Abstract
Background/Aim: Canine mammary gland tumors (MGTs), as a potential model of human breast cancer, have a well-defined histological classification system. MicroRNA (miRNA) expression is a key part of the molecular signatures of both MGTs and human breast cancer, although the signatures alone do not yet provide a sufficient basis for definitive diagnosis. In this study, we investigated the association between miRNA expression patterns and histological classification. Materials and Methods: Mammary gland tissue was collected from healthy dogs (n=7) and dog patients (n=80). Further samples (n=5) were obtained from established MGT cell lines. We targeted miRNAs differentially expressed in metastatic tumor tissue versus non-metastatic and normal tissue. A subset of samples was analyzed using small RNA next generation sequencing (NGS) with subsequent qPCR. Results: Six differentially expressed miRNAs were selected from the NGS analysis and submitted for large-scale qPCR. The large-scale qPCR analysis revealed greater alternations in miRNA expression. Large-scale analysis, based on 79 samples, revealed a hierarchical clustering based on selected miRNAs that did not strikingly match the histopathological subtype classification. Conclusion: We successfully investigated the large-scale miRNA expression pattern in canine MGT and provided the whole miRNA expression. The selected miRNA demonstrated that there is no straightforward mapping between molecular signatures and histological classification of canine MGTs at the miRNA level.
Canine mammary gland tumors (MGTs) are attracting attention in genetic research, with microRNAs (miRNAs) regarded as a fruitful line of investigation. Canine MGTs share many features with human breast cancer, including histological and biological behavior as well as molecular features, thus they should be fully established as a suitable model of human breast cancer (1, 2). Molecular signatures encompass miRNAs, small noncoding RNAs that play a role in gene regulation through mRNA silencing (3-5). Identifying miRNAs that are differentially expressed in certain disease conditions may reveal molecules of interest for the diagnosis and treatment of these conditions.
In veterinary oncology, miRNAs have been examined in other tumor types. Furthermore, similarities in miRNA expression pattern between human breast cancers and canine MGTs have been noted (6-8), and research on miRNAs in human breast cancer has identified many potential therapeutic and diagnostic targets.
In our previous study, we used NGS followed by large- scale qPCR to investigate miRNAs differentially expressed in canine MGTs relative to normal mammary gland tissue (9). Furthermore, differences in microRNA expression between non-metastatic and metastatic canine MGTs have already been reported in a microarray-based study (10), suggesting their expression profiles differ between at least some histological subtypes and indicating their potential as histological- subtype- specific biomarkers.
In this study, we identified miRNAs differentially expressed between non-metastatic and metastatic mammary gland tissue by using NGS. We then proceeded to investigate the relevant miRNA expression patterns through real-time PCR and hierarchical clustering.
Materials and Methods
Sample collection. The canine mammary gland samples targeted for real-time PCR and NGS analyses in this study were from the same pool of tissue samples described in our previous study (9), and relevant information on these samples covering the breed and age of the relevant dog, and the diagnosis for the relevant case is presented in Table I. Basically, this pool consisted of normal tissue (n=7) and clinical (n=80) samples from female dogs. The normal tissue samples were collected from laboratory animals (beagles) at Shin Nippon Biomedical Laboratories (Kagoshima, Japan). The clinical samples were collected from dogs undergoing treatment and care at the Kagoshima University Veterinary Teaching Hospital, Japan, or its affiliated hospitals. The sample donors were aged between four and 17 years (median age: 11 years). Further samples (n=5) were obtained from metastatic malignant mammary gland tumor cell lines [CIPp (RRID:CVCL_L149), CIPm (RRID:CVCL_L148), CTBp (RRID:CVCL_L151), CHMm (RRID:CVCL_L146), CHMp (RRID:CVCL_L147)] (11), and included as metastatic malignant mammary gland tumor samples. Collected samples were promptly submerged in RNAlater™ (Invitrogen; Thermo Fisher Scientific, Waltham, MA, USA), and incubated overnight at 4°C followed by storage at -80°C.
Demographic information on MGT and normal tissue donors.
Pathological diagnosis. Histopathological specimens were prepared by hematoxylin and eosin (H&E) staining of collected tissue, and H&E-stained specimens were submitted for microscopic examination by a certified veterinary pathologist. All pathologists were engaged in examining samples for this study have a diploma from the Japanese College of Veterinary Pathologists), and used common, published standards for the classification and grading of canine MGTs and Surgical Pathology of Tumors of Domestic Animals Volume 2: Mammary Tumors in making their diagnoses (12, 13). Specifically, these standards were used to classify the tumor tissue samples by type and/or subtype. Furthermore, tumor was regarded as metastasized when tumor cells were found in the removed lymph nodes, and/or imaging of the lung revealed regions of mass. Tumor tissue samples not meeting these criteria were regarded as non-metastatic for the purposes of this study. Each dog in this study underwent imaging with radiography and/or ultrasound, and in most of cases full-body Computed Tomography, for veterinary assessment.
Total RNA extraction and sequencing. Total RNA was extracted by using a mirVana™ miRNA isolation kit (Thermo Fisher Scientific), in accordance with the manufacturer’s instructions. RNA concentration was measured using NanoDrop 2000c (Thermo Fisher Scientific). The quality of RNA was assessed using the 2100 Bioanalyzer System (Agilent Technologies, Santa Clara, CA, USA), according to the manufacturer’s instructions. Samples with an RNA integrity number (RIN) exceeding eight with confirmed microRNA region in the 2100 Bioanalyzer System were used for the study.
The small RNA libraries were prepared and sequenced by Hokkaido System Science (Hokkaido, Japan). Small RNA libraries were constructed from 1 μg of total RNA using the TruSeq Small RNA Library Preparation Kit (Illumina, San Diego, CA, USA), in accordance with the manufacturer’s instructions. The libraries were subjected to 100-bp paired-end sequencing on an Illumina HiSeq 2500 System (Illumina).
Availability of data. Sequence reads for mammary gland (n=3), adenomas (n=3), adenocarcinomas (n=3), were submitted to the sequence read archive (SRA) (www.ncbi.nlm.nih.gov/sra) under the Bioproject accession number; PRJNA716131. Sequence reads for metastasized adenocarcinomas (n=6) were submitted to the SRA under the Bioproject accession number; PRJNA738308.
Processing small RNA sequencing data. Small RNA sequences were processed and analyzed using CLC Genome Workbench 10.1.1 (CLC bio, Cambridge, MA, USA). The sequence reads were annotated against miRBase (release 21) (14-16), Ensembl canine ncRNA database (Canis familiris.canfam3.1.ncrna) (17-19). miRBase was prioritized over other annotation resources. Small RNAs differentially expressed in the tumor tissue were identified with the empirical analysis of DGE tool in CLC Genome Workbench 10.1.1.
Quantification of miRNAs with qPCR. qPCR was performed as previously described (20). Total RNA (2 ng/μl) was reverse-transcribed to cDNA with TaqMan miRNA assays (Thermo Fisher Scientific), in accordance with the manufacturer’s protocol. qPCR was performed with a TaqMan Fast Advanced Master Mix kit and the StepOnePlus™ real-time PCR system (Thermo Fisher Scientific). RNU6B was used as the internal control, and expression levels were determined with the 2-ΔΔCt method. qPCRs with an undetermined Ct were assigned Ct ?40. TaqMan miRNA assays used for qPCR in this study and their assay IDs were as follows: cfa-miR-187-3p (ID: 001193), cfa-miR-202-5p (ID: 002362), cfa-miR-424-5p (ID: 002309), cfa-miR-450a-5p (ID: 001031), cfa-miR- 450b-5p (ID: 006407_mat), and cfa-miR-542-3p (ID: 001284).
Statistical analysis. qPCR data were analyzed to create a graphical representation using GraphPad Prism 7 (GraphPad Software, San Diego, CA, USA). The qPCR data were analyzed by One-Way ANOVA (nonparametric) and were subjected to a Kruskal-Wallis test. Furthermore, box & whiskers plots were drawn in accordance with Tukey’s definition to exclude outliers. Heatmap hierarchical cluster analysis was performed using the R statistical environment.
Ethics approval. The study design and experimental protocols were approved by Kagoshima University and the Kagoshima University Veterinary Teaching Hospital Ethics Committee (Approval No.: KV0004). All procedures in this study conformed with the ethics by laws and regulations of Kagoshima University. All samples were collected after obtaining the owners’ informed consent.
Results
miRNA expression profiles determined with NGS. In order to identify miRNAs that are potentially differentially expressed between canine MGT histological subtypes, we compared miRNA expression values between non-metastatic and metastatic canine mammary gland tissue, using NGS analysis. The samples targeted in this analysis were obtained from four different histological subtypes. The non-metastatic tissue comprised nine samples from our previous study (9); specifically, normal mammary gland (n=3; Sample Nos.: MG1-3), and adenoma (n=3; Aden1-3), and adenocarcinoma (n=3; AdCa1-3). The metastatic tissue consisted of six samples; specifically metastasized adenocarcinoma [n=6; AdCa(Meta) 1-6].
A total of 39 miRNAs were significantly differentially expressed in metastatic tissue vs. non-metastatic tissue (p<0.05), of which 17 were upregulated and 22 were downregulated. The 39 miRNAs are listed in Table II.
List of differentially expressed miRNAs of Non-Meta vs. Meta comparison in canine mammary gland tumor samples analyzed using Next Generation Sequencing.
Of these 39 miRNAs, we targeted those with a false-discovery-rate-adjusted p-value (FDP-p) <0.05 and detectable normalized expression value >50 in the NGS for inclusion in subsequent investigations and qPCR analysis.
NGS-based heatmap hierarchical clustering of differentially expressed miRNAs. To visualize patterns in miRNA expression elucidated by NGS, the canine mammary gland tissue samples were then submitted for heatmap hierarchical clustering analysis. Metastasized adenocarcinoma was the most widely scattered subtype across the dendrogram for mammary gland tumor tissues, and mammary gland tissue was the most tightly clustered within the dendrogram (Figure 1).
Heatmap hierarchical clustering of differentially expressed miRNAs in canine mammary gland tumor tissue samples using Next Generation Sequencing. The numbers on the Y axis are miRNAs in Table II. The samples on the X axis are tissue samples used in Non-Meta vs. Meta comparison. AdCa(meta) scatters in the different clusters, while all the mammary gland samples (MG 1, MG 2, MG 3) are classified into the same cluster. The expression values were normalized (reads per millions) and transformed by log for graphing this figure in R. MG: Mammary gland; Aden: adenoma; AdCa: adenocarcinoma; AdCa(meta): metastasized adenocarcinoma.
qPCR-based investigation of differential expression patterns of target miRNAs. We targeted some of the relevant miRNAs for further investigation in our full pool of tissue samples using qPCR analysis. The sample pool comprised 92 samples covering seven histological subtypes–mammary gland, tumor-adjacent tissue, adenoma, complex adenoma, benign mixed tumor, adenocarcinoma, and metastasized adenocarcinoma. This analysis included target miRNAs meeting the selection criterion for NGS results described above (normalized mean expression value>50 and FDR-p<0.05). Of these eight miRNAs, two (cfa-miR-133a-3p and cfa-miR-133c-3p) had already undergone the relevant analysis in our previous study (9). The other six target miRNAs (cfa-miR-187-3p, cfa-miR-202-5p, cfa-miR-424-5p, cfa-miR-450a-5p, cfa-miR-450b-5p, and cfa-miR-542-3p) were included in this analysis. The results obtained for each miRNA are presented graphically in box-and-whisker plots in Figure 2.
Relative expression values of cfa-miR-187-3p, cfa-miR-202-5p, cfa-miR-424-5p, cfa-miR-450a-5p, cfa-miR-450b-5p, and cfa-miR-542-3p in canine mammary gland tumors by real-time PCR. The six target miRNAs were found to be upregulated in the metastatic group using NGS. However, real-time PCR revealed more diverse expression across tumor subtypes that was predicted from the NGS data. *p<0.05, **p<0.01, ***p<0.001. MG: Mammary gland; TAT: tumor adjacent tissue; Aden: adenoma; CA: complex adenoma; BMT: benign mixed tumor; AdCa: adenocarcinoma; AdCa(meta); metastasized adenocarcinoma.
cfa-miR-187-3p yielded similar results to cfa-miR-133a- 3p and cfa-miR-133c-3p. However, cfa-miR-424-5p, cfa- miR-450a-5p, and cfa-miR-450b-5p showed more diverse expression patterns, with some sporadic differences between normal tissue and non-metastatic MGTs. miR-450b-5p and miR-542-3p showed a wider range of values for metastasized adenocarcinoma than other subtypes. No miRNA showed a significant difference in expression between normal mammary gland tissue and tumor-adjacent tissue.
qPCR-based heatmap hierarchical clustering of target miRNAs (cfa-miR-187-3p, cfa-miR-202-5p, cfa-miR-424-5p, cfa-miR-450a-5p, cfa-miR-450b-5p, and cfa-miR-542-3p). To visualize patterns in miRNA expression elucidated in realtime PCR, data from 79/92 samples in this study were then submitted for heatmap hierarchical clustering analysis based on adjusted threshold cycle (-ΔΔCt) values for the target. In the resultant dendrogram for the target miRNAs, clusters are labeled from A to E (Figure 3).
Heatmap hierarchical clustering of miRNAs in canine mammary gland tumor tissue samples analyzed using real-time PCR. The five clusters identified in the hierarchy were designated as Clusters A-E. The number of samples in each cluster is stated in Table III. The Ct values were adjusted using RNU6B, and the –ΔΔCt values were used for graphically representing figures in R. The samples that are also described in Figure 1 are marked in grey (the annotation for the samples is provided in Table IV). MG: Mammary gland; TAT: tumor adjacent tissue; Aden: adenoma; CA: complex adenoma; BMT: benign mixed tumor; AdCa: adenocarcinoma; AdCa(meta): metastasized adenocarcinoma.
The numbers of samples in the five clusters were not uniform: Cluster A contained 32 samples, Cluster B 7 samples, Cluster C 22 samples, Cluster D 17 samples, and Cluster E one sample (Table III). Some samples are also described in Figure 1 and Figure 3, where they are marked in grey. The annotations for the samples are provided in Table IV. These sample numbers imposed limitations on direct percentagewise comparisons between the subtypes in each cluster, so our evaluation focused on the absence of particular subtypes from a cluster (instances of n=0 for a particular miRNA).
Number of samples in the clusters of Figure 3.
Cluster A contained samples from all tumor subtypes, but no samples of normal or tumor-adjacent tissue, and resembled Cluster D in its tumor subtype composition, apart from the small proportion of benign mixed tumor samples, and the single mammary gland sample. In contrast, Cluster C contained mostly non-tumor samples (normal and tumor adjacent tissue) but lacked adenoma and complex adenoma samples. The composition of Cluster B was similar to that of Cluster C, although it contained seven samples, a smaller number than Cluster C. The most markedly distinct cluster was Cluster E, because it contained only one sample of metastasized adenocarcinoma.
The qPCR-based heatmap hierarchical clustering yielded a markedly greater scattering of samples than that based on NGS, with both non-tumor and tumor tissue subtypes represented across a wider range of clusters and in different proportions within clusters, thus indicating the existence of molecular diversity across the histological classification system for canine MGTs.
Discussion
To the authors’ knowledge, this is the first study to investigate whole alternative miRNA expression in a range of histological subtypes of canine MGT by NGS. We used NGS to establish an initial profile of miRNA expression in tissue from dogs histologically diagnosed with MGT and sought to confirm these expression patterns using qPCR with a large number of clinical samples.
As the major novel finding of this study, we identified unexpectedly diverse levels of miRNA expression across tumor subtypes, based on the results for large-scale profiling with qPCR analysis. With the origin of the cancer metastasis yet to be fully elucidated (21, 22) and considering that metastasis may occur in any cellular stage, we initially set out to identify miRNAs for analysis by histological subtype through a comparison of metastatic MGT (metastasized adenocarcinoma) tissue with non-metastatic MGT (adenoma and adenocarcinoma) and normal mammary gland tissue using NGS. However, when comparing expression patterns of the identified miRNAs using qPCR with a large number of the clinical samples, we found a much more diverse expression than could be predicted from the smaller dataset obtained through NGS analysis (with a wider range of histological subtypes; normal mammary gland and tumor- adjacent tissue as non-tumor tissue, adenoma, complex adenoma, benign mixed tumor, and adenocarcinoma as nonmetastatic tumors, in addition to metastasizedadenocarcinoma). The qPCR results were suggestive of a high level of miRNA expression diversity across a range of canine MGT subtypes.
Reports on similar variations in molecular expression already exist. Human breast cancer, a condition for which there is some consensus on molecular classification (ranging from luminal A, luminal B to HER2), reportedly shows such variation (23, 24), highlighting the importance for standardization of analytical methods for microarraybased breast cancer classification systems (25). In the veterinary field, a microarray-based study identified distinct metastasis-related differences in the miRNA profile of canine mammary gland tumors, but the differential expression was not as marked in qPCR as it was in the microarray analysis (10).
The limitation concerns the classification of nonmetastatic tissue. We used a definition of “non-metastatic” in this study that applies only to what we could see in the mammary gland tissue samples. Each dog in this study underwent radiography and ultrasound examinations, and full-body CT scanning was also performed in the majority of cases. The results of these image-based veterinary assessments were fully consistent with the pathological diagnoses of non-metastatic tissue. Therefore, our results can be regarded as valid for a comparison of metastatic-tumor-bearing and apparently non-metastatic-tumor-bearing mammary gland tissue, even without conclusively demonstrating the absence of metastasis.
In conclusion, we reported the whole miRNA expression pattern in metastatic and non-metastatic MGT. Then, we regard the miRNA expression diversity across histological canine MGT subtypes as the principal point of interest for this study. Our findings support future studies of miRNA analysis comparing canine and human breast cancers, and shed some light on the issue of molecular classification systems for canine MGTs established through miRNA profiling. Establishing molecular classification systems may contribute to advances in research on human breast cancers.
Acknowledgements
The Authors thank Ms. Ayako Masuda for her invaluable assistance with the experiments, and Henry Smith (Co-chair of the Veterinary Special Interest Group in the European Medical Writers Association), of the Joint Faculty of Veterinary Medicine, Kagoshima University, for his help with the English editing of this manuscript.
This study was supported by the Japan Society for the Promotion of Science KAKENHI (Tokyo, Japan): grant nos. 21H02366, 20K21375, 17H03926, and the first author especially wish to express their gratitude to the KOHNAN Asia Scholarship Foundation and the Japan-Taiwan Exchange Association for their role in individual sponsorship.
Footnotes
Authors’ Contributions
Hui-Wen Chen and Naoki Miura: Conceptualization, design of study, and methodology. Hui-Wen Chen, Md. Mahfuzur Rahman, Yu-Chang Lai, Al Asmaul Husna, Md. Nazmul Hasan, Hitoshi Hatai, Noriaki Miyoshi, Osamu Yamato: Formal analysis, investigation, validation, and visualization. Hui-Wen Chen and Naoki Miura: Writing original draft, review and editing. Naoki Miura: Supervision, project administration and funding acquisition. All Authors reviewed and approved the final manuscript before the submission.
Conflicts of Interest
The Authors declare no conflicts of interest in relation to this study.
- Received April 30, 2022.
- Revision received May 19, 2022.
- Accepted May 24, 2022.
- Copyright © 2022 The Author(s). Published by the International Institute of Anticancer Research.
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.