Canine invasive mammary carcinomas as models of human breast cancer. Part 1: natural history and prognostic factors

Purpose Dogs have been proposed as spontaneous animal models of human breast cancer, based on clinicopathologic similarities between canine and human mammary carcinomas. We hypothesized that a better knowledge of the natural history and prognostic factors of canine invasive mammary carcinomas would favor the design of preclinical trials using dogs as models of breast cancer. Methods The 2-year outcome of 350 female dogs with spontaneous invasive mammary carcinoma was studied. The investigated prognostic factors included age at diagnosis, pathologic tumor size, pathologic nodal stage, lymphovascular invasion, histological grade, and expression of Estrogen Receptor alpha (ERα), Progesterone Receptor, Ki-67, Human Epidermal Growth Factor Receptor 2, basal cytokeratins 5/6, and Epidermal Growth Factor Receptor. Multivariate survival analyses were performed using the Cox proportional hazards model. Results The overall survival after mastectomy was 11 months. Within 1 year post mastectomy, 41.5% of dogs (145/350) died from their mammary carcinoma. By multivariate analysis, the significant prognostic factors for overall survival included a pathologic tumor size larger than 20 mm [HR 1.47 (95% confidence interval 1.15–1.89)], a positive nodal stage [pN+, HR 1.89 (1.43–2.48)], a histological grade III [HR 1.32 (1.02–1.69)], ERα negativity [HR 1.39 (1.01–1.89)], a high Ki-67 proliferation index [HR 1.32 (1.04–1.67)], and EGFR absence [HR 1.33 (1.04–1.69)]. Conclusion The short natural history of spontaneous canine invasive mammary carcinomas and high rate of cancer-related death allow for rapid termination of preclinical investigations. The prognostic factors of invasive mammary carcinomas are remarkably similar in dogs and humans, highlighting the similarities in cancer biology between both species.


Introduction
Breast cancer represents the most prevalent cancer and the leading cause of cancer death in women worldwide [1]. Despite considerable progress in breast cancer management, prognosis in the metastatic setting remains poor. The 5-year specific survival after initial diagnosis was estimated 97% for stage I, 88% for stage II, 70% for stage III, and only 25% for stage IV breast cancer [2]. One of the current challenges is to define molecular tools and relevant models that can predict the response and potential resistance to therapies. The classic in vitro (tumor cell lines) and in vivo (xenografts) preclinical models have indeed limitations related to the difficulty to reproduce interactions with the microenvironment, the absent or incomplete metastatic pattern, and their inability to fully integrate the host immune response [3]. Spontaneous tumor models are thus of high interest, to study the pharmacokinetics of innovative therapeutics in vivo, their effect on tumor (pathologic response) and patient (metastasis, survival), and the interactions between tumor cells and their microenvironment. In this respect, canine spontaneous cancers seem particularly relevant to human oncology [4][5][6].
Here, we hypothesized that (1) knowledge of the natural history of CMCs would emphasize the aggressive and short course of the disease, and could be useful for the design of preclinical therapeutic trials in dogs with CMC, as translational models of human breast cancer; (2) knowledge of the prognostic factors of CMCs would highlight the biological similarities between spontaneous CMCs and breast cancers.
The aims of this study were thus to describe the natural history of invasive CMCs, i.e., cancer progression and mortality rates, in the largest cohort collected so far (350 female dogs); to describe invasive CMCs using human pathological criteria including immunohistochemical markers; and to validate these criteria as prognostic factors able to predict patients' outcome. In part 2 of this article, we evaluated the prognostic significance of the immunohistochemical classification of human breast cancer applied to dogs.

Patients and samples
This retrospective study included 350 female dogs with invasive mammary carcinoma, but free from other cancer, initially diagnosed in two laboratories of veterinary histopathology (Laboratoire d'Histopathologie Animale, Oniris, Nantes, and Laboratoire d'Anatomie Pathologique Vétérinaire d'Amboise, France) between 2007 and 2010. The owners' written consent and approval from the Oniris College of Veterinary Medicine local Animal Welfare Committee were obtained prior to inclusion. Dogs were eligible for inclusion when a histological diagnosis of invasive mammary carcinoma was established and confirmed by an absent layer of p63-positive myoepithelial cells (anti-p63 antibody, clone ab111449, Abcam) by immunohistochemistry (IHC) that differentiates invasive from in situ mammary carcinomas [44,45]. All dogs were treated surgically by their veterinarian, and none of them received any additional treatment before and/or after mastectomy. Age, breed, spay status, parity, contraception, prior benign mammary lesions, medical history, and outcome were obtained through written questionnaires or telephone interviews with referring veterinarians and owners. All 350 dogs were followed for at least 48 months with particular emphasis on the occurrence of locoregional relapse (time between mastectomy and the earliest local recurrence on the same mammary gland, new primary mammary tumor, or lymph node metastasis), distant metastasis-free interval (time from mastectomy to first evidence of distant metastases by medical imaging), and disease-free interval (interval from mastectomy to the first local recurrence, new primary tumor, lymph node metastasis, and/or distant metastasis). Overall survival was defined as the time between mastectomy and death from any cause. Specific survival was defined as the time between mastectomy and death attributable to the mammary carcinoma.

Pathological evaluation
Histological examination was performed on 3-μm-thick hematoxylin-eosin-saffron (HES) stained sections. The 350 tumors were classified according to the human breast cancer classification adapted to dogs (World Health Organization classification system) [46,47], and graded according to the criteria of Elston and Ellis [48] adapted to canine mammary carcinomas [38]. The pathologic tumor size (pT, measured on histological slides), lymphovascular invasion (LVI), dermal infiltration, cutaneous ulceration, muscle invasion, margin status, and central necrosis were recorded for each case. Peritumoral lymphohistiocytic inflammation was considered positive when moderate to severe. In case of multicentric CMC, the largest tumor and/or tumor of highest histological grade was considered for prognostic purposes.
ERα, PR, and Ki-67 were assessed based on the number of positive nuclei among > 500 neoplastic cells (manual image analysis, Image J software, National Institute of Health, Bethesda, Maryland, USA), and considered positive at threshold ≥ 10% for ERα and PR [45,49], CK5/6, and EGFR [50]. The > 33.3% threshold for Ki-67 was evaluated by the receiver-operator-characteristic curve calculated for 2-year cancer-specific mortality. HER2 was scored 0 for no staining at all or incomplete, faint/barely perceptible membrane staining in ≤ 10% of tumor cells; 1 + for incomplete and faint/barely perceptible membrane staining in > 10% of tumor cells; 2 + for circumferential and incomplete and/ or weak/moderate membrane staining in > 10% of tumor cells; or incomplete and circumferential membrane staining that is intense but within ≤ 10% of tumor cells; and 3 + for circumferential, complete, and intense membrane staining in > 10% of tumor cells. Carcinomas were considered HER2 positive only for a 3 + IHC score [45,51].
Negative controls were included in each IHC run, and consisted in replacing the primary antibody with normal mouse or rabbit serum (prediluted reagents, Roche Diagnostics). The positive controls were mostly internal (epidermis and hair follicles for Ki-67, CK5/6, and EGFR; nonneoplastic mammary gland surrounding the carcinoma for ERα and PR; sebaceous glands for ERα). For HER2, the pathway HER2 4-in-1 control slides (Roche Diagnostics) were chosen to assess the quality of staining for each HER2 score (0, 1 + , 2 + , 3 +).
Four veterinary pathologists (JA, FN, LP, AG) and 1 medical pathologist (DL) examined the HES and IHC slides blindly (i.e., without any information on the dog or on the other pathologists' interpretation). In case of discrepancy between evaluators, cases were collectively reviewed in order to achieve a consensual diagnosis, grade, and immunohistochemical scoring.

Statistical analyses
The MedCalc ® statistical software (Ostend, Belgium) was used. Continuous variables are expressed as median, [range], mean ± standard deviation. Correlations between categorical variables were analyzed using the Pearson Chi-square test. The Kaplan-Meier method and log-rank tests were used for univariate survival analyses, and Cox proportional hazards models for multivariate survival analyses, whose results are reported using the Hazard Ratio (HR), its confidence interval (95%-CI), and the p value of each covariate. For all statistical tests, a p value < 0.05 was considered significant.
The mean pathologic tumor size was 18 ± 7 mm [median 18 mm, range (2-49), n = 227 dogs]; in the other cases, the pathologic tumor size could not be precisely determined due to larger size and/or positive margins. In 236 dogs (67.4%), the pathologic nodal stage was pNX due to absence of lymph node sampling for histopathology. Nodal stage pN+ (with metastasis of any size) was confirmed in 75 cases (21.4%). Six dogs (1.7%) had evidence of distant metastasis (M1) at diagnosis.
All of the included cases correspond to invasive mammary carcinomas according to breast cancer classification. The predominant histological types were simple  Both immunohistochemical protocols used to assess HER2 expression were highly correlated (p < 0.0001, Chisquare test). HER2 score 0 was predominant (70.3% of the cases with clone 4B5, 74.9% with polyclonal A0485), followed by HER2 score 1 + ( Table 1). The cohort does not comprise any case with HER2 overexpression (score 3 +).

Locoregional relapse (LRR)
The median time to LRR was 26.4 months; the LRR probability was 34% at 1 year, and 47% at 2 years post diagnosis (Fig. 2a). At the end of the follow-up period, 76 dogs (21.7%) had experienced tumor recurrence at the site of prior mastectomy, 56 dogs (16.0%) a new primary mammary tumor, and 18 dogs (5.1%) more than one locoregional event. Fig. 1 Immunohistochemical markers of canine invasive mammary carcinomas. Positivity to a Estrogen Receptor alpha (ERα, nuclear), b Progesterone Receptor (PR, nuclear), c the proliferation index Ki-67 (nuclear), d score 2 + for Human Epidermal Growth Factor Receptor type 2 (HER2, membranous), and positivity to e basal cytokeratins 5 and 6 (CK5/6, cytoplasmic), and f Epidermal Growth Factor Receptor type 1 (EGFR, membranous) in 6 different canine invasive mammary carcinomas. Indirect immunohistochemistry, initial magnification × 400, bar = 50 micrometers By univariate analysis, 11 parameters were significantly associated with the LRR risk (Table 2), of which 4 remained as significant independent prognostic factors by multivariate analysis (p < 0.0001): the strongest predictor of locoregional relapse was ERα positivity (HR 0.48), followed by the pathological nodal stage pN + (HR 1.92), the presence of lymphovascular invasion, and positive margins (HR = 1.55 for each).

Distant Metastasis-Free Interval (DMFI)
The risk of distant metastasis was 17% at 1 year and 24% at 2 years post diagnosis (Fig. 2b), and was likely underestimated in this retrospective cohort, as the dogs' owners may have declined complete staging, for financial reasons.
By univariate analysis, six parameters were significantly associated with DMFI (Table 3), of which four remained as significant independent prognostic factors by multivariate analysis (p < 0.0001): the strongest was lymphovascular invasion (HR 2.66), followed by age at diagnosis (HR 2.16 for older dogs), multicentricity (HR 1.89), and the Ki-67 proliferation index (HR 1.0149).

Disease-free interval (DFI)
The median DFI was 34.4 months. Cancer progression (locoregional recurrence and/or distant metastasis) was recorded in 34% of dogs at 1 year post diagnosis, and 45% at 2 years.
By univariate analysis, 13 parameters were significantly associated with DFI (Table 4)

Specific survival
The median time to death attributable to cancer was 19.5 months [2 days-56 months] (Fig. 2d). The cancerrelated death rate was 41.5% at 1 year and 54.1% at 2 years post diagnosis. By univariate analysis, 15 clinicopathologic parameters were significantly associated with cancer-related death (Table 6), of which six were independent prognostic factors by multivariate analysis (p < 0.0001). The most significant predictors of cancer-related death were those that define the stage of invasive CMCs: the pathologic tumor size (pT ≥ 20 mm: HR 1.41), pathologic nodal stage (pN + : HR = 1.82), and the presence of distant metastases at diagnosis (M1: HR 2.61). Peritumoral inflammation (HR 1.54), ERα negativity (HR 1.56), and a high Ki-67 proliferation index (HR 1.67) were also associated with cancer-related death, independently of the stage of the carcinoma at diagnosis.
The epidemiological characteristics of CMCs in this cohort, although in agreement with some previous reports [20,33], are characterized by an older age at diagnosis, lower rate of positivity to ERα and PR, and higher Ki-67 index than previous descriptions [19,23,27,31,37,39]. These differences can be attributed at least in part to the systematic exclusion of mammary carcinomas in situ, which are diagnosed in younger dogs, and are more commonly ERα and PR positive compared to invasive CMCs (unpublished observations, manuscript in preparation), as described in human breast cancer [52].
Another particularity of this cohort is the absence of any HER2-positive CMC, as previously reported [53]. However, HER2-positive CMCs have been previously described by immunohistochemistry, with the polyclonal A0485 antibody [18,26,[54][55][56][57][58] or the CB11 clone [59]. In this study, the external positive controls (cytospins of breast carcinoma cell   lines, representative of the HER2 scores 0-3 +) ensured that HER2 expression was neither underestimated nor overestimated on the canine slides, a precaution that was rarely taken in veterinary oncology [26,54]. Of note, HER2 gene amplification was not previously found in CMCs [60], and thus the existence of HER2-positive mammary carcinomas in dogs is still uncertain [61,62]. The natural history of invasive CMCs is much shorter in dogs (54% cancer-related death at 2 years post diagnosis in this study) than in human breast cancer [2], probably in relation to shorter life expectancy in dogs, and the lack of adjuvant therapy, a situation that favors the setting of preclinical trials in the canine species. The effects of a given compound on patient survival, including in first-line regimen, is expected to be evaluable in short delays in dogs with CMCs, an advantage already highlighted for other canine cancers [4,5]. The prognostic factors of invasive CMCs, described here in the largest retrospective cohort described so far, include the pathologic tumor size, pathologic nodal stage, lymphovascular invasion, histological grade, and ERα positivity, which are all also strong prognostic factors in human breast cancer [63], confirming the similar biology of invasive mammary carcinomas in both species.

Conclusions
The results of the present study confirm that canine invasive mammary carcinomas have a short disease course, which is predictable with clinicopathologic criteria close to those of human oncology. In the second part of this article, we hypothesized that CMCs could be subdivided into luminal and triple-negative cases with different outcomes, as in human breast cancer.