Materials and Methods

Tumor samples—Canine and equine FFPE tumor samples were obtained from the University of Tennessee College of Veterinary Medicine pathology database. Samples were identified through a medical record search by use of the keywords melanoma, amelanotic melanoma, and melanomatosis. With the use of these search terms, 22 equine cases and 765 canine cases were identified between January 1, 2000, and December 17, 2007; 70 canine and 8 equine cases were selected for further review on the basis of the availability of adequate FFPE tissue blocks. From these tissue blocks, individual cases were then reviewed for tumor histologic classification and complete medical record with diagnosis, treatment, and, in some cases, follow-up information. Canine tumor histologic examination included sampling of both oral and nonoral melanomas. Nonoral locations included cutaneous benign melanomas, cutaneous malignant melanomas, and digital melanomas. Equine tumor histologic examination included benign cutaneous melanomas, malignant cutaneous melanomas, and melanomatosis (disseminated melanomas). Slides from all 78 tumors were reviewed by a board-certified pathologist (SJN). Ultimately, 39 canine and 8 equine tumors remained that contained adequate and appropriate tissue for further analysis.

Control samples—Control tissue samples were isolated from animals necropsied at the University of Tennessee College of Veterinary Medicine with no evidence of melanocytic tumors. Canine control tissues included pigmented and nonpigmented normal cutaneous and oral tissue. In addition, 2 anaplastic sarcomas (melan-A and S100 negative via immunohistochemical analyses) were included as tumor control tissues. Equine control tissues included pigmented and nonpigmented cutaneous tissue from both gray and nongray horses. These samples were FFPE by use of standard procedures. Histologic classification was confirmed by a pathologist (SJN).

RNA isolation—The RNA was isolated from FFPE tissues by use of a kit^a in accordance with the manufacturer's instructions. Paraffin-embedded tumor sample blocks were selected by a pathologist (SJN). Five 20-μm sections were then cut from each tissue block by use of a microtome. Blocks and sections were manipulated to include only tumor tissues and carefully avoid overlying normal cutaneous tissues. Sections were transferred to a microcentrifuge tube and deparaffined with xylene. Tissue was then washed with ethanol and air-dried. Cellular lysates were prepared by use of digestion buffer with proteinase K. Binding solution and ethanol were added to the lysates. Lysates were applied to RNA-binding columns, washed, and eluted. Final RNA quality and concentrations were determined by evaluating absorbance at optical densities of 260 and 280 nm. Although concentrations varied widely, typical values were > 300 μg/mL. Following RNA quantification, samples were stored at −80°C until analysis.

Gene expression assays—Tyrosinase and MHC I mRNA expression were evaluated by use of custom-made RT-PCR assays^b designed by use of the manufacturer's online design software.^c The canine tyrosinase assay was based on the clone CF02626293_m1, with the probe centered on the exon 1–2 boundary. On the basis of this information, the software designed a 72-bp assay. The equine assay was designed by use of the full-length equine tyrosinase mRNA sequence (XM_001492560) with the probe centered over the exon 1–2 boundary. On the basis of this information, the software designed a 73-bp assay. The canine MHC I assay was designed by use of the full-length canine MHC I clone (NM_001014378.1) with the probe centered at the exon 3–4 boundary and an amplicon length of 97 bp. The equine MHC I assay was designed by use of the full-length equine MHC I clone (NM_001082507.1) with the probe centered at the exon 4–5 boundary and an amplicon length of 64 bp. For an endogenous control, the eukaryotic 18S ribosomal RNA, which amplified a 69-bp target centered on nucleotide 40 of clone Hs03003631_g1, was used.

RT-PCR assay—Quantitative RT-PCR assay was performed by use of a commercially available 1-step kit.^d In brief, reaction mixtures contained 6.375 μL of RNA template (approx 1 μg), 0.375 μL of reverse transcriptase enzyme mix, 7.5 μL of RT-PCR mix,^d and 0.75 μL of the appropriate gene expression assay^b for a total volume of 15 μL. Reaction mixtures were assembled in a 96-well plate in duplicate. Reaction conditions were as follows: reverse transcription at 48°C for 15 minutes, activation of the DNA polymerase at 95°C for 10 minutes, 60 cycles of PCR amplification consisting of denature at 95°C for 15 seconds, and annealing and extension at 60°C for 1 minute. Real-time PCR reactions were performed on a 96-well RT-PCR detection system.^e Assay efficiency (90% to 105%) and linearity (r² = 0.980) were confirmed prior to analysis. Final reaction products were run on 2.0% agarose gels and stained with ethidium bromide for size verification.

Data analysis—Gene expression analysis was performed by use of commercial software.^f Baseline and threshold detection limits were automatically assigned via the system software,^g with their appropriate placement visually verified. Threshold cycle values for each sample were determined and normalized to the S18 reference gene C_t value to obtain ΔΔC_t. Standard deviations were obtained from duplicate runs and normalized similarly. Tyrosinase expression was assayed in all samples by use of the eukaryotic 18S ribosomal RNA as the reference gene. For all samples, relative (normalized) tyrosinase expression was determined. Residual RNA samples were then used to assay MHC I expression by use of the eukaryotic 18S ribosomal RNA as the reference gene; however, not all samples contained adequate residual RNA to perform this assay. The mean tyrosinase signal in normal tissue control samples from each species was set to a baseline equal to 1 and then used to determine relative tyrosinase expression in tumor samples. Relative MHC I expression was obtained similarly by use of the mean control signal.

To evaluate the association between tumor histologic classification and relative tyrosinase or MHC I expression (ΔΔC_t), canine tumors were grouped into benign, malignant, amelanotic, and nonmelanocytic categories by a pathologist (SJN). Similarly, equine tumors were grouped into benign, malignant, gray horse, and nongray horse categories. For canine malignant melanomas, tumor anatomic location (oral vs other) was also evaluated.^14,15 Normalized tyrosinase and MHC I expression for histologic tumor variants were analyzed by use of an ordinary least squares regression.^g All values were evaluated for normality by use of the Shapiro-Wilk test. Nonnormal distributions were logarithmically transformed prior to analysis. Factors found to be significant in univariate models were analyzed in a multivariate model. Tyrosinase or MHC I expression was described as the sole independent variable analyzed for each outcome. To assess for a significant correlation between the relative expression of tyrosinase and MHC I, a 2-step procedure was used. The data were first analyzed to document a normal distribution via the Shapiro-Wilk normality test. The Spearman correlation test was then used on the nonnormally distributed sample to assess correlation. Values of P < 0.05 were considered significant for all comparisons.

Results

Canine tumor samples—For descriptive purposes and tyrosinase expression analysis, canine melanocytic tumor samples were grouped into either primary oral or nonoral locations. Within each group, tumor samples were further classified as either pigmented or amelanotic. For nonoral locations, tumors were also classified as either benign or malignant on the basis of histologic analysis and mitotic index (mitotic figures/10 hpf, ≥ 3). Twenty canine oral melanoma samples were identified for analysis and included 3 labial, 4 maxillary, 3 lingual, 6 mandibular, 2 soft palate, and 2 tonsillar locations. Histologic findings varied from low-grade tumors to anaplastic and high-grade tumors. Six tumors had either minimal or no obvious pigmentation and were thus classified as amelanotic; this was confirmed with immunohistochemical analysis for s100, vimetin, or melan-A. Age at diagnosis ranged from 7 to 16 years; mean age of onset was 11.1 years, and median age of onset was 12.0 years. Dogs included 12 males and 8 females, and most were mixed breed.

Nineteen canine nonoral melanocytic tumor samples were selected, including 8 benign and 11 malignant samples. Benign melanocytic tumors included 7 cutaneous and 1 uveal location. Malignant tumors included 2 metastatic lymph node samples, 6 cutaneous (various locations) samples, and samples from 3 digital locations. Histologic findings in malignant tumors were consistent with high-grade (anaplastic) tumors with a mitotic index > 3 in all cases. Two of the malignant tumors (1 cutaneous and 1 digital) were further characterized as having minimal to no observable pigmentation and thus classified as amelanotic. Immunohistochemical analysis by use of combinations of S-100 or melan-A confirmed the diagnosis in poorly pigmented tumors. Age at diagnosis ranged from 1 to 13 years, mean age of onset was 8.7 years, and median age of onset was 10 years. Dogs included 8 males and 11 females, and most were mixed breed.

Control samples were obtained from 10 dogs (4 female and 6 male) with no evidence of melanocytic tumors. Both oral and cutaneous normal tissues were collected. The oral samples included 3 nonpigmented oral (buccal) mucosa samples and 1 pigmented oral mucosa sample. The cutaneous samples included 2 nonpigmented skin samples and 2 pigmented skin samples. Two additional control samples were included that represented nonmelanocytic tumors (anaplastic sarcomas). These tumors were negative for both melan-A and S-100 but positive for vimentin immunoreactivity. Mean age of all control dogs was 9.1 years, with a median age of 9.0 years (range, 7 to 11 years).

Equine tumor samples—Eight equine tumor tissues were evaluated for tyrosinase expression. The tumor samples were identified from both gray and nongray horses. The gray horse tumor samples included a uveal melanoma, 3 dermal melanomas (1 from a horse with multifocal disease), and a lymph node with metastases. The nongray horse samples included 2 benign dermal melanomas and 1 morphologically malignant dermal melanoma. Age of the gray horses ranged from 8 to 20 years (mean, 15.2 years; median, 16 years). Three horses were female and 2 were male. Age of the nongray horses ranged from 11 to 18 years (mean, 13.3 years; median, 11 years). All 3 nongray horses were male.

The control samples from normal equine tissues were obtained from gray and nongray horses with no external evidence of melanocytic tumors. Six equine control tissues were collected from 3 horses. These included 3 samples from pigmented skin (1 gray horse and 2 nongray horses) and 3 samples from nonpigmented skin (1 gray and 2 nongray horses).

Canine tumor tyrosinase expression—For canine oral melanomas, values ranged from 4.06-fold to 4,810.90-fold (mean, 629.44-fold; median, 170.22-fold) relative expression, compared with the mean control signal (1.0 ± 0.59). All tumors had high relative tyrosinase expression; the tumor with the lowest expression (4.06 ± 0.15) nevertheless had substantially higher expression than the control signal. This tumor was a low-grade oral melanoma located on the gingiva. The tumor itself consisted of a small population of melanocytes with minimal criteria of malignancy and invasion. The highest tyrosinase expression for an oral melanoma (4,810.90 ± 71.55) was found in a sparsely pigmented oral tumor with extremely aggressive malignancy (Figure 1). Results of immunohistochemical analysis were positive for melan-A, confirming the diagnosis of amelanotic melanoma.

View Full Size

Photomicrograph of a section of a poorly pigmented amelanotic melanoma in the oral mucosa of a dog; the tumor had high tyrosinase mRNA expression, and results of immunohistochemical analysis were positive for melan-A. H&E stain; bar = 200 μm.

Citation: American Journal of Veterinary Research 73, 2; 10.2460/ajvr.73.2.272

• Download Figure
• Download figure as PowerPoint slide

For canine nonoral melanocytic tumors, relative tyrosinase expression ranged from 21.65 to 2,135.32 (mean, 338.34; median, 59.05), relative to the mean control signal. The lowest tyrosinase expression (21.65 ± 4.13) was detected in a benign uveal melanoma. The tumor had marked local invasion into the sclera; however, cellular morphology was most consistent with a benign tumor, and the mitotic index was extremely low (none observed). Highest expression was found in a malignant digital melanoma (2,135.32 ± 51.26) with a high mitotic index, marked bone invasion, and vascular invasion. Values are expressed as mean ± SEM.

Tyrosinase expression in all canine control samples was predictably low, regardless of degree of pigmentation or anatomic location (oral vs cutaneous). Expression for all samples ranged from 0.34 ± 0.34 for non-pigmented oral mucosa to 2.28 ± 0.48 for an anaplastic periorbital sarcoma (mean, 1.07; median, 0.89). Relative expression of oral tumors ranged from 0.34 ± 0.34 to 1.07 ± 0.31 and was not significantly different from cutaneous samples that ranged from 0.71 for a non-pigmented cutaneous sample to 2.28 for the anaplastic sarcoma. The remaining oral anaplastic sarcoma had a similarly low expression (0.46 ± 0.19).

Ordinary least squares regression was used to evaluate the relationship between tyrosinase expression and several covariates, including histologic classification (benign vs malignant), location (oral vs nonoral), degree of pigmentation (melanotic vs amelanotic melanomas), and sample type (ie, melanocytic tumors vs control samples). No significant (P = 0.84) difference in tyrosinase expression was found between benign and malignant canine tumors. Similarly, there was no significant (P = 0.71) difference in tyrosinase expression between oral and nonoral tumor locations. No significant (P > 0.84) difference in tyrosinase expression on the basis of the presence of pigmentation was found between malignant melanotic (ie, heavily pigmented) and malignant amelanotic (nonpigmented) tumors. However, analysis of expression in melanocytic tumors, compared with control tissues, identified a significant difference between the groups. Melanocytic tumors had significantly (P < 0.001) higher tyrosinase expression, compared with control tissues.

Equine tumor tyrosinase expression—All tumors, including benign, malignant, and multifocal tumors, had high relative tyrosinase expression ranging from 8.29 ± 1.84 to 219.57 ± 13.21 (mean, 93.2; median, 61.29). The lowest expression was found in a sparsely pigmented (essentially amelanotic) melanoma from the ventral portion of the neck of a silver-colored horse. No special stains were performed on this tumor to confirm histologic classification. The 2 highest expressions were found in tumor samples obtained from gray horses with melanomatosis. Tyrosinase expression in control tissues was low in both gray and nongray horses, regardless of pigmentation (mean, 1.0; median, 0.83). The lowest tyrosinase expression was found in skin from a chestnut horse (0.22 ± 0.08), and the highest (1.96 ± 0.63) was found in the pigmented (perirectal) skin of a chestnut horse.

Similar to dogs, an ordinary least squares regression was used to evaluate the relationship between tyrosinase expression and, for horses, the covariates of coat color and histologic classification. Horses were classified as gray or nongray, and histologic classification was recorded as benign or malignant on the basis of mitotic index and morphology. Although the number of cases was small, no significant difference was found in relative tyrosinase expression attributable to either coat color (P = 0.13) or tumor histologic classification (P = 0.36). However, comparison of tyrosinase expression between tumor and control tissues identified the predictably higher tyrosinase expression in melanocytic tumors versus controls (P = 0.019).

MHC I expression—All samples with adequate residual RNA (following tyrosinase expression) were further assayed for MHC I expression. Canine tissue samples (n = 15) included 3 controls, 1 anaplastic sarcoma, 3 benign melanomas, and 8 malignant melanomas. Major histocompatibility complex I expression in the canine normal control tissues ranged from 0.4 ± 0.02 for a nonpigmented cutaneous sample to 2.15 ± 0.11 for a nonpigmented oral sample (mean, 1.0; median, 0.45). Expression in canine malignant tumor samples ranged from a low value of 0.02 ± 0.002 for a metastatic lymph node lesion to a high value of 8.88 ± 0.68 for an oral malignant melanoma (mean, 2.53; median, 0.67). Benign canine tumor samples had relatively low MHC I expression, ranging from 0.40 ± 0.03 for an ocular tumor to 1.06 ± 0.12 for a cutaneous tumor (mean, 0.67; median, 0.57). No significant associations were found between relative MHC I expression and tumor histologic classification (benign vs malignant; P = 0.77), tumor location (oral vs other; P = 0.45), or sample type (control vs tumor; P = 0.11). Furthermore, to assess correlation between relative tyrosinase and MHC I expression, a Spearman correlation test was performed. No significant (P = 0.61) correlation was detected between the expressions of these 2 genes in this canine sample population.

Equine tissue samples (n = 6) that were evaluated for relative MHC I expression included 3 control samples, 1 benign tumor, and 2 malignant tumors. Major histocompatibility complex I expression in the normal equine tissues ranged from 0.18 ± 0.04 for the pigmented skin in a gray horse to 1.72 ± 0.36 for the nonpigmented skin sample from a nongray horse (mean, 1.0; median, 1.1). Relative expression in the tumor tissues differed widely, ranging from 1.42 ± 0.51 for a malignant cutaneous tumor in a nongray horse to 1,990.7 ± 49.8 for a benign cutaneous tumor in a nongray horse. The small sample size and lack of values from gray horses precluded the evaluation of the relationships between MHC I expression and the described covariates.

Discussion

The goal of this study was to measure tyrosinase expression in a series of canine and equine tumor samples obtained from the University of Tennessee College of Veterinary Medicine pathology tumor database. The sample included 39 canine and 8 equine tumor samples, including benign, malignant, pigmented, and nonpigmented variants. The RT-PCR method was chosen because a prior study⁸ and the present authors' exhaustive efforts had both failed to detect tyrosinase protein antigen via immunohistochemical analysis by use of commercially available antibodies. The benefit of the RT-PCR method is that it is extremely sensitive and allows the use of FFPE tissues and fresh cytologic or histologic specimens to quickly determine both quantitative and relative tyrosinase expression.^16,17 Furthermore, unlike immunohistochemical analysis, RT-PCR assay can be used to determine the presence or absence of specific portions of target transcripts that may be immunogenic.^18,19

In the present study, RT-PCR assay and tyrosinase mRNA-specific gene expression assays were used to determine the presence or absence of appropriate targets in the samples. The canine and equine expression assays included the sequences orthologous to the human immunodominant region recognized by monoclonal antibody T-311, and those sequences were thus detected by each assay.²⁰ Importantly, this immunodominant region is thought to be required to elicit an antitumor response against the tyrosinase tumor antigen.^20–22 In the present study, a modified RNA isolation method that is useful in isolating RNA from FFPE tissues was also used. This resulted in isolation of considerable amounts of high-quality tumor RNA from each of the 48 tumor and 16 control samples. The RNA samples were quantified via spectrophotometry and assayed for quality via gel electrophoresis. Aliquots of each sample were then used for the tyrosinase-specific gene expression assays.

The RT-PCR assays revealed high tyrosinase expression for most tumor samples. Control tissues had extremely low tyrosinase expression. The high expression in melanoma tumors and the relative lack of expression in normal tissues suggests a tissue-restricted expression pattern for this gene in horses and confirms previous data in dogs.^9,23,24 On the basis of this information, tyrosinase-targeted immunotherapies may be considered in horses. However, future studies are needed to examine tyrosinase expression in additional control samples and other tumor types to more accurately define expression patterns. Furthermore, because of small sample size, the present study was not able to assess the prognostic importance of relative tyrosinase expression in melanocytic tumors. Evaluation of additional cases will be helpful to determine the prognostic importance of tyrosinase expression in general and in patients treated with the melanoma vaccine.^h

Identification of tyrosinase expression, specifically the immunodominant sequence of tyrosinase, provides proof of a target for immunotherapies targeting this protein in horses. The tyrosinase expression assays used in this project were designed to amplify the exon 1–2 boundary of tyrosinase and thus ensure the detection of transcripts that have undergone appropriate splicing in this region. Importantly, this assay would not recognize the presumably nonfunctional canine splice variant that lacks this boundary.⁹ The exon 1–2 region of the tyrosinase protein encodes the catalytically active domain required for tyrosinase function.¹ Inappropriate splicing or mutations within this region could lead to pigmentation defects through loss of function (ie, amelanotic tumors) along with loss of the important immunodominant region. In fact, many humans with disorders of pigmentation have mutations in this critical region.²⁵ Interestingly, in the present study, no correlation was found between tyrosinase expression and the degree of tumor pigmentation (ie, melanotic vs amelanotic tumors). In other words, similar expression of the intact (and presumably) catalytically active (immunodominant) region of tyrosinase was detected in pigmented and nonpigmented melanomas. This suggests that mutations in other proteins involved in melanogenesis may be responsible for the lack of pigmentation in canine amelanotic tumors. Additionally, the presence of the immunodominant portion of the tyrosinase transcript in both melanotic and amelanotic tumor samples supports the use of antityrosinase immunotherapies regardless of degree of pigmentation.

Appropriate expression and function of the antigen-processing machinery are required to elicit an effective immunologic response against tumor-specific antigens.^12,13 Because dysregulation of the antigen-processing machinery is common in tumors, we evaluated the expression of MHC I mRNA as a representative of the antigen-processing machinery.^12,13 Reduced or absent expression of the MHC I transcript in these melanocytic tumors, compared with normal tissue, may suggest an impediment to immunotherapies targeting the tyrosinase tumor antigen.^12,13 In the present study, which had a small sample size, no consistent evidence was found of reduced or absent MHC I expression in either normal or tumor samples. Unfortunately, sufficient residual RNA was not available to assess expression in all samples. Furthermore, MHC I is only 1 portion of the entire antigen-processing machinery.^12,13 Other proteins and genes that appear to be commonly altered or dysregulated in tumor cells include the TAP (transporter associated with antigen processing) gene and proteins, β-microglobulin, a variety of cellular chaperones, and other components of the antigen-processing machinery.^12,13 These genes and proteins were not assessed in the present study but would be logical candidates for subsequent studies. However, these initial data provide further support for tyrosinase as an immunologic target in dogs. Future research would be helpful to more fully assess the presence of expressional changes in genes or proteins involved in antigen processing.

Another potential use of this RT-PCR technology is to complement traditional staging tests (ie, blood and lymph nodes can be screened for the presence of tyrosinase). In human medicine, one of the first RT-PCR tests used in tumor-bearing patients was designed to identify tyrosinase expression in the blood of patients with melanoma.²⁶ Many studies^24,26–29 have evaluated the prognostic importance of tyrosinase expression in blood or lymph nodes and provided conflicting results. In general, results of most studies²⁷ suggest that either a single high expression or changes in tyrosinase expression found in the blood of patients with melanoma can provide important prognostic information related to the potential increased risk of metastatic spread rather than as a tumor burden marker. In light of the availability of an immunotherapy that targets tyrosinase,^h further research is required to determine the prognostic importance of tyrosinase expression in the blood of tumor-bearing dogs and horses.