Changes in Serum and Salivary Proteins in Canine Mammary Tumors.

Simple Summary The present study describes for the first time the differences in the serum and saliva proteomes between healthy bitches and bitches with mammary tumors using a high-throughput proteomic approach. More than 1000 proteins were identified and 35 in serum and 49 in saliva were significantly modulated. Additionally, their related pathways were discussed in order to improve understanding of the pathophysiology of the disease and one protein, serum albumin, was further validated. The results of the present study could be a source of potential biomarkers for canine mammary tumors in saliva and serum and increase the knowledge on the pathophysiology of the disease. Abstract The aim of this study was to evaluate changes in serum and saliva proteomes in canine mammary tumors (CMT) using a high-throughput quantitative proteomic analysis in order to potentially discover possible biomarkers of this disease. Proteomes of paired serum and saliva samples from healthy controls (HC group, n = 5) and bitches with CMT (CMT group, n = 5) were analysed using a Tandem Mass Tags-based approach. Twenty-five dogs were used to validate serum albumin as a candidate biomarker in an independent sample set. The proteomic analysis quantified 379 and 730 proteins in serum and saliva, respectively. Of those, 35 proteins in serum and 49 in saliva were differentially represented. The verification of albumin in serum was in concordance with the proteomic data, showing lower levels in CMT when compared to the HC group. Some of the modulated proteins found in the present study such as haptoglobin or S100A4 have been related to CMT or human breast cancer previously, while others such as kallikrein-1 and immunoglobulin gamma-heavy chains A and D are described here for the first time. Our results indicate that saliva and serum proteomes can reflect physiopathological changes that occur in CMT in dogs and can be a potential source of biomarkers of the disease.


Introduction
Mammary tumors are the most common cancer in intact female dogs, accounting for almost 50% of all tumors [1]. Approximately half of all canine mammary tumors (CMT) are diagnosed as malignant if they were spayed less than one year ago, if there was possibility of concurrent diseases based on physical examination and laboratory analyses, or if gingivitis was detected. None of the serum and saliva samples showed haemolysis. All dogs with TNM were treated by surgical removal of the neoplasm and were alive and without relapse 6 months after the surgery.
All the procedures were approved by the ethics committees of the University of Murcia and the Ministry of Agriculture, Livestock, Fishing and Aquaculture of the Region of Murcia (A13170503).

Serum and Saliva Sampling
Surplus serum remainings after routine clinical biochemical analyses were stored at −80 • C until their use for the present study. Venipuncture of the jugular or cephalic vein was performed for blood collection. Whole blood was stored in tubes containing a coagulation activator and a gel separator and kept at room temperature (25 • C) until visible clot reaction. Samples were centrifuged (3500× g, 10 min) and the supernatants were stored at −80 • C until analysis.
Saliva samples were obtained as previously reported [19]. In brief, a small sponge was placed into the mouth and, when it was thoroughly moistened, it was then removed and placed into collection devices (Salivette saliva collection tube/V-Bottom, Sarstedt, Aktiengesellschaft & Co, Nümbrecht, Germany). After the samples were centrifuged (3000× g, 10 min, 4 • C), the supernatants were stored at −80 • C until analysis [20]. At least 0.2 mL of saliva was collected from each patient.

Proteomics Study of Saliva and Serum Samples and LC-MS/MS Analysis
A total of 35 µg of protein per sample was subjected to reduction, alkylation, digestion and labelling using 6-plex Tandem Mass Tag reagents, according to manufacturer instructions (Thermo Scientific) as described previously [10,11]. In brief, after protein determination using Bradford assay [21], 35 µg of sample and internal standards (consisting of a pool of 35 µg of each sample) were reduced for 1 h with 200 mM DTT (Sigma-Aldrich), alkylated for 30 min with 375 mM iodoacetamide (Sigma-Aldrich) and precipitated with ice-cold acetone (VWR, Llinars del Vallés, Barcelona, Spain). The next day, the samples were centrifuged, the acetone was decanted and the pellets were resuspended with 50 µL of 100 mM TEAB buffer and subsequently digested with trypsin (Promega, Madison, WI, USA) overnight at 37 • C (2.5 µg of trypsin per 100 µg of protein). The TMT labelling reagents were resuspended in LC-MS-grade anhydrous acetonitrile (Thermo Scientific, Waltham, MA, USA) and mixed with each sample for 1 h at room temperature. The reaction was quenched by adding 5% hydroxylamine (Thermo Scientific) for 15 min and the samples were combined at equal amounts. Then, 6 µg of each mixed sample set was placed in a well of a microplate, vacuum-dried and stored at −80 • C before further LC-MS/MS analysis. LC−MS/MS analysis was performed using the Dionex Ultimate 3000 RSLC nano-flow system (Dionex, Camberley, UK) and an Orbitrap Q Exactive Plus mass spectrometer (Thermo Fisher Scientific), as described elsewhere [22]. The SEQUEST algorithm, Proteome Discoverer (version 2.0., Thermo Fisher Scientific, Waltham, MA, USA), was used for peptide identification and relative quantification. A NCBI database search for Canis Lupus FASTA files was performed considering two missed trypsin cleavage sites, a precursor tolerance of 10 ppm and a fragment mass tolerance of 0.02 Da. The Percolator algorithm within the Proteome Discoverer workflow was used to determine the false discovery rate (FDR) for peptide identification, which was set at 1%.

Validation of Serum Biomarkers
For validation of serum biomarkers detected by proteomic analysis, serum samples from 25 bitches that were presented to the Department and Clinic of Animal Reproduction, University of Life Sciences, Lublin, Poland, were employed-healthy controls (n = 10; aged 7.5-11) and dogs with CMT stage 1-2 tumors (n = 15; aged [8][9][10][11][12][13]. Albumin in serum was selected for validation since it is commonly performed in routine biochemistry analysis in our laboratory and, therefore, that data were already available. Serum albumin was determined using a commercially available kit (Albumin OSR 6102; Olympus Life and Material Science Europe GmbH, Irish branch, Ennis, Ireland), according to manufacturer instructions.

Statistical Analysis
All statistics were performed using R v3.2.2 [23]. First, proteins with fewer than two unique peptides and proteins with >90% missing data were removed from the analysis. Sample outliers were detected for each of the proteins using Dixon's test from R package outliers v0.14 [24]. If a sample outlier was significant (p < 0.05), it was removed from further analysis. As the majority of the analysed proteins did not follow normal distribution, as tested by the Shapiro-Wilk test, the Wilcoxon-Mann-Whitney test was performed in order to test for differences in protein abundance between groups. The fold change between the two groups was calculated as the mean protein abundance for the CMT group divided by the mean protein abundance for the HC group.
For the validation of serum biomarkers, the D'Agostino and Pearson omnibus normality test was employed to determine the distribution of data and, since data were not normally distributed, the non-parametric statistical Mann-Whitney U (two-way) test was used to compare between groups.

Bioinformatics Analysis
The proteins' GI accession numbers were converted into official gene symbols either using the DAVID conversion tool (https://david.ncifcrf.gov/conversion.jsp), UniProtKB ID mapping (https://www. uniprot.org/uploadlists/) or the SEQUEST search engine implemented into Proteome Discoverer [25]. Genes encoding the differentially abundant proteins in the CMT and HC groups were used to determine the GO terms overrepresented in CMT using the Protein Analysis Through Evolutionary Relationships (PANTHER) classification tool (http://www.pantherdb.org/).

Proteomic Analysis in Serum
After the removal of proteins with fewer than 2 unique peptides, NMT <5% FDR, missing data and outliers, 379 serum proteins remained for statistical analysis (Table A1). The Wilcoxon-Mann-Whitney test revealed statistically significant different abundances in the HC and CMT groups for 35 proteins, corresponding to 16 unique genes after the removal of duplicates and isoforms, as summarized in Table 1 and Figure A1. The 35 proteins differentially expressed in serum in CMT and HC were used for subsequent bioinformatics analysis in terms of functional clusters, according to the PANTHER classification system ( Table 2). The identified differentially modulated proteins in CMT and HC had three molecular functions: binding (40%), catalytic activity (30%) or molecular function regulators (30%). Three different biological processes were involved: 66% of proteins were involved in cellular process, 17% of proteins in biological adhesion and another 17% in localization. Regarding biological pathways, half of the proteins were involved in blood coagulation, while 25% were involved in the plasminogen activating cascade and in angiotensin II-stimulated signalling through G protein and beta-arrestin. Regarding protein class, 50% of proteins were enzyme modulators, followed by transfer/carrier proteins (17%), signalling molecules (17%), and receptors (16%). Table 1. Proteins downregulated and upregulated in serum. In serum, 35 proteins were initially identified with statistically significant different abundances in CMT and HC. After the removing of duplicates and isoforms, there were 9 downregulated and 7 upregulated proteins.

GI Accession
Gene Color represent relative abundance of the protein (red for downregulated, green for upregulated). Table 2. Molecular function, biological process, pathways, and protein class as the number of genes and gene percentage of the differentially expressed proteins in the HC and CMT groups in serum and saliva based on the Protein Analysis Through Evolutionary Relationships (PANTHER) classification system.

Serum Saliva
Gene % Gene Gene % Gene

Molecular Function
Binding

Proteomic Analysis in Saliva
After the removal of proteins with fewer than two unique peptides, NMT <5% FDR, missing data and outliers, 730 proteins remained for statistical analysis (Table A2). The Wilcoxon-Mann-Whitney test identified 49 proteins (corresponding to 28 unique genes) with different abundances between the HC and CMT groups (p < 0.05), as summarized after the removal of duplicates and isoforms in Table 3 and Figure A2. Proteins differentially expressed in saliva between CMT and HC were used for subsequent bioinformatics analysis in terms of functional clusters, according to the PANTHER classification system, as shown in Table 2. The identified differentially modulated proteins in CMT and HC had five molecular functions: binding (37%), catalytic activity (28%), structural molecule activity (21%), molecular function regulators (7%), and molecular transducer activity (7%). Furthermore, proteins were involved in nine different biological processes-with 35%, 22%, and 13% of proteins involved with cellular process, metabolic process, and localization, respectively. Most proteins were involved in nucleic acid binding (50%), although there were also cytoskeletal proteins, isomerases and transfer/carrier proteins. Regarding biological pathways, the differentially modulated proteins were involved in de novo purine biosynthesis, de novo pyrimidine deoxyribonucleotide biosynthesis, de novo pyrimidine ribonucleotide biosynthesis, glycolysis (P00024), and salvage pyrimidine ribonucleotides (20% of proteins for each).
Finally, when proteomic results were compared between saliva and serum, the abundance of no protein changed statistically significantly in both serum and saliva.

Discussion
In this report, serum and salivary proteomes showed changes in bitches with CMT when compared to healthy individuals. Some of the proteins that were modulated in the present study such as S100A4 or haptoglobin were also described in previous studies on canine mammary tumors [27] or human breast cancer [14,15,28,29], while other proteins including immunoglobulin gamma-heavy chains A and D or kallikrein-1 were described here for the first time as CMT-related proteins.
In serum, 35 of 379 proteins were differentially modulated in CMT. Fibrinogen A-alpha chain (FGA) was the most upregulated protein in serum in CMT. This protein is a degradation product of the alpha chain of fibrinogen [30] and its protective role against tumor growth and metastasis has been proposed previously [31]. Increases in fibrinogen degradation have been described in neoplasms and seem related to an increase in fibrin deposition in the tumor [32]. FGA has been proposed as a biomarker for HER2-positive breast cancer [33], although it can also appear increased in oral cancer [34] and in infectious diseases such as dirofilariosis or leishmaniosis [35].
The two most downregulated proteins in serum of bitches with CMT were interleukin-13 receptor subunit alpha-2 precursor and immunoglobulin gamma-heavy chains A and D. Interleukins are key immunoregulatory and anti-inflammatory cytokines in tumor cells, produced by different immune cells such as B, T, mast cells, dendritic or natural killer cells [36]. In dogs, the upregulation of interleukin-13 receptor subunit alpha-2 genes was observed in macrophages grown as a co-culture with canine mammary cancer cells [37]. In human breast cancer tissues, overexpression of interleukin-13 receptor subunit alpha-2 precursor has been suggested as an independent predictor of poorer outcome, playing important roles in cancer cell survival and progression, although it was dependent on the breast cancer subtype [38]. In the present study, dogs had cancers of low TNM grade and did not have relapses for at least 6 months, which could explain at least partially the observed low levels of this protein. Immunoglobulin gamma is the most abundant class of antibodies, predominant in the immune secondary response and participating in different important roles such as the activation of the complement cascade, or mediation of antibody-dependent cell cytotoxicity [39]. There is a lack of information on the possible relation between IgG and CMT and, therefore, further studies would be diserable in order to discern the possible reasons for the IgG downregulation observed in our study.
Additionally in serum, modulations of known acute-phase proteins (APP) in dogs with increases in haptoglobin and decreases in albumin were observed in the proteomic analysis of this study. While haptoglobin is considered a moderate APP in dogs that increases by approximately 2-to 5-fold, albumin is a negative APP that decreases during inflammation [40]. In agreement with our results, increases in serum haptoglobin concentrations were observed in dogs with mammary tumors when compared to healthy controls [41,42], as well as in dogs with hemolymphatic, mesenchymal, and epithelial tumors [40,43,44]. The downregulation of albumin in serum in CMT bitches was further verified previously by a commercially available kit using an automated analyser, showing a decrease in albumin in dogs with CMT in comparison to controls.
In saliva, three proteins of the S100A family, namely S100A2, S100A4, and S100A6, were among the most upregulated proteins in bitches with mammary tumors when compared to healthy controls. S100A proteins are a family of calcium-binding proteins with potential roles in different cancers including breast [45]. In breast cancer, the positive correlation between S100A4 expression and cancer progression has been described in several studies [28,46,47]. A controversial role in carcinogenesis has been described for S100A2, which has been proposed both as a tumor suppressor [48] and a tumor promotor [49]. Higher tissue expression of S100A2 was significantly correlated with worse overall survival in ovarian cancer patients [15], and S100A2 protein expression in saliva was observed to be higher in oral squamous cell carcinoma patients in comparison to other oral potentially malignant disorders. S100A6 expression has been associated with worse prognosis or the progression of several cancers including oncogynecology [50], colorectal [51], gastric [52], and renal [53]. In contrast, a recent study suggests a relation between S100A6 expression and better prognosis in breast cancer [54]. However, to the best of our knowledge, this is the first report in which the S100A family is described in saliva of bitches with mammary tumors. Based on our results and previous studies in women, evaluation of the S100A family as possible biomarkers of prognosis and evolution of mammary tumors in dogs could be of interest in order to evaluate whether dogs with poorer prognosis can have higher values of these proteins.
Calmodulin-like proteins, namely CALM2 and CALM3, were also upregulated in saliva of bitches with mammary tumors. These proteins are described here as modulated in saliva in CMT for the first time. Previous studies in humans reported overexpression of calmodulins in serum of patients with breast cancer [55] and the use of calmodulin antagonists as a therapeutic strategy [56].
Kallikrein-1-like (KLK1) protein was the only downregulated protein in saliva of bitches with CMT. Both kallikrein genes and proteins might promote or inhibit cancer cell growth, angiogenesis, invasion and metastasis by different mechanisms [57]. Therefore, the role of KLK in cancer is still controversial, with several studies associating KLK with better or worse income. For example, the in vitro inhibition of KLK1 suppresses the invasiveness of breast cancer cells, and therefore it could be a defence mechanism [58]. On the other hand, KLK-mediated degradation of extracellular matrix proteins facilitates tumor cell invasion and metastasis [59], and higher serum levels of other kallikreins such as KLK5 [60], KLK10 [61], and KLK14 [62] have been found in women with breast cancer when compared to healthy ones. Hormonal regulation of KLK1 expression was suggested in human prostate and breast tissues [63] and its salivary secretion was proposed as modulated by age [64]. Nevertheless, to the best of the author's knowledge, this is the first report describing a downregulation of KLK1 in saliva in CMT, and the possible causes for its modulation should be further studied.
When serum and salivary proteomes were compared, approximately twice as many proteins were identified in saliva compared to serum (730 in saliva vs. 379 in serum). This is in agreement with previous studies reporting a higher number of proteins in saliva [65], and indicates that saliva can provide opportunities for the identification of new non-invasive biomarkers. This fact also suggests that a high number of proteins that appear in saliva are not transported from blood but may come from salivary glands, nasal and bronchial secretions, gingival crevicular fluid among others, and, thus, their origin should be investigated in future studies. In our study, none of the proteins were significantly differentially modulated in both serum and saliva at the same time, which reinforces the possible different origin of many proteins in serum and saliva.
The results of the present study suffer from some limitations. First, the included animals were client-owned bitches of different breeds, which may have influenced our results and increased variability since, for example, higher expression of p53 (which is generally associated with poorer overall survival) has been reported in large-breed dogs [66]. In addition, although these biomarkers were not studied in our reports, cancer-related biomarkers including estrogen receptor alpha (ERα), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) differ between the type and subtype of CMT and could have a potential influence in the results [67,68]. Finally, the sample size was relatively small, although it was greater than the minimum of three biological replicates recommended for proteomic studies [69]. Therefore, this study should be considered as a pilot study, and further research with larger populations are needed in order to confirm our findings and to evaluate the potential influence of variables such as age, breed, tumor type based on immunophenotype, and hormonal influence in addition to other possible confounding factors such as other types of neoplasms, different diseases or the presence of oral alterations such as gingivitis.

Conclusions
Dogs with CMT have changes in serum and saliva proteomes as compared to healthy dogs. Bioinformatic analyses in serum and saliva showed that most modulated proteins have binding or catalytic activity molecular functions. Proteins such as kallikrein-1 and immunoglobulin gamma-heavy chains A and D were related to CMT in this study for the first time, and could be considered as potential novel CMT biomarkers. In addition, some of the modulated proteins found in the present study such as haptoglobin or S100A4 have been related to CMT or human breast cancer previously, which confirms the validity of our study design for CMT evaluation and the use of CMT as a model for breast cancer research. Overall, data presented in this report reflect the changes that occur in the proteome of serum and saliva in dogs with CMT and could be a potential source for the study of new biomarkers for this disease.

Acknowledgments:
The authors would like to thank the pet owners and their colleagues that kindly agreed to collaborate in this study.

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A
Animals 2020, 10, x FOR PEER REVIEW 11 of 47 Acknowledgments: The authors would like to thank the pet owners and their colleagues that kindly agreed to collaborate in this study.

Conflicts of Interest:
The authors declare no conflict of interest.