Eighty-five cases from canine patients in which melanoma was diagnosed were retrieved from the archive of the Comparative Pathology Laboratory, Department of Pathology, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil. Informed client consent was obtained for all patient samples used in this study. Of these, 29 cases were oral melanomas and 56 were cutaneous melanomas. To confirm the primary diagnosis each sample was re-examined independently by two pathologists (E.F. and T.S.) according to the criteria, established by the World Health Organisation (WHO), of the International Histological Classification of Tumours of Domestic Animals (19), using a Olympus BX41 microscope. Formalin-fixed and paraffin-embedded (FFPE) tissues were cut into sections (4 μm) using a Leica RM2245 microtome. Deparaffinised samples were used for Hematoxylin and Eosin (H&E) staining. Clinicopathological features were evaluated: (1) presence of ulceration; (2) degree of pigmentation; (3) morphology of neoplastic cells (spindle, epithelioid or mixed); (4) mitotic index (MI); and (5) tumoural vascular invasion. The MI was determined by counting all the mitoses within 10 random, non-overlapping high-power fields (HPFs) ( ×400), FN 22/40x objective (2.37 mm²) (20). A MI was attributed according to the number of mitoses: low MI (<4 mitoses per 10 HPFs); high MI (≥4 mitoses per HPFs) (21). The degree of pigmentation was estimated using a subjective scale from 0 (no pigmentation), 1 (pigmentation in 25% of neoplastic cells), 2 (pigmentation in 25–50% of neoplastic cells) to 3 (pigmentation in >50% of neoplastic cells).

FFPE tissue samples were cut into 4 μm sections and processed for immunohistochemical (IHC) analysis of S-100, melan-A, PNL-2, COX-2 and Ki-67. All immunostainings utilised the streptavidin-biotin-peroxidase complex method with a commercial detection anti-mouse/anti-rabbit system (Novolink Polymer Detection Sistem; Leica Biosystems, Newcastle upon Tyne, UK) according the manufacturer's instructions. Antigen retrieval was performed by Pascal® in citrate buffer pH 6.0 (DakoCytomation Target Retrieval Solution) at 125°C for 2 min, followed immediately by 20 min at room temperature (RT). All sections were incubated with the primary specific antibody: S-100 (1:100 dilution, Clone S100, Dako, Glostrup, Denmark), melan-A (1:100 dilution, Clone A103, Dako, Glostrup, Denmark), Melanoma Antigen (1:100 dilution, Clone PNL-2, Santa Cruz Biotechnology, Dallas, Texas, USA), COX-2 (1:80 dilution, Clone SP21, Thermo Fisher Scientific, Walthan, MA, USA) or Ki-67 (1:50 dilution, Clone MIB-1, Dako, Glostrup, Denmark) for 18 h at 4°C. The antibody reactions were visualised with chromogen 3,3′ -diaminobenzidine tetrachloride (DAB) diluent (Dako, Code K3468), incubated for 3 min at RT. The samples were washed 3x times in distilled water for 5 min at RT, then counterstained with Giemsa for 30 min at RT, dehydrated and mounted (22). Sections treated with isotype-matched primary antibodies mouse anti-human IgG were used as negative controls and this study also included adequate positive controls: canine mammary tumour for COX-2; and canine epidermis for melan-A, PNL-2, and Ki-67. Images were captured on BX41 Olympus microscope. A subjective scale was then used to establish the degree of positive staining for S-100, melan-A and PNL-2 (23, 24). For COX-2 staining a distribution score and intensity were multiplied to obtain a total score, which ranged from 0 to 12. Distribution was estimated as the percentage of positive cells in 10 HPFs (x400): 0 (absent); 1 (<10% stained cells); 2 (10–30% stained cells); 3 (>30–60% stained cells); and 4 (>60% stained cells). For staining intensity: 0 (absent); 1 (weak); 2 (moderate); and 3 (strong). The samples with a final score 0–5 were considered to have COX-2 low expression, and samples with a score of 6–12 were considered to have high COX-2 expression (24). The Ki-67 index was evaluated by counting 500 tumour cells within a microscopic grid at high magnification ( ×400) and the index was expressed as a percentage.

RNA was extracted from FFPE tissue samples using the RecoverAll™ Total Nucleic Acid Isolation Kit (Thermo Fisher Scientific, Walthan, MA, USA, Code AM1975), according to the manufacturer's instructions. Extracted RNAs were quantified and 260/280 nm absorbance determined by NanoDrop Nucleic Acid Quantification (Thermo Fisher Scientific, Walthan, MA, USA). Total RNA was reverse transcribed using the M-MLV kit (Invitrogen, CA, USA, Code 280250-13) according to the manufacturer's instructions. The RNA integrity was evaluated by conventional RT-PCR analysis with the amplification of a 120 base pair fragment of the dog GAPDH reference gene, using the forward 5′-TTCCACGGCACAGTCAAG-3 and reverse 5′-ACTCAGCACCAGCATCAC-3′ primers. For quantitative real-time PCR 90 ng of cDNA was used in a final qPCR reaction with SYBR Green® PCR Master Mix Kit according to manufacturer's instructions (Invitrogen, CA, USA) and ABI PRISM® 7500 Sequence Detection System (Carlsbad, CA, USA). For the RT-qPCR the cycling conditions used were: (stage 1) 1 cycle of 50°C for 2 min; (stage 2) 1 cycle at 95°C for 10 min.; (stage 3) 40 cycles of 95°C for 15 s then 55°C for 1 min. After completion of the amplification reaction, melt-curve analysis was carried out at 95°C for 15 s to produce a dissociation curve. Relative gene expression levels were obtained by normalisation to the expression level of a housekeeping gene (HPRT) (25). Primer sequences are shown in Table 1.

The panel of canine melanoma cell lines used in this study included CMGD2, CMGD-5, CML-10 and TLM1 (Kerafast, Boston, MA, USA), cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% foetal bovine serum, 100 μg/ml streptomycin (ThermoFisher Scientific) and 5% HEPES (ThermoFisher Scientific). Canine transitional cell carcinoma of the urinary bladder cell line (K9TCC) (a kind gift from Deborah Knapp and Jane Stewart, Purdue University) was grown in DMEM/F-12 (ThermoFisher Scientific) containing HEPES and L-glutamine and supplemented with 10% FBS and 100 μg/mL penicillin/streptomycin. All cells were maintained at 37°C in a humidified atmosphere with 5% CO₂.

RNA was extracted from cells using the RNeasy Kit (Qiagen, CA USA) according to the manufacturer's instructions. Real-time PCR was executed using 50 ng of amplified RNA and the Stratagene Mx3000p qPCR system (Aligent, CA, USA), using the Platinum® SYBR® Green qPCR SuperMix-UDG according to manufacturer's instructions (Invitrogen, CA, USA). For the RT-qPCR the cycling conditions used were: (stage 1) 1 cycle of 95°C for 2 min; (stage 2) 40 cycles of 95°C for 5 s followed by 60°C for 10 s. After completion of the amplification reaction, melt-curve analysis was carried out according to the instrument's instructions. Relative gene expression levels of COX-2 were obtained by normalisation to the expression level of a housekeeping gene (GAPDH) (25). Primer sequences are shown in Table 2.

The generation of inducible vectors expressing COX-2 miRNAs were constructed using custom designed complementary canine COX2 oligos (Invitrogen's RNAi Designer, USA), with each containing 4 nucleotide overhangs necessary for directional cloning. The complementary COX2 oligos were annealed and the resulting double-stranded oligo was cloned into pcDNA™6.2-GW/EmGFP-miR vector (pcDNAGFPmiRNACOX2-346) according to the Block-iT ™ PolII miR RNAi expression vector kit manual (Invitrogen). A negative control plasmid, pcDNATM6.2-GW/EmGFP-miRneg (included in the kit), containing a miRNA insert predicted not to target any known vertebrate gene, was also used for the subsequent cloning steps.

Custom designed canine double-strand COX2-346 oligo, with 4 nucleotide overhangs (bold), cloned into pcDNA™6.2-GW/EmGFP-miR vector:

The pcDNAGFPmiRNACOX2-346 and pcDNA™6.2-GW/EmGFP-miRNeg vectors are Gateway® compatible due to the presence of attB sites flanking the COX-2 miRNA insert and can be transferred to other Gateway® adapted destination vectors that contain attR sites. After removal of emGFP, canine COX-2 miRNAs and miRNeg expression cassettes were transferred to the pTRIPZdest destination vector (a kind gift from Dr Peter Hohenstein, Roslin Institute) using the rapid Invitrogen BP/LR protocol (Invitrogen). The destination vector pTRIPZdest is a lentiviral inducible vector engineered to be Tet-On. Following transfer of the COX2miRNA in the pTRIPZdest, the tetracycline response element (TRE), placed upstream of a minimal promoter, will drive the co-expression of a TurboRFP reporter gene and the COX2miRNA if the transactivator (rtTA3) binds to TRE which only occurs in the presence of doxycycline (DOX). The induced expression of TurboRFP allows tracking of transfection efficiency, TRE promoter activity and COX2miRNA expression. The pTRIPZdestmiRCOX2 vectors contain the puromycin resistance gene (puro) which allows for puromycin selection of any mammalian cell lines that are stably transfected with the vectors.

CMGD-2, CMGD-5, CML-10 and TLM-1 cells were seeded at 2 ×10⁵ in 6-well plates and incubated for 24 h. At 80% confluency cells were transfected with 2.5 μg DNA (pTRIPZdestmiRCOX2-346 expression vector) per well at a final concentration of 1 μg/μl using 12.5 μl (CMGD-2 and CMGD-5), 9 μl (CLM-10) and 6 μl (TLM-1) Lipofectamine® 2000 reagent (ThermoFisher Scientific, USA) according to the manufacturer's instructions. pTRIPZdestmiRNeg plasmid negative control plasmid was used as a negative control in subsequent experiments. To establish stable transfection, cells were exposed to 2 μg/ml of puromycin 48 h post-transfection. After 10–14 days of selection, puromycin-resistant colonies were pooled and expanded.

Transfected cells were seeded in triplicate in opaque 96-well plates (Corning, CA, USA) at 2,000 cells/well. Cells were incubated 37°C, 5% CO₂ and cell viability was assayed 48 h later using the CellTiterGlo® Luminescent Cell Viability Assay (Promega, Madison, USA, Code G7570) according to the manufacturer's instructions. Luminescence was recorded using a Viktor3 luminometer (PerkinElmer, Massachusetts, USA). Data was averaged, and then normalised against the average signal of untreated/negative control treated samples.

Transfected cells were seeded at 500 cells/10 cm plate and immediately treated with the indicated dose of DOX. Plates were incubated until colonies were visible at 37°C in a humidified CO₂ incubator. Growth media was changed approximately once a week. The colonies were fixed by incubating for 5 min with ice-cold methanol at RT. Colonies were air dried and then stained with Giemsa (Invitrogen, Paisley, UK) for 20 min at RT. The total number of colonies were counted.

Invasion assays were performed using transwell system with two compartments separated by 8 μm pore polycarbonate philtres coated with Matrigel (BD Biosciences). In the upper compartment ~2 ×10⁴ cells were seeded in serum-free media, and the lower compartment contained media spiked with 5% FBS. Cells were allowed to transmigrate for 24 h. Non-migrated cells on the upper side of the transwell were removed. Cells on the underside of the transwell were fixed with methanol and stained with 0.1% crystal violet. These represented cells that had migrated through the philtre. Ten different fields (x200) were counted in triplicate experiments.

Transfected cells were incubated at 37°C in humidified 5% CO₂ incubator grown until 80–90% confluent in 6-well plates. A scratch was made in the confluent monolayer of cells with a pipette tip. The gap was visualised using the Axiovert 40 CFL microscope coupled with an AxioCAM HRm camera (Carl Zeiss Ltd, Cambridge, UK) and measured at 6 points/gap and an average taken. Measurements were made over the indicated time course. Scratch width was converted to “relative migration distance” of cells, where the distance is a percentage of the initial wound width.

Data were expressed as a mean ± S.D. Statistical analysis was performed with GraphPad Prism version 5.0 for Windows (GraphPad software, CA USA) using Spearman's correlation for non-parametric analysis test, analysis of variance and Student's t-test or mann-whitney test. The criterion for significance was p < 0.05 for all comparisons.

A total of 85 melanoma tumours (29 orals and 56 cutaneous) were included in this study and classified accordingly: 23 spindle cell, 53 epithelioid cell and nine mixed. High pigmentation level is inversely correlated with overall survival time in human patients and is an important marker of the disease (26). Here the degree of pigmentation was estimated using a subjective scale from 0 (no pigmentation) to 3 (pigmentation in >50% of neoplastic cells). Of the 85 samples: 20 had no pigmentation; 42 scored 1; 7 scored 2; and 16 scored 3. Regarding the mitotic index (MI), 74 cases were evaluated (24 orals and 50 cutaneous), of those, 14 oral and 20 cutaneous melanomas had a high MI (≥4 mitoses per 10 HPF) (Table 3). The MI could not be evaluated on 11 other cases due to high pigmentation.

The immunoreactivity for S-100, melan-A, and PNL-2 was observed as brown cytoplasmic staining. All samples were positive for S-100 protein and PNL-2, and 75 of 85 were positive for melan-A. Of the 10 cases that were negative for melan-A the diagnosis was confirmed by positive staining for PNL-2 (Table 4). Overall, a high COX-2 expression revealed a statistically significant correlation with Ki-67 (p = 0.03), however when oral and skin melanomas were analysed independently only the skin tumours presented statistical significance (p = 0.01) (representative images of Ki-67, melan-A, PNL-2 and COX-2 staining are shown in Figure 1).

We found that COX-2 was located in the cellular membrane, cytoplasm, and nuclear membrane in a diffuse and homogeneous pattern. In the primary tumour the labelling intensity of COX-2 was comparable to that of neoplastic intravascular emboli, however, the extension of COX-2 staining in neoplastic intravascular emboli was always diffuse. These results confirmed that COX-2 was expressed, at some degree, in 90% of all tumours examined (Figures 1E,F) and that 34 of 85 of these tumours had a combined distribution and intensity score of 6–12. Of the 29 oral melanomas, 10 showed a COX-2 score of 6–12, whereas 24 of 56 of cutaneous melanomas scored 6–12 (Table 4).

The high COX-2 expression revealed a statistically significant correlation with both oral and cutaneous tumours with Ki-67 (r = 0.45; p = 0.03), MI (r = 0.49; p < 0.001), lymphatic invasion (r = 0.41; p = 0.04) and the histological type of tumour (epithelioid and mixed) (r = 0.51; p = 0.003). In the cutaneous tumours the presence of skin ulceration (r = 0.49; p = 0.02) was also statistically correlated with high COX-2 levels. Thus, these results suggest that high COX-2 protein levels could be associated clinicopathological aggressiveness parameters. To correlate this data to COX-2 gene expression we isolated RNA from the FFPE samples and determined COX-2 gene expression by qRT-PCR. The samples were divided according to the previous IHC results into low and high COX-2 protein levels. The gene expression revealed a statistically significant inverse correlation with the COX-2 protein expression (r = 0.64; p = 0.0008) (Figure 1G).

Artificial miRNAs were engineered to have 100% homology to the target COX-2 sequence and result in cleavage. To test the candidate miRNAs for suppression of COX-2 expression, three canine COX-2miRNA vectors were transiently transfected by lipofection into a canine bladder cancer cell line (K9TCC) that highly expresses COX-2. The vector contains a CMV promoter-driven puromycin resistance gene and a green fluorescent protein (GFP) gene to select and identify transfected cells. EmGFP/COX-2miRNA and EmGFP/miRNeg expressing cells were sorted 48 h post-transfection by fluorescence-activated cell sorting (FACS). COX-2 down-regulation was assessed by qRT-PCR using the comparative CT (2ΔΔCT) method. The pcDNAemGFPmiRCOX2 vectors are Gateway® compatible (Figure 2A(i)). After removal of emGFP, canine COX-2miRNAs and miRNeg expression cassettes were transferred to the pTRIPZdest destination vector (a kind gift from Dr Peter Hohenstein, Roslin Institute) (Figure 2A(ii)). This is an inducible system activated by DOX. COX-2 down-regulation was assessed by qRT-PCR in K9TCC inducible clones grown over a week in DOX (2 μg/ml) (Data not shown).

The basal gene expression of COX-2 in each cell line was determined by qRT-PCR. COX-2 expression was highest in CMGD2, whereas the lowest level COX-2 expression was detected in CML10 cells (Figure 2B). The CMGD2 and TLM1 melanoma cell lines were stably transfected with the pTRIPZdestmiRCOX2-346 vector and control cells with the pTRIPZdestmiRNeg plasmid. Puromycin was applied 48 h post-transfection to select for stable expression of the plasmid. CMGD2 and TLM1 puromycin-resistant colonies were pooled 10–14 days following transfection, expanded and frozen for subsequent experiments. To test the inducible system, CMGD2 and TLM1 puromycin-resistant pools expressing the pTRIPZdestmiRCOX2-346 (miRNACOX2-346) vector or miRNeg control vector were exposed to 1 μg/ml of DOX. The pTRIPZdestmiRCOX2-346 vector contains a tetracycline-dependent promoter which in the presence of DOX drives the expression of the miRNA and the expression of a TurboREP reporter, allowing assessment of promoter activity. Here TurboRFP expression was assessed by fluorescence microscopy. We confirmed that COX-2 gene expression was decreased in CMGD2 cells expressing miRNACOX2-346 +DOX and TLM1 cells expressing miRNACOX2-346 +DOX compared to cells expressing the miR negative control (Figure 2C(i) and (ii), respectively).

To determine if COX-2 has an effect on cell viability of melanoma cell lines we used the CMGD2 and TML1 cell lines stably expressing the miRNACOX2-346 or the miR negative control. Here we induced knock-down of COX-2 by treating cells with 1 μg/ml DOX (+) or water (–) as a vehicle control. Cell viability was assayed 72 h after treatment. Here we showed that cell viability is COX-2 dependent in both CMGD2 (Figure 3A(i)) and TML1 cell lines (Figure 3A(ii)). Clonogenic survival analysis, as a measure of cell reproductive death after treatment, shows that only cells expressing the miRNACOX2-346 vector and treated with DOX have a reduction in colony-forming ability (Figure 3B). Taken together, these results indicate that COX-2 may be an important driver of melanoma cell survival.

The invasive capacity of CMGD2 and TLM1 cells stably expressing miRNACOX2-346 was determined using a Boyden chamber assay. CMGD2 miRNACOX2-346 and TLM1 miRNACOX2-346 cells treated with 1 μg/ml DOX (+) showed a significant reduction in invasiveness compared to both the vehicle control and the miRNeg control (Figures 4A,B). Similarly, we assessed migration potential using a standard scratch assay and showed that induced silencing of COX-2 reduced the ability of both TML1 (Figure 5) and CMGD2 (data not shown) to close the wound. These results indicate that COX-2 enhances the migration and invasion capabilities of canine melanoma cells.