A. E. L. and M. S. S. A. contributed equally for the manuscript.
Plasma biomarkers profile of female dogs with mammary carcinoma and its association with clinical and pathological features
Article first published online: 12 OCT 2013
© 2013 John Wiley & Sons Ltd
Veterinary and Comparative Oncology
Volume 14, Issue 1, pages 88–100, March 2016
How to Cite
Estrela-Lima, A., Araújo, M. S. S., Soares, R. P., Ribeiro, L. G. R., Damasceno, K. A., Costa, A. T., Teixeira-Carvalho, A., Martins-Filho, O. A. and Cassali, G. D. (2016), Plasma biomarkers profile of female dogs with mammary carcinoma and its association with clinical and pathological features. Veterinary and Comparative Oncology, 14: 88–100. doi: 10.1111/vco.12070
- Issue published online: 25 JAN 2016
- Article first published online: 12 OCT 2013
- Manuscript Accepted: 29 AUG 2013
- Manuscript Revised: 24 JUL 2013
- Manuscript Received: 5 JUN 2013
- Conselho Nacional de Desenvolvimento Científico e Tecnológico
- Fundação de Amparo a Pesquisa de Minas Gerais
- Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)
- clinical pathology;
- comparative oncology;
The immunological biomarkers profiles were evaluated using Luminex as putative measures to monitor canine mammary carcinomas (MCs). Forty female dogs were categorized into benign mixed tumour (MC-BMT = 28) and mammary carcinoma (MC=12). The ascendant biomarker signatures were used to compare the groups. For example, a higher frequency of MC-BMT animals producing IL-6, CXCL-8 and CXCL-10 was observed, whereas for the MC group IL-2 and CXCL-8 were detected. MC-BMT animals without metastasis had an increase in the levels of IL-2, CXCL-8, CXCL-10, IL-6, TNF-α, IL-15 and a decrease in IL-10 and CXCL-8. MC-BMT animals with metastasis showed only an increase in CXCL-10 and a decrease in IL-18. After comparing the ascendant signatures following the presence of metastasis in both groups, a higher frequency of dogs exhibiting IL-10 production was observed. Pearson correlation (P = 0.0273) and receiver operating characteristic (ROC) curve analysis revealed that this pattern was associated with worse outcome and lower survival rates in MC animals.
The concept that the immune system recognizes and controls cancer was first postulated over a century ago. In breast cancer, inflammation often correlates with increased invasiveness and poor prognosis. Strong evidence in humans and mice supports the involvement of cytokines in tumour initiation, growth and metastasis.
Over the last decades, many advances have occurred in the understanding of the role of cytokines in breast cancer. The cytokines can alter growth behaviour, the function of immunocompetent cells and activate/modulate specific or non-specific anti-tumour responses. More importantly, cytokines are probably involved in the tumour cell evasion mechanisms from the immunosurveillance system.
Cytokines functions are complex during tumour development. Some (IL-1, IL-6, IL-11, TGF-β) stimulate while others (IL-12, IL-18, IFNs) inhibit breast cancer proliferation and/or invasion. Similarly, high circulating levels of some cytokines seem to be favourable (soluble IL-2R), whereas others are unfavourable (IL-1β, IL-6, CXCL-8, IL-10, IL-18, gp130) prognostic indicators. High cytokine levels in primary breast cancers and in the circulation of affected patients have been associated with poor outcome.
Recent studies suggest that IL-10 possesses immune-stimulatory activity that enhances anti-tumour immunity, affecting growth and peritoneal dissemination of ovarian cancer. This would occur due to the inhibition of angiogenesis and increasing survival. Although IL-10 usually exerts anti-tumour activity, it might also promote tumour development. Its direct effects stimulating tumour cell growth have been reported. These opposing effects might depend on interactions with other cytokines/factors found in the tumour microenvironment. However, most of the mechanisms underlying this phenomenon are still unclear.
In dogs, mammary cancer is one of the most prevalent of tumours leading to death after metastasis. Recently, CXCL-8 was a non-invasive prognostic marker during follow-up in female dogs bearing mammary gland cancer. Its increased levels acts as an independent prognostic marker of survival and the identification of animals with the poor prognostic. However, IL-1 and IL-6 are abundantly expressed in malignant and metastatic canine mammary tumours through interaction with Tumour-infiltrating lymphocytes (TILs).
The aim of this study was to investigate the chemokine/cytokine plasma levels in female dogs with carcinoma in benign mixed tumours (MC-BMT) or mammary carcinoma (MC). Furthermore, a correlation with clinic-pathological parameters and clinical evolution was determined as a prognostic value in those two groups of dogs.
Materials and methods
All procedures in this study were according to the guidelines set by the Colégio Brasileiro de Experimentação Animal (COBEA). This study was approved by the Ethical Committee for the use of Experimental Animals of the Universidade Federal de Minas Gerais, Brazil (CETEA).
Groups of animals
Fifty purebreds or mixed-breed female dogs were selected for this study (age ranging from 8 to 18 years) including 40 dogs with MCs and 10 healthy controls. Animals were admitted at the Veterinarian Hospital of the Universidade Federal da Bahia (UFBA) from October, 2008 to August, 2009. They were submitted to clinical examination by evaluation of physiological parameters, analysis of historical evolution and reproductive records and haematological analysis. The animals with mammary neoplasm were submitted to macroscopic evaluation to determine tumour features (including size, presence of inflammatory reaction and/or ulceration), location and analysis of regional lymph nodes by palpation. Mammary tumours ≥ 3 cm in size were removed and classified based on the histopathological diagnosis.[10-12] Following histopathological analysis two groups were selected: (1) carcinoma in benign mixed tumours (MC-BMT, n = 29) and (2) MC (n = 20). Further histological analyses of inguinal and axillaries lymph node were used to categorize MC-BMT and MC into two subgroups referred as (1) (−) without lymph node metastasis, including clinical stage II-III [MC-BMT (−), n = 17; MC (−), n = 10] and ii) (+) with lymph node metastasis, including clinical stage IV-V [MC-BMT (+), n = 12; MC (+), n = 10]. The clinical stage is determined from information on the size tumour (T), involvement of regional lymph node (N) and the presence or absence of distant metastases (M). Stages # II, III, IV and V were defined based on TNM system previously described. Fifty animals were elected for the survival analysis (MC-BMT, n = 28; MC, n = 12 and controls, n = 10). These animals were submitted to quarterly follow-ups during twelve months, including systematic clinical evaluation, radiological examinations along with biochemical and haematological analysis. Survival rates were expressed in days between the surgical excisions of the end of follow-up. None of the female dogs had prior use of anti-neoplastic drugs, anti-inflammatory drugs or antibiotics within the 30 days prior to surgical excision. The control group (C) consisted of health dogs attended for routine clinical appointments with no evidences of mammary tumour or other disease detected during clinical examination and historical records analysis.
Whole blood samples were collected during clinical appointment and before mastectomy in the dogs with mammary neoplasias. Blood samples constituted of 5 ml of peripheral blood using EDTA as the anticoagulant (final concentration of 1 mg mL−1). The blood samples were centrifuged at 50 g for 30 min at 4 °C, followed by 10 min at 900 g. The plasma was collected and stored at −80 °C until analysis.
A Milliplex MAP Canine specific kit was used. The following chemokine/cytokines were assayed (CXCL-10, IL-2, IL-4, IL-6, CXCL-8, IL-10, IL-15, IL-18, IFN-γ and TNF-α). Each 96-well filter plate (Millipore) was blocked with 200 µL of Wash buffer and mixed on a plate shaker for 10 minutes at room temperature (20–25 °C) followed by filtration under vacuum. It was added 25 µL of each Standard or Control into the appropriate wells. Then, 25 µL of Assay Buffer and 25 µL of appropriate matrix solution were added into the background, standards, and control wells. The samples were added to the filter plates (plasma samples were diluted 1:3 with Assay Buffer provided by the kit). The plates were mixed and 25 µL of the Mixed or Premixed Beads were added to each well. The bead mix is light sensitive and was stored in the dark. The plates were placed on a shaker at 4 °C overnight (16–18 h). The following day, the medium was vacuum-filtered, and 25 µL of the detection antibody was added to each well. The plates were incubated with agitation on a shaker at room temperature (20–25 °C) for 1 h. Streptavidin-PE (50 µL) was added to each well. The plate was shaken for 30 min at room temperature. The wells were washed three times with 200 µL/well of wash buffer. The beads were resuspended in 150 µL of Sheath Fluid prior to analysis. Plates were read on a Bio-Plex™ 200 System (Bio-Rad Laboratories, CA, USA). Data for 100 beads per cytokine were collected for each standard and sample dilution. Results were expressed in pg mL−1 and mean fluorescence intensity as provided by the manufacturer (Fig. 1).
Cytokine signature analysis
The cytokine profile was first assessed to identify low and high cytokine producers, as previously suggested. Briefly, after the establishment of the global median of cytokine levels/control index, each cytokine from all animals was tagged as they display low (white) or high proinflammatory (black) indexes (Fig. 2). The percentage of animals showing high cytokine indexes was calculated for each chemokine/cytokine type. The ascendant frequency of high cytokine indexes was then used as the reference cytokine curves to identify changes in the overall cytokine patterns from all other groups.
Statistical analysis was performed using the GraphPad Prism 5.0 software package (San Diego, CA, USA). The comparative analyses of nonparametric data between groups (C, MC-BMT and MC) were performed by Kruskal–Wallis test followed by Dunns post-test to compare all pairs of samples. The survival curves were estimated with the Kaplan–Meier estimation method followed by Log-rank test. In all cases, the differences were considered significant at P < 0.05. The analysis of cytokine signatures was performed using the profile of cytokine signature as the reference curve, and significant differences were considered when the values emerged outside the quartile of the reference signature. Performance indexes were used to determine whether the selected plasma levels of IL-10 in MC would contribute as prognosis biomarker for canine MC. The receiver operating characteristic (ROC) curve was used to select the best cut-off value for IL-10 in MC to discriminate distinct evolution to death or survival. The performance analysis included the following: the global accuracy was also evaluated by taking the area under the ROC curve (AUC). Co-positivity (Co-pos) = [true positives/(true positive samples + false negative samples)] × 100; Co-negativity (Co-neg) = [true negatives/(true negative samples + false positive samples) × 100; Positive Likelihood ratio (LR+) = Co-positivity/(1 − Co-negativity); Negative Likelihood ratio (LR+) = (1 − Co-positivity)/Co-negativity].
Quantitative analysis of plasma immunological biomarkers
The quantitative plasma immunological biomarkers levels were quantified by Luminex bead-based assay and the results expressed in pg mL−1 or mean fluorescence intensity, as provided by the manufacturer (Fig. 1). Data analysis did not demonstrate any significant differences amongst the clinical groups during the quantitative analysis (Fig. 1). We have noticed that all clinical groups displayed a very wide range of values, leading to a broad scattering and interquartile range intervals. This large spreading of values had a major impact of the conventional statistical tests and therefore, no significant differences could be found between groups, even though the median values would suggest such differences. To further address the immunological biomarkers profile in canine MC, we have then applied a novel analytical tool to evaluate cytokine patterns previously reported by Luiza-Silva et al.
Plasma biomarkers profile
The biomarker profile was first assessed to identify the dogs with ‘low’ and ‘high’ plasma levels, as previously reported. Briefly, after the establishment of the global median for all animals, they were tagged as they display low or high biomarker indexes. The percentage of animals showing high cytokine indexes was calculated for each clinical groups (Control, MC-BMT and MC). Additional analysis was carried out to characterize the biomarker balance defined as the predominant proportion of ‘low’ or ‘high’ plasma levels on each clinical group (Fig. 2). The ascendant frequency of high cytokine indexes for all dog groups was then used as the reference (Fig. 2, left). Following this methodology, more than 50% of the control dogs exhibited higher levels of IL-10 and IL-18. For the MC-BMT group, a higher frequency of animals producing IL-6, CXCL-8 and CXCL-10 was observed, whereas for the MC group IL-2 and CXCL-8 was detected (Fig. 2, right).
To evaluate the changes in the cytokine profiles of MC-BMT and MC dogs, the assembling of the ascendant frequency of animals with high biomarkers levels as a control reference signature curve was created. Our findings demonstrated that a higher frequency of MC-BMT dogs showed an increase in CXCL-8, CXCL-10 and IL-6 and a decrease in the IL-10 and IL-18 levels (Fig. 3, arrows). However, the MC group exhibited an increased frequency of dogs producing IL-2, CXCL-8 and lower levels of CXCL-10 and IL-18.
The association of plasma biomarker profiles with metastasis occurrence in MC-BMT and MC dogs was also performed (Fig. 4). After comparison with the control signature group it was observed that MC-BMT animals without metastasis had an increase in the levels of IL-2, CXCL-8, CXCL-10, IL-6, TNF-α, IL-15 and a decrease in IL-10 and CXCL-8. MC-BMT animals with metastasis showed only an increase in CXCL-10 and a decrease in IL-18. In the MC group without metastasis an increase of IL-2 and CXCL-8 and a decrease of IL-10 and IL-18 was observed. However, in the MC with metastasis group showed higher levels of IL-2, IL-6 and TNF-α and lower levels of IL-18.
Relationship between plasma biomarkers with tumour metastasis
With the objective of evaluate the changes in the chemokine/cytokine profiles associated with metastasis, an ascendant curve was created for comparison within MC-BMT and MC groups (Fig. 5). The data showed that in the MC-BMT group with metastasis exhibited an increase in the IL-10 and a decrease in IL-2, TNF-α, IL-15, IL-6 and CXCL-8. Regarding the MC group, an increase in TNF-α, IL-10 and IL-6 was noticed and a decrease in CXCL-8.
While comparing the ascendant chemokine/cytokine signatures according to the presence of metastasis in MC-BMT and MC dogs, a different frequency in the cytokine levels was detected (Fig. 6). In both MC-BMT and MC groups without metastasis, a higher frequency of dogs showed an increase in IL-2 and CXCL-8. In the MC-BMT group without metastasis showed also an increase in IL-10, TNF-α, IL-15 and IL-6. In the metastatic animals from both groups a higher frequency of dogs exhibiting IL-10 production was observed. In addition, the higher frequency of animals producing CXCL-10 (MC-BMT) and TNF-α, IL, 6 and IL-2 (MC) was also detected.
Survival curves comparison
Since a higher frequency of animals in the metastatic group exhibited an increase in IL-10 production, an association with survival was performed (Fig. 6). Only one surviving dog (#6) (11%) presented higher levels of IL-10 (≤15), whereas in the other group 73% of the animals with higher IL-10 levels (>15) died (Pearson's correlation analysis, P = 0.0273; r = −0.5053). Confirming those data, the survival curves comparison (Kaplan–Meier) stratified based on the IL-10 levels, demonstrated significant (P < 0.05) lower levels of IL-10 (<15 MFI) in the dogs that died. Analysis of performance indexes of plasma IL-10 production as a prognosis survival biomarker supported our results, showing a high global accuracy value (AUC = 0.84) (Fig. 7).
The use of prognostic markers for mammary cancer is important for routine diagnosis and research since this disease is one of the most commonly found in female dogs.[17-20] One of the most relevant challenges in canine carcinomas follow-up is the low accuracy of biomarkers for postsurgical prognosis. Usually, prognosis is based on several tumour and host characteristics, including the histological type, tumour grade, stage, together with distinct aspects of host immune response.[17, 21-26] Our recent reports have demonstrated the role of circulating leukocytes as clinically relevant prognostic biomarkers for MCs in dogs.[16, 27] First, a high intensity of lymphocytic infiltrate and the relative abundance of the CD4+ and CD8+ T-lymphocytes might represent an important survival prognostic biomarker for this disease. Later, in order to determine the leucocyte population involved, it was observed a decreased percentage of B-cells and an increased frequency of NK-cells, CD8+T-cells and CD8+CD5Low+T-cells. More importantly, the CD4+/CD8+ T-cells ratio was elected as a valid parameter to predict survival in MC-BMT dogs.
The presence of inflammatory cells in human tumours has been correlated to an inefficient immune response in preventing tumour development.[28-30] There is evidence that some inflammatory cytokines (IL-1β, IL-6, IL-23 and TNF-α) promote tumour development by acting directly or indirectly on tumour cells.[28, 31, 32] However, the role of chemokine/cytokines during the course of MCs in dogs is still a missing step in the immunopathology of the disease.
To improve our understanding of the immunological heterogeneity of canine MC, it was determined the cut-off scores among the clinically different groups (Fig. 2). In our study, we have observed a distinct chemokine/cytokine profile depending on the group, where higher levels of IL-10 and IL-18 were observed in more than 50% of the control dogs. In the MC-BMT group, a higher frequency of animals producing IL-6, CXCL-8 and CXCL-10 was observed, whereas for the MC group IL-2 and CXCL-8 were detected (Fig. 2, right, Fig. 3). It was observed a higher frequency of CXCL-8 producers in both MC-BMT and MC groups.[16, 27] Those data are consistent with our previous results showing a high lymphocytic infiltrate in the canine MCs. The role of inflammation and inflammatory cells in tumour development and progression has been increasingly studied. Both tumour cells and those participating in inflammation release cytokines and chemokines. Interleukin-8 (CXCL-8) is a chemokine (chemotactic cytokine) produced by monocytes, lymphocytes, endothelial cells, epithelial cells, fibroblasts and a range of other cells in response to stimuli such as lipopolysaccharide and other cytokines (e.g., TNF-α and IL-1β). Besides, CXCL-8 is involved in angiogenesis and as a mediator of chemotactic and proliferative activity in endothelial cells and NSCLC tumour lines. Consistent with the studies with human breast, our data also indicate an increase in this cytokine in MC-BMT and MC female dogs. Interestingly, a higher frequency of CXCL-10 producers was observed in the MC-BMT group. CXCL-10 has been described as a potent anti-tumour agent being able of inhibiting tumour growth of Burkitts' tumour.[35, 36] Besides CXCL-8, several studies have reported that mammary gland cancer patients exhibited increased serum levels of IL-6 and IL-10 compared to healthy individuals. These cytokines stimulate many signalling pathways and thereby regulate the transcription of target genes involved in cell proliferation, differentiation and survival. IL-6 is suggested to have a pivotal role in the pathogenesis of Kaposi sarcoma and MM. Recent studies also suggest an association between circulating IL-6 and elevated risk of developing Hodgkin lymphoma. Furthermore, a promoter polymorphism study suggests that IL-6 is a predisposing genetic factor that contributes to breast cancer prognosis. This is a result of a G/C polymorphism within the promoter region of the IL-6 gene triggering high levels of IL-6 production leading to a worse prognosis. IL-6 properties may include anti-tumour activity, induction of T-cell and B-cell differentiation, stimulation of cytotoxic cells and help in producing limphokine-associated killer cells. Our data also indicated an increase in IL-6 in the MC-BMT and this may result in a better prognosis compared to the MC group. However, IL-2, considered an anti-tumour agent, its levels were higher in the MC group, thus representing a worse prognosis.
To observe if the type of cytokine could be associated with metastasis in either MC-BMT or MC dogs, the levels of plasma cytokines were compared. A distinct profile was observed in the dogs with metastasis in both groups. It is interesting to notice that IL-10 levels were increased in both groups with metastasis (Figs 5 and 6). However, the effects of IL-10 are dramatically opposed to those of IL-6, as IL-10 is immunosuppressive and anti-inflammatory. IL-10 inhibits NF-kB activation through ill-defined mechanisms[44, 45] and consequently inhibits the production of proinflammatory cytokines, including TNF-α, IL-6 and IL-12. Under normal circumstances, IL-10 inhibits tumour development and progression and higher levels of this cytokine were observed in normal dogs (Fig. 2, right). IL-10 has also been shown to modulate apoptosis and suppress angiogenesis during tumour regression.[47, 48] Expression of IL-10 in mammary and ovarian carcinoma xenografts inhibits tumour growth and spread.[7, 47] One mechanism by which IL-10 inhibits tumour growth was suggested to depend on down-regulation of MHC class I expression, leading to enhanced NK cell-mediated tumour cell lysis. Inhibition of the tumour stroma was suggested to contribute to the anti-angiogenic activity of IL-10. The ability of IL-10 to down-regulate VEGF, TNF-α, and IL-6 production by TAMs might also account for its inhibitory effect on the tumour stroma. Although IL-10 usually exerts anti-tumour activity, its biological effects are not all that simple, and consistent with its ability to activate STAT3, it might also promote tumour development. Direct effects of IL-10 on tumour cells that might favour its growth have been reported. For example, an IL-10 autocrine and/or paracrine loop might have an important role in tumour cell proliferation and survival. An elevated amount of IL-10 in the plasma has been correlated with poor prognosis in diffuse large B cell lymphoma patients. A role for IL-10 in the progression of B cell malignancies is also seen in IL-10−/− mice, in which B cell tumours grow more slowly. In a B16-melanoma xenograft model, IL-10-transfected cancer cells develop more vascularized tumours and exhibit further growth. In addition to direct growth modulation of cancer cells, the ability of IL-10 to suppress adaptive immune responses has also been suggested to favour tumour escape from immune surveillance. In summary, IL-10 has complex effects on tumour development. In many experimental systems, IL-10 is found to exert anti-tumour activity, but in other cases it can be pro-tumourigenic. These dramatically opposing effects of IL-10 might depend on interactions with either cytokines or factors found in the tumour microenvironment. A better understanding of IL-10 signalling is needed before its effects on tumour growth and anti-tumour immunity can be fully explained. However, in this work, the fact that IL-10 was observed in the different groups with metastasis in our analyses raised the question if this cytokine could be related to a worse prognosis in canine MC.
In this sense, we have also observed a higher percentage of MC dogs expressing higher levels of IL-10, suggesting a possible non-protective anti-tumoural role for this cytokine. Previous studies have demonstrated that a higher percentage of NK-cells in MC-BMT were correlated to a better prognosis. However, the lower expression of MCHI and MCHII was observed in the MC group with worse prognosis. The evaluation of performance indexes for IL-10 plasma levels and survival showed co-positivity (sensitivity) of 73%, co-negativity (specificity) of 89%. The data analysis showed that IL-10 expression (>15 MFI) is a parameter highly specific, indicating its ability to detect in female dogs with mammary tumours histologically classified as MC whose disease will progress to the death of an animal.
In conclusion, mammary gland tumours in female dogs are an excellent model for the clinic-pathological, diagnostic and prognostic investigation of mammary neoplasias. Prognostic and predictive markers are effective in research and routine diagnosis. Interleukins play a fundamental role in cancer, with a particular function in tumour growth, invasion and metastatic potential. In the lack of studies devoted to the diagnostic, therapy and prognostic markers in mammary gland cancer in canines, this work defines a potential prognostic marker for routine use in the clinic. Our data showed a different cytokine profile depending on the type of canine MC (MC-BMT or MC) and also a possible involvement of IL-10 in a worse prognosis for this disease.
This study was supported by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Fundação de Amparo a Pesquisa de Minas Gerais, (FAPEMIG) and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES). A. E. L. is a post-doctoral fellowship from CNPq. ATC, OAMF, RPS and GDC are grateful for CNPq research fellowship (PQ). The authors thank the programme for technological development in tools for health—PDTIS—FIOCRUZ for use of its facilities.
- 62013; 64: 45–57.and . Cytokines, obesity, and cancer: new insights on mechanisms linking obesity to cancer risk and progression. Annual Review of Medicine
- 101999., , and . Histological classification of mammary tumors of the dog and the cat. Armed Forces Institute of Pathology and the American Registry of Pathology and The World Health Organization Collaborating Center for Worldwide Reference on Comparative Oncology, Washington DC, pp 11–29,
- 12Consensus for the diagnosis, prognosis and treatment of canine mammary tumors. Brazilian Journal of Veterinary Pathology 2011; 4: 153–180., , , et al.
- 13TNM classification of tumors in domestic animals. Geneva, World Health Organization, 1980..
- 17Tumors of the mammary gland. In: Tumors in Domestic Animals. 4 edn., DJ Meuten Ed., Iowa, Iowa State Press, 2002: 575–606..
- 18Diseases of the mammary glands. In: Handbook of Small Animal Practice. 3 edn., RV Morgan Ed., Philadelphia, 1994: 615–625.and .
- 19Canine and feline mammary tumors. In: Cancer in Dogs and Cats: Medical and Surgical Management. WB Morrison Ed., Baltimore, Williams and Wilkins, 1998: 591–598..
- 20Treatment of mammary tumors in dogs and cats. In: Proceedings of the North American veterinary conference. Orlando, 2003..
- 22Small animal surgery. St. Louis, Mosby-Year Book, 1997: 539–544..
- 23Predictors of tumor response. In: Proceedings of the North American veterinary conference, Orlando, Florida, 679–681, 2003..
- 28The relationship between the systemic inflammatory response, tumour proliferative activity, T-lymphocytic and macrophage infiltration, microvessel density and survival in patients with primary operable breast cancer. British Journal of Cancer 2008; 99: 1013–1019., , , , et al.
- 34Tumors of the mammary gland. In: Tumors in Domestic Animals. 3 edn., California, University of California Press, 1990: 518–550..