Introduction

Mammary tumors represent the majority of neoplastic diseases in female dogs, constituting over 40% of all female canine tumors [1]. Of these approximately 50% are malignant and in majority being of epithelial origin (simple adenoma/carcinoma). In contrast to human breast cancer 20% of the canine tumors exhibit myoepithelial cell proliferation (complex carcinoma). Canine mammary tumors (CMT) have been proposed as a model for human breast cancer and increasing knowledge about the molecular differences and similarities will support the use of CMT in breast cancer research [2–4]. However, molecular research in canine mammary gland tumors in the recent years seemed to focus primarily on genes with well established roles in human carcinogenesis (for review see: [2]). In order to gain a deeper understanding of molecular processes that can lead to neoplastic transformation of the canine mammary gland a broader spectrum of investigated genes is required.

Recently, our group identified a recurrent deletion of the proximal part of cfa27 by high-throughput sequencing analyses of five canine malignant mammary tumors (MMTs). Within this region PFDN5 appeared as the most plausible candidate with cancer-driving functional relevance. [5].

The Prefoldin Subunit 5 gene (PFDN5) maps to the proximal end of the canine chromosome (cfa) 27 (1,906,133–1,909,613 bp, Broad CanFam 3.1). The encoded protein is part of the prefoldin complex that functions as chaperone and is involved in the posttranslational folding of high-order protein structures such as actin and microtubules [6]. Furthermore, prefoldin and its subunits can localize to the nucleus and act as transcriptional regulators by binding to the DNA or transcription factors [6]. PFDN5 in particular is known to repress the c-MYC transcription factor [7]. Three different c-MYC repression mechanisms are described that are mediated by PFDN5: i) repression of the transcriptional activity by binding to the N-terminal region of c-MYC ii) recruiting of the ubiquitin ligase complex to c-MYC that drives proteasome dependent degradation iii) repression of the Wnt4 expression that leads to the repression of c-MYC expression [6–11]. Therefore, PFDN5 has the functional potential to act as a tumor suppressor gene. Indeed, a frequent amino acid substitution (A157R) that led to a loss of its c-MYC transcriptional repression activity was found in 50–60% of leukemia/lymphoma cells and in more than 75% of human tongue cancers [8]. Furthermore, tumor-specific alternative splicing variants were detected in thyroid and head/neck cancer [12, 13]. One variant was generated by the retention of intron 2 that generated a premature stop-codon and another detected splice isoform led to the insertion of 101 amino acids [12, 13]. Both splice-variants are expressed at higher levels in tumor as compared to normal tissues or benign lesions [12, 13].

In human breast cancer PFDN5 was amongst 102 genes for which differential expression in tumor versus normal tissue was detected; 69% of the analysed tumor samples showed a decreased expression [14].

However, PFDN5 cannot be considered as a well-known established cancer-driver gene, neither in human nor in canine cancers. Nevertheless, significantly higher Ki-67 scores, indicative of a higher proliferative activity, were detected in the tumors with PFDN5 deletion [5], which can serve as hint for a functional association. Since the sample number in our first explorative study was small the primary goal of this study was to confirm the deletion of PFDN5 in a larger number of canine MMTs compared to benign tissue.

Material and Methods

Results

Overall, the relative copy numbers of PFDN5 in malignant tumors (n = 102) was significantly lower (p<0.05) compared to benign tissue (n = 51), while, no significant difference was present in PBMC genomic DNA for the same groups (p = 0.16). If those tumors with high amount of benign tissue (carcinoma in benign tumors) and the carcinoma in situ were omitted from this analysis, the remaining 74 cases showed a highly significant deviation from benign controls (p<0.01).

Hence, by using a z-score delimiter of -3.0 the frequency of the PFDN5 deletion was higher in malignancies (n = 102), yielding an odds ratio of 4.9 (95% Cl: 1.4–17.2, p = 0.004 Fisher’s exact test).

At that delimiter of z = -3.0, the deletion was detected in 26% of complex carcinomas (7|27), in 21% of tubulopapillary carcinomas (3|14) and in 67% of solid carcinomas (6|9). Furthermore the deletion was detected in one simple carcinoma (1|10), one anaplastic carcinoma (1|3), one spindle cell carcinoma (1|5), one osteosarcoma (1|1) and in one carcinosarcoma (1|4). In the most frequently occurring carcinoma/sarcoma in benign tumors (3|28) the deletion was detectable in 11%. In 9% (3|32) of the BMTs the deletion was detected, which were one complex adenoma and two benign mixed tumors. In the 19 non-neoplastic mammary tissues, including the six MH and the two DE the deletion was not detected (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. PFDN5 deletion in different tumor groups.

https://doi.org/10.1371/journal.pone.0131280.t001

In samples that were histopathologically categorized as malignancy within benign tumors, the content of DNA from malignant cells is usually low in total tissue, which may result in failure to detect the deletion, even if present in malignant cells. To prove this, in a set of 6 tumors, the malignant and the benign parts were macrodissected from FFPE samples, which were scored negative of for the PFDN5 deletion, when homogenized total tissue was used. In contrast to the results in tissue, the deletion was found in three of the predominantly malignant parts (50%) but not in the areas, which were defined as predominantly adenomas (0|6).

Immunolabelling with Ki-67 antibodies was performed in a subset of 100 tissue samples and revealed a clear nuclear labelling that showed a variable intensity. Due to the low content of malignant cells in the so called carcinomas in benign tumors (n = 21), the KI-67 was reduced, congruent with the low abundance of the PFDN5 deletion in the whole tissue of that tumor type. This type was therefore excluded from statistical analysis to avoid a bias in favour of an association between KI-67 and the somatic deletion. The percentage of Ki-67 expressing cells was evaluated in 79 tumor samples (MMT n = 50; BMT n = 21; MH n = 6 and DE n = 2) and ranged between 4% and 81%. Ki-67 expression was significantly higher (p = 0.0001; Mann-Whitney U-test) in MMT (median: 27%, 25^th-75^th percentile: 21.3%-33%) compared to benign neoplasms (median: 15%, 25^th-75^th percentile: 13%-24%) (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Ki-67 score and SD in different tumor groups.

https://doi.org/10.1371/journal.pone.0131280.t002

In the MMTs a significant higher Ki-67 expression between the PFDN5-deleted (median: 29%, 25^th-75^th percentile: 25%-35%) and non-deleted group (median: 26%, 25^th-75^th percentile: 20%-31%) was present (p = 0.05; Mann-Whitney U-test; Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Box plot of Ki-67 scores.

Boxplots of Ki-67 score in all malignant mammary tumors (MMT), all benign mammary tumors (BMT) (*** p<0.0001), in malignant mammary tumors with (w/) and without (w/o) the PFDN5 deletion and without the carcinoma in benign tumor (* p<0.05)

https://doi.org/10.1371/journal.pone.0131280.g001

Discussion

This study was intended to validate our earlier finding [5] on the presence of a recurrent somatic deletion of the PFDN5 gene on CFA27. The results confirm that this somatic deletion is present in canine MMT and represents one potential new tumor marker. Furthermore, the somatic presence of the deletion in 23% of the analysed malignant tumors supports PFDN5 as being a potential tumor driver gene. Cancer drivers are defined as genes that when somatically altered contribute to the development of cancer [19, 20]. It is thought that several different mutations are required to drive oncogenesis, the actual number, however, can only be estimated and is variable between different cancers [20]. Cancer driver genes are commonly thought to be altered at a certain level of recurrence [20]. An amplification of the c-MYC oncogene for example is found in 15.7% of human breast cancers [21].

In the current study the highest deletion frequency (67%) of canine PFDN5 was detected in the group of solid carcinomas followed by 26% in complex carcinomas. On the other hand, only 10% of the analysed simple carcinomas carried the deletion. These differences might be indicative of different carcinogenic mechanisms that give rise to the different carcinoma types, as recently suggested [3].

The very low percentage of detected deletions in the special types of carcinoma or sarcoma in benign tumors is most likely due to the "dilution” of the often low numbers of malignant cells by the benign fraction of the tumor. In these tumors foci of malignant-appearing cells or distinct nodules of malignant cells reside within complex adenomas or benign tumors [22]. Our results in macrodissected malignant areas, where the deletion was found in 3 of 6 samples that were not detectable when homogenized total tissue was used, confirmed such a dilution effect of the DNA from malignant cells by the benign fraction.

Interestingly, the deletion was also found in a few BMTs, though they did not show histopathological evidence of malignancy at the time of surgery. However, those three cases were in complex and mixed tumors, where the histopathological evaluation is extremely difficult, a recent survey unveiled a considerable false negative rate [23], furthermore focal differences between the section used for microscopy and for molecular analyses can never be excluded.

On the other hand, benign mammary tumors (BMT) have been considered as precancerous lesions and carry an increased risk of malignant transformation [24]. In this line it could also be hypothesized that a deletion of the c-Myc repressor gene PFDN5 may be a risk indicator for precancerosis.

Further indication for the proposed role of PFDN5 as tumor-suppressor is the detected association between CFA27 deletion and higher Ki-67 scores. Ki-67 antigen is a non-histone protein, which is only expressed in proliferating cells and enabling to determine the growth fraction of a cell population [25] and cell proliferation in mammalian tumors e.g. humans and dogs [4]. In CMT increased cell proliferation may serve as an indicator of malignancy since Ki-67 was reported to be significantly different between BMT and MMT [2]. High Ki-67 scores were associated with increased tumor grade, larger tumor size, lymph node involvement and metastasis and the survival time was significantly decreased for bitches with high Ki-67 score tumor [2]. Even though a statistical significant difference between Ki-67 scores of malignant and benign tumors was seen in our sample set, the large overlap (cf. Fig 1) seems to compromise a general use as prognostic marker in dogs. Similarly, Ki-67 scores in malignant tumors are significantly higher in those with the deletion compared to tumors without the deletion, but again both groups largely overlap.

Due to the location of PFDN5 close to the centromere it can not be excluded that its frequent deletion in canine mammary tumors is solely related to its position. Centric fusions that result from reciprocal translocation appear frequently in canine tumor cells and are thought to occur randomly and with greater frequency at later stages of tumor progression [26]. Reciprocal translocations often lead to large deletions at the breakpoints [27]. Therefore, it is possible that the recurrent deletion of PFDN5 in canine tumors is attributable to its position rather than to its function during carcinogenesis. Nevertheless, the fact that such a somatic telomeric deletion was not found for other chromosomes and that the PFDN5 deletion has a substantial frequency in malignant tumors makes a functional role well conceivable.

Conclusion

The present study confirms and extends our previous findings of the abundance of a deletion of the proximal part of CFA27 harbouring the PFDN5 gene in canine mammary tumors. This confirmation warrants further studies including the assessment of biochemical and in vitro data that evidently prove the functional link to PFDN5 and oncogenic transformation. With the current knowledge, the question as to whether the observed PFDN5 deletion is a primary or secondary driver event during tumorgenesis cannot be answered.

Acknowledgments

The authors acknowledge Sarah Bierau, Stefan Balzer (Chronix Biomedical GmbH, Göttingen), Klaus-Peter Kuhlmann, Bettina Buck and Kerstin Schöne (Department of Pathology, University of Veterinary Medicine Hannover) for their excellent technical assistance.