Cell lines, primary cell cultures, primary tumor samples

Canine OS cell lines OSA8 and OSA16 [5] were maintained in RPMI-1640 (Gibco Life Technologies, Grand Island, NY, USA) supplemented with 10 % fetal bovine serum, non-essential amino acids, sodium pyruvate, penicillin, streptomycin, L-glutamine, and HEPES (4-(2-dydroxethyl)-1-piperazineethanesulfonic acid) at 37 °C, supplemented with 5 % CO₂ (media supplements from Gibco). Normal canine osteoblasts (catalog no. Cn406-05) Cell Applications Inc, San Diego, CA, USA) were cultured in canine osteoblast medium (Cell Applications Inc, catalog no. Cn417-500).

Primary canine osteoblast cultures were generated from trabecular bone isolated from the femoral heads of dogs undergoing total hip arthroplasty or femoral head ostectomy at the Ohio State University Veterinary Medical Center (OSU-VMC) as previously described [39]. Briefly, femoral heads were washed in buffered saline and trabecular bone was curetted to remove bone chips. Bone chips were washed and digested in serum-free Dulbecco’s modified Eagle medium (DMEM)/F12K medium (Gibco) supplemented with 239 U/mL collagenase type XI (Sigma, St. Louis, MO, USA), 2 mM L-glutamine, 50 μg/mL pencillin-streptomycin and transferred to a spinner flask in a humidified incubator at 37 °C with 5 % CO2 for 3–4 h. Following digestion of cellular material, the bone fragments were washed with buffered saline and plated into T25 flasks in calcium-free DMEM/F12 medium supplemented with 10 % fetal bovine serum, 50 μg/mL ascorbate (Sigma), 50 μg/mL pencillin-streptomycin, and 2 mM L-glutamine with changes of medium every 3–4 days. Osteogenic induction of confluent monolayer cultures was accomplished using DMEM/F12 (Gibco) medium supplemented with 10 % fetal bovine serum, 0.1 μM dexamethasone (Sigma), 10 mM β-glycerophosphate (Sigma), 50 μg/mL ascorbate (Sigma), 50 μg/mL pencillin-streptomycin, and 2 mM L-glutamine for 21 d with medium changes every 3–4 days [40]. Control cultures were maintained without osteogenic supplements. Cultures were evaluated for alkaline phosphatase expression using the Leukocyte Alkaline Phosphatase Kit (Sigma) according to the manufacturer’s instructions. The protocol for generation of canine osteoblasts was approved by the OSU Institutional Animal Care and Use Committee (IACUC, protocol 2009A0184). Normal canine tissue collections were approved by the OSU IACUC (protocol 2010A0015). Fresh frozen canine OS tumor samples were obtained from dogs presenting to the OSU-VMC and from Dr. Jaime Modiano at the University of Minnesota (UMN) Veterinary Medical Center. Tumor sample collections were performed in accordance with established hospital protocols and approved by the respective IACUCs at both OSU and UMN. Clinical patient data, including age, sex, breed, histopathological diagnosis, and primary tumor location is detailed in Additional file 1: Table S1.

RNA isolation, cDNA synthesis, RT-PCR and quantitative real-time PCR

Real-time PCR was performed using the Applied Biosystems StepOne Plus Detection System (Applied Biosystems, Foster City, CA, USA). Human Taqman miRNA assays (Applied Biosystems) were used according to manufacturer’s instructions to quantify mature miRNA levels in available canine cell lines and tissues (miR-1, miR-9, miR-10b, miR-29a, miR-122, miR-126, miR-199b, miR-200c, miR-451; all mature miRNAs share 100 % sequence homology between dogs and humans). MiRNA-specific primers were used to convert 50 ng total RNA to first-strand cDNA, followed by real-time PCR with TaqMan probes. All samples were normalized to U6 snRNA. To validate changes in mRNA expression for selected genes affected by miR-9 expression, total RNA was collected and cDNA was generated as described above. Canine GSN and TGFBI mRNA was detected using Fast SYBR green PCR master mix (Applied Biosystems) according to the manufacturer’s protocol and primer sets are detailed in Table 1. Normalization was performed relative to 18S rRNA. All reactions were performed in triplicate and included no-template controls for each gene. Relative gene expression for all real-time PCR data was calculated using the comparative threshold cycle method [41]. Experiments were repeated 3 times using samples in triplicate.

Quantitative real-time PCR analysis of formalin-fixed paraffin-embedded canine primary osteosarcoma tumor samples

Formalin-fixed paraffin-embedded (FFPE) primary canine OS tissues were obtained from the OSU-VMC Biospecimen Repository. H&E stained sections from a single random block from each patient were reviewed by a pathologist (OHI) to define and mark representative OS tumor regions. Using the marked H&E stained glass slide as a map, the corresponding areas of the unstained FFPE tissue block were identified and 15 targeted core samples of each cancerous tissue region were obtained. Tumor cores were then processed and RNA was isolated using the RecoverAll™ Total Nucleic Acid Isolation Kit for FFPE (Applied Biosystems) according to manufacturer’s recommendations. To quantify miR-9 expression, cDNA was generated and real-time PCR was performed using Human Taqman miRNA assays (Applied Biosystems) as described above.

MiRNA expression profiling

MiRNA expression profiling of normal canine tissues (brain cortex, liver, lymph node, kidney, skeletal muscle, spleen, thyroid), 72 fresh primary OS tumors, 2 primary osteoblast cultures, and canine osteoblast cells (Cell Applications) was performed at the OSU Comprehensive Cancer Center Genomics Shared Resource using the multiplexed nanoString nCounter miRNA system (nanoString Technologies, Seattle, WA, USA) according to manufacturer’s protocol [42]. Total RNA (100 ng) was used as input material. Small RNA samples were prepared by ligating a specific DNA tag onto the 3’ end of each mature miRNA according to manufacturer’s instruction (nanoString Technologies). These tags normalized the melting temperatures of the miRNAs and provided identification for each miRNA species in the sample. Excess tags were then removed and the resulting material was hybridized with an nCounter Human (V2) miRNA Expression Assay CodeSet containing a panel of miRNA:tag-specific nCounter capture and barcoded reporter probes. Hybridization reactions were incubated at 65 °C overnight. Hybridized probes were purified and immobilized on a streptavidin-coated cartridge using the nCounter Prep Station (nanoString Technologies). nCounter Digital Analyzer was used to count individual fluorescent barcodes and quantify target RNA molecules present in each sample. For each assay, a high-density scan (600 fields of view) was performed.

NanoString data analysis

Abundances of miRNAs were quantified using the nanoString nCounter gene-expression system [42]. Boxplot analysis did not detect obvious batch effect or poor sample integrity; therefore, all data were used for analysis. Raw data was normalized using internal positive control probes included in each assay and then a filtering step was applied. Internal negative control probes were used to determine a background threshold (2 standard deviations above the mean negative control probe count value) and if more than 90 % of the samples had miRNA expression lower than the background threshold cutoff value, those miRNAs were filtered out. After data filtering, a total of 519 miRNAs were used for analysis. Filtered data was quantile normalized and linear regressions were used to compare miRNA expression between tumor samples and normal osteoblast samples. A p-value of 1/519 = 0.0019 was used as a cutoff to claim for significance if controlling 1 false positive among the 519 tested miRNAs. Differential miRNA expression was determined by one-way analysis of variance (ANOVA) and p-values of <0.0019 were considered statistically significant.

miR-9, anti-miR-9, and shGSN lentivirus infection

Lentiviral constructs obtained from Systems Biosciences (Mountain View, CA, USA) were packaged using the pPACKH1 Lentivector Packaging KIT (catalog no. LV500A-1) according to manufacturer’s instructions. Canine osteoblast cells (Cell Applications Inc.) and OSA16 cells (5 × 10⁵) were transduced with negative control empty lentivirus (catalog no. CD511B-1) or pre-miR-9-3 lentivirus (catalog no. PMIRH9-3PA-1). For reciprocal knock-down experiments, canine OSA8 cells (5 × 10⁵) were transduced with pGreenPuro Scramble Hairpin Control lentivirus (catalog no. MZIP000-PA-1) or miRZip-9 anti-miR-9 lentivirus (catalog no. MZIP9-PA-1). Briefly, 5 × 10⁵ cells were plated and left overnight in 10 % serum-containing medium. The following day, the medium was changed to Stemline (Gibco) with transfection agent TransDux (Systems Biosciences) and either empty control or pre-miR-9-3 virus (osteoblasts) or miRZip-9 or negative control scrambled virus (OSA8) was added to cells according to manufacturer’s protocol. FACS-mediated cell sorting based on GFP expression was performed 72 h post-transduction and miR-9 expression was evaluated by real-time PCR (Applied Biosystems).

Stable knock down of gelsolin (GSN) was performed using short hairpin RNA (shRNA) lentiviral constructs (pLKO.1:Hygro-shScramble and pLKO.1:Hygro-shGSN) and high-titer lentiviral stocks were generated as described in the Addgene's pLKO.1 protocol. pLKO.1:hygro plasmid was a kind gift from Bob Weinberg (Addgene plasmid #24150). Briefly, 1 × 10⁵ OSA8 cells were plated and left overnight in 10 % serum-containing medium. The following day, the medium was replaced with serum-free medium and target cells were infected with transfection agent TransDux (Systems Biosciences) and either pLKO.1:Hygro-Scramble or pLKO.1:Hygro-shGSN virus or TransDux alone for 24 h. Cells were cultured for 7–10 days in 10 %-serum-containing medium supplemented with 75 ug/mL Hygromycin-B (Life Technologies) for plasmid selection. Knockdown of GSN was confirmed by quantitative real-time PCR and Western blotting Cells were collected and processed for Western blotting as described below to detect levels of GSN and efficiency of knock down. Sequences of template canine DNA were as follows: pLKO.1-shGSN.1 (5’-CCCGCTGTTCAAGCAGTTCTT-3’) and pLKO.1-shGSN.2 (5’-CTGCAGTATGACCTCCACTAC-3’).

Matrigel invasion assay

To assess the effects of miR-9 and GSN on invasion, cell culture inserts (8-μm pore size; Falcon) were coated with 100 μL of diluted (1:10) Matrigel (BD Bioscience, San Jose, CA, USA) to form a thin continuous layer and allowed to solidify at 37 °C for 1 h. Canine osteoblasts, OSA8 and OSA16 cell lines (5 × 10⁴/mL) transduced with control lentivirus, pre-miR-9-3 lentivirus, miRZip-9 lentivirus, or shGSN lentivirus were prepared in serum-free medium and seeded into each insert (upper chamber) and medium containing 10 % fetal bovine serum was placed in the lower chamber. The cells were incubated for 24 h to permit invasion through the Matrigel layer. Cells remaining on the upper surface of the insert membrane were wiped away using a cotton swab, and cells that had migrated to the lower surface were stained with crystal violet and counted in ten independent 20× high powered fields for each Matrigel insert. Experiments were repeated 3 times using samples in triplicate.

Wound healing assay

To evaluate the effects of miR-9 on cell migration, canine osteoblasts transduced with control lentivirus or pre-miR-9-3 lentivirus and OSA8 cells transduced with scrambled control lentivirus or miRZip-9 lentivirus were seeded in complete medium and grown until confluent in 6-well plates. A gap was introduced in the cells by scraping with a P200 pipette tip and cells were placed in fresh medium containing 10 % fetal bovine serum. After 20 h (OSA8) or 24 h (osteoblasts), migration across the gap was evaluated by digital photography. Each experiment was repeated 3 times.

Cell proliferation

Canine osteoblasts and OSA16 cells (2.5 × 10³) transduced with control lentivirus or pre-miR-9-3 lentivirus were seeded in triplicate in 96-well plates; non-transduced cells served as negative controls. After 24, 48, or 72 h of culture, media was removed and plates were frozen at −80 °C overnight before processing with the CyQUANT® Cell Proliferation Assay KIT (Molecular Probes, Eugene, OR, USA) according to the manufacturer’s instructions. Fluorescence was measured using a SpectraMax microplate reader (Molecular Devices, Sunnyvale, CA, USA). Cell proliferation was calculated as a percentage of non-transduced control cells. Each experiment was repeated 3 times.

Detection of Apoptosis/Caspase 3/7 activity

Induction of apoptosis was assessed using the Senso-Lyte® Homogeneous AMC Caspase- 3/7 Assay KIT (Anaspec Inc., San Jose, CA, USA) as previously described [43]. Canine osteoblasts and OSA16 cells (2.5 × 10³) transduced with either empty lentivirus or pre-miR-9-3 lentivirus were plated in triplicate in 96-well plates for 24 and 48 h prior to analysis. Fluorescence was measured on a SpectraMax microplate reader (Molecular Devices) and caspase 3/7 activity was reported after subtraction of background fluorescence elicited by medium alone. Each experiment was repeated 3 times.

2D-DIGE and protein identification by LC-MS/MS

Protein lysates prepared from canine osteoblast cells transduced with either empty lentivirus (n = 4) or pre-miR-9-3 lentivirus (n = 4) were purified using the 2-D Clean-Up Kit (GE Healthcare, Uppsala, Sweden). Samples were suspended in 100 μL of lysis buffer (30 M Tris pH 8.5, 7 M Urea, 2 M Thiourea, 4 % CHAPS) and quantitated by Bradford assay. Two-dimensional difference gel electrophoresis (2D-DIGE) was performed as previously described [44]. Briefly, internal control samples were prepared by mixing a portion of all individual samples. Pooled internal control standards (50 μg) were labeled with Cy2 dye and individual samples (50 μg) were labeled with the appropriate Cy3 or Cy5 dye (GE Healthcare) according to the manufacturer's instructions, mixed and separated by two-dimensional difference gel electrophoresis. The first dimension separation was achieved using IPG isoelectric focusing strips (24 cm length, pH 3–10; GE Healthcare). The second dimension separation was achieved by SDS-PAGE on large-format gels (20x24cm) using a Dalt 12 electrophoresis system (GE Healthcare). Gels were scanned using a Typhoon 9400 variable mode scanner (GE Healthcare) at appropriate wavelengths. Gel images were analyzed and relative protein abundance was performed using SameSpots software (TotalLabs). Background subtraction, quantification and normalization were automatically applied with low experimental variation. The student’s t-test was used to compare protein expression for each spot between empty vector and miR-9-transduced osteoblast samples and p-values of < 0.05 were considered significant.

Protein spots of interest were located and excised from separate preparative gels using the Ettan Spot Handling Workstation (GE Healtchare) according to manufacturer’s instructions. Gel pieces containing protein spots were cored, digested with trypsin (Promega, Madison, WI, USA), and subjected to capillary-liquid chromatography-nanospray tandem mass spectrometry (Nano-LC/MS/MS) using a Thermo Finnigan LTQ mass spectrometer equipped with a nanospray ion source. The LC system used was an UltiMate™ Plus from LC-Packings A Dionex Co (Sunnyvale, CA, USA) with a Famos autosampler and Switchos column switcher. Mascot Daemon software (version 2.2; Matrix Science, Boston, MA, USA) was used to search for the mass of the peptide ion against other mammalian proteins in the NCBI database (1,391,110 sequences). Protein identifications were checked manually and proteins with a Mascot score of 100 or higher with a minimum of two unique peptides from one protein having a -b or -y ion sequence tag of five residues or better were accepted.

RNA Sequencing

Total RNA was extracted from canine osteoblast cells transduced with either empty lentivirus (n = 3) or pre-miR-9-3 lentivirus (n = 3) using the TRIzol method and RNA sequencing was performed at the OSU Comprehensive Cancer Center Genomics Shared Resource. Briefly, total RNA was treated by Ribo-Zero Gold Subtraction reagents from TruSeq stranded total RNA (Illumina; RS-122-2201) to remove cytoplasmic and mitochondrial rRNAs and for the subsequent construction of the RNA-seq library according to the manufacturer’s instructions. Stranded total transcriptome libraries were quantified and qualified by Qubit and RIN analysis, respectively. Sequencing was performed on an Illumina HiSEq. 2500 instrument at a depth of ∼ 40 million paired-end, 100 bp long, strand-specific reads per sample. AdapterRemover was used to trim adapter sequences and the remaining reads were aligned using STAR to the canFam3 genome. Quality control was done using RNA-SeQC, in order to check for samples needing additional sequencing reads. Differential expression was determined using two different pipelines. First, gene expression counts were generated by HTSeq-Count and fed into DESeq2, which compared several sample groups using a likelihood-ratio test (LRT). Second, expression of individual transcripts was quantified by cuffquant with comparisons between pairs of conditions performed by cuffdiff.

Statistical analysis relative to mRNA expression data was performed using CuffDiff Software. Differential gene expression was determined by student’s t-test and p-values of <0.05 were considered statistically significant. Prediction of miR-9 binding to the 3’-UTR of genes down-regulated by miR-9 was performed with computer-aided algorithms obtained from TargetScan (http://www.targetscan.org), PicTar (http://pictar.mdc-berlin.de), miRanda (http://www.microrna.org), and miRWalk (http://www.umm.uni-heidelberg.de/apps/zmf/mirwalk).

Immunoblotting

Protein lysates from canine osteoblasts transduced with control or pre-miR-9-3 lentivirus and OSA8 cells stably expressing scramble or miRZip-9 lentiviral constructs were prepared and quantified, separated by SDS-PAGE, and western blotting was performed as previously described [43]. The membranes were incubated overnight with anti-gelsolin antibody (D9W8Y, catalog no. 12953, Cell Signaling Technology, Danvers, MA) then incubated with the appropriate horseradish peroxidase linked secondary antibodies, washed, and exposed to substrate (SuperSignal West Dura Extended Duration Substrate, Pierce, Rockford, IL). Blots were stripped, washed, and reprobed for β-actin (Santa Cruz Biotechnology, Santa Cruz, CA).

Statistics

Whenever possible, experiments were performed in triplicate and repeated 3 times. Data were presented as mean plus or minus standard deviation. Real time PCR miRNA or gene expression data was first normalized to internal control (U6 snRNA and 18S, respectively) and the delta delta Ct method [41] was used to compare miRNA expression by one-way ANOVA. For analysis of invasion assay data, a linear mixed effects model was used to take account of the correlations among observations run in the same biological replicate. Group comparisons in the CyQUANT® proliferation assays, caspase 3/7 activity, and invasion assays were analyzed by ANOVA. Values of p < 0.05 were considered statistically significant.

A unique miRNA expression signature is associated with primary canine OS

To characterize miRNA expression in canine OS and evaluate the role of miRNA dysregulation in OS pathogenesis, it was first necessary to validate a method to analyze differential expression. The Human (V2) miRNA Expression Assay CodeSet was used to profile a panel of seven normal canine tissues (brain cortex, liver, lymph node, kidney, skeletal muscle, spleen, thyroid; N = 3 per tissue). Because of the high degree of sequence conservation of miRNAs across humans and dogs, 168 miRNAs on the human panel have sequences identical to canine miRNAs (miRBase v.15). A high abundance of tissue-specific miRNAs (see Additional file 2: Figure S1) were detected and real-time PCR was used to validate this finding (see Additional file 3: Figure S2). These miRNA abundance data from normal canine tissues establish that the nanoString nCounter platform is a valid high-throughput methodology to study miRNA expression in canine samples.

To generate normal cells for comparison to osteosarcoma cells, primary osteoblast cultures were differentiated in vitro from canine mesenchymal stem cells. These cells were confirmed to express alkaline phosphatase and other bone-specific markers (Additional file 4: Figure S3). Global miRNA expression in fresh primary canine OS tumor samples (N = 72), primary osteoblast cultures (N = 2), and normal osteoblast cells (commercially available, N = 1) was evaluated using the nanoString nCounter platform. A distinct miRNA expression signature composed of 70 differentially expressed miRNAs was identified in primary canine OS tumor samples compared to canine osteoblast cells or primary osteoblast cultures (Fig. 1). We found 26 miRNAs that were significantly overexpressed in canine OS tumor samples compared to canine osteoblast cells or primary osteoblast cultures, while 44 miRNAs were downregulated in OS tumor tissues (Table 2). To validate these findings, real-time PCR confirmed differential expression for 4 of the 70 significant OS miRNAs among a random sampling of 16 fresh OS specimens and 5 control osteoblast cultures (Fig. 2). Specifically, we verified the overexpression of miR-126, miR-199b, miR-451 in OS samples relative to normal osteoblasts. In contrast, miR-29a showed significant down-regulation in primary canine OS tumor tissues compared to normal osteoblasts. Furthermore, several differentially expressed miRNAs identified in our canine OS tumor samples show similar alterations in human OS tumors (Table 2, bolded).

miRNA	Fold-change expression OS vs Ob	p-value	miRNA	Fold-change expression OS vs Ob	p-value	miRNA	Fold-change expression OS vs Ob	p-value
Upregulated miRNAs			Downregulated miRNAs			Downregulated miRNAs
hsa-miR-126	22.8	1.13E-13	hsa-miR-600	−2.2	1.75E-08	hsa-miR-1277	−1.6	7.94E-05
hsa-miR-450a	7.2	1.55E-08	hsa-miR-22	−4.1	3.37E-08	hsa-miR-98	−2.3	0.00011
hsa-miR-451	93.1	1.50E-07	hsa-miR-34a	−6.4	5.39E-08	hsa-miR-539	−1.7	0.00011
hsa-miR-30e	3.5	9.21E-06	hsa-miR-640	−1.8	2.27E-07	hsa-miR-1	−7.4	0.00011
hsa-miR-423-3p	3.0	1.93E-05	hsa-miR-655	−2.0	3.93E-07	hsa-miR-664	−1.7	0.00013
hsa-miR-532-3p	3.3	2.21E-05	hsa-miR-518f	−1.8	5.51E-07	hsa-miR-770-5p	−2.2	0.00015
hsa-miR-92a	2.6	2.89E-05	hsa-miR-143	−5.5	6.43E-07	hsa-miR-29a	−2.7	0.00020
hsa-miR-25	2.6	3.13E-05	hsa-miR-145	−5.4	6.79E-07	hsa-miR-365	−2.0	0.00020
hsa-miR-142-3p	6.8	6.52E-05	hsa-miR-1279	−2.1	1.48E-06	ebv-miR-BART11-3p	−1.9	0.00024
hsa-miR-497	4.5	7.69E-05	hsa-miR-200a	−2.3	4.35E-06	hsa-miR-133a	−2.1	0.00026
hsa-miR-499-5p	2.0	8.22E-05	hsa-miR-1283	−2.1	5.00E-06	hsa-miR-617	−1.6	0.00035
hsa-miR-181c	2.8	0.00013	hsa-miR-744	−2.7	6.23E-06	hsa-miR-1262	−1.6	0.00036
hsa-miR-342-3p	2.2	0.00013	hsa-miR-193b	−2.2	8.09E-06	hsa-miR-1274b	−5.2	0.00049
hsa-miR-362-5p	2.7	0.00015	hsa-miR-490-5p	−1.9	1.47E-05	hsa-miR-943	−1.6	0.00051
hsa-miR-144	20.8	0.00018	hsa-miR-1275	−2.5	2.16E-05	hsa-miR-1255a	−1.6	0.0012
hsa-miR-195	3.7	0.00026	hsa-miR-193a-3p	−2.2	2.18E-05	hsa-miR-579	−1.9	0.0013
hsa-miR-16	2.3	0.00037	hsa-miR-148b	−1.9	2.28E-05	hsa-miR-892a	−1.5	0.0013
hsa-miR-362-3p	3.0	0.00039	hsa-miR-1178	−2.4	3.22E-05	hsa-miR-125a-3p	−1.5	0.0014
hsa-miR-532-5p	2.5	0.00040	hsa-miR-29b	−2.7	3.59E-05	hsa-miR-523	−1.6	0.0015
hsa-miR-1225-3p	1.6	0.00050	hsa-miR-346	−1.9	4.18E-05	hsa-miR-1206	−1.5	0.0017
hsa-miR-26b	2.5	0.00052	hsa-miR-192	−1.9	4.28E-05	hsa-miR-506	−1.5	0.0018
hsa-miR-199b-5p	6.2	0.00054	hsa-miR-526a+	−1.8	3.10E-05	hsa-miR-1179	−1.5	0.0019
hsa-miR-197	2.2	0.00067	hsa-miR-518d-5p+
hsa-miR-9	9.2	0.00087	hsa-miR-520c-5p
hsa-miR-592	1.7	0.0014
hsa-miR-19a	2.0	0.0018

Bold indicates miRNAs similarly altered in human OS

1.3.2 miR-9 is up-regulated in canine OS tumor tissues and cell lines

Of the miRNAs found to be dysregulated in canine OS, miR-9 expression levels were significantly higher in primary canine tumor samples as compared to normal canine osteoblasts. This finding was independently validated by real-time PCR for miR-9 in fresh primary canine OS tumors, canine OS cell lines, normal canine bone, osteoblast cell lines, and primary osteoblast cultures (Fig. 3), demonstrating significantly higher levels of miR-9 expression in the tumors and OS cell lines relative to normal bone or osteoblasts. Primary osteosarcoma tumor specimens exhibit significant cellular heterogeneity; it was therefore possible that expression levels of miR-9 in tumor samples were influenced by the proportion of tumor cells to stroma/inflammatory cells. To assess the contribution of tumor microenvironment on miR-9 expression in primary canine OS tissues, we identified homogenous OS tumor cells regions in FFPE primary OS tumor specimens and isolated RNA from targeted tumor core samples. We found that miR-9 expression was markedly increased in OS tumor cells as compared to normal canine bone, normal canine osteoblasts and primary osteoblast cultures (Fig. 3). These findings demonstrate that the observed overexpression of miR-9 in canine OS tumor samples is not secondary to non-neoplastic cells infiltrating the tumor microenvironment, but derived directly from the malignant osteoblasts. These data are concordant with published data demonstrating overexpression of miR-9 in human OS [45].

Overexpression of pre-miR-9 does not alter cellular proliferation or caspase-3,7 dependent apoptosis in normal canine osteoblasts or the OSA16 cell line

To assess the biological consequences of miR-9 expression in normal osteoblasts or malignant OS cell lines, canine osteoblasts and the OSA16 cell line that exhibits low endogenous expression of miR-9 were transduced with a pre-miR-9-3 lentiviral expression vector. Stably transduced GFP + cells were sorted and real-time PCR was used to confirm miR-9 overexpression (Fig. 4a). To determine the impact of miR-9 expression on normal or malignant osteoblast cell proliferation and apoptosis, canine osteoblasts and the OSA16 cell line expressing control or pre-miR-9-3 lentiviral constructs were cultured for 24, 48, and 72 h and cell proliferation and caspase-3,7 activity was assessed. Overexpression of miR-9 had no observed effects on cell proliferation or apoptosis in either normal osteoblast cells or malignant OS cell lines (Fig. 4b, c).

miR-9 expression enhances invasion and migration in normal osteoblasts and the OSA16 cell line

To investigate the effects of enforced miR-9 expression on invasive capacity, a standard Matrigel Invasion assay was performed to evaluate cell invasion. As shown in Fig. 5a, overexpression of miR-9 in normal osteoblasts or malignant OSA16 cells significantly enhanced their invasion after 24 h of culture compared to cells expressing empty vector. Cell migration activity was assessed in normal osteoblasts overexpressing miR-9 using the wound-healing assay (scratch test). Fig. 5b demonstrates that miR-9 enhanced cell motility and scattering following gap formation in normal osteoblasts compared to osteoblasts expressing control vector (Fig. 5b). Collectively, these findings demonstrate that miR-9 promotes an invasive phenotype in normal and malignant canine osteoblasts.

Anti-miR-9 expression decreases cell invasion and migration in OSA8 cells

To determine whether inhibition of miR-9 would impair cell migration and invasion, the canine OSA8 cell line that expresses high basal levels of miR-9 was transduced with miRZip-9 (anti-miR-9) or control lentivirus. Transduced cells were sorted based on GFP expression and miR-9 expression was assessed by quantitative PCR to confirm mature miR-9 knockdown (Fig. 5c). In concordance with our findings in normal canine osteoblasts and OSA16 cells overexpressing miR-9, inhibition of miR-9 in OSA8 cells significantly decreased cell invasion and migration compared to control cells (Fig. 5c, d), providing further support for the role of miR-9 in OS invasion.

2D-DIGE electrophoresis and RNA sequencing identifies miR-9-induced alterations to the proteome and transcriptome of canine osteoblasts

To gain further mechanistic insight into miR-9-dependent cell signaling events that may promote the invasive phenotype of osteoblasts, we analyzed the proteomic and gene expression profiles of canine osteoblasts expressing control or miR-9 lentiviral constructs. Two-dimensional difference-in-gel electrophoresis (2D-DIGE) identified 10 protein spots that were differentially expressed in osteoblasts overexpressing miR-9 compared to cells expressing empty control vector (Additional file 5: Figure S4). Determination of the proteins located at these spots was undertaken using in-gel trypsin digestion followed by tandem mass spectrometry (Table 3). Four of the protein spots were unable to be definitively identified, and 2 of the proteins found to be significantly down-regulated following miR-9 overexpression did not have putative miR-9 binding sites within their 3’-UTR, implying that miR-9 may indirectly regulate their expression. Interestingly, miR-9 induced up-regulation of several proteins involved in actin dynamics and cytoskeletal remodeling, including gelsolin and cofilin-1. To independently validate these changes in protein expression, western blotting was performed for gelsolin, an actin binding protein implicated in neoplastic transformation and metastasis [46, 47]. Consistent with our 2D-DIGE results, gelsolin (GSN) was up-regulated in miR-9 expressing osteoblasts (Fig. 6a). Concordant with these results, GSN protein expression was substantially reduced following down-regulation of miR-9 in OSA8 cells transduced with anti-miR-9 vector as compared to cells expressing control vector (Fig. 6a). Furthermore, real time PCR demonstrated an increase in GSN mRNA expression in osteoblasts overexpressing miR-9 compared to empty vector controls, which was further validated with RNA sequencing (0.4-fold increase, p = 0.05) (Fig. 6b).

Spot #	t-test	Av Δ	Acc #	Description	pI	Mass	Sa	Mb
291	0.038	−1.4	gi|73997540	PREDICTED: ELKS/Rab6-interacting/CAST family member 1 isoformX1 [Canis lupus familiaris]	5.69	128164	147	2
919	0.007	−1.1	gi|73980394	PREDICTED: protein disulfide-isomerase A6 [Canis lupus familiaris]	4.97	48667	906	11
961	0.021	−1.3	No ID
1080	0.034	1.1	No ID
1384	0.033	1.2	No ID
1609	0.027	1.4	gi|5031635	cofilin-1 [Canis lupus familiaris]	8.52	19715	343	3
1617	0.03	1.2	No ID
1872	0.029	1.2	gi|478533920	PREDICTED: 60S acidic ribosomal protein P2 [Ceratotherium simum simum]	4.44	11772	364	4
1925	0.052	1.2	gi|350582724	PREDICTED: 14-3-3 protein theta isoform 2 [Sus scrofa]	4.67	28075	983	12
1935	0.008	1.2	gi|545518174	PREDICTED: gelsolin [Canis lupus familiaris]	8.49	95129	609	8

^aThe protein score is derived from Mascot and provides an indication of how well the peptides matched the indicated protein sequence. The actual score is calculated by the following equation: protein score = −10*Log(P), where P is the probability that the protein match is a random event. Scores above 100 indicate that p < 0.05

^bThe protein match score indicates the number of unique peptides that matched the sequence of the identified protein. Two unique peptide matches to a protein sequence confirms the identity of a prot

We compared the gene expression profile of osteoblasts possessing enforced miR-9 expression to that of cells expressing empty control vector and observed significant differences in transcript expression. RNA sequencing identified 55 transcripts that were significantly up-regulated (>2-fold) and 139 transcripts were significantly down-regulated in osteoblasts overexpressing miR-9 (Additional file 6: Table S2). Consensus binding sites for miR-9 were identified within the 3’-UTR of 37 genes that were significantly downregulated following miR-9 overexpression, suggesting that miR-9 may regulate the expression of these putative target genes (Additional file 6: Table S2, bolded). The transcripts identified as predicted targets of miR-9 are involved in a variety of cellular processes, including transcription (transcription factor 19, TCF19 and homeobox B6, HOXB6), RNA methyltransferases (NOL1/NOP2/Sun domain family, member 7, NSUN7), cytokinesis/microtubule assembly (protein regulator of cytokinesis 1, PRC1; kinesin family member 23, KIF23; stathmin 1, STMN1; and cancer susceptibility candidate 5, CASC5), endopeptidase activity (membrane metallo-endopeptidase, MME), and extracellular matrix organization (TGF-β-induced, TGFBI and collagen type IV, alpha 4–1 and −2, COL4A1, COL4A2). Interestingly, one of the most significantly down-regulated genes was TGFBI, an extracellular matrix protein and known mediator of osteoblast adhesion. TGFBI has several highly conserved predicted miR-9 binding sites within the 3’UTR indicating direct regulation of expression by miR-9. Concordant with our RNA sequencing results, real time PCR demonstrated downregulation of TGFBI transcript expression in canine osteoblasts overexpressing miR-9 (Fig. 6c).

GSN shRNA decreases cell invasion and migration in canine OSA8 cells

Our previous findings support the notion that miR-9 promotes cell invasion and migration in canine osteoblasts, in part, through up-regulation of GSN. Given the role of gelsolin in the regulation of actin polymerization and cycling, we designed lentiviral-shRNA for canine GSN to determine the impact of GSN downregulation on cell invasion and migration in canine OSA8 cells. Expression of GSN was significantly reduced in OSA8 cells transduced with GSN shRNA as evidenced by Western blotting and quantitative real-time PCR (Fig. 7a, b). Furthermore, downregulation of GSN correlated with a significant decrease in cell invasion in canine OSA8 cells transduced with GSN shRNA as compared to those transduced with scrambled control shRNA (Fig. 7c) demonstrating a direct role for GSN in contributing to the invasive properties of osteosarcoma cells.