Differential somatic coding variant landscapes between laser microdissected luminal epithelial cells from canine mammary invasive ductal solid carcinoma and comedocarcinoma Authors: Vivi Deckwirth, Sruthi Hundi, Marjo K Hytönen, Sari Hannula, Pekka Ellonen, Pia Björkenheim, Antti Sukura, Hannes Lohi Published: 2024 ABSTRACT AbstractBackgroundBreast cancer (BC) is the most common cancer in women. Likewise, canine mammary tumors (CMT) represent the most common cancer in intact female dogs and develop in the majority spontaneously. Similarities exist in clinical presentation, histopathology, biomarkers, and treatment. However, CMT subtype-specific genomic background is less investigated. Here, we assess the genetic etiology of two histomorphological (HM) subtypes with BC counterparts, the CMT invasive ductal simple solid carcinoma (SC) and comedocarcinoma (CC), and compare the results with BC data.MethodsGroups of 11–13 transformed ductal luminal epithelial cells were laser-capture microdissected from snap-frozen invasive mammary SC and CC subtypes of one intact female dog. HM unaffected lobular luminal epithelial cells were controls. Single-cell whole genome libraries were generated using PicoPLEX and sequenced to compare the subtypes’ somatic coding variant landscapes with each other and with BC data available in COSMIC-CGC and KEGG. Furthermore, HM and immunohistochemical (IHC) subtype characteristics were compared with the genomic results.ResultsThe CC had six times more variants than the SC. The SC showed variants in adherens junction genes and genes of the MAPK, mTOR and NF-kappa-B signaling pathways. In the CC, the extracellular matrix (ECM) receptor interaction, cell adhesion, PI3K-Akt and cGMP-PKG pathways were enriched, reflecting the higher cellular malignancy. Affected pathways in both CMT subtypes overlapped with BC pathways in KEGG. Additionally, we identified ATP6V1C2, GLYATL3, CARMIL3, GATAD2B, OBSCN, SIX2, CPEB3 and ZNF521 as potential new subtype-distinct driver genes. Furthermore, our results revealed biomarker alterations in IHC in the basal/myoepithelial cell layer without respective genetic mutations, suggesting changes to their complex signaling pathways, disturbed regulative feedback loops or other silencing mechanisms.ConclusionsThis study contributes to understanding the subtype-specific molecular mechanisms in the canine mammary invasive ductal simple SC and CC, and revealed subtype-specific molecular complexity for phenotypically similar characteristics. Several affected genes and signaling pathways overlapped with BC indicating the potential use of CMT as model for BC. Our findings emphasize the need for thorough characterization of cancer specimens with respect to translational cancer research, but also how insight into tumor heterogeneity will be crucial for the development of targeted prognostics and therapeutic interventions.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12885-024-13239-w.Keywords: Canine, Breast cancer, Single-cell sequencing, Cancer pathway, Laser microdissection, Comedocarcinoma, Solid carcinoma, Intratumor heterogeneity Background Breast cancer (BC) is the most common cancer in women and surpassed lung cancer as the most fatal cancer in women in 2020 [1]. Research on BC has identified distinct molecular subtypes and therapeutically important biomarkers such as human epidermal growth factor receptor-2 (HER2), estrogen receptor-alpha (ERα) and progesterone receptor (PR) [2]. Sequencing technologies have revolutionized our understanding of the genetic and epigenetic etiology of BC with increasing amounts of data being cataloged in human cancer databases, such as the TCGA Portal and COSMIC [2–6]. Accumulation of variants across genomic coding and non-coding regions has provided insight into tumor progression, and has been important for the discovery of affected molecular driver pathways [7]. Furthermore, identifying BC subtype-specific somatic mutations has lead towards molecular targeted therapy and individual precision medicine [5, 8]. Canine mammary tumors (CMT) are the most common cancer in intact female dogs [9, 10]. Similarities exist in epidemiology, histopathology, molecular subtypes, pathogenesis, clinical presentation and therapy, suggesting that the dog is an important translational model for BC [11–19]. Recent CMT sequencing studies reported variants in genes such as BRCA1, TP53, USH2A, FOXC2, cMYC and in pathways such as PIK3-Akt, KRAS and HDR, as well as described molecular congruence of genes and pathways between BC and CMT [20–23]. Overlapping molecular etiologies are also described for other canine cancers, indicating that the tumor mutation burden is subtype-specific [24]. However, the suitability of the CMT model needs further exploration, including the somatic mutational landscapes of various tumor subtypes and their cell populations. We targeted the CMT invasive ductal simple solid carcinoma (SC) and comedocarcinoma (CC) subtypes, because they have corresponding BC phenotypes [16, 25–27]. In both species, the malignant transformed ductal luminal epithelial cells proliferate to form either a solid sheet of pleomorphic cells filling the duct lumen (solid carcinoma) or display a peripheral solid sheet of pleomorphic cells with central necrosis (comedocarcinoma). In women, these phenotypes represent invasive ductal carcinomas mimicking ductal carcinoma in situ, i.e. with often unobvious invasion and metastasis in the histopathological evaluation [28–34]. Likewise, canine SC and CC associate with high vascular invasion and regional/distant metastasis rates at the time of diagnosis as well as shorter survival [16, 25, 26, 35–41]. We aimed to identify somatic variant landscapes from transformed ductal luminal epithelial cells from these two carcinoma subtypes, correlate with corresponding HM and IHC, and to compare the findings with recent discoveries in CMT and BC for possible shared etiology and differences. Materials and methods Canine clinical data The samples originate from one intact, nulliparous 17-year-old mixed breed female dog. The right caudal 5th (inguinal) gland presented one mass of 20 × 30 mm in size at the pre-mastectomy clinical examination. The mass was noticed 2–3 weeks previously with following rapid growth and no previous history of mammary tumors. Therapeutic mastectomy comprised the 3rd-to-5th mammary glands en-bloc together with the inguinal fat pad. Additionally to the sampled neoplasia in the 5th mammary gland, the HM diagnosis revealed one simple tubular adenoma (ca. 3 mm in diameter) in the 3rd gland and one simple tubulo-papillary adenoma and one complex adenoma (both ca. 4 mm in diameter) in the 4th gland. Concurrent with the mastectomy, the dog was ovariohysterectomized. The right uterine horn displayed a well-delineated benign leiomyoma. According to the owner, the dog had no other illnesses or any medication. Sample collection Directly following therapeutic mastectomy, fresh tissue samples (ca. 5 × 10 mm in size) from macroscopically and palpable unaffected normal mammary gland and the 5th mammary gland neoplasia were collected for genetic analysis, snap-frozen in liquid nitrogen and stored at −80 °C (Fig. 1A). The remainder mastectomy tissue specimen together with two lymph nodes from the inguinal fat pad was submerged into 10% neutral-buffered formalin for 24 h at + 4 °C. The tissues were sampled and processed by automated (Tissue-Tek VIP Jr., Sakura) paraffin embedding (FFPE) and storage at RT (Fig. 1A) [42]. Sections adjacent to the samples for genetic analysis were specified for further investigation. Flow diagram summarizing sample collection, whole genome sequencing, and data analysis. A Sample collection and tumor classification. Tumor sample from mastectomized right 5th (inguinal) canine mammary gland for genomic analysis and tumor characterization using histochemistry and immunohistochemistry (IHC). FFPE: Formalin-fixed paraffin-embedded. B Laser-capture microdissection of cells for genomics. Targeted laser-capture microdissection (LCM) of luminal epithelial cell groups from normal lobular (control) and simple solid carcinoma as well as comedocarcinoma subtypes with documentation in three consecutive steps: step 1: Cellular capture area freehand marked with green. Scale bar 75 μm; step 2: Area after LCM. Scale bar 75 μm; step 3: Cap check of collecting tube for Cresyl-violet-stained sample material. Scale bar 150 μm. C Library construction and sequencing. DNA library construction, using the PicoPlex toolkit, from collected normal and carcinomatous cell groups. Analysis of libraries with Preseq2 and selection of further deep sequencing. WGS of selected libraries with Illumina Novaseq. D Discovery of somatic variants, genes and pathways. 1: FASTQ data aligned to the reference genome with BWA-MEM2 to generate BAM files. 2: Bioinformatics pipeline to identify somatic variants from invasive ductal solid and comedocarcinoma samples, including SNPs and indels with GATK Mutect2 and structural variants with Delly2, using a panel of normals comprising the matched normal lobular sample and 56 in-house samples. 3: Annotation of filtered variants to identify non-silent variants and further pathway analysis from KEGG using the DAVID online tool Histopathology and pathologic staging Canine and human classification criteria were used for histopathologic evaluation of mammary gland and lymph node samples from hematoxylin and eosin (HE) tissue sections of 4 μm [25, 27]. For this the slides were deparaffinized and rehydrated in xylen and alcohol series, followed by a rinse for 2 min in distilled aqua prior staining for 10 min in Meyer’s Hematoxylin (Cat. 180210 Reagena) and subsequent bluing step under running tap-water for 10 min. After another rinsing step in distilled aqua, eosin (Eosin Y, C.I. 45380, Cat. 1.15935, Merck) counterstain was applied for 2 min, followed by dehydration in alcohol series, clearing in xylene and mounting with Pertex (Cat. 00811, Histolab). The procedure was automated (Leica AutoStainer XL, Leica Biosystems). The neoplasia focality, extent, type of invasion and resection margins were assessed, and the histologic grade and pathologic stage determined using semiquantitative tripartite scoring methods as described previously [15, 25, 42–44]. The evaluation was performed by VD. Shortly, for the histologic grade were considered glandular (acinar) and tubular differentiation, nuclear pleomorphism and mitotic count. For the glandular/tubular differentiation, the percentage of area with structures showing central lumen surrounded by polarized cells within the tumor was evaluated. Cut-off points were at 75% for discriminating between majority of tumor area and moderate differentiation, and at 10% for little or no glandular/tubular formation. For the nuclear pleomorphism, the size and uniformity of the neoplastic cells were assessed at a location displaying highest degree of pleomorphism compared to adjacent normal tissue. For the mitotic count, only clear mitotic figures were counted from 10 high-power fields (40 × objective; field diameter 0.57 mm). For the pathologic stage (pTNM), the extent of the primary tumor (categories: not possible to assess / no primary tumor / in situ carcinoma / tumor size in greatest dimension / tumor with direct extension to chest wall or skin), regional lymph node status (categories: not possible to assess / no regional lymph node metastasis / regional micrometastases or metastases) and distant metastasis (categories: not present / present) were determined. Possible lymph node affection by macrometastases (> 0.2 cm), micrometastases (> 0.2 mm to 0.2 cm and/or > 200 cells), or isolated epithelial tumor cells (ITC; ≤ 0.2 mm and ≤ 200 cells) was evaluated using pan-cytokeratine (Clone AE1/AE3) and high-molecular-weight cytokeratine (Clone 34βE12) IHC [43]. Further classification based on Masson’s trichrome stain and IHC biomarkers including hormone receptors (e.g. ERα, PR, HER2), luminal epithelial (e.g. CK8, CK18, CK19, E-Cadherin) and basal/myoepithelial markers (e.g. CK5, p63, Calponin, α-SMA) as well as the proliferation indicator Ki-67. Neuroendocrine differentiation was excluded using Chromogranin A and Synaptophysin IHC [45–48]. Immunohistochemistry FFPE serial Sects. (4 µm) were prepared on Menzel Superfrost Plus Adhesion microscope slides (Cat. J1800AMNZ, Thermo Fisher Scientific) and baked at + 37 °C o/n, deparaffinized and rehydrated in xylen and alcohol series (Leica AutoStainer XL, Leica Biosystems) before antigen retrieval. Used primary antibodies and conditions are detailed in Table S1. After cooling to RT, the slides were blocked for 10 min with 3% hydrogen peroxide in PBS (Cat. BP339-4, Fisher Scientific) and rinsed twice with TBS (Cat. J640-4L, VWR Chemicals) + Tween (Cat. P1379, Sigma). The BrightVision Poly-HRP-Anti Ms/Rb/Rt IgG Kit (Cat. DPVO110HRP, Immunologic) was used for staining at RT, and 3.3´-diaminobenzidine-tetrahydrochloride (Cat. VWRKBS04-110, Immunologic) was applied for 5 min for visualization. Harris hematoxylin (Cat. HX57998853, Merck) counterstain was used for 10 s, followed by dehydration, and mounting with Pertex (Cat. 00811, Histolab). Positive controls for ChromograninA and Synaptophysin were canine pancreatic tissue. For all other antibodies, adjacent canine normal mammary was an internal positive control. For negative controls antibody diluent only was used. Evaluation of ERα, PR, HER2 and Ki-67 followed human and canine guidelines, with consideration of HER2 scores 2 + and 3 + as positive [9, 49–53]. Thresholds for ERα and PR positivity were ≥ 1% of neoplastic cells displaying nuclear staining and for HER2 positivity > 10% of neoplastic cells showing complete membrane staining. Other biomarkers were evaluated negative or positive according to absence or presence of staining, and the positive reaction was further categorized as high, moderate, or weak. Cell isolation using laser-capture microdissection Samples for genetic analysis were stored at −80 °C until LCM. Prior to cryosectioning, Polyethylene Naphthalate 1.0 (PEN)-MembraneSlides (Cat. 415190–9041-000, Zeiss, Göttingen, Germany) were UV-radiated for 30 min, and tissue samples were equilibrated to cryosection temperature (−27 °C) and embedded into Optimal Cutting Temperature Compound (Cat. 361603E, VWR International, Radnor, USA). Sections of 5 μm were cryosectioned (CM1900 Cryostat, Leica, Nussloch, Germany) onto the PEN-MembraneSlides followed by cresyl violet staining. For this, the slides were immersed in 70% ethanol, followed by a dip for 10 s into 1% cresyl violet acetate solution (Cat. 1 A 396, Chroma-Gesellschaft Schmid & Co., Stuttgart-Untertürkheim, Germany). The slides were dehydrated with dipping into 70% ethanol and 100% ethanol and dried completely before LCM. Separate solutions were prepared for normal and pathologic samples. The cresyl violet stain allowed reliable histomorphologic identification of the respective sample features and manual selection of the cell populations for genomic analysis under visual control using a 40 × objective (Zeiss PALM MicroBeam instrument, Zeiss, Munich, Germany) (Fig. 1B, Table 1). The laser-pressure catapulted microdissected cell populations were dry-collected into sterile PALM opaque AdhesiveCap microfuge tubes (Cat. 415190- 9201–000, Zeiss, Göttingen, Germany) for immediate next generation sequencing (NGS) library construction. Stepwise microphotographs and data of collected cells were documented (Fig. 1B, Table 1). Table 1Laser-capture microdissection (LCM) sample data (PALMRobo element list, Version V 4.5.0.9)Sample IDTypeNumber of cellsLCM area (μm2)Histopathological diagnosis79Freehand13506Lobular normal luminal epithelium87Freehand12532Lobular normal luminal epithelium142Freehand13900Ductal solid carcinoma, transformed intraluminal epithelium194Freehand11793Ductal comedocarcinoma, transformed intraluminal epitheliumSample ID refers to the sample number indicated in stepwise microphotographs shown in Fig. 1B. Type – Freehand stands for freehand marking of LCM sample on tissue using a 40 × objective Laser-capture microdissection (LCM) sample data (PALMRobo element list, Version V 4.5.0.9) Sample ID refers to the sample number indicated in stepwise microphotographs shown in Fig. 1B. Type – Freehand stands for freehand marking of LCM sample on tissue using a 40 × objective DNA library construction The LCM cell material was immediately processed into NGS libraries using the PicoPLEX DNA-seq Kit (Cat. R300381, Rubicon Genomics, Ann Arbor, MI USA) according to the manufacturer’s protocol (Fig. 1C). This was executed by the Institute for Molecular Medicine Finland (FIMM, Helsinki, Finland). Libraries chosen for whole genome sequencing are listed in Table 2. Table 2PicoPLEX (Rubicon Genomics, Ann Arbor, MI USA) genomic libraries chosen for deeper whole genome sequencing and their respective NCBI Sequence Read Archive (SRA) identifications (ID)LCM IDPicoPLEX library IDNCBI SRA IDHistopathological diagnosis79DOG_MAMMA_28SRR22765702Lobular normal luminal epithelium87DOG_MAMMA_32SRR22765701Lobular normal luminal epithelium142DOG_MAMMA_1SRR22765700Ductal solid carcinoma, transformed intraluminal epithelium194DOG_MAMMA_10SRR22765699Ductal comedocarcinoma, transformed intraluminal epitheliumPicoPLEX library ID refers to the sample numbers in Additional File 2. LCM stands for laser-capture microdissection PicoPLEX (Rubicon Genomics, Ann Arbor, MI USA) genomic libraries chosen for deeper whole genome sequencing and their respective NCBI Sequence Read Archive (SRA) identifications (ID) PicoPLEX library ID refers to the sample numbers in Additional File 2. LCM stands for laser-capture microdissection Whole genome sequencing NGS libraries were initially sequenced at low coverage (~ 1x) to test for yield and utility for further sequencing. The acquired FASTQ files were aligned using BWA-MEM2 (version 2.2.1), and the resulting BAM files were evaluated for WGS using Preseq2 lc_extrap (version 2.0.0) [54] (Fig. S1). Two normal mammary lobular luminal epithelial libraries and one library each for the transformed luminal epithelium from ductal simple SC and CC subtypes were selected for deeper sequencing (Table 2). The normal lobular libraries were sequenced at 15X and 12X coverages, and the SC and CC at 9X and 15X coverages, respectively. Sequencing was performed using Illumina Novaseq S1 PE151, and FASTQ files were produced using the bcl2fastq tool (version 2.20). Sequencing was executed by the Institute for Molecular Medicine Finland (FIMM, Helsinki, Finland). See Figs. 1C and D for a flowchart of the bioinformatic approach. Somatic variant calling The produced library FASTQ files were aligned with GSD1.0 (Accession ID: GCA_011100885.1) [55] as the base reference and supplemented with Y chromosome from ROS_Cfam1.0 (Accession ID: GCA_014441545.1) with BWA-MEM2 to make BAM. Best practices guideline from GATK4.2 Mutect2 were followed to identify somatic variants from the BAM files of SC and CC with a false discovery rate 0.1 using the BAM from unaffected normal mammary lobule samples as a matched normal since they originated from the same dog [56]. A panel of normal (PON) samples was used to capture technical artifacts and improve variant calling analysis results. To build the PON, Mutect2 was used to call somatic variants from 56 in-house WGS (Illumina) blood sample data. This was followed by LearnReadOrientationModel and FilterMutectCalls from GATK to filter variants due to sequencing artifacts. Variants with an allelic depth greater than or equal to 0.3 and a minimum read depth of 10 were selected to exclude the false positive risk due to sequencing errors. An in-house germline database was provided as a germline resource for Mutect2. The FASTQ data used to create the germline database has been made available in the NCBI SRA BioProject PRJNA913118. Structural variants (SVs) were called using a somatic structural variant pipeline from DELLY2 (Version 1.1.6) [57] and included insertions, deletions, duplications and intrachromosomal rearrangements. A PON was created to filter false positives and germline structural variants efficiently using BAM files from normal mammary lobules and 56 in-house samples. The FASTQ data used to create the PON has been made available in the NCBI SRA BioProject PRJNA913118. Imprecise SVs were filtered out according to the DELLY pipeline. A precise set of SVs was selected after validation using the IGV [58] and Ribbon genome browsers [59]. The final set of precise variants that included single nucleotide polymorphisms (SNPs), indels and SVs were annotated with gene information from RefSeq databases using ANNOVAR (Version: $Date: 2020–06–07) [60] with annotation database built from GSD1.0 (Accession ID: GCA_011100885.1) as the base reference and supplemented with Y chromosome from ROS_Cfam1.0 (Accession ID: GCA_014441545.1). See Figs. 1C–D for the bioinformatic approach. Variant pathogenicity prediction Missense variants predicted to be possibly damaging (less confident prediction) variants and probably damaging (more confident prediction) variants by PolyPhen-2 [61, 62] and other loss-of-function variants, such as frameshift, non-frameshift, start- / stop-loss and stop-gain variants by ANNOVAR (Version: $Date: 2020–06–07) [60] annotations were considered as pathogenic variant set. Pathway enrichment analysis The genes with non-silent variants were subjected to functional analysis using DAVID v6.8 (https://david.ncifcrf.gov/), with the selected identifier as OFFICIAL_GENE_SYMBOL, species as Canis lupus familiaris and list type as ‘Gene List’. Functional annotation summary results for KEGG_PATHWAY with EASE score /p-value set to 0.05 were downloaded. The p-value of the pathways represent the raw significance level of pathways being overrepresented in this dataset providing initial insights into potential enriched pathways. Driver gene prediction The non-silent variant set was further analyzed with OncodriveCLUSTL (version 1.1.4) to identify cancer drivers [7]. The tool was adapted for the canine genome with the above-mentioned reference and ran with parameters set to k-mer = 3, smooth_window = 11, cluster_window = 11, n_simulations = 1000, simulation_mode = mutation_centered and simulation_window = 31 to identify driver genes. Determination of ERBB2 amplification status VarScan2 (version 2.4.2) was employed to identify amplification and deletion of ERBB2 in SC and CC with two matched normal samples as control [63]. Comparison with known literature and databases Variant gene lists were compared with CMT and BC literature retrieved from the PubMed -database (https://pubmed.ncbi.nlm.nih.gov/) based on gene specific search and used for the discussion. To determine the number of genes from our precise variant list that have been previously reported in BC and CMT, we compared our variant gene list with the COSMIC Cancer Gene Census (CGC) and the KEGG (Kyoto Encyclopedia of Genes and Genomes) CMT pathway (cfa05224). Human gene orthologs were retrieved with Ensembl BioMart [64] for comparison with COSMIC-CGC [65] gene list, including Tier 1 and Tier 2 categories. Results Canine tumor classification as invasive ductal simple solid carcinoma and comedocarcinoma The tumor comprised two HM subtype components determined as solid carcinoma and comedocarcinoma based on HE and Masson’s trichrome (Figs. 2A-C, 3A-C). Transformed epithelial cells were confined within the luminal area with occasional focal discontinuity and invasion (Figs. 2A and 3A). Expanded ducts showed cells in solid growth pattern and multifocal solid cell areas with central necrosis. Thin stromal support localized in-between (Figs. 2A, C, 3A, C). Overlapping transformed cells displayed high pleomorphy with increased size and large pleomorphic nuclei, peripheral cytoplasm and increased number of nucleoli (Figs. 2B, 3B). Mitotic figures were increased and showed aberrancies (Figs. 2B and 3B). Apoptosis was more frequent in the CC. For both subtypes, the nuclear grade was high, and the cellular grade poorly differentiated (Figs. 2B, 3B). Multifocal peripheral moderate chronic inflammatory infiltrate was observed (Data not shown). Lymphovascular embolization was present (Fig. 3R). Inguinal lymph nodes were negative for isolated tumor cells, micro- and macrometastases (Data not shown). Pre-surgical clinical examination included no thorax radiographs for metastasis evaluation. The pathologic stage of the neoplastic disease was determined as T2N0M- according to the canine and human systems [15, 44]. Histopathology and biomarker immunohistochemistry of canine invasive ductal solid carcinoma phenotype. Structures have been annotated as follows: LE=luminal epithelium, S=stroma. A Solid carcinoma with scant supporting stroma. HE. Scale bar 50 μm. B High-grade morphology and high mitotic activity (3–11/HPF 40x objective). Some mitoses are marked by arrows. HE. Scale bar 20 μm. C Solid growth pattern with scant intervening stroma (blue). Masson’s trichrome stain. Scale bar 100 μm. D Moderate epithelial E-Cadherin expression. Scale bar 50 μm. Counterstain Harris hematoxylin. E Moderate luminal epithelial CK8 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. F Moderate luminal epithelial CK19 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. G Partial weak-to-moderate luminal epithelial CK18 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. H Peripheral loss of basal / myoepithelial nuclear p63 expression. Up right residual entrapped normal ductal nuclear expression. Scale bar 50 μm. Counterstain Harris hematoxylin. I Loss of basal / myoepithelial CK5 expression. Single positive entrapped cells are present. Scale bar 100 μm. Counterstain Harris hematoxylin. J Loss of basal / myoepithelial CK14 expression. Single entrapped CK14-positive cells are present. Scale bar 50 μm. Counterstain Harris hematoxylin. K Loss of basal / myoepithelial and increased stromal myofibroblastic α-SMA expression. Scale bar 100 μm. Counterstain Harris hematoxylin. L Loss of basal / myoepithelial Calponin expression. Entrapped single positive cells are present. Scale bar 50 μm. Counterstain Harris hematoxylin. M Negative ERα expression. Scale bar 50 μm. Counterstain Harris hematoxylin. N Negative PR expression. Scale bar 50 μm. Counterstain Harris hematoxylin. O Weak-to-moderate complete membrane HER2 expression in >10 % of tumor cells. Score 2+. Scale bar 20 μm. Counterstain Harris hematoxylin. P Weak-to-moderate complete membrane EGFR expression in >10 % of tumor cells. Scale bar 50 μm. Counterstain Harris hematoxylin. Q High score nuclear Ki-67 expression. Scale bar 50 μm. Counterstain Harris hematoxylin Histopathology and biomarker immunohistochemistry of canine invasive ductal comedocarcinoma phenotype. Structures have been annotated as follows: LE=luminal epithelium, S=stroma, asterix=necrosis. A Comedocarcinoma with central necrosis and scant supporting stroma. HE. Scale bar 50 μm. B High-grade morphology and intermediate mitotic activity (2–6/HPF 40x objective). Some mitoses are marked by arrows. HE. Scale bar 20 μm. C Solid growth pattern with central necrosis and scant intervening stroma (blue). Masson’s trichrome stain. Scale bar 100 μm. D Moderate epithelial ECadherin expression. Scale bar 50 μm. Counterstain Harris hematoxylin. E Moderate luminal epithelial CK8 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. F Moderate luminal epithelial CK19 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. G Partial weak luminal epithelial CK18 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. H Peripheral loss of basal / myoepithelial nuclear p63 expression. Single positive cells are present. Scale bar 50 μm. Counterstain Harris hematoxylin. I Loss of basal / myoepithelial CK5 expression. Single entrapped CK5-positive cells are present. Scale bar 100 μm. Counterstain Harris hematoxylin. J Loss of basal / myoepithelial CK14 expression. Single entrapped CK14-positive cells are present. Scale bar 50 μm. Counterstain Harris hematoxylin. K Loss of basal / myoepithelial and increased stromal myofibroblastic α-SMA expression. Scale bar 100 μm. Counterstain Harris hematoxylin. L Loss of basal / myoepithelial Calponin expression. Single positive cells are present, partially entrapped. Scale bar 50 μm. Counterstain Harris hematoxylin. M Negative ERα expression. Scale bar 50 μm. Counterstain Harris hematoxylin. N Negative PR expression. Scale bar 50 μm. Counterstain Harris hematoxylin. O Weak-to-moderate complete membrane HER2 expression in >10 % of tumor cells. Score 2+. Scale bar 20 μm. Counterstain Harris hematoxylin. P Weak-to-moderate complete membrane EGFR expression in >10 % of tumor cells. Scale bar 50 μm. Counterstain Harris hematoxylin. Q Intermediate score nuclear Ki-67 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. R Peritumoral neoplastic lymphovascular embolization. Tumor cell emboli are marked by arrows. HE. Scale bar 20 μm E-Cadherin positivity confirmed ductal origin for the transformed cells in both subtypes [66]. The staining intensity was peripheral strong and centrally weaker (Figs. 2D, 3D). In both subtypes, the transformed cells were cytokeratin (CK) 8 and 19 positive and multifocal positive for CK18, but negative or with single entrapped positive cells for CK5 and CK14, indicating luminal epithelial origin of the transformed cell populations as compared to basal epithelial cell origin (Figs. 2E–G, I–J, 3E–G, I–J) [67]. In the peripheral basal cellular compartment both subtypes presented complete or almost complete loss of the basal/myoepithelial markers CK5, CK14, p63 and Calponin (Figs. 2H–J, L, 3H–J, L). Furthermore, suggestive of myoepithelial loss, α-SMA expression was decreased or lost basally. However, the biomarker was increased in the stroma, indicative of myofibroblastic reaction (Figs. 2K, 3K). Masson’s trichrome showed only suggestive rudimentary basement membrane (BM) (Figs. 2C, 3C). Invasion was determined as the absence of myoepithelial cells and/or extension through the BM [25, 27, 68]. IHC for PR and ERα were negative and HER2 was determined as equivocal (2 +) positive (Figs. 2M–O, 3M–O) [9, 50–52]. Unfortunately, we could not assess possible ERBB2 amplification using in situ hybridization according to BC diagnostic recommendations [51, 52]. However, the implementation of VarScan 2 did not detect any amplification or deletion across the subtypes (Table S2). Manual inspection of ERBB2 in IGV using BAM files showed some coverage difference between control and tumor samples but did not reveal any strong signs of duplication (Figure S2). EGFR was membranous and weak-to-moderate in both subtypes (Figs. 2P, 3P). Negativity for Chromogranin A and Synaptophysin excluded carcinoma subtypes with neuroendocrine differentiation as described for both species (Data not shown) [45–48]. Expression of Ki-67 showed higher proliferative activity in transformed cells of both subtypes compared with normal tissue (Figs. 2Q, 3Q). Interestingly, the percentage of Ki-67-positive tumor cells was higher in SC (49%, high) than CC (13%, intermediate) as determined using the global scoring method [49, 51, 53]. Normal mammary IHC biomarker expression patterns are shown in Figure S3. According to the canine classification system, the tumor consisted of invasive ductal simple carcinoma of the comedo subtype with areas of solid subtype [25]. According to human breast tumor classification criteria, this corresponds to the subtype of invasive carcinoma of no special type (ductal carcinoma NST) comprising a mixed phenotype with comedo and solid components [48]. Considering molecular subtypes, the tumor was negative for hormone receptors ERα and PR, but equivocal positive (2 +) for HER2 and positive for luminal epithelial biomarkers CK8, CK18 and CK19 [13, 69]. Differential somatic variant landscapes between invasive ductal solid and comedo subtypes Altogether, 227,962 somatic variants were identified in the WGS of transformed luminal epithelial cells from SC and CC when filtered against the matched non-tumorigenic genome. Combining both subtypes, we found 3100 somatic coding variants, including SNVs and indels (Fig. 4). The SC carried 524 exonic variants of which 364 were non-silent variants (Fig. 4). Among the 364 non-silent variants were 318 (87.36%) non-synonymous (NSV), 12 (3.30%) frameshift deletion, four (1.10%) frameshift insertion, four (1.10%) non-frameshift insertion, two (0.55%) non-frameshift deletion, one (0.27%) non-frameshift substitution, 19 (5.22%) stop-gain, three (0.82%) stop-loss and one start-loss (0.27%) variants. The CC had 2576 exonic variants, including 2094 non-silent variants, about 6 times more than the solid subtype. Of the 2094 non-silent variants in CC, 1746 (85.42%) were NSV, 189 (9.03%) frameshift deletion, four (0.19%) frameshift insertion, four (0.19%) non-frameshift deletion, one (0.05%) non-frameshift insertion, seven (0.33%) non-frameshift substitution, 133 (6.35%) stop-gain, nine (0.43%) stop-loss and one (0.05%) start-loss variant (Fig. 4). Somatic coding variants in canine invasive ductal solid and comedocarcinoma. Total exonic variants includes non-silent and silent variants. Amounts of the non-silent variant categories of non-synonymous, frameshift / non-frameshift and stop / start-gain / -loss variants and numbers of predicted pathogenic variants are presented According to the PolyPhen-2 prediction analysis of the 318 NSVs in SC, 197 were predicted as benign, 78 as possibly damaging and 43 as probably damaging. Of the 1746 NSVs in CC, 1110 were predicted as benign, 412 as possibly damaging and 224 as probably damaging. Thus, the SC had 31.87% (167 out of 524) of exonic variants predicted as pathogenic (probably damaging, possibly damaging, frameshift and non-frameshift variants, stop-loss, stop-gain and start-loss), and the CC had 38.19% (984 out of 2576) of exonic variants predicted as pathogenic (Tables S3A, S3B). We also found 196 SVs, of which 36 belonged to SC and 160 to CC. Of these, four inversions and one duplication localized within coding regions in the SC, and two duplications, two deletions and four inversions in the CC (Tables S4A, S4B). Differentially affected pathways between invasive ductal solid and comedo subtypes Non-silent coding variants affected 344 genes in the SC and 1863 genes in the CC, with more than one variant identified in 17 genes of the SC and in 202 genes of the CC (Tables S3A and S3B). We found 73 genes with non-silent variants in both subtypes (Table S5). We conducted gene enrichment analysis to find subtype-specific affected pathways. The KEGG pathways enriched (p < 0.05) gene lists showed 15 enriched pathways (gene count ≥ 5) for SC and 24 enriched pathways (gene count ≥ 5) for CC as shown in Fig. 5 and listed in Tables S6A and S6B. These results indicate that more genes and pathways are affected in the CC than in the SC, with only a minor number of mutual genes involved. KEGG pathway analysis identified distinct enriched pathways for canine invasive ductal solid and comedocarcinoma. Color shade intensity increases with p-value significance. More significant pathways localize to the top. Note different scales in the left and right panels Shared and unique predicted driver genes between invasive ductal solid and comedo subtypes Using OncodriveCLUSTL, we found shared and unique driver genes in the two subtypes (Table S7). NEO1 was identified as a driver in both subtypes, ATP6V1C2 as a driver in SC and GLYATL3, CARMIL3, GATAD2B, OBSCN, SIX2, CPEB3 and ZNF521 as drivers in CC. These results show more driver genes in the CC than in the SC. Data comparison to annotated human and canine databases We mapped the gene sets against available databases to assess whether the affected genes and pathways were previously associated with BC and CMT. We found 23 genes overlapping with the CMT pathway (cfa05224) in the KEGG database [70]. Of these, two genes were shared by both canine subtypes; three were distinct for SC and 18 for CC. Next, we retrieved the human orthologs for our genes and compared them with the Tier 1 and Tier 2 lists of 736 genes in COSMIC-CGC [71]. About 7.27% (25 out of 344) of the SC genes and 6.39% (119 out of 1863) of the CC genes matched (Table 3). These results show that the few overlapping pathways and genes are mostly subtype specific (Table S8). Table 3The number of overlapping genes identified with CMT KEGG-cfa05224 and BC COSMIC-CGC databasesSolid carcinomaComedocarcinomaCommon in both subtypesCOSMIC-CGC251198KEGG-cfa052245202Total in this study344186373 The number of overlapping genes identified with CMT KEGG-cfa05224 and BC COSMIC-CGC databases Taken together, our results demonstrate that despite the observed close similarities in HM and IHC biomarker panel expression, the investigated CMT subtypes are molecularly distinct from each other, with genomic alterations more abundant and diverse in the CC than in the SC. Moreover, both subtypes’ gene and pathway affection overlap with BC. Discussion CMTs are suggested as a translational model for BC research. However, the dog subtype specific genetic etiology is less investigated. Here, we compared the somatic variant landscape of two CMT subtype components of an invasive mixed phenotype tumor in groups of transformed ductal epithelial cells from a single dog with each other and human data. Furthermore, we performed histopathological classification and IHC to characterize both subtypes’ phenotypes and compared these with the genomic data. Phenotypically, both subtypes displayed ductal origin (E-Cadherin+) with aberrant ductal size, morphology and cell components, suggesting dysfunction in ductal morphogenesis and epithelial cell lineage differentiation (Figs. 2D and 3D) [67]. The expanded transformed epithelial cell populations expressed the luminal epithelial biomarkers CK8+, CK18± and CK19+ with a hormone receptor status of ERα−, PR−, HER2+ and EGFR+ (Figs. 2E-G, M-P, 3E-G, M-P). We observed no alterations to ESR1 encoding for ERα or PGR encoding for PR in the investigated subtypes, inferring some other mechanism(s) leading to their loss of expression. The pleomorphic cells displayed increased proliferation (Ki-67) with atypical mitoses. The basal/myoepithelial cellular compartment was incomplete or lost, as p63, CK5, CK14, Calponin and α-SMA IHC demonstrated. These results implicate aberrant function in various pathways associated with amongst other ductal morphogenesis, epithelial cell lineage differentiation, and proliferation. Local and lymphovascular invasion were present (Figs. 2A, 3A and R), but the investigated lymph node samples were negative for ITC, micro- or macrometastases. The ECM showed modification (e.g., α-SMA IHC and Masson’s trichrome stain, Figs. 2C, K, 3C and K) and multifocal immune cell infiltration. The total number of genomic variants was fivefold higher in the CC than in the SC. The number of non-silent variants (missense, stop-gain/loss, frameshift) was sixfold higher in the CC than in the SC. While 69% of exonic variants were non-silent in SC, 81% of exonic variants were non-silent in CC. The observed increase in the number of mutations, as well as the higher proportion of non-silent and pathogenic variants among the exonic variants, along with a greater number of driver genes in CC compared to SC, indicate a higher level of genomic instability in CC relative to SC [72, 73]. The subtypes did not share identical variants but had variants in 73 mutual genes suggesting intratumor heterogeneity. According to COSMIC, these 73 genes include cancer hallmark genes such as NOTCH2 and DICER1 with tumor suppressor gene (TSG) and oncogene (OG) function, DNMT3A and FUS with TSG function only, as well as the OGs IKBKB, NSD3 and NUP98. NOTCH2 encodes a receptor of the Notch family, of which the members have temporally regulated function and expression in the development and maintenance of the normal mammary morphology and influence several hallmark genes in a cancer-specific manner, including in BC [74, 75]. Our subtypes carried different somatic variants of this gene: the SC, a probably damaging NSV and the CC, a stop-gain variant with loss-of-function, suggesting independent mutational incidence for each subtype. NOTCH2 expression increases in the normal mammary gland during lactation and after ovariectomy, implicating a hormonal dependency and role in luminal epithelial differentiation [76]. The CC carried another Notch receptor family variant: a probably damaging NSV in NOTCH1, a gene with TSG and OG function, and listed among the BC Tier 1 genes with annotated variants in COSMIC. NOTCH1 mediates asymmetric cell division of mammary progenitor cells, resulting in self-renewal of stem cells [74]. In cell experiments, activation of NOTCH1 leads to luminal cell expansion, and its inhibition reduces stem cell population and mammosphere formation [74]. In the CC, a noteworthy stop-gain variant with loss-of-function located in KRT15. Cytoplasmic keratin intermediate filaments (KRT or CK) have essential intra- and intercellular functions with tissue and cell differentiation status-specific expression [77–79]. KRT15 positive epithelial cells in the basal layer are described as progenitor cells with contribution to tissue regeneration through the capacity of self-renewal and pluripotent epithelial lineage differentiation [80, 81]. In women KRT15 has been suggested as a marker to differentiate between DCIS and invasive BC, with a correlation of KRT15 expression with histological grade, tumor node stage and tumor node metastasis stage [82]. In invasive BC low KRT15 expression associates with poor prognosis [83]. To our knowledge, no previous reports exist on KRT15 in the normal or tumorous canine mammary gland, and its expression and role need further exploration. Both investigated subtypes displayed an expansion of transformed cells of the luminal epithelial differentiation program with CK8, CK18 and CK19 expression as well as loss of the basally located stem cells expressing only CK5 without luminal or myoepithelial markers and of progenitor cells expressing CK5 or CK14 simultaneous with luminal or myoepithelial markers (Figs. 2E-G, I-J, 3E-G and I-J) [67]. As a result, we observed in both investigated subtypes the loss of functionally differentiated α-SMA+ myoepithelial cells (Figs. 2K and 3K) [67]. Basal/myoepithelial cells maintain tissue homeostasis by participating in BM production, mediation of luminal epithelial cell polarity and cell-to-BM attachment, as well as secretion of protease inhibitors, down-regulation of matrix metalloproteinases and paracrine decrease of angiogenesis [67, 84, 85]. In line with the observed changes in ductal morphology and size, we observed loss of the p63, CK5, CK14, α-SMA and Calponin expressing basal/myoepithelial cell populations (Figs. 2H-L and 3H-L). Interestingly, we did not identify somatic mutations in the corresponding genes (TP63, KRT5, KRT14, ACTA2, CNN3), suggesting possible modifications in upstream, reciprocal, or other regulatory components [67]. Loss of basal/myoepithelial cell markers and alterations in the otherwise strictly regulated BM and ECM composition and production facilitate aberrant luminal epithelial expansion, invasion, and metastasis [85–87]. In line with this, we identified in HE and Masson’s trichrome stains rudimentary BM and altered ECM, infiltrative growth and intravascular tumor cell emboli together with several SNVs in associated individual genes and the ECM-receptor interaction pathway, including various laminin (such as LAMA1, LAMB1, LAMB2, and LAMB4) and collagen subunits (COL2A1, COL4A3, COL5A3, COL12A1, COL16A1 and COL22A1), fibronectin (FN1), tenascin (TNC), thrombospondin (THBS1, THBS3), integrins (ITGA2B, ITGA3, ITGA8, ITGA10 and ITGB7) and the ECM serine protease reelin (RELN) as well as the BM-specific heparan sulfate proteoglycan (HSPG2) (Figs. 2A, C, K, 3A, C, K and R; Tables S3B and S6B). These kind of genomic alterations were more abundant in CC compared to SC, despite close similarities in HE, Masson’s trichrome and α-SMA IHC stainings. The CC presented somatic variants in several subunit genes of laminin and collagen and the proteoglycans perlecan and versican. Of these, LAMA1, LAMB1 and LAMB4 showed loss-of-function frameshift deletions, LAMB2, COL5A3, COL12A1 and COL16A1 predicted benign NSVs, whereas COL2A1 and COL22A1 probably damaging NSVs, and VCAN one benign and another probably damaging NSV. COL2A1 encodes the fiber-forming type II collagen, an ECM component providing the tissue tensile strength [88, 89]. Type II collagen fibrils join other collagen fibers and other ECM molecules through association with collagen type XII and link to the cellular cytoskeleton through integrins [88, 90]. Alteration of the normal ECM composition into carcinoma-associated ECM affects tissue homeostasis, tumorigenesis and metastatic progression [89]. COL2A1 is down-regulated in canine simple carcinoma-associated mammary stroma, and moderately expressed in canine simple carcinomas and up-regulated in canine complex carcinomas [91–93]. In BC primary tumors and cell lines up-regulation and down-regulation of COL2A1 are known, with down-regulation associated with migration and invasion [90, 91]. Furthermore, collagen type II acts as a ligand for the inhibitory leukocyte-associated immunoglobulin-like receptor-1 (LAIR-1), thus participating in immune response regulation [89]. We noticed only in the CC variants in the network-forming collagen IV family, which constitutes approximately 50% of the collagen in the BM of the mammary gland and provides it with structural support, functions as macromolecular diffusion barrier and exerts various collagen IV subtype-specific functions through integrin and non-integrin receptors, including the regulation of ERα expression and its function [87, 94–98]. Recently, BC subtype-specific expression patterns of collagen IV subtypes were suggested [98]. In our CC sample the collagen IV subtype COL4A3 displayed two possibly damaging NSVs and one stop-gain mutation with loss-of-function. Interestingly, in the normal adult human mammary gland ducto-lobular BM expression of only COL4A1, COL4A2, COL4A5 and COL4A6 are demonstrated, but not that of COL4A3 [99, 100]. However, COL4A3 has been identified as an angiogenesis-related gene and was recently determined as key prognostic gene in BC through its involvement in the regulation of epithelial cell differentiation and endothelial cell proliferation [101]. Additionally, expression of COL4A3 seems to correlate with a specific immune cell response in BC [101]. The loss of collagen IV subunit COL4A5 and COL4A6 expression has been demonstrated in invasive BC and their presence in the mammary BM is linked to functional α-SMA positive myoepithelial cells [100]. Our samples did not show alterations in COL4A5 and COL4A6, but the observed loss of α-SMA positive myoepithelial cells in IHC in our samples could thus represent one further mechanism explaining the functional loss of the BM. BM function is additionally affected by the observed NSVs of LAMA3 and HSPG2 which were predicted as probably damaging to the respective proteins in our CC. LAMA3 encodes the alpha-3 subunit of laminin-5 (also laminin-332) in the mammary gland participating in cell adhesion (hemidesmosomes), migration and branching morphogenesis [102]. In BC, LAMA3 expression is partially or totally lost [103]. This can for example result following LAMA3 promoter methylation as described in BC in association with increased tumor size and stage [104] or LAMA3 mutations as previously described for inflammatory BC [105]. HSPG2 encodes the protein perlecan which localizes in the normal mammary to the BM where it networks with collagen IV and laminin, to affect and participate in various functions such as cell–matrix adhesion, growth factor mediation, cell differentiation and angiogenesis [106]. In contrast to normal mammary, perlecan is fragmented or lost from the BM in BC and its synthesis translocated to stromal tumor cells and myofibroblasts facilitating stromal angiogenesis [106]. Additionally, CC had probably damaging NSVs in the matricellular protein genes TNC, THBS1 and THBS3. TNC modulates epithelial cell interactions with the stromal ECM and is expressed in mammary epithelial cells and stroma [107]. Increased combined epithelial and stromal expression in BC associates with poorer prognosis and lymph node metastases [107, 108]. Similar expression is reported in CMTs [109–111]. TNC promotes tumor progression in several ways (e.g., metastasis and decreased immune response), and knock-down (KD) of TNC impaired BC metastasis and colonization to lungs and bone [112–114]. THBSs support calcium-dependent cell attachment, and THBS1 is anti-angiogenetic and induces microvascular endothelial cell apoptosis, thus inhibiting tumor growth [115, 116]. However, in BC THBS1 is also reported to promote invasion and metastatic colonization through interactions with other proteins, and high expression is associated with poor prognosis [117–119]. THBS3 is less investigated, and its significance remains to be explored. The observed invasive nature could be further explained by alterations in focal and cell adhesion molecules. In our SC adherens junction genes such as CREBBP, INSR, ACTN4 and MAPK3 display NSVs with probable damaging pathogenicity, and CSNK2B has one frameshift deletion and one stop-loss NSV, both with loss-of-function. The CC shows variants in several genes related to ECM receptors, focal adhesion, axon guidance and cell adhesion pathways, of which some were already discussed. Interestingly, neither subtype had somatic mutations in CDH1, which encodes the transmembrane adherens junction protein E-Cadherin. However, the IHC expression pattern in the transformed intraluminal epithelial population was biphasic, with differences in peripheral compared to centrally located cells (Figs. 2D and 3D). Diagnostically, E-Cadherin positivity corresponds to the ductal origin of the carcinoma compared to the lobular origin, and variant expression patterns are associated with epithelial dedifferentiation, invasion and metastasis [66, 120, 121]. Disruption in cell–cell junction is a prerequisite for cell proliferation. Increased proliferative activity was HM examined in HE and using IHC against Ki-67 and determined as high score in the SC but intermediate in the CC (Figs. 2Q and 3Q). Atypical mitoses were present (Figs. 2B and 3B). These findings could correspond to genetic alterations in the CC pathways of base excision repair, apoptosis and actin cytoskeleton regulation. However, the HM observed differences in the proliferation activity in SC and CC need further exploration. Likewise, MAPK pathway cascades are essential in the regulation of cellular processes such as differentiation, proliferation, survival, apoptosis as well as motility, and they comprise a multitude of sequential components [122–124]. In our material only the SC shows a NSV of MAP3K1 with probably damaging pathogenicity. TSG and OG functions attribute to MAP3K1, which is also a Tier 1 BC gene [124]. Other genes in the MAPK signaling pathway show alterations in the investigated subtypes. MAPK3 and MAP4K4 had probably damaging NSVs in SC, and MAP2K2 in CC. The CC displayed additionally benign NSVs in MAP4K2 and MAPKAPK5, but loss-of-function in MAP3K5 with a frameshift deletion and in MAPK8IP3 a stop-gain variant. Of other Tier 1 BC genes in COSMIC, the CC also carried variants in ERBB2, NTRK3, and TBX3 with TSG and OG function, and in NCOR1 with TSG function only. Two frameshift deletions with loss-of-function were identified in NCOR1, which encodes an estrogen receptor co-repressor and represents an independent prognostic factor in BC for disease-free and overall survival [125]. In their BC study, Zhang and colleagues observed lower NCOR1 mRNA expression with higher histological grade, negative ERα and PR statuses, but positive HER2 status and a tumor diameter larger than 2 cm [125]. This seems to align with our HM and IHC results. NTRK3 encodes TRKC, which activates the PI3K/Akt pathway, increases cell survival and prevents apoptosis [126]. NTRK3 fusions are common in BC, but reports on other genomic alterations, such as copy number gain and mutation, are rare [126–129]. In CC we found one NSV that was predicted to be benign whereas no variations were seen in SC. To our knowledge, this is the first report of a NSV in canine NTRK3. A probably damaging NSV was determined in the CC ERBB2, encoding for HER2, a well-known BC biomarker with diagnostic, prognostic and therapeutic significance. Interestingly, our CC showed weak-to-moderate positive complete membrane reaction of HER2 in IHC in > 10% of the transformed luminal cells (Fig. 3O). However, it also is possible that our laser microdissection for WGS targeted a cluster of cells that display the observed NSV, indicating possible intrasubtype cellular heterogeneity, which could explain the observed HER2 IHC expression pattern. Contradicting data exists on HER2 expression in CMTs due to variations in definition of HER2-positive tumors and in applied biomarker panels for the identification of molecular subtypes [16, 130–135]. However, several SNPs are reported for the canine ERBB2, some associated with specific tumor subtypes [136, 137]. Furthermore, its loss is reported in malignant CMT [138]. Noteworthy, HER2 acts as a PIK3CA-Akt pathway activator [139]. PIK3CA is a significantly mutated driver gene and OG in BC and CMT, encoding a protein necessary for the enzyme PI3K [22–24, 135, 139–142]. A somatic variant in at least one PI3K pathway gene was reported in 50% of dogs with MTs [24]. Furthermore, Arendt and colleagues identified PIK3CA mutations in 25% of their investigated CMTs. These included a recurrent missense mutation homologous to hotspot mutation in various BC subtypes [139]. Although no specific alterations were found in PIK3CA in our subtypes, 51 genes involved in the PI3K-Akt pathway had NSVs in the CC, including the hallmark genes BRCA1, ERBB2 and COL2A1 with probably damaging prediction, as well as MTOR and JAK3 with benign prediction, and MYC with a loss-of-function frameshift deletion. Of these genes, the TSG and cell cycle regulator BRCA1 is identified with germline and somatic mutations in BC [143, 144]. Likewise, various alterations of BRCA1 are associated with CMT, especially with the loss of proper expression and function in malignant tumors [138, 139, 145–153]. Our CC showed an amino acid substitution and could possibly correspond with the reported cases with loss of BRCA1 expression in association with high Ki-67 expression and ERα negativity in IHC [145]. MTOR regulates multiple cellular events such as cell growth, homeostasis, survival and autophagy, cytoskeleton remodeling and cell motility through mTORC1 and mTORC2 complexes [154]. Usually, MTOR is dysregulated in cancer through upstream events, and only a few reports exist on human cancer-associated mutations in MTOR itself [155–157]. In BC, subtype differences are reported with activated p-mTOR localizing more often nuclear in triple-negative breast cancer (TNBC) than non-TNBC [158]. In previous canine studies, p-mTOR is not expressed in normal mammary tissues, but in 78% of mammary carcinomas, and the expression of the mTORC2 component Rictor is positively correlated with lymphatic invasion and poorer survival [159–161]. To our knowledge, no previous reports exist on somatic variants in the canine MTOR. We determined one loss-of-function frameshift deletion in MYC in our CC. MYC is usually amplified in BC [162], with reports of amplification also in CMT [135, 138]. One MYC germline mutation with causative amino acid change of uncertain pathogenicity is reported in association with BC [163]. MYC has significant regulatory importance in cellular functions such as growth, differentiation, proliferation, apoptosis, and metabolism [162]. One NSV with benign prediction was determined in JAK3 in our CC. Somatic mutations, amplifications, deep deletions and hypermethylation of JAK3 exist in BC [164–166]. JAK3 is a member of the tyrosine kinase family with importance in cytokine signaling and plays a significant role in BC tumor cell invasion and transendothelial migration as well as an effector of a JAK3-distinct immune cell response [165, 167–169]. JAK3 regulates the PI3K-Akt pathway activation through association with IL-2 [166]. This is the first report of a NSV in JAK3 associated with a CMT subtype. Additionally, to the PI3K-Akt pathway, the CC displayed enrichment in the cGMP-PKG pathway with 25 affected genes, and predicted effects ranging from benign (e.g. SRF and CREB3), probably damaging (e.g. MYLK, CREB5 and MAP2K2) to loss-of-function (e.g. RAF1 and PRKCE). In BC cell culture studies, Fallahian and colleagues demonstrated that cGMP-PKG activation inhibits proliferation and induces apoptosis [170, 171]. PKG isoforms are downregulated in various BC subtypes [172]. Likewise, the apoptosis pathway of the CC was enriched with 21 affected genes, including NFKB1 (predicted benign), ACTB and FAS (possibly damaging), CTSB, MAP2K2 and IKBKB (probably damaging), as well as RAF1 and MAP3K5 (loss-of-function). A further affected pathway in the SC was the NF-kappa-B pathway with several probably damaging variants (e.g., TAB2, IL1R1 and IKBKB) and one loss-of-function variant (ERC1). The pathway has a significant role in the regulation of proliferation, apoptosis and inflammation, and activated NF-kappa-B is essential for a specific subset of ERα− BC for proliferation and escapement of apoptosis [173, 174]. We also identified three cardiomyopathy-related pathways (dilated, hypertrophic and arrhythmogenic right ventricular) enriched in CC (Table S6B). A recent BC stratification study found dilated and hypertrophic cardiomyopathy pathways enriched together with other cancer related pathways [175]. The biological mechanisms and significance should be explored further. Driver genes often promote tumor progression, and some are implicated prognostic in BC [176, 177]. While drivers are increasingly identified in BC, their exploration in CMT remains low, with previous reports of only PIK3CA as such [22, 142]. We predicted NEO1, ATP6V1C2, GLYATL3, CARMIL3, GATAD2B, OBSCN, SIX2, CPEB3 and ZNF521 as drivers. The TSG NEO1 was identified in both subtypes, and displayed a loss-of-function stop-gain NSV in the SC and a probably damaging NSV in the CC. The cell adhesion molecule Neogenin-1 is significant in mammary morphogenesis together with the laminin-related Netrin-1 [178, 179]. Decreased or lost neogenin expression is reported in BC [179]. Loss of the Netrin-1/Neogenin complex is associated with aberrant ductal morphogenesis and breaks in the BM [178]. Furthermore, NEO1 plays a role in tumor progression, angiogenesis, apoptosis and migration in colorectal cancer and neuroblastoma cells [180, 181]. The V-ATPase-enzyme subunit gene ATP6V1C2 was predicted as driver only in the SC and displayed a probably damaging NSV. Differences in expression statuses and prognostic significance are reported in various cancer cell lines, with a less significant role in BC, present in 7% of cases by overexpression or genomic amplification [182, 183]. In a colon adenocarcinoma study, ATP6V1C2 knock-down (KD) affected WNT pathway-related genes and epithelial-to-mesenchymal transition (EMT)-related genes, with a decrease in fibronectin-1, vimentin and increase in E-Cadherin, thus attenuating metastasis [182]. Of the predicted drivers in CC, a probably damaging NSV was demonstrated in GATAD2B, a transcription factor and essential component of the Nucleosome Remodelling and Deacetylation (NuRD) complex that regulates gene expression through chromatin remodeling, and thus plays a role in transcription regulation, chromatin assembly, genomic stability and cell cycle progression [184–187]. Depending on the cellular context, NuRD can promote or suppress tumorigenesis, and through the complex members MTA1 and MTA2 affect tumor progression and metastasis [186, 188]. One paper identified GATAD2B to enhance tumor growth through direct interaction with MYC [188]. However, the specific role of GATAD2B in cancers is poorly investigated. CC harbored a probably damaging NSV in the predicted driver CPEB3. CPEBs regulate cytoplasmic translation through mRNA binding [189]. Data on CPEB3 in the mammary gland and BC is limited [190]. CPEB3 is suggested as TSG with implications of regulatory involvement in various pathways associated with tumor progression [189, 191]. Down-regulation in digestive tract cancers is demonstrated, and for example CPEB3 KD increased proliferation, migration and invasion in colorectal cancer cells via the JAK/STAT pathway activation [189]. One NSV, predicted to be benign, was detected in CC CARMIL3. CARMIL3 regulates cytoskeletal actin organization, actin-based mobility, and via ZEB2 the expression of E-Cadherin, with CARMIL3 loss leading to attenuated focal adhesions, cellular polarization as well as impaired cellular migration, invasion and metastatic colonization [192]. Of the predicted other drivers, SIX2 regulates BC metastasis by downregulating E-Cadherin, and KD of SIX2 decreases distant metastasis without affecting primary tumor growth or the associated (lymph)angiogenesis [193]. Our CC displayed a possibly damaging NSV. OBSCN is frequently mutated in BC and is implicated as TSG with cell morphology, adhesion and migration functions, preventing EMT and metastasis [194–196]. Our CC presented with two probably damaging NSVs. Expression of the stem cell-associated transcription co-factor ZNF521 associates with a protective role in invasive BC [197]. Its aberrant expression is reported in several cancers [198]. Our CC displayed a benign NSV. The role of the enzyme GLYATL3 in cancer is less explored [199, 200]. The relevance of the probably damaging NSV in the CC remains to be investigated. Following Li and colleagues, our results suggest that some driver genes might be subtype-specific and accumulate with increasing malignancy [142]. Unlike previous CMT studies, we did not find any variation in TP53, but our CC had NSV and frameshift deletion variants in TP53RK and TP53BP1, respectively [135, 138, 201, 202]. Both play a role in p53 activation [203–205]. Furthermore, loss of TP53BP1 function increases angiogenesis and is linked to poor prognosis in cancer [205]. Nor did we identify mutations in PTEN, which are previously reported for benign and malignant CMTs [137–139]. Our methodological approach allowed precision comparison between transformed ductal luminal epithelial cells from two intratumoral phenotypic distinct subtype components within a single canine individual, and to investigate their HM, immunophenotypical and genomic relationships. We demonstrated differences between the subtypes and correlation of HM and IHC with the identified genomic alterations, but also observed changes in IHC without modifications in the respective genes. The number of shared mutations in the SC and CC samples was relatively low, with CC displaying a higher overall mutation frequency. Also, the pathways involved differed significantly between the subtypes, as did the specific mutations in the genes that were mutated in both samples, inferring differences in clinical behavior. For example, genes affected in CC were often correlated to BM function and ECM composition, indicating possibly a higher ability to escape and metastasize. Interestingly, neither of the subtypes displayed mutations in the basal/myoepithelial biomarkers TP63, KRT5, KRT14, ACTA2 and CNN3 despite their loss of expression in IHC (p63, CK5, CK14, α-SMA, Calponin). Nor did we detect any mutations in the genes encoding the hormone receptors ERα (ESR1) and PR (PGR) but identified several ways in which the expression of ERα could be affected. These results remind us of the complexity of interdependencies in the expression of cellular components and the diversity of possible genetic, epigenetic and transcriptomic alterations in cancer. With our sample material it is difficult to determine if the observed intratumor heterogeneity could indicate an independent origin for the subtypes. Intratumor heterogeneity is taken into consideration in the human classification system with the category of NST tumors with mixed histological pattern [48]. In the dog, intratumor heterogeneity has been described but to our knowledge its frequency has not been further investigated. We acknowledge the limitations of only investigating subtypes from a single dog and the low sequencing depths accomplished with respect to our findings. Investigation of additional subtype-specific cell clusters and their microenvironment with known spatial location in this kind of heterogenic tumor would improve our insight of intratumor heterogeneity, clonal expansion and tumor evolution [72, 73, 206–208]. Moreover, functional cell culture studies would assess the relevance and significance of our observations. Alteration in IHC biomarker expression can be a direct effect of gene modification or result from changes in complex cell regulatory events. Furthermore, many of our findings suggest significant alterations to the BM and ECM. Tumorigenesis, progression, invasion and metastasis are complex and multifactorial. Complex reciprocal interactions between transformed epithelial cells and tumor microenvironment provoke further exploration and characterization into subtype-specific pathogenesis and progression as the basis for individualized precision medicine. We did not have tissue samples from equivalent BC subtypes to analyze, but we used available literature and databases such as COSMIC and KEGG to detect subset overlap with BC. Finally, our data demonstrate the importance of thoroughly characterizing research specimens suggested for translational cancer studies. However, further exploration with a larger CMT sample size and comprehensive cell culture studies are needed to verify our findings. Conclusions We demonstrated differential coding variant landscapes and pathway affections between the canine invasive SC and CC ductal luminal epithelial cells of a mixed phenotype CMT with immunophenotypic similarities. We identified new driver genes with probable subtype specificity. Furthermore, we uncovered subtype molecular mechanism profiles also affecting BM and ECM homeostasis. Significantly, some involved genes and pathways overlapped with BC data indicating a potential use of CMT as model for BC for certain subtypes. Additionally, we observed alterations in biomarker IHC expression without respective genetic mutations, indicating changes to their complex signaling pathways, disturbed regulative feedback loops or other silencing mechanisms. This demonstrates the continuing significance of IHC as a clinical diagnostic tool. In all, our data support subtype-distinct molecular biological disease entities and intensify the need to develop individualized subtype-specific diagnostic and therapeutic intervention methods. Our data are essential for translational research and are worth further exploration and validation using larger sample sizes and cell cultural experiments of humans and canines. Supplementary Information Additional file 1. Table S1. List of applied IHC antibodies. Additional file 2. Figure S1 Preseq2 complexity estimation and yield prediction for PicoPLEX libraries. Black line in A–C represents standard library, and colored lines low-coverage libraries sequenced at 1x. A Normal lobular. B Solid carcinoma. C Comedocarcinoma. Both normal sample libraries (IDs DOG_MAMMA_28 and DOG_MAMMA_32), one solid carcinoma sample library (ID DOG_MAMMA_1), and one comedocarcinoma sample library (ID DOG_MAMMA_10) were selected for further sequencing. Additional file 3. Figure S2 IGV (Integrative Genome Viewer) coverage plot for gene ERBB2 (NM_001003217) and region (chr9:22,713,320–22,736,532). An increase in coverage of reads from the BAM file is observed in solid carcinoma (ID DOG_MAMMA_1) and comedocarcinoma (ID DOG_MAMMA_10) samples compared to the two normal samples (IDs DOG_MAMMA_28 and DOG_MAMMA_32) in regions chr9:22,723,000–22,725,00. Additional file 4. Figure S3 Histomorphology and biomarker immunohistochemistry of normal canine mammary gland. Structures have been annotated as follows: S=stroma, L=lobule, asterix=duct. A Normal ductolobular morphology. HE. Scale bar 100 μm. B Ductolobular units red, collagenous basement membrane and fibrovascular stroma blue. Masson’s trichrome. Scale bar 100 μm. C High epithelial E-Cadherin expression. Scale bar 50 μm. Counterstain Harris hematoxylin. D High luminal epithelial CK8 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. E Weak-to-moderate luminal epithelial CK19 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. F High luminal epithelial CK18 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. G High basal/ myoepithelial nuclear p63 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. H High basal / myoepithelial CK5 expression. Scale bar 100 μm. Counterstain Harris hematoxylin. I High basal / myoepithelial CK14 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. J High basal / myoepithelial and vascular α-SMA expression. Occasional stromal myofibroblasts are positive. Scale bar 100 μm. Counterstain Harris hematoxylin. K High basal / myoepithelial Calponin expression. Occasional stromal myofibroblasts are positive. Scale bar 50 μm. Counterstain Harris hematoxylin. L Weak-to-moderate nuclear epithelial ERα expression. Scale bar 50 μm. Counterstain Harris hematoxylin. M Moderate-to-high nuclear epithelial PR expression. Scale bar 50 μm. Counterstain Harris hematoxylin. N High membranous (and cytoplasmic) HER2 expression adjacent to investigated tumor mass. Scale bar 50 μm. Counterstain Harris hematoxylin. O Weak-to-moderate membranous EGFR expression. Scale bar 50 μm. Counterstain Harris hematoxylin. P Occasional nuclear epithelial Ki-67 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. Additional file 5. Table S2. Amplifications and deletions identified by VarScan 2 in canine invasive ductal solid carcinoma and comedocarcinoma compared with normal samples. Table S3A. Somatic variants and predicted pathogenicity in canine invasive ductal. solid carcinoma. Table S3B. Somatic variants and predicted pathogenicity in canine invasive ductal comedocarcinoma. Table S4A. Structural variants in canine invasive ductal solid carcinoma. Table S4B. Structural variants in canine invasive ductal. comedocarcinoma. Table S5. Mutual genes with variants in canine invasive ductal solid carcinoma and comedocarcinoma. Table S6A. KEGG pathways in canine invasive ductal solid carcinoma. Table S6B. KEGG pathways in canine invasive ductal. comedocarcinoma. Table S7. Driver genes identified by OncodriveCLUSTL from canine invasive ductal solid carcinoma and comedocarcinoma. Table S8. List of genes with non-silent variants in canine invasive ductal solid carcinoma and comedocarcinoma compared with COSMIC-CGC and KEGG databases. Acknowledgements We wish to thank Outi Vapaavuori (Veterinary Teaching Hospital, University of Helsinki, Finland) for the CMT sample, Mikael Niku (Department of Veterinary Biosciences, Section of Anatomy, University of Helsinki) for assistance in LCM-related technical aspects, and Krista Weber (Department of Veterinary Biosciences, Section of Pathology, University of Helsinki) for assistance in sample preparation for LCM. We acknowledge the CSC—IT Center for Science, Finland for providing high-performance computing infrastructure. Abbreviations Alpha-Smooth muscle actin Necrosis (Fig. 3) and duct (Figure S3) Binary Alignment Map Human breast cancer Comedocarcinoma Cytokeratine Canine mammary tumor Catalogue of Somatic Mutations in Cancer, Cancer Gene Census 3.3´-Diaminobenzidine-tetrahydrochloride Database for Annotation, Visualization and Integrated Discovery Deoxyribonucleic acid Epidermal growth factor receptor Epithelial-to-mesenchymal transition Estrogen receptor Formalin-fixed paraffin-embedded Human epidermal growth factor receptor-2 Histomorphology Integrative Genomics Viewer Immunohistochemistry Isolated tumor cells Knock-down Kyoto Encyclopedia of Genes and Genomes Keratine Next generation sequencing Lobule (Figure S3) Luminal epithelium (Figs. 2 and 3) No special type Laser-capture microdissection Non-synonymous variant Oncogene Phosphate-buffered saline Panel of Normal Progesterone receptor Pathologic stage comprising tumor extent, regional lymph node status and distant metastasis Room temperature Stroma (Figs. 2, 3 and S3) Single Nucleotide Polymorphism Single nucleotide variant Solid carcinoma Structural variant Tris-buffered saline Tumor suppressor gene Triple-negative breast cancer Whole genome sequencing Authors’ contributions VD participated in the study conceptualization and experiment design, collected and diagnosed canine mammary samples, performed LCM, IHC staining and evaluation, participated in the evaluation of genomic library analysis, drafted the manuscript and acquired funding; SHu performed genomic library analysis, whole genome data sequence analysis, and drafted the manuscript; MKH provided genetic expertise and edited the manuscript; PE provided WGS expertise; SHa generated genomic libraries, performed genomic library analysis and WGS; PB was involved in CMT sample collection and provision of clinical patient data; AS was involved in the study design, sample collection, provided pathological expertise and edited the manuscript; HL participated in the conceptualizing, design and overall supervision of the study, drafted and edited the manuscript, and acquired funding. All authors read and approved the final manuscript. Funding Open Access funding provided by University of Helsinki (including Helsinki University Central Hospital). This study was partially supported by the Finnish Veterinary Foundation (VD), The Foundation of Veterinary Research (VD), the Cancer Foundation Finland (HL) and the Jane and Aatos Erkko Foundation (HL). FIMM Genomics NGS Sequencing unit at University of Helsinki is supported by HiLIFE and Biocenter Finland. Data availability The whole genome sequencing datasets generated and analysed from two canine mammary normal lobular libraries, one simple solid carcinoma and one comedocarcinoma subtypes have been submitted to NCBI SRA BioProject – PRJNA913118 and SRA accession IDs SRR22765699, SRR22765700, SRR22765701 and SRR22765702. The samples to create PON and in-house canine database are added to the NCBI SRA BioProject PRJNA913118. Declarations Ethics approval and consent to participate Not applicable. There were no human and/or animal experiments in this study. All canine sample material was collected from tissue previously removed during therapeutic surgery performed according to veterinary medicine best practices. The owner of the dog was informed of the research by the veterinary surgeon and agreed to sample collecting. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Vivi Deckwirth and Sruthi Hundi contributed equally to this work and share first authorship. References Associated Data Supplementary Materials Additional file 1. Table S1. List of applied IHC antibodies. Additional file 2. Figure S1 Preseq2 complexity estimation and yield prediction for PicoPLEX libraries. Black line in A–C represents standard library, and colored lines low-coverage libraries sequenced at 1x. A Normal lobular. B Solid carcinoma. C Comedocarcinoma. Both normal sample libraries (IDs DOG_MAMMA_28 and DOG_MAMMA_32), one solid carcinoma sample library (ID DOG_MAMMA_1), and one comedocarcinoma sample library (ID DOG_MAMMA_10) were selected for further sequencing. Additional file 3. Figure S2 IGV (Integrative Genome Viewer) coverage plot for gene ERBB2 (NM_001003217) and region (chr9:22,713,320–22,736,532). An increase in coverage of reads from the BAM file is observed in solid carcinoma (ID DOG_MAMMA_1) and comedocarcinoma (ID DOG_MAMMA_10) samples compared to the two normal samples (IDs DOG_MAMMA_28 and DOG_MAMMA_32) in regions chr9:22,723,000–22,725,00. Additional file 4. Figure S3 Histomorphology and biomarker immunohistochemistry of normal canine mammary gland. Structures have been annotated as follows: S=stroma, L=lobule, asterix=duct. A Normal ductolobular morphology. HE. Scale bar 100 μm. B Ductolobular units red, collagenous basement membrane and fibrovascular stroma blue. Masson’s trichrome. Scale bar 100 μm. C High epithelial E-Cadherin expression. Scale bar 50 μm. Counterstain Harris hematoxylin. D High luminal epithelial CK8 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. E Weak-to-moderate luminal epithelial CK19 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. F High luminal epithelial CK18 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. G High basal/ myoepithelial nuclear p63 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. H High basal / myoepithelial CK5 expression. Scale bar 100 μm. Counterstain Harris hematoxylin. I High basal / myoepithelial CK14 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. J High basal / myoepithelial and vascular α-SMA expression. Occasional stromal myofibroblasts are positive. Scale bar 100 μm. Counterstain Harris hematoxylin. K High basal / myoepithelial Calponin expression. Occasional stromal myofibroblasts are positive. Scale bar 50 μm. Counterstain Harris hematoxylin. L Weak-to-moderate nuclear epithelial ERα expression. Scale bar 50 μm. Counterstain Harris hematoxylin. M Moderate-to-high nuclear epithelial PR expression. Scale bar 50 μm. Counterstain Harris hematoxylin. N High membranous (and cytoplasmic) HER2 expression adjacent to investigated tumor mass. Scale bar 50 μm. Counterstain Harris hematoxylin. O Weak-to-moderate membranous EGFR expression. Scale bar 50 μm. Counterstain Harris hematoxylin. P Occasional nuclear epithelial Ki-67 expression. Scale bar 50 μm. Counterstain Harris hematoxylin. Additional file 5. Table S2. Amplifications and deletions identified by VarScan 2 in canine invasive ductal solid carcinoma and comedocarcinoma compared with normal samples. Table S3A. Somatic variants and predicted pathogenicity in canine invasive ductal. solid carcinoma. Table S3B. Somatic variants and predicted pathogenicity in canine invasive ductal comedocarcinoma. Table S4A. Structural variants in canine invasive ductal solid carcinoma. Table S4B. Structural variants in canine invasive ductal. comedocarcinoma. Table S5. Mutual genes with variants in canine invasive ductal solid carcinoma and comedocarcinoma. Table S6A. KEGG pathways in canine invasive ductal solid carcinoma. Table S6B. KEGG pathways in canine invasive ductal. comedocarcinoma. Table S7. Driver genes identified by OncodriveCLUSTL from canine invasive ductal solid carcinoma and comedocarcinoma. Table S8. List of genes with non-silent variants in canine invasive ductal solid carcinoma and comedocarcinoma compared with COSMIC-CGC and KEGG databases. Data Availability Statement The whole genome sequencing datasets generated and analysed from two canine mammary normal lobular libraries, one simple solid carcinoma and one comedocarcinoma subtypes have been submitted to NCBI SRA BioProject – PRJNA913118 and SRA accession IDs SRR22765699, SRR22765700, SRR22765701 and SRR22765702. The samples to create PON and in-house canine database are added to the NCBI SRA BioProject PRJNA913118.