The Clinical and Pathological Characteristics of Mammary Neoplasms with Malignant Mesenchymal Components in Female Dogs

Mammary neoplasms with malignant mesenchymal components are not common in female dogs, and they are poorly understood. As such, this study aimed to describe the clinical presentation, histological ndings, and the COX-2 immunohistochemical expression of mammary neoplasms in female dogs with malignant mesenchymal components, as well as verify the relationships between the different neoplasm types and these aspects. We selected 41 female mammary neoplasms (23 carcinosarcomas, 16 sarcomas, and 2 sarcomas in mixed tumors). Medical records were reviewed to obtain clinical data. Subsequently, histological slides were analysed to establish histological parameters, and immunohistochemistry was used to assess the expression of COX-2 receptors. Carcinosarcomas and sarcomas developed as large tumours, mainly in the abdominal and inguinal mammary glands, with frequent intratumoral necrosis and a low frequency of nodal metastasis. Fifty-eight percent of the cases of malignant mesenchymal proliferation were identied as osteosarcomatous, and 24.5% chondrosarcomatous and brosarcomastous each. The osteosarcomatous pattern was the most predominant type in sarcomas and carcinosarcomas, and was the only one that resulted in vascular invasion, regional lymph node metastases, and higher histologic grades. High COX-2 expression was detected in 10% of the carcinosarcomas and 25% of the sarcomas. In conclusion, sarcomas and carcinosarcomas showed similar results regarding the clinical and pathological aspects. Discovering carcinosarcomas and sarcomas with high COX-2 expression suggests that, in some cases, these neoplasms may respond to therapy with COX-2 inhibitors.


Introduction
Neoplasms with malignant mesenchymal components in the mammary glands of female dogs are called sarcomas, carcinosarcomas, and sarcomas in mixed tumors. They are characterized by neoplastic cells with poorly differentiated connective tissue (Cassali et al. 2011;Goldchmidt et al. 2017). Sarcomas comprise approximately 1% of mammary neoplasms in female dogs, while carcinosarcomas may account for 4% (Nunes et al. 2018) and their presence generally points toward a poor prognosis and a high occurrence of metastases (Hellmen et al. 1993 Clinical and pathological characteristics, such as size, lymph node status, distant metastasis, vascular invasion, and the expression of immunohistochemical markers like cyclooxygenase-2 (COX-2), have been used to assess and issue the prognoses of female dogs with malignant mammary neoplasms (Queiroga et al. 2007; Nunes et al. 2018).
COX-2 is an inducible enzyme that is involved in in ammatory processes, neoplastic transformation, and tumour progression (Heller et al. 2005). COX-2 overexpression has been recognized as a negative prognostic marker in canine mammary malignant tumours (Millanta et  Nd enhance cell motility and adhesion (6). Overexpression of COX-2 causes tumorigenesis in animal models, and its inhibition has significant effect on reducing the incidence and progression of tumors in animal models and in the treatment of some cancer patients (7,8). Evaluation of COX-2 expression and its roles in treatment and prevention has been performed mostly on tumors of epithelial origin; little is known about the roles of COX-2 in tumors of mesenchymal origin, especially GIST and enhance cell motility and adhesion (6). Overexpression of COX-2 causes tumorigenesis in animal models, and its inhibition has significant effect on reducing the incidence and progression of tumors in animal models and in the treatment of some cancer patients (7,8). Evaluation of COX-2 expression and its roles in treatment and prevention has been performed mostly on tumors of epithelial origin; little is known about the roles of COX-2 in tumors of mesenchymal origin, especially GIST However, due to the considerable histologic variety of malignant mesenchymal proliferation, and the low frequency of occurrence of malignant mesenchymal mammary neoplasms, very little is known about COX-2 expression and the clinical and pathological pro les of these tumours.
As such, this study aimed to describe the clinical presentation, histological ndings, and the COX-2 immunohistochemical expression of malignant mesenchymal mammary neoplasms in female dogs, as well as verify the relationships between different types of neoplasms and these aspects.

Case selection
Formalin xed para n embedded canine neoplasms with malignant mesenchymal components sampled between 2014 and 2018 were selected from pathological archives of the Setor de Patologia Veterinária da Universidade Federal do Rio Grande do Sul, SPV-UFRGS. The criteria of inclusion were diagnoses of carcinosarcoma, sarcoma, or sarcoma in a mixed tumour. We excluded cases with diagnoses of other types of malignant mammary neoplasms; cases for which the slides could not be accessed; and cases with histologic sections that were not representative of the lesions (Araújo et al. 2017).
Exam request records were examined to collect the following information: age, breed, affected mammary glands, lesion size, and regional lymph node, and distant metastases.

Histological assessment
The selected cases were histologically reviewed on Hematoxylin and Eosin stained sections. Standard criteria were followed for the classi cation of neoplasms such as sarcomas and (Misdorp et al. 1971;Goldschmidt et al. 2011) and carcinosarcomas (Misdorp et al. 1999;Cassali et al. 2020). The mammary origin of the neoplasms was determined based on the presence of mammary acini in the histological section.
The histological parameters evaluated were: type of sarcomatous proliferation, intratumoral necrosis (present or absent), vascular invasion (lymphatic and hematogenic), lymph node invasion (present or absent), and grade. Lymphatic invasion was de ned as the presence of neoplastic cells in spaces with an endothelium lining. Vascular invasion was de ned as the presence of neoplastic cells in spaces with an endothelium lining, muscle and elastic bres, and intraluminal erythrocytes.
We followed the established criteria for grading soft tissue sarcomas in humans as a grading scheme for our study (Trojani et al. 1984). This has been previously done to grade mammary sarcomas in female dogs (Dolka et al. 2013). The most representative area of the sarcoma was used to examine the predominant tumour type. This assessment took into consideration: the degree of differentiation in the malignant mesenchymal component, the mitotic rate (that was calculated by counting of mitotic gures in ten high-power elds (400x, 0.239 mm 2 area), and intramural necrosis. The mitotic rate was established by counting at the periphery of the neoplastic proliferation. Areas affected by necrosis, haemorrhage, in ammation or artifacts were excluded.
We used the Nottingham system, modi ed by Elston and Ellis (1998), to grade the carcinoma components of carcinosarcomas.
Histological sections of the lymph nodes were reviewed to assess for neoplastic cells and the type of metastatic proliferation (epithelial, mesenchymal, or both). Lymph nodes were considered positive for metastasis when isolated cells or groups of non-lymphoid neoplastic cells were detected in the subcapsular sinus or lymph node parenchyma.

Immunohistochemistry (IHC)
Histological sections 4 μm thick were sliced from each para n block and put on gelatinized slides that were dewaxed in xylol and rehydrated in a series of progressively diluted alcohols. For antigenic recovery, the sections were incubated in a water bath containing citrate buffer (Recovery solution, Dako Cytomation, Carpinteria, CA, USA) pH 6.0 for 20 minutes. Endogenous peroxidase was blocked by incubating the slides in 0.3% hydrogen peroxide with absolute methanol for 15 minutes at room Sections of canine breast tumours that had previously tested positive for vimentin and cytokeratin AE1/AE3 were used as positive controls. For COX-2, sections of canine kidney that had previously tested positive were used as a positive control. For negative control, PBS was used to omit the primary antibody in the three markers.

Assessing immunomarkers in tumoral cells
The intracytoplasmic staining patterns of antibodies were used to examine vimentin, cytokeratin AE1/AE3, and COX-2 expression.
The expression of COX-2 in the tumour cells was evaluated in a semi-quantitative way, taking into account the distribution parameters and intensity of positive markings in the ve high-power elds [19]. To assess distribution, markings were evaluated at: 0 (no immunostaining); 1 (less than 10% immunostaining); 2 (between 10% to 30% immunostaining); 3 (between 31% to 60% immunostaining); 4 (more than 61% immunostaining). The intensity of the immunostaining was assessed as follows: 0 (absent); 1 (+); 2 (++); 3 (+++). Additionally, for each eld, the scores were multiplied, and the result was the average of all elds. The nal scores were divided into two extracts: from 0 to 5 and from 6 to 12 with scores of 0 to 5 considered low and scores from 6 to 12 considered high (Lavalle et al. 2009).

Statistical analysis
The data were recorded in an electronic Excel ®2010 spreadsheet. Subsequently, we analysed the absolute frequencies and percentages to relate variables. We evaluated the possible association between histological types (sarcomas and carcinosarcomas) and clinical and pathological variables (size, intratumoral necrosis, intratumoral vascular invasion, the histologic grade of the mesenchymal component, and lymph node mesenchymal metastasis). We also examined the possible association between mesenchymal subtypes (osteosarcomatous versus non-osteosarcomatous) and the morphological variables (necrosis, intratumoral vascular invasion, mesenchymal lymph node metastasis, and the histologic grade of the mesenchymal component). All associations were analysed using Fisher's exact test (p <0.05) and STATA 14.0 software (Stata Corp LP).

Results
There were 1621 diagnosed cases of mammary neoplasms in female dogs diagnosed in the study period and 12.2% (198) presented a malignant mesenchymal component. Carcinosarcomas accounted for 7.8% (126/1621) of the total number of neoplasms, while sarcomas made up 3.2% (52/1621) and sarcomas in mixed tumours 1.2% (20/1621). We selected 41 of these cases for analysis, of which 16 were sarcomas ( Fig. 1a), 23 were carcinosarcomas (Fig. 1b), and 2 were sarcomas in mixed tumors.
In total, 29.2% of the female dogs (12/41) were mixed breed dogs (MBD) and 70.7% were purebred (29/41). The mean age was 11.3 years (ranging from 5 to 17 years). Non-spayed female dogs made up 60.9% (25/41) of the cases. Tumour sizes varied between 2.5 cm and 20 cm, with an average of 10.5 cm as the largest dimension. We had access to the radiologic exam results for distant metastasis for 26 cases; of these, 7.6% (2/26) presented images that suggested pulmonary metastasis.
The most frequent type of mesenchymal proliferation was osteosarcomatous; it made up 58.5% (24/41) of the sample. Pure sarcomas presented malignant mesenchymal differentiation of the following types: osteosarcomatous 62.5% (10/16), brosarcomatous 31.2% (5/16), and liposarcomatous 6.2% (1/16 The osteosarcomatous component was characterized by the proliferation of highly pleomorphic and poorly differentiated osteoblasts of an ovoid and fusiform shape, intermixed with varying quantities of cartilaginous, collagenous, or osteoid matrix (sometimes mineralized), frequent mitotic gures, and occasional giant multinucleated cells. Chondrosarcomas were characterized by a proliferation of pleomorphic and poorly differentiated ovoid cells, intermixed with islands of a disorganized and moderately to intense cellularized hyaline matrix. The brosarcomatous component was characterized by the proliferation of fusiform cells, with moderate pleomorphism, arranged in multidirectional bundles, with elongated or ovoid nuclei and infrequent mitoses. In turn, the liposarcoma component presented polygonal, elongated, or fusiform cells, interspersed with lipid vacuoles of varying sizes and peripheral nuclei and few mitotic gures.
There was no statistical association between the osteosarcomatous and non-osteosarcomatous mesenchymal subtypes (chondrosarcomatous, brosarcomatous, and liposarcomatous) or the histologic grade of the mesenchymal component (p = 0.274). The association between the other variables (intratumoral necrosis, mesenchymal vascular invasion, and lymph node metastasis of the mesenchymal component) and the mesenchymal subtype was not tested due to the insu cient number of samples for statistical analysis. The distribution of variables according to the type of mesenchymal proliferation is shown in Table 1.
Adjacent lymph node samples were available for 58.5% (24/41) of the cases. Whenever there was a metastasis of the mesenchymal component, the proliferation was of the osteosarcomatous type. The frequency of lymph node metastases for sarcomas and sarcomas in mixed tumours was calculated at 30% (3/10), while 21,5% (3/13) of carcinosarcomas presented mesenchymal metastasis. There was lymph node metastasis of carcinomatous components in 7.1% (1/14) of the carcinosarcomas. This was of the anaplastic subtype, but the primary tumour exhibited a simple papillary pattern. None of the lymph nodes we assessed presented metastasis with both components. There was intratumoral vascular invasion by the mesenchymal component in 19.5% (8/41) of the cases; of these four cases of invasion occurred in blood capillaries and two cases in lymphatic vessels. In two cases, lymphatic and blood invasion occurred (4.7%).
Low-grade mesenchymal components were well-differentiated, with less than 50% of the parenchyma affected by necrosis, and with up to nine atypical mitoses in 10 high-power elds. The intermediate grade components presented poorly differentiated parenchyma, with variable degrees of intratumoral necrosis and ten to nineteen atypical mitoses. The high histologic grade components were undifferentiated, with necrosis in more than 50% of the parenchyma, and more than 20 atypical mitoses in 10 high-power elds. A high degree of the mesenchymal component was observed in 78% (32/41) of the sarcomas and carcinosarcomas, an intermediate degree in 9.7% (4/41), and a low degree in 12.2% (5/41) of the cases. Carcinomatous components were classi ed as grade I in 62.5% (15/24) of the cases, and grade II in 37.5% (9/24).
Results revealed high COX-2 expression in 10% (2/10) of the carcinosarcomas and in 25% (4/16) of the sarcomas and sarcomas in mixed tumors. (Table 2). One carcinosarcoma presented high COX-2 expression in the carcinomatous component that was cytoplasmatic and multifocal. Another carcinosarcoma showed high expression in the sarcomatous component. In the sarcomas, this expression was cytoplasmatic and multifocal and was detected in osteoblasts, broblasts, chondroblasts, and occasionally in multinucleated giant cells. COX-2 expression was also observed in areas affected by in ammation and necrosis. However, these were not considered in the analyses.
There was no statistically signi cant association between histological subtypes (sarcoma and carcinosarcoma) and the clinical and morphological characteristics we evaluated ( Table 2).
In immunohistochemical tests for cytokeratin AE1/AE3, intracytoplasmic staining was observed in carcinomatous components (Fig. 2a) and lymph node metastasis of the anaplastic type. Intracytoplasmic staining for vimentin was observed in the sarcoma component of the four histologic types: brosarcoma (Fig. 2b), chondrosarcoma (Fig. 2c), osteosarcoma (Fig. 2d), and liposarcoma. The same pattern of vimentin marking was noted in cases of lymph node mesenchymal metastasis of the osteosarcomatous component.

Discussion
This study allowed us to comprehensively explore the histological variations of sarcomas and carcinosarcomas, as well as possible relations between these variations, clinical presentations, and morphological criteria of aggressiveness.
Few reports have addressed the frequency of mammary sarcomas in female dogs. In the female dogs we assessed, the frequency of mammary neoplasms with malignant mesenchymal components was similar to data from a previous report (13%) (Misdorp et al. 1971). However, in a recent survey with a fuller description, other authors reported that neoplasms with malignant mesenchymal components accounted for 5% of the total number of tumours ( The rst challenge in performing a histologic diagnosis of mammary neoplasms with mesenchymal components is determining the benign or malignant character of the mesenchymal proliferation. Carcinosarcomas must be histologically distinguished from carcinomas in mixed tumors that are characterized by carcinomatous and benign mesenchymal components. To do so, the key is to identify malignant characteristics in the mesenchymal cells, such as the level of differentiation in the matrix, high cellularity, elevated nuclear and cellular pleomorphism, karyomegaly, irregular nuclear membrane, prominent nucleoli, and multinucleation Goldschmidt et al. 2017). The same aspects should be considered when distinguishing between sarcomas and benign mesenchymal tumours.
For all the cases in our study, we performed an immunohistochemical analysis of the expression of intermediate vimentin and cytokeratin AE1/AE3 laments. This made it possible to discriminate between epithelial and mesenchymal proliferation. All of the mesenchymal components of the sarcomas and carcinosarcomas presented vimentin markers (cell markers of mesenchymal origin), and all the epithelial components of the carcinosarcomas presented cytokeratin AE1/AE3 markers (cellular markers of epithelial origin) (Boos et al. 2011;Dolka et al. 2013). Besides this, in some cases, the use of immunohistochemistry made it possible to detect small foci of sarcomatous and carcinomatous proliferation in lymph nodes. These were not detected with H&E staining.
The correct classi cation of sarcomas by the type of proliferation is another challenge in morphological diagnosis. This is due to the varying nature of the matrix in the same tumour. Different combinations of mesenchymal proliferation (osteosarcomatous, chondrosarcomatous, brosarcomatous, and liposarcomatous) and epithelial proliferation were observed. Similarly, this has previously been described in human (Yakan et al. 2014) and canine mammary glands ). The histologic variety of the mesenchymal components of neoplasms has been attributed to myoepithelial cell transformations ) epithelial-mesenchymal transition (EMT) (Kokkinos et al. 2007), or totipotent cells with a high differentiation capacity (Hellmen and Lindgren 1989).
We classi ed all tumours that presented a malignant osteoid matrix as osteosarcomas because, among the sarcomas found in different anatomical locations, this type exhibits the most aggressive type of proliferation (Goldschmidt et al. 2017).
Although the most frequent mesenchymal subtype in our study was osteosarcoma, another report on sarcomas in the mammary glands of female dogs demonstrated a higher frequency of brosarcomas (Dolka et al. 2013). On the other hand, another investigation revealed that, among extraskeletal osteosarcomas in dogs, the mammary gland is the most frequent location (Kokkinos et al. 2007). There were no pure chondrosarcomas in our ndings. However, this histologic type has already been found in the mammary glands of female dogs (Serin and Aydogan 2009; Tavasoly et al. 2013).
Whether they are located in the mammary glands or other anatomical locations, osteosarcomas can be classi ed into different types, due to the heterogeneity of their cellular populations and diverse extracellular matrix formation. At times, the same neoplasm can simultaneously present different types and this can make them di cult to classify. Moreover, it corroborates the highly pleomorphic behaviour of this component. Any correlations between histologic types and prognoses are still uncertain, but skeletal broblastic osteosarcomas have been associated with better prognoses while the opposite has been reported for the telangiectatic type (Thompson and Dittmer 2017). Other researchers have found no correlation between the histologic subtypes of skeletal osteosarcomas and histologic grades (Nagamine et al. 2015). Similarly, we found no correlation between histologic subtypes and other clinical or pathological variables. However, other studies with larger numbers can give answers about the importance of this correlation.
Our ndings concerning the highest frequency of the mesenchymal lymph node metastasis corroborate those of previous studies that related mammary gland sarcomas in female dogs with greater metastatic potential and a worse prognosis (Hellmen et al. 1993). However, the similarity between the frequency of lymph node metastases in sarcomas and carcinosarcomas (30% and 21,5% respectively), substantiated the malignancy of both histologic types (Goldschmidt et al. 2017). The prevalence of mesenchymal lymph node metastases with osteosarcoma components con rmed its malignancy and capacity to disseminate along the lymphatic pathway (Hellmen 2014) and reinforces the importance of assessing regional lymph nodes.
Our nding of a low frequency of cases with radiographic evidence of distant metastasis was different from other results in the literature. These studies examining the mammary glands of humans and female dogs related mammary gland sarcomas to the frequent occurrence of metastasis to other organs (Benjamin and Lee 1999;Yakan et al. 2014). The lack of radiographic exam results weakens knowledge regarding the behaviour of the tumours, from the time of the rst diagnosis onwards, and this divergence was probably due to issues of access to the results of pre-surgical radiographic exams, as well as the ones carried out during the clinical history.
Determining the histologic grade is important for issuing a prognosis and deciding on a therapeutic plan for invasive mammary carcinomas (Karayannoupoulou et al. 2005). However, there is no established system for grading the mesenchymal components of mammary sarcomas in female dogs.
A previous study used a grading system to assess mammary sarcomas in female dogs, and most of the cases demonstrated a high histologic grade (Dolka et al. 2013). By using a grading system, the authors also observed that sarcomas with a higher histologic grade were associated with more proliferative activity (Ki67 index). This suggests there may be implications regarding the grade and prognosis of these neoplasms. We used the same grading system and also observed a prevalence of high-grade sarcoma components among the sarcomas and carcinosarcomas we assessed. However, further prognostic studies with larger sample sizes are needed to validate this grading system. There are few published reports regarding COX-2 expression in the mesenchymal components of the mammary glands in female dogs. One study found intense expression in mammary brosarcoma (Arenas et al. 2016). Another investigation discovered signi cant COX-2 expression in appendicular osteosarcomas in dogs (Millanta et al. 2012).
The percentage of carcinosarcomas with a high COX-2 score was lower than the number previously described by Queiroga et al. (2010) and Carvalho et al. (2016). However, it was greater than the percentage reported by Lavalle et al. (2012). Probable justi cations for these differences in results include the use of dissimilar antibodies, inherent distinctions in components and antibodies, or inherent variations in immunohistochemical exams.
Recently, COX-2 overexpression was associated with more aggressive histological types, such as carcinosarcomas (Pastor et al. 2020). This overexpression in carcinosarcomas and other mammary carcinomas was also linked to immunohistochemical markers for angiogenesis, proliferation and in ammation, shorter disease-free survival, and shorter global survival (Queiroga et al. 2010;Carvalho et al. 2016).
In conclusion, sarcomas and carcinosarcomas were found to be clinically similar and developed as large tumours, mainly in the abdominal and inguinal mammary glands, with frequent intratumoral necrosis but a low frequency of nodal metastasis. Sarcomatous histological patterns varied and, when compared with the other types of mesenchymal proliferation, the osteosarcomatous subtype was the only one that resulted in vascular invasion, metastasis in regional lymph nodes, and a higher histologic grade. Besides this, the ndings of carcinosarcomas and sarcomas with high COX-2 expression suggest that, in some cases, these types of neoplasms may respond to therapy with COX-2 inhibitors. However, the usefulness of COX-2 inhibitors in treating these neoplasms remains to be further investigated.

Declarations
Funding: The authors are thankful to the Conselho Nacional de Desenvolvimento Cientí co e Tecnológico (CNPq) for nancing this study (Proc. 454297/2014-9) and concession of scholarship on scienti c initiation.
Con icts of interest: The authors declare that there is no con ict of interest.
Availability of data and material: The datasets generated during the current study are available from the corresponding author on request.   Vimentin markers in an accentuated proliferation of osteocytes. Note the absence of markers in the epithelial cells. DAB, 40X (d)