase. MVs isolated in 0.9% NaCl were pooled together, the protein content, as determined by micro BCA assay, normalized to 45 μg/mL, and samples then aliquoted out into micro centrifuge tubes (six samples per group). The treatment was added to each tube (0.9% NaCl, Tween 20 + RNase, 1α,25(OH)2D3, or 1α,25(OH)2D3 + RNase) bringing the total sample volume to 100 μL (with 10−8 M 1α,25(OH)2D3, 0.05% Tween 20, and/or 10 μg/mL RNase) and the tubes were incubated in a 37˚C water bath for three hours. 700 μL of Qiazol lysis reagent was added to each sample and processed with Qiagen’s miRNeasy micro kit. RNA was eluted in 30 μL of nuclease free water per sample.

RNA was quantified with a RNA 6000 pico chip on a BioAnalyzer. The generated pseudo gel images were analyzed for specific band region using Licor’s Image Studio Lite software (LICOR, Lincoln, NE).

MV Endocytosis

MVs and 0.9% NaCl were stained with the lipophilic membrane dye PKH26 (Sigma) as follows. MVs were normalized to 75 μg/mL of protein and 120 μL of MV solution or 0.9% NaCl was added to 400 μL of diluent C (Sigma) in each ultracentrifuge tube. 2 μL of PKH26 dye mixed with 500 μL of diluent C was added to each ultracentrifuge tube and the mixture was allowed to sit at room temperature for 4 min (pipetting every 90 s to mix). 1 mL of sterile filtered 1% BSA in PBS was added to quench the remaining PKH26 and then 8 mL of DMEM FM (without ascorbic acid) was added. The solution was ultracentrifuged (100,000 g for 70 min at 4 °C), the supernatant was discarded and the remaining red pellet resuspended in 100 μL sterile DPBS.

GC cells were grown to third passage and then passed onto 8 well glass chamber slides for the fourth passage. At 100% confluence the media were replaced with either 10−8 M 1α,25(OH)2D3 or vehicle media and 11 μL of the stained MV solution or NaCl were added and the cultures were incubated for 24 h. All media were removed, wells were washed twice with 1X PBS, fixed with 300 μL 10% neutral buffered formalin for 30 min at room temperature, and counterstained with 100 μL 5 μM DRAQ-5 in 1X PBS for 27 min at 37 °C. The chamber and gasket were removed from each slide, Fluoro Gel with DABCO (EMS, Hatfield, PA) mounting media were added, and then a coverslip attached. One image per well was taken on a LSM 880 confocal microscope (Zeiss, Jena, Germany) using a 20 × objective for quantification and a 63 × objective coupled with an airyscan detector for high resolution imaging.

Quantification was performed by two independent and blinded observers. The images were overlaid with a 3 × 3 grid and each observer counted the number of DRAQ5 stained nuclei and the number of cells with six or greater distinct PKH26 spots. The percent stained cells was calculated for each image and the average of both observers used for quantification. Additionally, CellProfiler 3.0 was used for automated quantification of the images [43].

Statistical Analysis

MV miRNA release and alkaline phosphatase activity are presented as mean ± standard error of the mean for six samples for each group. An ANOVA with Tukey HSD post-hoc test was used to examine differences between the groups. Significance was determined by a p ≤ 0.05.

Results

1α,25(OH)2D3 Regulates Chondrocyte Cellular and MV miRNA

We analyzed three separate MV isolations using a NanoSight NS300 in order to obtain a nanometer-based size distribution. The average of the three runs had two primary peaks at 98 and 143 nm (Fig. 1A) with 64.4% of the total particles falling between 50 and 150 nm. Western blot characterization of isolated MVs, plasma membranes (PM), and cell pellets (CP) identified CD63 and CD81 only in the MV sample. Alix was found in both CP and MV samples. B-actin and annexin V were found in the CP sample. Na+/K+ ATPase was visible in all samples with MV exhibiting the greatest intensity (Fig. 1B). Greater than twofold enrichment of alkaline phosphatase specific activity in MVs compared to that of the CP and PM isolations was used to validate the MV preparation (Fig. 1C).

Fig. 1
figure 1

A Size distribution of particles within MV isolate at 1:50 dilution. Mean of three samples run on NanoSite NS300. B Western blot results for cell pellet (CP), plasma membrane (PM), and matrix vesicle (MV) samples. C Alkaline phosphatase activity in CP, PM, and MV samples with either vehicle or 1α,25(OH)2D3 treatment. Mean ± SE, ‘a’ indicates significance to Vehicle with student T-Test (α = 0.05)

Full size image

A comparison of the 14 samples (3 cells + vehicle, 3 cells + 1α,25(OH)2D3, 3 MVs + vehicle, 3 MVs + 1α,25(OH)2D3, and 2 exosome-depleted FBS) with principle component analysis (PCA) and a heatmap indicated that there were three distinct clusters, one for each sample type (cell, MV and exosome-depleted FBS) (Fig. 2). There was visible difference between the main sample types (cell, MV and exosome-depleted FBS) in the heatmap with the cell samples clustering in the first six columns, the MV samples in the following six and the exosome-depleted FBS samples in the final two (Fig. 2A). The miRNA could be divided into four distinct clusters as indicated on the y-axis. The sample grouping was also clearly visible in a principle component analysis with the sample types clustering on the x and y axes that together accounted for 90% of the variance between samples (Fig. 2B). The exosome-depleted FBS samples formed distinct clusters separate from both cell and MV groups and were not included in the differential expression analysis.

Fig. 2
figure 2

A Heatmap of sample z-scores clustered with the complete linkage method and euclidean distance measure. B PCA plot of components 1 and 2 for all 14 samples

Full size image

Examination of the differentially expressed (p value < 0.05 and absolute log twofold change > 1) miRNA between parent cells and MVs indicated a selectively exported population of miRNA in the MVs with 211 differentially expressed in the vehicle group and 198 in the 1α,25(OH)2D3 group (Fig. 3A–G). Of these miRNA, 44 were uniquely differentially expressed in the vehicle group and 31 in the 1α,25(OH)2D3 group (Fig. 3D, F). The 167 MV miRNA that were shared between the1α,25(OH)2D3 and vehicle treatment groups were found to be selectively exported in identical fashion between vehicle and 1α,25(OH)2D3 groups (same set of miRNA enriched in the MV as compared to cell samples between vehicle and 1α,25(OH)2D3 groups) (Fig. 3E). Comparing MV samples from 1α,25(OH)2D3 and vehicle groups directly resulted in 18 differentially expressed miRNA (Fig. 3G). The differentially expressed miRNAs from these comparisons (cell vehicle to MV vehicle unique, cell 1α,25(OH)2D3 to MV 1α,25(OH)2D3 unique, cell vehicle to cell 1α,25(OH)2D3, and MV vehicle to MV 1α,25(OH)2D3) were selected for pathway analysis (Table 1). The 167 miRNAs shared between vehicle and 1α,25(OH)2D3 groups when comparing cells to MVs were not included.

Fig. 3
figure 3

Differentially expressed microRNA (p value < 0.05 & absolute log twofold change > 1) being exported by chondrocytes into MVs with either vehicle or 1α,25(OH)2D3 treatment. Only differentially expressed miRNA are shown. A The differentially expressed miRNA between cell & MV with vehicle treatment. B Venn diagram displaying overlap of differentially expressed microRNA depending on treatment (vehicle or 1α,25(OH)2D3) with 167 miRNA found to be differentially exported under both conditions. C The differentially expressed miRNA between cell & MV with 1α,25(OH)2D3 treatment. D 44 uniquely differentially expressed miRNA between cell & MV with vehicle treatment that are not differentially expressed when treating with 1α,25(OH)2D3. E 167 differentially expressed miRNA that are shared in common between vehicle and 1α,25(OH)2D3 treatments. F 31 uniquely differentially expressed miRNA between cell and MV with 1α,25(OH)2D3 that are not differentially expressed when treating with vehicle. G Differentially expressed miRNA when comparing the two MV populations produced by chondrocytes treated with vehicle or 1α,25(OH)2D3

Full size image

miEAA analysis of the differentially expressed miRNA (between cell vehicle and cell 1α,25(OH)2D3, MV vehicle and MV 1α,25(OH)2D3, miRNA uniquely packaged following 1α,25(OH)2D3 treatment, and miRNA uniquely packaged following vehicle treatment), resulted in large numbers of potentially impacted pathways, targeted mRNA, related diseases and tissue locations for three of the four comparisons (no results were returned for the first comparison) (Fig. 4A–C). Searching the initial results, which included information pertaining to the entire organism, for musculoskeletal specific terms (Supplemental Table 1) significantly reduced the pathways and mRNA involved (Fig. 4D–F).

Fig. 4
figure 4

Visualization of miEAA search results using only the differentially expressed miRNA and the number of pathways, mRNAs, tissues, and diseases/drugs they are known to impact. A, B and C. Unfiltered results. A 18 differentially expressed miRNA between MV groups with vehicle or 1α,25(OH)2D3 treatment. B The 31 uniquely differentially expressed miRNA from cell to MV with 1α,25(OH)2D3 treatment. C The 44 uniquely differentially expressed miRNA from cell to MV with vehicle treatment. D, E and F. Results filtered for musculoskeletal search terms. D Between MV with vehicle or 1α,25(OH)2D3. E Exported from cell to MV with 1α,25(OH)2D3 treatment. F Exported from cell to MV with vehicle treatment

Full size image

Examining the miRNA exported in MVs with a p-value < 0.05 and a log twofold change > 1 that are under the regulation of 1α,25(OH)2D3 treatment on MEINTURNET returned hundreds of genes. When investigating the first 10 genes with a minimum of 2 miRNA interactions sorted by ascending p-value, the majority (9) of genes in the 1α,25(OH)2D3 group have known musculoskeletal activity while only half of the genes in the vehicle treatment group do (Table 2).

Table 2 First ten genes returned by MIENTURNET target-scan search for miRNA exported in MVs with and without 1α,25(OH)2D3 treatment along with details about known musculoskeletal activity associated with each gene
Full size table

1α,25(OH)2D3 Does Not Cause miRNA Release from MVs

Treatment of MVs with Tween 20 followed by RNase demonstrated that loss of membrane integrity and protein denaturing exposed smallRNA to enzymic degradation, as demonstrated in RNA pseudo gels from a BioAnalyzer. Examination of the smallRNA region of the pseudo gel showed that the RNA bands for the tween 20 + RNase group had clear degradation when compared to negative control (0.9% NaCl), 1α,25(OH)2D3, and 1α,25(OH)2D3 + RNase groups with an arrow highlighting the smallRNA band that has both shifted down and lost definition in the tween 20 + RNase group, lanes 1 through 3 (Fig. 5A). Quantifying the smallRNA region of the gels using Licor’s Image Studio Lite confirmed that there was a significant reduction in RNA for the tween 20 + RNase group when compared to all other groups (Fig. 5B and Supplemental Fig. 1).

Fig. 5
figure 5

MicroRNA release from MVs with 1α,25(OH)2D3 treatment. A Bioanalyzer gel results from MV samples. Groups with an arrow highlighting the smallRNA band that has both shifted down and lost definition B Quantification of small RNA signal region. Mean ± SE, groups not sharing a letter are statistically different (α = 0.05). NC negative control, no treatment

Full size image

1α,25(OH)2D3 Promotes MV Endocytosis

The PKH26 dye was able to successfully stain GC cells in all treatment groups, visible at 20X (Fig. 6A). Manual and automated quantification produced results with the same statistical trends with MV + 1α,25(OH)2D3 staining more cells than NC + 1α,25(OH)2D3 (Fig. 6B, C). When increasing the threshold for a stained cell up to 30 distinct spots on the automated quantification data, MV with 1α,25(OH)2D3 was greater than both MV + vehicle and NC + 1α,25(OH)2D3 (Fig. 6D). High-resolution view of stained MV + 1α,25(OH)2D3 cell with a cloud of distinct green spots (Fig. 6E).

Fig. 6
figure 6

MV Endocytosis. A Representative images from each of the four groups: PKH26 stained MVs with 1α,25(OH)2D3, PKH26 stained MVs with vehicle, PKH26 stained saline with 1α,25(OH)2D3 & PKH26 stained saline with vehicle (scale bar = 30 µm). B Percent of cells stained with 6 + PKH26 spots from human counting. C Percent of cells stained with 6 + PKH26 spots using cell profiler pathway. D Percent of cells stained with 30 + PKH26 spots using cell profiler pathway. E High magnification images of chondrocytes stained treated with PKH26 stained MVs with 1α,25(OH)2D3 (scale bar = 50 µm, large image, and 10 µm magnified images). Mean ± SE, groups not sharing a letter are statistically different (α = 0.05). NC NaCl treated with PKH26. MV MVs treated with PKH26

Full size image

Discussion

1α,25(OH)2D3 has been known to have a strong regulatory impact on the growth plate and this study further expands on its regulatory capabilities with respect to GC cells, the production of GC MVs, and the fate of MVs in the ECM. Recent work in our lab has also demonstrated that selectively exported miRNAs found in MVs are able to regulate growth plate chondrocytes [9]. In this study we have demonstrated that miRNA was packaged into MVs under the regulatory influence of 1α,25(OH)2D3. These packaged miRNA are predicted by various pathway and target analysis tools to have a wide-ranging impact on cartilage and bone tissues. In addition to the regulatory role with respect to miRNA production and packaging, 1α,25(OH)2D3 was demonstrated to play a role in the endocytosis of MVs by GC cells.

Characterization of the GC MVs found them to be in accordance with previous studies with the two primary size peaks at 98 and 143 nm and the 50 to 150 nm size range encompassing 64.4% of the total particles in solution. The MVs had significantly increased alkaline phosphatase specific activity when compared to the CP and PM isolations. The only unexpected result was the low to indistinguishable band for annexin V in the western blot of the MV sample, although this could be due to its role in binding to collagen within the matrix and subsequent shearing off during MV release from the ECM using trypsin and a cell scrapper. The vesicle markers CD 36, CD 81, and Alix were all concentrated in the MV sample with Alix having a slightly dimmer band in the CP sample. Alix also has numerous established roles within cells including the involvement in miRNA packaging into extracellular vesicle biogenesis and may play a similar role for MVs [61, 62]. Na+/K+ ATPase enzyme is an established component of MVs and located in the plasma membrane resulting in it being found in the CP, PM, and MV samples [4].

The role of 1α,25(OH)2D3 in modulating miRNA within the growth plate appears to be primarily through the coordinated production of microRNA by GC cells followed by the selective packaging of miRNA into MVs. The lack of differential expression noticeable when directly comparing the cell miRNA population between vehicle and 1α,25(OH)2D3 treatments highlights just how targeted the regulation of 1α,25(OH)2D3 is as miRNA are differentially expressed when comparing the MV populations directly and when examining the miRNA being packaged into MVs by chondrocytes. To only discover one miRNA differentially expressed between the cell populations indicates that these miRNA are being produced and packaged into MVs efficiently.

1α,25(OH)2D3 is likely acting through both genomic and nongenomic avenues within GC chondrocytes to achieve this regulatory effect. Chondrocytes undergo maturation in the growth plate, and new GC cells continuously exit the proliferating cell zone. Altering the MV miRNA population enables 1α,25(OH)2D3 to modulate a wide array of pathways in the growth plate well downstream of the current location. Identification of impacted pathways via bioinformatic analysis can point to mechanisms, but further validation is needed in order to determine which ones are being regulated and in which regions of the growth plate. Depending on growth plate stage, this regulation may well be exerted on bone forming cells during vascular invasion and matrix turnover as a defining characteristic of MVs is binding to collagen in the ECM.

A total of 242 miRNA were found to be differentially expressed between the MV and their parent cells in both the 1α,25(OH)2D3 and vehicle groups. Even after taking into account the 167 shared miRNA, there were 31 and 44, respectively, groups of unique miRNA found to be under the control of 1α,25(OH)2D3.

When examining the pathways that are potentially impacted by the set of differentially expressed miRNA using miEAA analysis we found thousands of pathways that we were able to trim down to hundreds of pathways using a manually curated set of musculoskeletal specific search terms. This resulted in a set of miRNA with associated pathways and/or mRNA, another set of miRNA that were just associated on a tissue level and had no pathways or mRNA associated with them, and a final set of microRNA that were differentially expressed in our study but have no known associations with musculoskeletal terms. All three groups of miRNA are clear avenues of further investigation in order to understand what pathways they are in fact modulating in chondrocytes and potentially bone forming cells. The suggested pathways and mRNA are also valid areas of further investigation. The ability to narrow down the miEAA results from thousands of pathways to dozens is a good first step towards fully understanding what these specific miRNA may be doing in the growth plate.

Using the MIENTURNET tool to examine predicted miRNA-target genes, we discovered that 9/10 of the genes examined in the 1α,25(OH)2D3 group had known cartilage activity with a focus on arthritis and matrix turnover. Genes in the 1α,25(OH)2D3 group would result in both an increase in certain matrix degrading proteases and an increase in ECM component production following a down regulation of these specific genes. ECM turnover is an important and well-regulated process within the growth plate and one that MVs could be playing an even greater role in. Only 5/10 of the genes from the vehicle group were related to musculoskeletal research, and their primary functions were associated with arthritis and initial limb development.

The MV population is a heterogeneous one indicating the likelihood that only a subset of MVs are involved in the transport of regulatory miRNA. Whether these MVs are distinct from the MVs involved in tissue mineralization was not addressed in this study. Transmission electron micrographs of matrix vesicles in the growth plate demonstrate initial calcium phosphate crystals along the inner leaflet of the MV membrane in some, but not all [63]. There is also variability in size, which we noted as well in our NanoSight analysis. Moreover, there is a broad array of miRNA populations, suggesting that there may be subsets of miRNA within individual MVs.

These miRNA populations have a broad range of tissue specific effects that they can exert on cells. How the miRNA makes its way back into cells for this task remains unclear. It could occur as part of matrix turnover within the growth plate either during the maturation process or when the mineralized tissue is being converted over to bone. Since MVs containing microRNA are being produced by both RC and GC chondrocytes there are multiple points at which the MV microRNA may be released. Additionally, the MVs could be triggered by specific compounds to destabilize. The various populations of MVs and the process by which they release their cargo is a vital area of additional study.

1α,25(OH)2D3 has been demonstrated to release the content of MVs by destabilizing the membrane, releasing matrix processing enzymes into the ECM. In previous studies, we showed that miRNA was present in MV lysates and could be digested with RNase [15], suggesting that they would be released by the membrane destabilizing action of 1α,25(OH)2D3. However, in the present study, we observed that MVs are able to safeguard the miRNA cargo. Treatment of isolated MVs with 1α,25(OH)2D3 plus RNase did not result in increased RNA degradation indicating that the miRNA may be bound to protective carrier proteins within the MVs, bound to the inner leaflet of the MVs, or the MV population may be heterogeneous with some MVs reacting to 1α,25(OH)2D3 and releasing their contents into the matrix and others remaining intact with their miRNA cargo. The ability of the RNA to remain safe from degradation following 1α,25(OH)2D3 treatment further bolsters the notion of the miRNA being actual regulatory molecules that are able to return to the cells.

Examining MV endocytosis by chondrocytes following fluorescent staining, we observed an apparent increase in cells with distinct spots of PKH26 staining in the MV group as compared to the negative control. This was not statistically significant in the quantification of the images as there was a high degree of variability in the results and some background PKH26 staining in all groups. The background staining did not show up in any control images taken without PKH26 stain and so is not likely to be autofluorescence. Both the NC and MVs had a visible red pellet that was resuspended for the cell treatment, likely exposing all cells to a mix of bound and unbound PKH26. Results between the automated and manual image counting approaches using 6 spots as a threshold were statistically identical and since the automated approach was developed independently of the manual counting results we felt confident in being able to adjust the threshold value in the automated scoring system. We ran the automated approach with an increase threshold value set at 30 to see if certain groups had more intensely clustered cells. This resulted in greater reductions for the NC + vehicle, NC + 1α,25(OH)2D3 and MV + vehicle groups compared to MV + 1α,25(OH)2D3. Treatment with 1α,25(OH)2D3 may make the cells and/or a specific subpopulation of the MVs more available for endocytosis and as a result increasing the percent of cells with high concentrations of stained MVs. Alternatively, the MVs may be rupturing as a result of 1α,25(OH)2D3 treatment, increasing the number of distinct spots observed on the cells. A high-resolution view of the cells demonstrates the distinct spots of PKH26 staining visible in the area around the counterstained nuclei that was consistent throughout the images and treatment groups and at varying focal planes.

With the high-resolution images that involved stained MVs, we were able to see what appeared to be MVs or vesicle fragments (the miRNA may be bound to the interior of the lipid bilayer) adhering to the cell membrane. The stained MVs were found a distance away from the stained nucleus and appeared to be bound to or passing through the cell membrane. Regardless of the focal plane we typically didn’t see significant PKH26 staining overlaying the nucleus with the vast majority of points being found in clusters around the nucleus.

Our results were not able to resolve the question of how the miRNA gain access to the chondrocytes to perform their regulatory role, although we were able to demonstrate that the miRNAs packaged within MVs in the ECM are protected from RNase even with 1α,25(OH)2D3 treatment. The apparent protection of the miRNA provided by the MVs, gives additional emphasis to the intentional packaging of miRNA in this format.

We have now demonstrated that 1α,25(OH)2D3 regulates the production and packaging of miRNA into MVs by GC chondrocytes. 1α,25(OH)2D3 not only regulates the populations of miRNA found in MVs, but also the populations of miRNA found in their parent cells. In both cases, the miRNA populations are associated with musculoskeletal related pathways and mRNAs.

Our understanding of 1α,25(OH)2D3’s role continues to become more complex, as it is integral to the overall functioning of the growth plate. The cellular population of miRNAs experienced virtually no observable impact as a result of 1α,25(OH)2D3 treatment, indicating that the effect of the vitamin D metabolite on the miRNA is primarily on their packaging into MVs and release into the ECM. The subsequent role of 1α,25(OH)2D3 is directly on the MVs in the ECM. Future studies are needed to resolve how MVs protect their miRNA payload following breakdown of the MV membrane in response to 1α,25(OH)2D3, how 1α,25(OH)2D3 mediates the uptake of MV miRNA, and how MV heterogeneity contributes to the overall physiology of the growth plate.