Introduction

MicroRNAs (miRNAs) are a rapidly expanding class of important, noncoding RNA that have proven regulatory roles in both plants and animals (Ambros 2004; Bartel 2004). Known to work in a post-transcriptional fashion, miRNAs are active in development, apoptosis, differentiation, and tumorigenesis (Ambros 2004; Bartel 2004; Hwang and Mendell 2006). Bioinformatic analysis suggests that there are approximately 1000 miRNA genes in a mammalian genome, and miRNAs are estimated to regulate approximately 30% of all genes (Berezikov et al. 2005; Lewis et al. 2005). In particular, some miRNAs have been labeled “oncomirs” because of their associations with various cancers. This is highlighted by the fact that 50% of the computationally predicted and experimentally proven miRNAs are located within human genomic fragile regions with already proven links between breaks and oncogenesis (Calin et al. 2004b). More specifically, 15 miRNAs are at fragile sites at which alterations and breaks occur frequently, and these aberrations are linked with human breast cancers (Calin et al. 2004b). Other studies have successfully used microarrays and quantitative real-time PCR (qRT-PCR) to “profile” the expression patterns of several different cancers (Blenkiron et al. 2007; Iorio et al. 2005; Jiang et al. 2005; Mattie et al. 2006; Volinia et al. 2006). Breast cancer is currently one of the main foci of these investigations because it is the most often diagnosed cancer in American women (American Cancer Society 2007). The American Cancer Society estimates that 240,510 women would be diagnosed with some form of breast cancer in 2007 and another 40,460 women would die (American Cancer Society 2007).

Breast cancer is not limited to humans. Mammary gland tumors represent almost half (41.7%) of all tumor types of the intact female dogs (Dorn et al. 1968). In fact, a population study found that canine mammary tumors represent 13.4% of all tumors found in the dog (Dorn et al. 1968). Approximately 50% of all canine mammary tumors are malignant. For these malignant variants, surgery is the common treatment and appears to be the only successful treatment option (Lana et al. 2007). Depending on various prognostic factors and tumor-free surgical margins, surgical excision can be curative in many dogs. Key prognostic factors for overall survival time include but are not limited to multiple masses, tumor grade, lymphatic invasion, and distant metastasis (Lana et al. 2007). Perhaps the best characterized prognostic factor is tumor size. For tumors smaller than 3 cm, the median survival time in one study was 22 months. Dogs with a tumor larger than 3 cm had a median survival time of only 14 months (Philibert et al. 2003). Experience with radiation therapy for canine mammary tumors is limited and no definitive data exist. Chemotherapy with agents such as doxorubicin and docetaxel has not radically increased overall survival time (Simon et al. 2006). Hormonal therapy with agents such as tamoxifen does not improve survival and is currently not recommended for use in dogs (Lana et al. 2007). Thus, at this time there are no effective systemic treatment options for dogs with mammary tumors.

There is considerable debate as to whether canine mammary tumors represent a strong model for human breast cancers. Both tumor types are primarily epithelial in nature and both occur spontaneously in the normal population (Dickson et al. 2005; Lana et al. 2007). The biological behavior of both tumors includes local invasion with regional lymph node metastasis and distant metastasis to the lungs (Dickson et al. 2005; Lana et al. 2007). Furthermore, the highly aggressive inflammatory carcinoma is reported in both species. Aberrant expression patterns of proliferating cell nuclear antigen, Bcl-2, cytokeratins, calponin, and the genes p53, Cox-2, and BRCA1 have been found in malignant canine mammary tumors (De Las Mulas et al. 2004; Dore et al. 2003; Kumaraguruparan et al. 2006; Nieto et al. 2003). Many of these cellular markers are used in diagnosis and prognosis of human breast cancers (Dickson et al. 2005). Furthermore, canine mammary tumors and human breast tumors have similar antigenic phenotypes (Mottolese et al. 1994). These similarities support the use of canine mammary tumors as a genetic model for human breast cancers.

In particular, the dog may prove an ideal model for the study of oncomirs. Because mammary cancer is a spontaneously occurring disease in the dog (as opposed to induced rodent models or manipulated in vitro models), the functional genetics can be explored more accurately. The genetic similarities found between the two species suggests conservation of function in that miRNAs may be targeting genes conserved between the human and dog to initiate tumorigenesis or metastasis. Presented here are the expression levels of oncomirs in canine mammary cancer using the highly sensitive qRT-PCR approach.

Materials and methods

Ten human precursor miRNA sequences (hsa-miR-15a, hsa-miR-16, hsa-miR-17-5p, hsa-miR-21, hsa-miR-29b, hsa-miR-125b, hsa-miR-145, hsa-miR-155, hsa-miR-181b, and hsa-let-7f) were obtained from the Sanger miRNA Registry (Griffiths-Jones et al. 2006). These precursor sequences were used to search for the homologous canine sequences using NCBI’s BLAST search program (http://www.ncbi.nlm.nih.gov/genome/seq/BlastGen/BlastGen.cgi?taxid=9615). Identity percentage was calculated by dividing the number of matched bases by the total number of bases in the human precursor. Mature percentages were calculated by dividing the number of matches at the mature sequence by the total number of bases in the mature sequence.

Six primary canine mammary tumors were collected and used in this study (Table 1). Surgical excision of all tumors followed routine surgical guidelines for neoplastic excision. Tumor samples were then stored in RNAlater ® (Ambion Inc., Austin, TX) immediately after surgical excision. Biopsy reports were available for five of the six tumors. Ten normal mammary tissue samples (Table 2) were collected from healthy adult female dogs that underwent an elective ovariohysterectomy (OHE) through Texas A&M University’s Veterinary Medical Teaching Hospital General Surgery Department. Harvesting of normal mammary tissue was approved by the university’s Clinical Research Review Committee (#07-02) and via a signed owner consent form. At the conclusion of the OHE, a 1-cm incision was made over the most prominent mammary gland. A sample of mammary tissue approximately 3 mm × 10 mm was collected and stored in RNAlater. Total RNA was isolated from all samples using the mirVana™ miRNA Isolation Kit (Ambion) following the manufacturer’s recommended protocol. Samples were evaluated for quality and quantity using the NanoDrop® ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE).

Table 1 Canine mammary cancer samples
Full size table
Table 2 Canine normal mammary samples
Full size table

TaqMan® MicroRNA Assays (hsa-miR-15a, hsa-miR-16, hsa-miR-17-5p, hsa-miR-21, hsa-miR-29b, hsa-miR-125b, hsa-miR-145, hsa-miR-155, hsa-miR-181b, and hsa-let-7f) were used to reverse transcribe ten mature miRNA sequences to cDNA (Applied Biosystems, Foster City, CA). Each 15-μl reverse transcription reaction contained 5 μl total RNA diluted to a 1.4-ng/μl concentration, 1 × TaqMan MicroRNA RT primer, 0.25 U RNase inhibitor, 1 × RT buffer, 3.33 U MultiScribe™ reverse transcriptase, and 1 mM dNTPs (ABI). The reverse transcription reactions were run in a BioRad thermocycler (Bio-Rad Laboratories, Hercules, CA) for 30 min at 16°C, 30 min at 42°C, 5 min at 85°C, and then held at 4°C. Controls included reactions with no template, no primer, and the RNU6B TaqMan assay (ABI), which codes for a small RNA that has been validated by ABI as an appropriate control.

The reverse transcription products were diluted 1:15. The 20-μl real-time reaction contained 1.33 μl diluted RT product, 1× TaqMan Universal PCR Master Mix (No AmpErase® UNG), and 1× TaqMan assay, which contained the forward and reverse primers as well as the TaqMan probe. Each reaction was run in triplicate on a 96-well plate. Reactions containing no reverse transcription products served as negative controls. Reactions were incubated in an ABI 7500 Prism Sequence Detection System for 10 min at 95°C, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min.

After completion of the qRT-PCR, the threshold value was automatically configured above the baseline displayed in the amplification plot. Relative quantification of gene expression was evaluated by utilizing the comparative critical threshold (CT). The CT values for each mature miRNA reaction were subtracted from the respective CT value of the RNU6B control, resulting in the ∆CT value. The largest ∆CT value was arbitrarily used as a constant that was subtracted from all other ∆CT values to determine ∆∆CT value. Fold changes were then generated for each mature miRNA by calculating 2-∆∆CT. Expression levels of normal samples were compared to cancerous samples using a standard MANOVA.

Results

Bioinformatics data revealed the ten oncomirs of interest (miR-15a, miR-16, miR-17-5p, miR-21, miR-29b, miR-125b, miR-145, miR-155, miR-181b, let-7f) to have substantial sequence identity between human and dog at the precursor level and complete conservation at the mature level. Sequence identity is presented in Table 3. Based on previous research, the identity found at the mature level permits the use of the ABI TaqMan MicroRNA assays with canine input despite being designed for the corresponding miRNA genes of the human (Boggs et al. 2007). qRT-PCR was chosen as the method of choice to quantify miRNA expression due to its need for very little RNA input as well as the specificity and sensitivity of TaqMan.

Table 3 Mammary oncomir conservation comparison
Full size table

A brief comparison (Fig. 1) of the overall expression of the ten miRNAs studied with qRT-PCR and analyzed with the traditional ΔΔCT method reveals five miRNAs with tumor suppressor activities (miRs-15a, -16, -17-5p, -125b, and -155) and four with oncogenic activity (miRs-21, -29b, -181b, and let-7f). MiR-145 does not seem to differ in expression between normal and malignant mammary tissues. The statistical analysis to compute p values (listed in Table 4) was performed using the multivariate analysis of variance (MANOVA) calculation, a more robust test than the standard t test. MANOVA takes into account that all ten miRNAs were evaluated on the same six samples, and it revealed miR-21 and miR-29b to be significantly different (p ≤ 0.05). However, the biological trends cannot be ignored and are further addressed in the Discussion.

Fig. 1
figure 1

Mammary oncomir expression. Expression levels of ten oncomirs were studied from normal and cancerous canine breast tissue and compared using a MANOVA analysis. (a) Comparison of all six tumor samples versus ten normal samples. (b) Comparison of two tubular papillary carcinoma samples to the ten normals. (c) Comparison of three ductal carcinoma samples to the ten normal samples

Full size image
Table 4 p values are listed for each miRNA and were calculated using MANOVA for total and subtype analysis
Full size table

Comparison of the samples based on two types of histology (ductal carcinomas and tubular papillary carcinomas) showed that miR-15b and miR-16 have a significantly lower expression in canine ductal carcinoma diagnoses versus normal samples. The ductal carcinomas were identified by neoplastic cells invading the mammary gland. Alternatively, tubular papillary carcinomas invaded the mammary glands, producing abnormal tubular projections into the duct. The tubular papillary tumors had a higher expression of let-7f and miR-181. This is in addition to miR-21 and miR-29b and indicates that these four genes may have oncogenic roles in this particular type of malignancy.

Discussion

It was not surprising to find that the oncomirs of interest are highly conserved between the dog and the human. Previous analysis of the two species revealed that the canine sequence, as a whole, shows great sequence identity to the human, even more than the popular mouse model (Kirkness et al. 2003). The physiology of the dog enables larger and easier-to-obtain specimen collections compared to rodent models. The companion animals also have extraordinary health care, second only to that of the human medicine. The vast similarities in both genetics and physiology and considerable overlap in treatment regimens make the dog an ideal model for many human diseases, including malignant breast cancer.

Like the human, the dog is known to have several risk factors for the development of mammary cancer. All canine mammary tumors appear to be promoted by sex hormones (Lana et al. 2007). Studies have shown that there is a direct correlation between the total amount of estrogen exposure and the risk for mammary tumor development. Dogs that are spayed before their first, second, or third estrus cycle have a risk of 0.5, 8, and 26%, respectively, for developing a mammary tumor (Schneider et al. 1969). This is intriguing because early-aged menopausal women have a significant decrease in risk for development of breast cancer compared to those women who reach menopause at biologically normal times (Dickson et al. 2005).

It must be noted that what is known about the role of estrogen receptors in canine mammary tumors indicates that they may not be good biological models of human mammary cancer. That is, while estrogen clearly promotes the growth of breast cancer in both species (Dickson et al. 2005; Schneider et al. 1969), multiple studies have shown that in true canine mammary malignancies, estrogen receptors are downregulated compared to normal mammary tissue (Illera et al. 2006; Millanta et al. 2005; Nieto et al. 2000). This finding is markedly different when compared to the majority of human postmenopausal breast cancer (Dickson et al. 2005). Consequently, estrogen receptor blockers such as tamoxifen have little effect in the dog. Interestingly, dysplasia and benign canine tumors are characterized by elevations in the estrogen receptors similar to women. Thus, the potential of dogs with mammary tumors to serve as models for study of human breast cancers is not without complication. Even so, the conflicting data pertaining to estrogen receptors is outweighed by the aforementioned shared characteristics between canine and human mammary cancers. These similarities suggest that while treatment response may vary due to the lack of estrogen receptors, the biological development and progression of mammary cancers in these two species is similar enough to enhance the understanding of human breast cancer through the use of the canine model. To this end, evaluation of miRNAs in canine mammary tissue may be a highly productive avenue of investigation.

Previous work grouped oncomirs into three categories: tumor suppressors, oncogenes, and those that fulfill both functions (Calin and Croce 2006; Esquela-Kerscher and Slack 2006; Wijnhoven et al. 2007; Zhang et al. 2007). Because miRNAs can bind to more than one mRNA (Fig. 2), it is not only possible but more than likely that they will bind to several mRNAs. Likewise, the miRNAs studied in this article can be placed into one of the three categories.

Fig. 2
figure 2

MicroRNA biogenesis. The primary miRNA (pri-miRA) is transcribed from the genomic DNA. Drosha processes the pri-miRNA into a precursor miRNA (pre-miRNA) which is transported to the cytoplasm by Exportin-5. Dicer further cleaves the hairpin and leaves a short 18–22-nucleotide double-stranded RNA duplex. The RNA-induced silencing complex (RISC) then incorporates the mature strand of the miRNA and binds to the target mRNA in one of two fashions displayed above. The miRNA can then act as a tumor suppressor or oncogene dependent upon the mRNA it targets

Full size image

Tumor suppressors

The first oncomirs, miRs-15a and 16, were initially discovered to play a role in chronic B-cell lymphocytic leukemia (Calin et al. 2004a). They are clustered together at a known cancer fragile site and act to suppress tumors by repressing BCL2, an anti-apoptotic oncogene (Cimmino et al. 2005). The loss of these two miRNAs results in increased expression of BCL2 and subsequent prosurvival states, a documented event in the majority of chronic cases of B-cell lymphocytic leukemia cases (Cimmino et al. 2005). Decreased expression and a loss in copy number of miR-15a and miR-16 were found in other hematopoetic and solid cancers, including human mammary cancers (Bottoni et al. 2005; Iorio et al. 2005; Mattie et al. 2006; Volinia et al. 2006). The work reported here provides evidence that both miR-15a and miR-16 are generally suppressed in canine mammary carcinomas. However, when the ductal carcinoma phenotype is singled out, the different patterns of expression become significant thereby suggesting that miR-15a and miR-16 may be responsible for the differentiation of the malignancy.

Another known tumor suppressor, miR-125b, also acts to control cell proliferation and regulates the oncogene Lin28 in the human (Lee et al. 2005; Wu and Belasco 2005). Expression of this miRNA in the dog is reduced in malignant samples compared to normal mammary tissues. This is in agreement with microarray, Northern blotting, and comparative genomic hybridization studies done on primary and in vitro samples from the human (Blenkiron et al. 2007; Iorio et al. 2005; Mattie et al. 2006; Volinia et al. 2006; Zhang et al. 2006).

Differential expression of miR-145 was not detected in canine mammary cancer, but it is described as a tumor suppressor and has been found to target FLJ21308 in the human (Kiriakidou et al. 2004). This oncomir is more often associated with colorectal tumors (Bandres et al. 2006), but studies have documented a differential expression in mammary cancer (Iorio et al. 2005; Volinia et al. 2006). It is interesting to note that one group described the expression of both the mature and precursor expression levels of miR-145 in colorectal tumors, but found a difference only at the mature level (Michael et al. 2003). This finding underscores the importance of choosing expression tools that target the “functional element” of miRNAs, i.e., the mature product.

Oncogenes

The oncogenes miR-21 and miR-29b are associated with numerous cancers, including human mammary cancer (Blenkiron et al. 2007; Esquela-Kerscher and Slack 2006; Fabbri et al. 2007; Frankel et al. 2007; Iorio et al. 2005; Mattie et al. 2006; Meng et al. 2007; Mott et al. 2007; Pekarsky et al. 2006; Si et al. 2007; Silveri et al. 2006; Volinia et al. 2006; Zhang et al. 2006; Zhu et al. 2007). Consistently, these two miRNAs are upregulated in analyses of mammary cancers using microarrays, and Zhang et al. (2006) confirmed that these two oncogenes showed an increase in copy number in mammary tumor tissues. In particular, miR-21 targets both PTEN and TPM1, and the loss of this miRNA results in increased caspase activity and subsequent apoptosis (Meng et al. 2007; Zhu et al. 2007). Tcl1 and MCL1 are targets of miR-29b (Mott et al. 2007; Pekarsky et al. 2006). Because these miRNAs have roles in so many cancers, it is reasonable to hypothesize that they have a general role in tumorigenesis rather than a specific role in metastasis or development of mammary cancers. Both miRNAs were significantly upregulated in tumors for this study (p ≤ 0.05). They remained significant even when the tubular papillary carcinoma phenotype was separately analyzed. Because of the increased expression, miR-21 and miR-29b may be ideal targets for in vitro knockdown studies.

Combination tumor suppressors and oncogenes

Some miRNAs serve opposing roles, specifically as both tumor suppressor and oncogene. This is the most interesting and complicated category of miRNAs, and as functional studies continue and more data are generated, it is likely that many more oncomirs will be found to have these dual roles.

Known to suppress the oncogene AIB1, miR-17-5p would be classified in the tumor suppressor category (Hossain et al. 2006). However, it is also known to target E2F1, the transcription factor for c-Myc (Hossain et al. 2006; O’Donnell et al. 2005). Previous work showed increased expression of miR-17-5p in several cancers, and this miRNA is known to specifically accelerate C-MYC-induced tumorigenesis in the mouse (He et al. 2005). Both Iorio et al. (2005) and Volina et al. (2006) used microarray data to suggest that miR-17-5p acts as an oncogene in mammary cancers while many other studies used a variety of techniques to show evidence of miR-17-5p acting as tumor suppressor (Hossain et al. 2006; Lu et al. 2005; O’Donnell et al. 2005; Zhang et al. 2006). In fact, this miRNA is deleted, due to its fragile site position, in 21.9% of breast cancer samples (Zhang et al. 2006). The data from the dog correlate with the majority of the findings when all mammary tumors and tubular papillary carcinomas were analyzed. There does not seem to be a difference in expression in the ductal carcinomas. Because the differences were not statistically significant, additional tumor samples would need to be collected to further investigate the role of miR-17-5p in canine mammary cancer.

In chronic lymphocytic leukemia, miR-181b targets Tcl1, another known oncogene, suggesting that the miRNA is a tumor suppressor (Pekarsky et al. 2006). Other work implicated the gene in differentiation of B cells (Chen and Lodish 2005). In human glioblastoma and pituitary adenomas, miR-181b seems to act in a tumor-suppression fashion (Bottoni et al. 2007; Ciafre et al. 2005). However, Zhang et al. (2006) reported an increased copy number of miR-181b in mammary tumors, and microarray studies showed overexpression of this miRNA in mammary cancer tissue samples. Overall, the canine mammary tumors appear to exhibit increased expression of miR-181b. In the tubular papillary carcinomas, the miRNA has significantly increased expression, indicating that miR-181b may be important in differentiation of mammary cells as well.

Located within the BIC gene, miR-155 is generally thought of as an oncogene because of its increased expression in hemopoietic cancers, thyroid cancers, and lung adenocarcinomas (Costinean et al. 2006; Mattie et al. 2006; Volinia et al. 2006; Zhang et al. 2006). Microarray data showed an increase in expression in multiple tumor samples of different origins, giving further evidence for this oncogenic role (Iorio et al. 2005; Volinia et al. 2006). However, a study of pancreatic cancer showed miR-155 to be downregulated (Roldo et al. 2006). In canine mammary cancers, the miRNA takes on both roles, dependent upon tumor differentiation. The average expression in tubular papillary carcinomas is lower compared to normal mammary tissue, while the ductal carcinomas display a slightly higher expression when compared to the healthy counterparts. It is important to note that only the averages are being compared because the difference of expression was not considered statistically significant. Interestingly, Mattie et al. (2006) found that miR-155 is differentially expressed when comparing estrogen receptor-positive and -negative tumors.

A part of the let-7 family of miRNAs, let-7f is described as a tumor suppressor because it knocks down the oncogene ras (Johnson et al. 2005). The let-7f miRNA is conserved in as distantly related species as the worm, plant, and fruitfly (Pasquinelli et al. 2000). By targeting K-ras and N-ras mRNA for degradation, let-7f’s role in cancer cell proliferation was clear (Johnson et al. 2007). In addition, multiple studies have shown a reduced expression of this miRNA in lung cancer (Jiang et al. 2005; Johnson et al. 2005, 2007). However, one study used the highly sensitive qRT-PCR and showed an upregulation of let-7f in breast cancer cell lines (Jiang et al. 2005). Only the tubular papillary carcinomas of canine mammary cancer showed a statistically significant difference in expression of let-7f, which supports the findings of the qRT-PCR study.

The data reported here were generated using a qRT-PCR approach that has proven specificity and sensitivity. This technique is particularly useful because there is very little RNA obtained from the mammary biopsies and the miRNAs being quantified differ by only one nucleotide. While Northern blots were not performed to confirm the data, previous research has shown the qRT-PCR approach to be more accurate. In addition, RNU6B was chosen as the “housekeeping gene” for two reasons: previous researchers have used the U6 control, making the data comparable, and the expression of RNU6B was found to be relatively consistent among several cancers (Jiang et al. 2005).

Relatively few samples were used in this study because of the difficulty in obtaining tumor and normal tissue from client-owned animals. This means that while most of the miRNAs did not show statistically significant differences, they still may play vital roles in cancer. The analysis of more tumor samples to improve the power of the study is required before any definitive statements on the role of miRNAs in canine mammary cancer are made. This point underscores the need for future collaborations between veterinarians and researchers as well as the establishment of tumor banks.

The fact that the dog appears to follow the trends described in previous literature on human breast cancer research is encouraging as it provides theory for future research that utilizes the canine species as a model. As more samples are collected and in vitro cell lines are established, more canine phenotypes can be explored and compared to the human forms of malignancy beyond just breast cancer. The cell-specific functions should be further investigated and predicted targets can be tested in cell lines derived from canine tumor samples. Elucidating the actions of miRNAs will advance cancer diagnostics, treatment, and survival by providing gene therapy targets of the future. Ideally, the miRNAs found to be involved in tumorigenesis or tumor suppression could serve as targets for gene therapy in both the dog and the human.