Five markers useful for the distinction of canine mammary malignancy
© Pawłowski et al.; licensee BioMed Central Ltd. 2013
Received: 21 December 2012
Accepted: 9 July 2013
Published: 11 July 2013
Spontaneous canine mammary tumors constitute a serious clinical problem. There are significant differences in survival between cases with different tumor grades. Unfortunately, the distinction between various grades is not clear. A major problem in evaluating canine mammary cancer is identifying those, that are “truly” malignant. That is why the aim of our study was to find the new markers of canine malignancy, which could help to diagnose the most malignant tumors.
Analysis of gene expression profiles of canine mammary carcinoma of various grade of malignancy followed by the boosted tree analysis distinguished a `gene set`. The expression of this gene set (sehrl, zfp37, mipep, relaxin, and magi3) differs significantly in the most malignant tumors at mRNA level as well as at protein level. Despite this `gene set` is very interesting as an additional tool to estimate canine mammary malignancy, it should be validated using higher number of samples.
The proposed gene set can constitute a `malignancy marker` that could help to distinguish the most malignant canine mammary carcinomas. These genes are also interesting as targets for further investigations and therapy. So far, only two of them were linked with the cancer development.
Spontaneous mammary tumors are the most common type of cancer in women and in female dogs. The annual incidence in dogs is 3 times more frequent than in humans. Thus, it constitutes a serious clinical problem . Unfortunately, the distinction between various grades of tumors can be considerably less clear than one would like . A major problem in evaluating canine mammary cancer is identifying those, that are “truly” malignant . The most significant criteria for the diagnosis of malignant mammary tumors in the dog are as follow: tumor type, significant nuclear and cellular pleomorphism, mitotic index, presence of randomly distributed areas of necrosis within the neoplasm, peritumoral and lymphatic invasion, and regional lymph node metastasis. However, classification based on these criteria sometimes leads to the over-diagnosis of mammary carcinoma . However, good pathological diagnosis constitute basis to yet proper prognosis . Karayannopoulou et al.  found significant differences in survival between cases with different tumor grades: survival was worse in dogs with grade III carcinomas than in those with grade I or grade II.
The advent of DNA microarrays - a modern and powerful tool for investigation of biological processes through the widespread analysis of genes from a particular cell, tissue, or organism - enabled investigators to identify characteristic expression patterns of groups of genes that are associated with specific tumor traits. They are a great hope of finding new cancer markers, markers of clinical outcome, and targets for anticancer therapy .
The purpose of this preliminary study was to point out find new markers of canine mammary malignancy (that could help to improve pathological diagnosis of the tumor grade) based on the gene expression. We built the molecular classifier, which was able to distinguish the most malignant canine mammary tumors from the lowest estimated malignant tumors. As of the highest grade of malignancy is associated with an increased risk of death within 2 years after mastectomy , proper diagnosis may lead to prediction to the clinical outcome.
Tumor samples of canine mammary cancers were obtained from patients subjected to surgery. The tumors then, were divided into two halves, one of them was fixed in 10% neutral buffered formalin and routinely embedded in paraffin to perform histological assay (n = 78 samples in total). Whereas the other one was snap frozen in liquid nitrogen and stored in −80°C (n = 18). Four μm samples from paraffin blocks were fixed on glass slides, stained with haematoxylin – eosin (HE) and had been examined by certified pathologists (prof. Dr. hab. Elżbieta Malicka and Dr. Izabella Dolka, both from the Warsaw University of Life Sciences, Poland). The immunohistochemical examination of cytokeratin, vimentin, smooth muscle actin, s100 protein and p63 protein expression was performed in order to diagnose these tumors (data not shown). The tumor types of specimens were classified based on the World Health Organization (WHO) Histological Classification and Mammary Tumors of the Dog and Cat classification . Histological tumor grading was conducted on HE-stained sections using a Misdorp classification . The mammary carcinoma grading was assessed in respect to tubule formation, degree of differentiation and mitotic index as grade I, grade II, or grade III tumor. The tumors dedicated to this study were non-metastatic (no metastasis detected in lymph nodes or lungs of the patients). Clinical information about the outcome of these patients is unknown.
Statistical analysis and building of the classifier of canine mammary malignancy
Twenty the highly significant genes in the most malignant tumors
Prior to building the classifier we calculated expression of the selected genes relative to the median expression the reference genes as measured on the same microarray (rps19, cgi-119, ctbp1 and b2m were used as the reference genes) [8, 9]. We built the classifier using this relative signal calculated for features listed in Table 1. The decision tree classifier was built with the use of boosted C5.0 algorithm.
The gene function was identified with the use of NCBI database and PANTHER pathway analysis software . The pathway analyses were conducted using one-way ANOVA with binominal Bonferroni statistic test (PANTHER) with the cut-off value p < 0.05.
Primers used for real-time qPCR
Optimum annealing temp. (°C)
Optimum annealing time (sec)
Four μm sections from paraffin blocks containing tumour tissue (n = 60 in total; 20 tumor tissues from each grade of malignancy) were baked in 37°C overnight. After dewaxing in xylene and rehydration in ethanol, for antigen retrieval, the slides were placed in 0.02 M citrate buffer, pH 6.0 and boiled in the decloaking chamber. The samples were incubated in the Peroxidase Blocking Reagent (Dako, Denmark) for 10 min at room temperature prior to the antibody incubation. After 30 min incubation in 5% bovine serum albumin (Sigma Aldrich, Germany), the following mouse/rabbit primary antibodies were used (diluted 1:100 in 1% bovine serum): anti MAGI3, ZFP37, SERHL, Relaxin, MIPEP (all of them obtained from Abcam, UK). The slides were incubated with antibodies overnight at +4°C. For the staining the EnVision kit (Dako) was used (Labelled Polymers consist of secondary anti-mouse or anti-rabbit antibodies conjugated with the Horseradish peroxidase HRP enzyme complex). To develop the colored product, the 3,3`-Diaminobenzidine (DAB) substrate was used. Finally, the haematoxylin was taken for nuclei counterstaining. To confirm the differences in expression of these antigens between tumors of the various grade, n = 60 samples per each antibody were examined.
For each immunohistochemical experiment the negative control was used (the staining without the use of primary antibodies).
The statistical analysis of Real-time qPCR and immunohistochemistry was conducted using Prism version 5.00 software (GraphPad Software, USA). The one-way ANOVA, and ANOVA + Tukey HSD (Honestly Significant Difference) post-hoc test were applied. The p-value <0.05 was regarded as significant whereas p-value <0.01 and p-value <0.001 as highly significant.
Selection of gene set used for diagnosis of malignant tumor
In the first analysis we built the classifier based on the training data. The algorithm selected five genes: mipep, serhl, relaxin, magi3 and zfp37 which expression can allow us to diagnose the most malignant tumors. The boosted tree rules were as follow:
Rule 1. estimated accuracy 95.65% [boost 100%]
sehrl < = 4.046 [ Mode: 1 ] = > 1
sehrl > 4.046 [ Mode: 3 ] = > 3
Rule 2. estimated accuracy 96.17% [boost 100%]
zfp37 < = 3.377 [ Mode: 1 ] = > 3
zfp37 > 3.377 [ Mode: 3 ] = > 1
Rule 3. estimated accuracy 96.57% [boost 100%]
zfp37 < = 2.914 [ Mode: 1 ] = > 3
zfp37 > 2.914 [ Mode: 3 ] = > 1
Rule 4. estimated accuracy 95.48% [boost 100%]
mipep < = 1.257 [ Mode: 1 ] = > 1
mipep > 1.257 [ Mode: 3 ] = > 3
Rule 5. estimated accuracy 88.44% [boost 100%]
magi3 < = 1.458 [ Mode: 3 ] = > 3
magi3 > 1.458 [ Mode: 1 ] = > 1
Rule 6. estimated accuracy 100% [boost 100%]
sehrl < = 4.046 [ Mode: 1 ]
relaxin1 < = 4.203 [ Mode: 1 ] = > 1
relaxin1 > 4.203 [ Mode: 3 ] = > 3
sehrl > 4.046 [ Mode: 3 ] = > 3
Its efficacy assessed using the learning data was 100% in case of the most malignant tumors.
Confirmation at mRNA level of the expression patterns of `diagnostic` set of genes in canine mammary carcinomas of various grade of malignancy
The highest relative expression of serhl showed grade III tumors (0.0095 ± 9.75e-005), whereas the lowest expression showed grade I tumors (0.0044 ± 0.0002). The relative mean expression of serhl in grade II tumors was 0.0057 (±0.0002). The relative expression of serhl differed highly significant between the most malignant tumors and other groups (p < 0.01) (Figure 2A).
The highest relative expression of zfp37 showed grade I tumors (0.0031 ± 0.0008), whereas the lowest expression exposed the most malignant tumors (0.0012 ± 5.5e-005). The relative mean expression of zfp37 in grade II tumors was 0.0022 (±0.0004). The relative expression of zfp37 differed significantly between grade III tumors and others (p < 0.05) (Figure 2B).
The highest relative expression of mipep presented grade III tumors (0.025 ± 0.0044), whereas the lowest expression showed grade I tumors (0.009 ± 0.0005). The relative mean expression of mipep in grade II tumors was 0.015 (±0.0038). The relative expression of mipep differed significantly between most malignant tumors and others (p < 0.05) (Figure 2C).
The highest relative expression of relaxin gave us the result grade III tumors (0.0019 ± 3.25e-005), whereas the lowest expression showed grade I tumors (0.0006 ± 4.65e-005). The relative mean expression of relaxin in grade II tumors was 0.0013 (±3.5e-005). The relative expression of relaxin differed highly significant between grade I and grade III tumors (p < 0.001) and between grade I and grade II tumors as well as grade II and grade III tumors (p < 0.01) (Figure 2D).
The highest relative expression of magi3 showed grade I tumors (0.0011 ± 0.0001), whereas the lowest expression was revealed in grade III tumors (0.0002 ± 5.5e-006). The relative mean expression of magi3 in grade II tumors was 0.0004 (±1.3e-005). The relative expression of magi3 differed significantly between grade III tumors and others (p < 0.001) (Figure 2E).
Expression of selected set of markers at protein level
The antibodies (described in Materials and Methods section) were dedicated to be used in various species, indicating their wide cross-specificity. However, we have performed the analysis of protein sequence in species which specificity was guaranteed by manufacturer and in canine by blasting them using NCBI tool. Additionally, we used the BLOSUM62 amino acid substitution matrix to score alignments between evolutionarily divergent protein sequences. Our results revealed a high similarity between these proteins in dogs and reactivity species, and thus based on the results we came up to use these antibodies in our experiments. The analysis of whole MAGI3 protein sequence showed 100% identity (protein accession numbers: Q9JK71 and J9P535), 94% identity in MIPEP (protein accession numbers: Q99797 and F1PWG7), 80% identity in ZFP37 (protein accession numbers: Q9Y6Q3 and J9P9T4), 64% identity in SERHL (protein accession numbers: Q9EPB5) and 41% identity in relaxin (protein accession numbers: P04808 and Q9TRM8). The identity of the epitope recognized by the antibodies or the peptide fragment used for immunization in the antibody`s manufacturing process was significant. We used similar approach in our previously published study .
Optical density related with the expression of examined proteins in canine mammary carcinoma at various grade of malignancy
Grade I tumors
Grade II tumors
Grade III tumors
103.0 (±1.87) *
The mean optical density of the brown precipitates related to the examined antigen expression was showed in Table 3.
Analysis of gene expression profiles of canine mammary carcinoma of various grade of malignancy (n = 18) followed by the boosted tree analysis distinguished a `gene set` that might be helpful in diagnosis of canine mammary malignancy. This `set` consisted of five genes which expression have significantly changed in the most malignant tumors: sehrl, zfp37, mipep, relaxin, and magi3. Thus, based on their expression pattern the pathologist could distinguish between grade III tumors as well as others. In case of canine mammary tumors, the proper diagnosis is very important because it constitutes basis to the prognosis. The increased histological grade of malignancy is associated with an increased risk of death within 2 years after mastectomy . Despite the proposed `gene set` seems to be useful in malignancy diagnosis, its usefulness should be validated with higher number of samples. So far, these genes and their products have not been treated as `malignancy markers` in breast cancer. Due to similar epidemiological and clinical characteristics of such tumors in these both species, similar environmental conditions, epithelial nature of these tumors, and similar composition of tumor microenvironment, the use of canine mammary tumors as an experimental model for human breast cancers is relevant . Thus, in our opinion these results could be assessed in humans in order to validate the influence of these genes on cancer malignant transformation. So far, only two of these five genes were linked to cancer. These genes have not yet been described as malignancy markers in humans. In our opinion this `gene set` could be also a interesting target for further investigation.
An increased expression of SERHL was observed during passive stretch of skeletal muscles. It is suggested that its up-regulation can be related with skeletal muscle growth in response to mechanical stimuli . Similar conditions (stretching and mechanical stimuli) are observed in fast growing tumors. As being recognized the most malignant tumors are significantly larger than less malignant or benign tumors , increased expression of SEHRL in the grade III tumors is not surprising (Figure 2, Figure 3, Table 3).
The protein encoded by zfp37 belongs to the huge family of zinc finger proteins, which are regulators of transcription. High ZFP37 expression was shown during neurons and chondrocyte development, however in adults it was detected only in brain and testes [14, 15]. ZFP37 is supposed to take part in development and differentiation of the cell. The results of the presented study showed significantly decreased expression of this gene and its protein in the most malignant tumors (Figure 2, Figure 3, Table 3). It can indicate the role of ZFP37 in cancer differentiation due to decrease in differentiation during the malignant transformation process in cancer .
The other selected gene is mipep which is involved in cellular metabolism. The product of this gene performs the final step in processing of nuclear-encoded proteins targeted to the mitochondrial matrix or inner membrane . This gene may contribute to the functional effects of frataxin deficiency and the clinical manifestations of Friedreich ataxia (FA). The exact function of frataxin is still unclear, however it has been implicated in iron homeostasis, protection from oxidative stress and apoptosis . Conflicting results have been reported regarding the role of frataxin in cellular growth: both frataxin knockdown and frataxin over-expression impair cell growth. Despite the role of frataxin in cancer is unknown, the relation between FA and malignancy has been suggested .
The next gene which increased expression was characteristic for the most malignant canine mammary tumors was relaxin. This peptide hormone, which is known as an inhibitor of uterine contractility and promoter of lengthen of the interpubic ligament before partition , promotes growth and development of mammary gland [21, 22]. Moreover, it increases expression and catalytic activities of matrix metalloproteinase (MMP) . It has been also found to enhance invasiveness of cancer cells by MMPs modulation . The results of the present study showed up-regulation of relaxin in the most malignant tumors (Figure 2, Figure 3, Table 3) which are usually the most invasive .
The last significant gene for canine mammary malignancy is magi3. The MAGI protein is usually detected in the tight junctions between cells . This protein reacts with LPA2 receptors which expression is increased in tumors and during inflammation [25, 26], but inflammation often accompanies cancer development [27, 28]. Moreover, in normal cells the activation of this receptor is not required . LPA2 can be activated by MAGI3 or NHERF-2. Knockdown of MAGI3 increases expression of NHERF-2, which enhances cell growth, COX-2 expression and ensures cancer resistance to chemotherapy [30, 31]. These data correlates with our own findings that MAGI3 expression was significantly down regulated in the most malignant tumors (Figure 2, Figure 3, Table 3).
A group of Prof. Klopfleisch published a very interesting paper exposing presentation of changes in proteome at different stages of canine mammary cancer progression . They found 48 proteins which expression significantly changed between benign and malignant tumors. These proteins do not overlap with our findings. Differences may be related with the fact that a group of Prof. Klopfleisch examined general differences between benign and malignant tumors, whereas we have analyzed specific differences between various grades of canine mammary malignancy. However, both studies showed that tumor progression is associated with a stepwise change in protein expression levels, not only between benign and malignant tumors, but also in a relation between less and more malignant ones.
The genes given above can altogether constitute a `set of malignancy markers` used to distinguish the most malignant canine mammary carcinomas from the others. These genes are also interesting as targets for further investigations and therapy. So far, only two of them were linked with the cancer development.
This work was supported by grant no N N308 574940 from Ministry of Sciences and Higher Education.
- Pawłowski KM, Maciejewski H, Dolka I, Mol JA, Motyl T, Król M: Gene expression profiles in canine mammary carcinoma of various grades of malignancy. BMC Vet Res. 2013, 9: 78-10.1186/1746-6148-9-78.PubMedPubMed CentralView ArticleGoogle Scholar
- Polton G: Mammary tumors in dogs. Irish Vet J. 2009, 62 (1): 50-56.Google Scholar
- Goldschmidt M, Pena L, Rasotto R, Zappulli V: Classification and grading of canine mammary tumors. Vet Pathol. 2011, 48: 117-10.1177/0300985810393258.PubMedView ArticleGoogle Scholar
- Karayannopoulou M, Kaldrymidou E, Constantiinidis TC, Dessiris A: Histological grading and prognosis in dogs with mammary carcinomas: application of a human grading method. J Comp Pathol. 2005, 133: 246-252. 10.1016/j.jcpa.2005.05.003.PubMedView ArticleGoogle Scholar
- Król M, Pawłowski KM, Otrębska D, Motyl T: DNA microarrays – future in oncology research and therapy. JPCCR. 2008, 2: 091-096.Google Scholar
- Midsorp W: Tumors of the mammary gland. Tumors in domestic animals. Edited by: Meuten DJ. 2002, Ames: Iowa State Press, 575-606.Google Scholar
- Smyth GK: Linear models and empirical Bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol. 2004, 3: 3-Google Scholar
- Brinkhof B, Spee B, Rothuizen J, Penning LC: Development and evaluation of canine reference genes for accurate quantification of gene expression. Anal Biochem. 2006, 356: 36-43. 10.1016/j.ab.2006.06.001.PubMedView ArticleGoogle Scholar
- Etschmann B, Wilcken B, Stoevesand K, Von der Schulenburg A, Sterner-Kock A: Selection of reference genes for quantitative real-time PCR analysis in canine mammary tumors using the GeNorm algorithm. Vet Pathol. 2006, 43: 934-942. 10.1354/vp.43-6-934.PubMedView ArticleGoogle Scholar
- Schmittgen TD, Livak KJ: Analyzing real-time PCR data by the comparative Ct method. Nature Prot. 2008, 3: 1101-1108. 10.1038/nprot.2008.73.View ArticleGoogle Scholar
- Manuali E, Giuseppe De A, Feliziani F, Forti K, Casciari C, Marchesi MC, Pacifico E, Pawłowski KM, Majchrzak K, Król M: CA15-3 cell lines and tissue expression in canine mammary cancer and the correlation between serum levels and tumour histological grade. BMC Vet Res. 2012, 8: 86-10.1186/1746-6148-8-86.PubMedPubMed CentralView ArticleGoogle Scholar
- Sadusky TJ, Kemp TJ, Simon M, Carrey N, Coulton GR: Identification of Serhl, a new member of the serine hydrolase family induced by passive stretch of skeletal muscle in vivo. Genomics. 2001, 73: 38-49. 10.1006/geno.2000.6483.PubMedView ArticleGoogle Scholar
- Sorenmo KU, Kristiansen VM, Cofone MA, Shofer FS, Breen AM, Langeland M, Mongil CM, Grondahl AM, Teige J, Goldschmidt MH: Canine mammary gland tumours; a histological continuum from benign to malignant; clinical and histopatological evidence. Vet Comp Oncol. 2009, 7: 162-172. 10.1111/j.1476-5829.2009.00184.x.PubMedView ArticleGoogle Scholar
- Payen E, Verkerk T, Michalovich D, Dreyer SD, Winterpacht A, Lee B, Zeeuw De CI, Grosveld F, Galjart N: The centromeric/nuclear chromatin protein ZFP-37 may function to specify neuronal nuclear domains. J Biol Chem. 1998, 10: 9099-9109.View ArticleGoogle Scholar
- Dreyer SD, Zhou L, Machado MA, Horton WA, Zabel B, Winterpacht A, Lee B: Cloning, characterization, and chromosomal assignment of the human ortholog of murine Zfp-37, a candidate gene for Nager syndrome. Mamm Genome. 1998, 9: 458-462. 10.1007/s003359900796.PubMedView ArticleGoogle Scholar
- Baba A, Catoi C: Mammary gland tumors. Comparative Oncology. 2007, Bucharest: The Publishing House of the Romanian Academy, Chapter 11Google Scholar
- Chew A, Sirugo G, Alsobrook JP, Isaya G: Functional and genomic analysis of the human mitochondrial intermediate peptidase, a putative protein partner of frataxin. Genomics. 2000, 15: 104-112.View ArticleGoogle Scholar
- Guccini I, Serio D, Condo I, Rufini A, Tomassini B, Mangiola A, Maira G, Anile C, Fina D, Pallone F, Mongiardi MP, Levi A, Ventura N, Testi R, Malisan F: Frataxin participates to the hypoxia-induced response in tumors. Cell Death Dis. 2011, 2: e123-10.1038/cddis.2011.5.PubMedPubMed CentralView ArticleGoogle Scholar
- Kidd A, Coleman R, Whiteford M, Barron LH, Simpson SA, Haites NE: Breast cancer in two sisters with Friedreich`s ataxia. Eur J Surg Oncol. 2001, 27: 512-514. 10.1053/ejso.2000.1093.PubMedView ArticleGoogle Scholar
- Binder C, Hagemann T, Husen B, Schulz M, Einspanier A: Relaxin enhances in vitro invasiveness of breast cancer cell lines by up-reulation of matrix metalloproteases. Mol Human Rep. 2002, 8: 789-796. 10.1093/molehr/8.9.789.View ArticleGoogle Scholar
- Zhao L, Roche PJ, Gunnersen JM, Hammond VE, Tregear GW, Wintour EM, Beck F: Mice without a functional relaxin gene are unable to deliver milk to their puppies. Endocrinology. 1999, 140: 445-453. 10.1210/en.140.1.445.PubMedGoogle Scholar
- Zhao L, Samuel CS, Tregear GW, Beck F, Wintour EM: Collagen studies in late pregnant relaxin null mice. Biol Reprod. 2000, 63: 697-703. 10.1095/biolreprod63.3.697.PubMedView ArticleGoogle Scholar
- Simon D, Schoenrock D, Baumgartner W, Nolte I: Postoperative adjuvant treatment of invasive malignant mammary gland tumors in dogs with doxorubicin and docetaxel. J Vet Intern Med. 2006, 20: 1184-1190.PubMedGoogle Scholar
- Adamsky K, Arnold K, Sabanay H, Peles E: Junctional protein MAGI-3 interacts with receptor tyrosine phosphatase beta (RPTP beta) and tyrosine-phosphorylated proteins. J Cell Sci. 2003, 116: 1279-1289. 10.1242/jcs.00302.PubMedView ArticleGoogle Scholar
- Fang X, Gaudette D, Furui T, Mao M, Estrella V, Eder A, Pustilnik T, Sasagawa T, Lapushin R, Yu S, Jaffe RB, Wiener JR, Erickson JR, Millis GB: Lysophospholipid growth factors in the initiation, progression, metastases, and management of ovarian cancer. Ann N Y Acad Sci. 2000, 905: 188-208.PubMedView ArticleGoogle Scholar
- Peter C, Waibel M, Keppeler H, Lehmann R, Xu G, Halama A, Adamski J, Schulze-Osthoff K, Wesselborg S, Lauber K: Release of lysophospholipid ‘find me’ signals during apoptosis requires the ATP-binding cassette transporter A1. Autoimmunity. 2012, 45: 568-573. 10.3109/08916934.2012.719947.PubMedView ArticleGoogle Scholar
- Król M, Pawłowski KM, Majchrzak K, Gajewska M, Majewska A, Motyl T: Global gene expression profiles of canine macrophages and canine mammary cancer cells grown as a co-culture in vitro. BMC Vet Res. 2012, 8: 16-10.1186/1746-6148-8-16.PubMedPubMed CentralView ArticleGoogle Scholar
- Pawłowski KM, Homa A, Bulkowska M, Majchrzak K, Motyl T, Król M: Expression of inflammation-mediatted cluster of genes as a new marker of canine mammary malignancy. Vet Res Commun. 2013, 37: 123-131. 10.1007/s11259-013-9554-1.PubMedPubMed CentralView ArticleGoogle Scholar
- Lin S, Wang D, Iyer S, Ghaleb AM, Shim H, Yang VW, Chun J, Yun CC: LPA2 promotes tumor formation in an experimental model of colitis-associated colon cancer. Gasteroenterology. 2009, 136: 1711-1720. 10.1053/j.gastro.2009.01.002.View ArticleGoogle Scholar
- Mills GB, Moolenaar WH: The emerging role of lysophosphatidic acid in cancer. Nat Rev Cancer. 2003, 8: 582-591.View ArticleGoogle Scholar
- Yun CC, Sun H, Wang W, Rusovici R, Castleberry A, Hall R, Shim H: The LPA2 receptor mediates mitogenic signals in human colon cancer cells. Am J Physiol-Cell. 2005, 289: C2-C11. 10.1152/ajpcell.00610.2004.View ArticleGoogle Scholar
- Klose P, Weise C, Bondzio A, Multhaup G, Einspanier R, Gruber AD, Klopfleisch R: Is there a malignant progression associated with a linear change in protein expression levels from normal canine mammary gland to metastatic mammary tumors?. J Proteome Res. 2011, 10 (10): 4405-4415. 10.1021/pr200112q.PubMedView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.