The frequency of malignancies such as breast cancer has been constantly increasing. This pathological tendency has now become the most common form of malignancy among women in almost the whole Europe and North America continents. In bitches the incidence of mammary tumour has been found to be at the level of even three times more frequent than in human [1]. About 50% of all mammary tumours are malignant [2]. The aetiology of breast cancer is very complex and not completely comprehend. Genetic, hormonal, dietary, environmental and carcinogenic factors are known as being mediators of tumourigenesis in human and canine [1–3].

Although extensive research on breast cancer has been in the process for over decades, yet still, challenges prevail in early diagnosis and management of cancer patients. Furthermore, molecular mechanisms underlying breast cancer progression remain poorly understood [4]. This deficit has led to an increased interest in diverse and complex molecular biology of the malignancy. Investigators are focused on discovering novel predictive markers in mammary cancer. Such bio-markers play the pivotal role in terms of efficiency in patients management. The conventional approach to cancer therapy provide treatment according to actual organ or tissue in which the cancer originates. However, different intracellular signalling pathways are perturbed in various types of cancer. Thus, the patients with the same type of cancer often have dissimilar genetic defects in their tumours and respond in a heterogeneous manner to anticancer treatment [4].

That is why the objective of this study had been targeted to assess gene expression profiles of canine mammary carcinoma (in unsupervised manner) diagnosed by pathologists as grade 1 (well differentiated), grade 2 (moderately differentiated) or grade 3 (poorly differentiated) and to be compared against their molecular and pathological classifications. The molecular classification reflects biological processes and pathways within the tumour cell, not only the morphological features. It is also helpful in better understanding of cancer biology (to find genes and pathways responsible for tumourigenesis or to explore molecular interactions within the tumour). Our analysis have exposed, that some pathologically distinct tumours may have similar gene expression and vice versa. Probably that is why sometimes patients with similar pathological type of cancer have various outcome and respond in a heterogeneous manner to anticancer agents. Despite there is a significant correlation between pathological diagnosis and prognosis in canine patients with mammary tumour, there is still a need to improve treatment strategies [5, 6]. Therefore, the molecular interaction within the tumour should be kept explored. That is why in our opinion pathological and molecular examination could significantly complement one another.

Canine mammary cancer constitute more than 40% of all malignancies in dogs [13] thus this is a serious clinical problem. Unfortunately, its molecular biology has not been fully recognized yet. So far, there are only three reports of wide gene expression analysis in canine mammary tumour tissues [14–16]. The most interesting study was conducted by Klopfleisch et al. [15]. The authors identified a gene expression profile of canine mammary tumours which was associated with an early metastatic spread to the lymph nodes. This “expression profile” can be used as a marker of metastasis prediction. In the opposite reflection, as far as we are aware, there have not been published any results regarding comparison between pathological and molecular diagnosis in canine mammary malignancy.

Thus, the objectives of our study were three-fold: 1) to assess gene expression profiles of canine mammary carcinomas of various grade of malignancy, 2) to perform unsupervised hierarchical classification based on their molecular portraits, and 3) to compare their molecular classification with the pathological diagnosis.

The results of our unsupervised analysis allowed the classification of the examined tissues into three groups. The first one (tumour group no. 1) significantly differed from the others and consisted of four grade 3 carcinomas and one grade 2 carcinoma (these tumours showed classical pathological features of higher biological aggressiveness). The second group (tumour group no. 2) consisted of four grade 1 carcinomas. The last one (tumour group no. 3) was very heterogeneous (based on the pathological diagnosis of these tumours) and consisted of two grade 1 carcinomas, two grade 3 carcinomas and five grade 2 carcinomas. Hierarchical clustering presented the view that the gene expression in the tumour groups no. 2 and no. 3 was relatively similar, whereas gene expression in the tumour group no. 1 (the tumour group showing classical pathological features of higher biological aggressiveness) differed significantly (Figure 1; Additional file 1: Tables S1–S6). Our results showed that pathologically different tumours may have similar gene expression and vice versa (Figure 1, Figure 3) what can be particularly observed in tumour group no. 3. These results can explain why sometimes very similar (pathologically) tumours respond for treatment in a different manner. We also consider these results as partially related with the fact that the tissue sample taken to the pathological examination is not the same that was taken to the microarray analysis. However, neighbour tissue samples were collected for both analyses. Moreover, within the four differently diagnosed tumours only one sample was of a mixed nature (tumour no. 88). Thus, in our opinion the risk of probes contamination in the other three cases is unlikely.

For clinicians, the most important seems to be the tumour group no. 1, as these tumours have shown classical pathological features of higher biological aggressiveness. The gene expression of these tumours also differed significantly from the other groups, thus only these data files are being as the discussion in this documentation sheet. From our perspective a few of up-regulated genes seem to be particularly interesting.

Among 25 up-regulated genes in tumours classified into group no. 1, compared to the other two groups, PANTHER software recognized two of them as responsible for chemotherapy resistance (ATP2C1 and ABHD7), two associated with myeloid cells infiltration of tumour mass (SIGLEC10 and SIGLEC11), one related to hypoxia (HYOU1), four genes related to cancer cell motility (MYL6, MYL6B, GPI and ARHGAP), one associated with DNA-repair (XRCC2), one with p53-signalling (RFWD2) and three transcriptional factors (NR6A1, RDH5, CYP19).

As of this matter what we do believe in, is the fact that important up-regulated genes in tumour group no. 1 were: SIGLEC10 and SIGLEC11. These sialic acid-binding Immunoglobulin-like lectines are expressed on the myeloid-origin cell surface [17]. All of the identified SIGLEC proteins are very similar to CD33 protein, which is a marker of myeloid cells. For example, SIGLEC10 expression was found in eosinophils and B lymphocytes, whereas expression of SIGLEC11 was found mainly in tissue macrophages. Expression of these proteins in cancer may be related with increased recruitment of myeloid cells into the tumour mass. The tumour is composed of various cells depending on the tumour type, but myeloid cells form a major component [18]. Clinical studies have shown a correlation between the number of myeloid cells (mainly macrophages) and poor prognosis in following cancers: breast, prostate, ovarian, cervical, endometrial, oesophageal, urinary bladder [19–23]. Our previous studies had pointed out the significantly higher expression level of myeloid-specific antigens in grade 3 canine mammary tumours [16, 24]. It indicates that in dogs the number of tumour-infiltrating myeloid cells or expression of myeloid-specific antigens by cancer cells may constitute a new marker of malignancy. Based on the previous outcomes it can be suggested that cancer cells may acquire phenotype of myeloid cells which, as the matter of fact, was demonstrated by expression of myeloid cell antigens on their surface [25]. One of the explanation why cancer cells may exhibit myeloid cell-specific antigens is that these cells fuse, forming hybrids that adopt phenotypic features of both parental cells [26–28]. The tumour cells that express myeloid antigens may also exhibit other phenotypic characteristics of myeloid cells, such as capabilities of cell rolling, spreading, dissociation, diapedesis, migration and matrix invasion [25].

The most malignant tumours are usually more resistant to anticancer therapy [29]. Thus, from our perspective two up-regulated genes in the tumour group no. 1 may be important: ATP2C1 and ABHD7. In breast cancer ATP2C1 ensures resistance to paclitaxel [30]. A product of the other gene (ABHD7) is responsible for cellular detoxication after drug administration (mainly compounds containing aromatic rings) [31]. Obviously, in cancer cell it can be responsible for the metabolism of anti-cancer drugs (e.g. anthracyclines which contains aromatic rings), decreasing their activity [31]. These results correlate with a clinical picture of canine mammary malignant tumours [32].

Interestingly, in the tumour group no. 1 we highlighted a significant over-manifestation of genes involved in the β1 and β2 adrenergic receptors signalling, H2 histamine receptor signalling, and acetylcholine M2 and M4 receptor signalling. The possible important role of neurotransmitters in cancer progression and metastasis had been previously described [33, 34]. Neurotransmitters are regulators of cell migration in a similar manner as chemokines. Unfortunately, only a few reports are available on the neurotransmitter signalling in tumour tissues. Our previous study [34] had indicated that carcinoma-associated fibroblasts increased the expression of neurotransmitters in cancer cells. A role of stress hormones in carcinogenesis and tumour progression is commonly known. It has been shown that norepinephrine strongly induced migration of tumour cells [35, 36], whereas epinephrine modulated carcinogenesis in the lungs [37]. Thus, stress is considered as a major risk factor of cancer development [34].

Molecular classification of canine mammary tumours is not identical with pathological classification. The tumours classified to the group no. 1 (showing pathological parameters classically associated to high biological aggressiveness) constituted a molecularly homogeneous group which differs significantly from the others. The most important up-regulated genes in these tumours were mainly involved in myeloid cell phenotype, neurotransmitters pathways as well as in anti-cancer drug resistance.