VISTA Is a Diagnostic Biomarker and Immunotherapy Target of Aggressive Feline Mammary Carcinoma Subtypes

Simple Summary Mammary tumors are common in cats, showing aggressive behavior and few therapeutic options. Recently, feline mammary carcinomas have become a reliable cancer model for human breast cancer studies, due to the similarities between the two species. Thus, the identification of new tumor biomarkers and therapeutic targets to improve cat’s prognosis is needed. VISTA is an important immune checkpoint protein that has gained importance over the past few years in women’s cancers. In this study, the serum VISTA levels and tumor expression were analyzed in cats with mammary tumors, being correlated with other immune checkpoints. In the diseased animals, VISTA is overexpressed in more aggressive tumor subtypes (HER2-positive and triple-negative), showing a positive correlation with the expression of VISTA in tumor-infiltrating lymphocytes, and is associated with an immunosuppressive status, suggesting that VISTA could be a promising non-invasive prognostic biomarker and therapeutic target in cats with mammary carcinomas, as reported in humans. Abstract Feline mammary carcinoma (FMC) is a common neoplasia, showing aggressive clinicopathological features, without viable therapeutic options. The study of tumor microenvironment has gained importance, due to the ability to control tumor progression by regulating the immune response. Considering the lack of knowledge, feline serum VISTA levels from cats with mammary carcinoma were compared with healthy controls, and with serum levels of PD-1/PD-L1, CTLA-4, LAG-3, IL-6, and TNF-α. In parallel, VISTA tumor expression was evaluated in FMC samples. The obtained data revealed that serum VISTA levels were significantly higher in cats presenting HER2-positive (p = 0.0025) or triple-negative subtypes (p = 0.0019), with higher serum levels in luminal A (p = 0.0025) correlated to the presence of metastasis (p = 0.0471). Furthermore, in HER2-positive or triple-negative tumors, correlations were obtained between serum VISTA levels and the serum levels of the above-mentioned molecules. In tumors, VISTA expression revealed a stronger intensity in cancer cells, when compared to TILs (p < 0.0001). Stratifying the samples by subtypes, a higher number of VISTA-positive TILs was observed in the HER2-positive subtype, compared with triple-negative tumors (p = 0.0138). In conclusion, results support the development of therapeutic strategies for HER2-positive and triple-negative FMC subtypes, reinforcing the use of cats as a human oncology model.


Introduction
Feline mammary carcinoma (FMC) is the third most common tumor in cats, representing 12% to 40% of all neoplasms [1,2] with an aggressive and infiltrative behavior [1][2][3], as in human breast cancer [4,5]. At the molecular level, human and feline breast cancer could be categorized into different subtypes, listed by its malignant behavior: luminal A, luminal B, HER2-positive, and triple-negative [6,7]. Briefly, considering the most aggressive (DFS) and the overall survival (OS) rates were also recorded, being 8.9 ± 7.5 months (n = 45; 95% CI: 6.5-11.2 months) and 14.3 ± 9.6 months (n = 42; 95% CI: 11.3-17.3 months), respectively. As controls, 14 serum samples were collected from healthy cats that underwent elective ovariohysterectomy at the Teaching Hospital of the Faculty of Veterinary Medicine, University of Lisbon, Portugal, presenting a mean age of 1 year (range 0.5-1.5 years). The animals were anesthetized before surgery and blood samples collection was performed without interfering with their wellbeing, with all the manipulation procedures being consented to by the owners.
In addition, the clinicopathological data for all the animals were registered. Briefly, the following features were recorded: breed; age; reproductive status; applied treatment; presence of multiple tumors, its size, and histopathological classification; hormonal (ER and PR) and HER2 status [6] and Ki-67 index [38] of the tumor masses; malignancy grade, scored based on the Elston & Ellis system [39]; presence of tumor necrosis, positive lymphatic invasion or lymphocytic infiltration, and/or cutaneous ulceration; involvement of the regional lymph node; and clinical stage (TNM system; Table 1). Moreover, the molecular subtyping of FMC [2,6] were characterized, with cats being stratified into four different groups: luminal A (n = 7), luminal B (n = 16), HER2-positive (n = 10) and triple-negative (n = 13) subtypes. The collected tumor tissue samples were subjected to a controlled fixation process for 24 to 48 hours using 10% buffered neutralized formalin (pH 7.2) and embedded in paraffin. In parallel, blood samples were centrifuged (1,500× g, 10 min, 4 • C), in order to separate the serum from clotted blood, and stored at −80 • C until further use. As recommended, after thawed, all samples were observed and discarded, when presented hemolysis [2,40].
Additionally, the homology between the immunogen location in human and feline VISTA proteins is 93.0% (UniProt, accession numbers: Homo sapiens Q9H7M9; Felis catus M3WI46).

Measurement of Serum VISTA Levels
To quantify the serum VISTA levels a commercial ELISA-based kit (R&D Systems, Minneapolis, USA; DY7126-05) was used and the manufacturer's instructions were followed. Firstly, using the recombinant proteins provided in the kit, a standard curve was plotted with 12-fold serial dilutions, and by a quadratic regression, the r 2 value was calculated (r 2 = 0.9959). Summarizing the technique, a 96-well plate was coated with the capture antibody and incubated overnight. After this initial step, and to prevent non-specific protein binding, plates were incubated for 1 hour with 1% bovine serum albumin (BSA) in phosphate-buffered saline (PBS). Then, standards and diluted serum samples of controls and diseased animals were added to the plate and incubated for 2 hours, followed by another 2-hour incubation step with the detection antibody, both at room temperature (RT). To conclude the procedure, the streptavidin conjugated to horseradish peroxidase (HRP) was added to the plate and incubated for 20 min at RT, and then the substrate solution in 1:1 H 2 O 2 and tetramethyl-benzidine (20 min at RT in the dark). To interrupt the reaction, a stop solution (2NH 2 SO 4 ) was mixed in each well and the absorbance was read by a spectrophotometer (FLUOStar OPTIMA, Microplate Reader, BMG, Ortenberg, Germany), using a 570 nm reference wavelength and a primary wavelength of 450 nm.
Furthermore, to conclude about the systemic response of the diseased queens, data on immune checkpoint molecules and cytokines (PD-1/PD-L1, CTLA-4, LAG-3, IL-6 and TNF-α), were considered in this study. The measurement of this molecules presented a similar procedure, as previously reported [19,22].

Assessment of VISTA Expression by Immunohistochemistry (IHC)
Initially, to define a representative tumor area, the formalin-fixed paraffin-embedded (FFPE) tumor tissue samples (n = 46) were stained with hematoxylin-eosin. Then, 3 µm slices of the selected FFPE tumor areas were sectioned (Microtome Leica RM135, Newcastle, UK) and mounted on a superfrost glass slide (SuperFrost Plus, Thermo Fisher Scientific, MA, USA). Samples were deparaffinized, hydrated, and antigen retrieval, using citrate buffer pH 6.0 (EnVisionTM Flex Target Retrieval Solution High pH, DAKO) was performed on a PT-Link module (DAKO, Agilent, Santa Clara, CA, USA) for 20 min at 96 • C. Thereafter, slides were cooled until reaching RT and immersed twice for 5 min in distilled water. For the IHC technique solutions from the commercial kit, the Novolink TM Max Polymer Detection System Kit (Leica Biosystems, Newcastle, UK) was used. Firstly, samples were incubated for 15 min with Peroxidase Block Novocastra Solution (Leica Biosystems), followed by blocking the unspecific antigenic recognition for 10 min using Protein Block Novocastra Solution (Leica Biosystems). Then, the anti-VISTA primary polyclonal antibody (Ref.: ab254581; Abcam, Cambridge, UK) diluted at 1:200 was incubated was added in a humidified chamber overnight at 4 • C, as recommended [28,29]. After this period, the detection polymer was incubated at RT for 30 min, followed by the detection solution of diaminobenzidine (DAB substrate buffer and DAB Chromogen, Leica Biosystems) for 5 min. During this technique, between all the incubation steps, the slides were washed twice for 5 min, using a PBS solution pH 7.4. To finish the IHC, samples were counterstained for 2 min with Gills hematoxylin (Merck, NJ, USA), dehydrated in an ethanol gradient and xylene, and mounted using Entellan mounting medium (Merck). As positive controls, to validate the technique, human tonsil and feline lymph nodes were used, whereas feline lymph nodes were also used as internal controls. In parallel, healthy tissue sections in the analyzed samples were evaluated as negative controls.
VISTA expression was scored according to the guidelines described by the International TILs Working Group 2014 applied to breast cancer. In sum, tumor tissue sections were evaluated, considering the area occupied by lymphocytes [41] in the stromal compartment, and excluding zones with necrosis and the TILs outside the tumor areas. Additionally, VISTA immunostaining was evaluated in membranes and the nuclei of the TILs, which were identified by their characteristic morphology. Furthermore, cancer cells were evaluated for VISTA immunostaining. The percentage of stained cells was defined as: 0% meaning the absence of staining to 100% implying that all cells were stained [29,42], with a cut-off value established at 5% [43] for a positive sample. Additionally, VISTA intensity was evaluated in the tumor samples to assess the immunostaining distribution. As previously reported [28,44,45], the staining intensity was classified by the following scores: 0 (no signal), 1+ (weak), 2+ (moderate) and 3+ (intense). The IHC slides were classified by two pathologists, in a blind scoring evaluation.

Data Statistical Analysis
The statistical analysis of the data collected in this study was performed using the GraphPad Prism software, version 5.04 (California, USA). Statistically significant results were considered when two-tailed p-values < 0.05 were observed, considering a 95% confidence interval (* p < 0.05, ** p < 0.01 and *** p < 0.001). The calculated mean values were reported with the proper standard error (SEM). The non-parametric Mann-Whitney test was used to compare the serum VISTA levels with several clinicopathological features. In order to choose the optimal cut-off values for serum VISTA levels, and to determine the specificity and sensitivity of the technique to properly identify tumor-positive animals, the receiver-operating characteristic (ROC) curves were calculated. Furthermore, using the Kaplan-Meier test, survival analysis was performed to evaluate the DFS in diseased queens. In parallel, to infer cats' immunosuppressive status, correlations between the previously reported serum PD-1/PD-L1 [19], CTLA-4, IL-6, TNF-α [22], and LAG-3 levels and the serum VISTA concentrations determined in this study were investigated applying the Spearman's rank correlation coefficient. To end the analysis and compare VISTA serum levels and tissue expression status in cats with mammary carcinoma, the non-parametric Kruskal-Wallis test was applied.

Serum VISTA Levels Are Higher in Cats with Luminal A, HER2-Positive or Triple-Negative Tumor Subtypes
Serum VISTA levels were evaluated in a healthy/control group and compared to those of a population of queens diagnosed with mammary carcinoma, stratified by its molecular subtypes ( Table 2). The statistical analysis revealed differences between the mean ranks of at least one pair of groups (p = 0.0016; Figure 1A). In fact, cats showing luminal A, HER2-positive or triple-negative tumor subtypes presented higher serum VISTA levels than the healthy population (10.31 ± 1.09 pg/mL. vs. 5.90 ± 0.51 pg/mL, p = 0.0025; 12.90 ± 3.73 pg/mL vs.
Furthermore, the correlation of serum VISTA levels with the recorded clinicopathological features, revealed that for the luminal A tumor subtype, an increase in serum VISTA levels was associated with the metastization process (p = 0.0471; Figure 2).

The Tumor VISTA Analysis Revealed a Heterogeneous Intensity along the Cancer Cells, While TILs Presented a Homogeneous Intensity in the General of the Tumor Samples
Tissue VISTA expression was evaluated in TILs and cancer cells of FMC samples, stratified by different subtypes (Table 4). In order to evaluate feline VISTA expression, human tonsil and feline lymph nodes were used as positive controls, revealing a similar staining pattern, with a strong IHC score (3+) in lymphocytes (Figure 4 A,B, respectively). Furthermore, normal mammary cells, as well as the tumor interstitial compartment, namely fibroblasts and collagen cells revealed no VISTA expression (0 score; Figure 4 C,D, respectively). The tumor tissue areas showed a heterogeneous intensity in the VISTA staining, in the general of the cancer cells, contrary to a homogenous VISTA intensity in TILs, independent of the tumor subtype ( Figure 4E,F).
In addition, FMC samples presented a different VISTA expression pattern along the tumor. A stronger VISTA intensity was observed in the epithelial luminal zones (3+ score) of the cancer cells ( Figure 5), when compared to the mesenchymal areas. Furthermore, a stronger VISTA intensity was observed in the membrane, when compared with the cell cytoplasm.
Moreover, the results revealed that VISTA-positive cancer cells and serum VISTA levels were found in 100% of tumors, independent of the molecular subtype. In parallel, VISTA-positive TILs and serum VISTA levels occurred in 82.61% (38/46 samples) of the studied population. Thus, by analysis of the tumor tissue samples, evaluating VISTA expression in TILs and cancer cells, and comparing it to the serum VISTA levels measured, differential and complementary information could be obtained.

VISTA Expression Is Higher in Cancer Cells, with VISTA-Positive TILs Being Associated with Less Aggressive Clinicopathological Features
A higher percentage of VISTA-positive cancer cells were observed in tumor areas, when compared to TILs (p < 0.0001; Figure 6A). Interestingly, when VISTA intensity is multiplied by the percentage of positive cells, a correlation was observed in the VISTA expression between cancer cells and TILs (p = 0.0022, r = 0.4543; Figure 6B). Moreover, the percentage of VISTA-positive TILs occurred higher in malignancy grade II (p = 0.0025; Figure 6C), with no TILs present in malignancy grade I.  In addition, FMC samples presented a different VISTA expression pattern along the tumor. A stronger VISTA intensity was observed in the epithelial luminal zones (3+ score) of the cancer cells ( Figure 5), when compared to the mesenchymal areas. Furthermore, a stronger VISTA intensity was observed in the membrane, when compared with the cell cytoplasm. Moreover, the results revealed that VISTA-positive cancer cells and serum VISTA levels were found in 100% of tumors, independent of the molecular subtype. In parallel, VISTA-positive TILs and serum VISTA levels occurred in 82.61% (38/46 samples) of the studied population. Thus, by analysis of the tumor tissue samples, evaluating VISTA expression in TILs and cancer cells, and comparing it to the serum VISTA levels measured, differential and complementary information could be obtained. when compared to TILs (p < 0.0001; Figure 6A). Interestingly, when VISTA intensity is multiplied by the percentage of positive cells, a correlation was observed in the VISTA expression between cancer cells and TILs (p = 0.0022, r = 0.4543; Figure 6B). Moreover, the percentage of VISTA-positive TILs occurred higher in malignancy grade II (p = 0.0025 Figure 6C), with no TILs present in malignancy grade I.

VISTA Expression Is Higher in HER2-Positive Tumors Than in Triple-Negative Carcinomas
Stratifying tumors by their subtype, a significant difference of VISTA-positive TILs expression was observed between HER2-positive and triple-negative tumors (p = 0.0138; Figure 7A), with no differences in VISTA-positive cancer cells (p = 0.4677; Figure 7B).
Cancers 2021, 13, 5559 11 of Stratifying tumors by their subtype, a significant difference of VISTA-positive TI expression was observed between HER2-positive and triple-negative tumors (p = 0.01 Figure 7A), with no differences in VISTA-positive cancer cells (p = 0.4677; Figure 7B). Furthermore, considering the HER2-positive tumor subtype and evaluating t cancer cells, a higher VISTA expression was obtained in tumors with a metastizati process (p = 0.0200; Figure 8). Furthermore, considering the HER2-positive tumor subtype and evaluating the cancer cells, a higher VISTA expression was obtained in tumors with a metastization process (p = 0.0200; Figure 8).
Furthermore, considering the HER2-positive tumor subtype and evaluati cancer cells, a higher VISTA expression was obtained in tumors with a metast process (p = 0.0200; Figure 8).

Discussion
The majority of feline mammary tumors are malignant, with the HER2-positi triple-negative subtypes presenting a poor prognosis [6,46]. Therefore, early detecti effective treatments will have a major impact on survival time [47], with the develo of more efficient therapies being imperative. Furthermore, the immune chec proteins produced by cancer cells, as a strategy to escape from the immune s downregulating the systemic response, could reveal as important therapeutic t Currently, there are several proteins from the immune system, targeted as adjuv the treatment of human breast cancer, with some of these already proposed to be

Discussion
The majority of feline mammary tumors are malignant, with the HER2-positive and triple-negative subtypes presenting a poor prognosis [6,46]. Therefore, early detection and effective treatments will have a major impact on survival time [47], with the development of more efficient therapies being imperative. Furthermore, the immune checkpoint proteins produced by cancer cells, as a strategy to escape from the immune system, downregulating the systemic response, could reveal as important therapeutic targets. Currently, there are several proteins from the immune system, targeted as adjuvants in the treatment of human breast cancer, with some of these already proposed to be tested in diseased cats [19,22]. However, resistance mechanisms to immune checkpoint inhibitors have already been described, e.g., to anti-PD-L1 and anti-CTLA-4 compounds [48,49], being important to have different therapeutic options for the improvement in cats' prognosis. Thus, a better understanding of the VISTA expression in diseased animals and FMC samples could reveal not only new non-invasive diagnostic or prognostic biomarkers but also improved options as therapeutic targets for cats with mammary carcinoma.
For the first time, serum VISTA levels were evaluated in a diseased population, and queens diagnosed with the most aggressive mammary carcinoma subtypes, HER2-positive or triple-negative [6], revealed higher serum VISTA expression levels (p = 0.0025 and p = 0.0019, respectively). These results are in concordance with the findings described for human breast cancer [21,29,50], suggesting a conserved role of the VISTA protein between both species. Interestingly, in cats with luminal A tumor subtype, an increase in serum VISTA levels (p = 0.0025) was also observed, being associated with the metastization process (p = 0.0471). This result suggests the need for the VISTA protein for the escape and evasion of cancer cells from the immune system [44,[51][52][53].
The VISTA protein is known as being responsible for the resistance to anti-PD-1/anti-CTLA-4 treatments, reported in prostate cancer [54,55], with an overall success rate of 30% of these targeted therapies [17,28,56], suggesting that blocking VISTA could be a good strategy to improve patients' prognosis. Thus, and as reported in the tumor microenvironment (TME) of distinct human cancers [21,28,29,50,57], positive correlations were observed between serum VISTA levels and several immune inhibitory checkpoint proteins [58], such as serum PD-1 (p < 0.0001), PD-L1 (p < 0.0001), CTLA-4 (p < 0.0001), and LAG-3 (p = 0.0002), as well as with the cytokines IL-6 (p = 0.0016) and TNF-α (p < 0.0001), proposing not only a co-regulation among these molecules [20,55,59,60] but also an immunosuppressive status of these animals, reinforcing previously reported data [19,22]. Furthermore, cats with triple-negative carcinomas presented a positive correlation with all the immune checkpoint proteins and cytokines described, suggesting that this tumor is highly immunogenic, as in humans [43].
In human cancers, VISTA plays a crucial role in the anti-tumor immunity and potentiates a local immunosuppressive status [28,30], its overexpression occurring in the TME [30], presenting immunostaining in the membrane and/or cell cytoplasm [43]. In parallel, in feline tumor samples, TILs present a VISTA-positive expression in the cell membrane and cytoplasm, while in VISTA-positive cancer cells, the protein's expression occurs mainly in the cell membrane, when compared to the cell cytoplasm. The cytoplasmic staining may represent intracellular stores of VISTA that will be released with the appropriate stimulation [44,61]. Furthermore, VISTA had a specific spatial/geographical pattern [25], similar to what was obtained in the feline tumors. In these samples, a homogenous VISTA expression was observed in TILs, while in cancer cells a heterogeneous VISTA expression was obtained, according to the findings in human cancers. Thus, this result indicates different cancer cell subpopulations that vary in their genetic, phenotypic, or behavioral characteristics, being associated with cancer progression and therapeutic resistance [62,63].
The statistical analysis of the immunostaining in the FFPE FMC samples revealed a higher expression of VISTA in cancer cells when compared to TILs (p < 0.0001), contrary to what is reported in a variety of human cancers [29,32,43,44,63,64]. Interestingly, a positive correlation was obtained by the increase in VISTA expression in cancer cells and TILs (p = 0.0022), which suggests a paracrine mechanism used by cancer cells to communicate with the immune system, as reported for PD-L1 in breast cancer patients [65][66][67]. Furthermore, a higher expression of VISTA-positive TILs occurs in grade II FMC, compared to grade III (p = 0.0025), which is in accordance with findings in human breast cancer, with VISTA TILs' expression being associated with a good prognosis [28,29]. Furthermore, TILs-positive VISTA expression was higher in HER2-positive carcinomas, contrasting with triple-negative subtype (p = 0.0138), as reported in human breast cancer [21,28,29,50]. In addition, previous studies in breast cancer patients revealed that VISTA leads to a local immunosuppressive microenvironment [44,[51][52][53], strengthening the results reported in this study, and the systemic immunosuppressive status described in queens diagnosed with HER2-positive tumors [6,19,22]. Interestingly, in the HER2-positive subtype, VISTApositive cancer cells were associated with the tumor metastization process (p = 0.0200), suggesting that cancer cells release VISTA in order to downregulate the tumor immune response [44,[51][52][53] and to evade the primary tumor location.
Notably, despite the lower concentration of serum VISTA levels found in the diseased animals, when compared to other proteins reported [19,22], an optimal cut-off value for VISTA was possible to determine, presenting higher sensibility, when compared, e.g., to PD-1 and PD-L1 [19]. Furthermore, the high concordance rate found between serum VISTA levels and VISTA-positive cancer cells in the FMC cases suggests VISTA protein as a non-invasive diagnostic biomarker, as reported for breast cancer patients [42,68].

Conclusions
In conclusion, these results demonstrated that higher serum VISTA levels were correlated with the most aggressive tumor subtypes (HER2-positive or triple-negative) and poor clinicopathological features, as the association to the metastization process in luminal A subtype. Furthermore, in cats presenting HER2-positive mammary carcinomas, an association was observed comparing the VISTA and PD-1 levels in circulation. In parallel, queens with triple-negative mammary tumors demonstrated correlations between different immune checkpoint proteins, such as PD-1/PD-L1, CTLA-4, and LAG-3, as well as between the cytokines IL-6 and TNF-α, suggesting a systemic immunosuppression of these cats. Additionally, the microenvironment of HER2-positive tumors revealed a high VISTA expression in TILs when compared to the triple-negative subtype, being the VISTAcancer cells expression associated with the metastization process. Altogether, the results