The discovery also contributes to a better understanding of the molecular mechanisms of pathological angiogenesis, the aberrant proliferation of blood vessels that occurs during cancer and other diseases.
The research, published in the journal PLOS Genetics, combined a study of the genes involved in retinopathy, as a model of angiogenesis, with analysis of transcriptomic gene expression profiles from public breast cancer databases.
Conducted by researchers at the University of São Paulo’s Chemistry Institute (IQ-USP), in collaboration with the Ontario Institute for Cancer Research (OICR) in Toronto, Canada, the study was supported São Paulo Research Foundation—FAPESP.
“We identified a set of genes whose expression in breast cancer correlates with the degree of pathological angiogenesis in the tumor, so that it serves as a genetic signature of angiogenesis that is prognostic and more robust than the signatures identified previously, given the correlation found between angiogenesis and tumors generally,” said Ricardo Giordano, a professor at IQ-USP, head of its Vascular Biology Laboratory and a co-author of the study.
In the study the Brazilian researchers identified 153 altered genes in both healthy and diseased retinas in mice.
From this list they identified 149 equivalent human genes. The result served as the basis for a genetic signature study in partnership with the Canadian team, using a database with information on breast cancer patients.
The conclusion was that 11 key genes involved in pathological angiogenesis performed best in terms of prognostic value.
Pathological angiogenesis is common to breast cancer and retinopathy. “The fact that these two diseases share this process and that angiogenesis is fundamental to the development of cancer in general led us to try to build a bridge between retinopathy and breast cancer,” said João Carlos Setubal, head of USP’s Bioinformatics Laboratory and also a co-author of the article.
According to the researchers, the study focused on breast cancer because of the large amount of data available on the disease.
“We had to have access to a vast quantity of public data, given the considerable variation between one patient and another.
This is the case for breast cancer. Genetic profiles are available for some 2,000 patients,” Setubal said.
Bioinformatics was crucial to finding the genetic signature, he added. The data generated in the laboratory was submitted to sophisticated computational processing in partnership with researchers at OICR in Canada.
Another co-author of the study, Rodrigo Guarischi de Sousa, then a Ph.D. student, wrote the program that tested the 149 human genes as possible components of the signature for breast cancer.
In this part of the study he was supported by a scholarship from FAPESP for a research internship abroad supervised by Paul C. Boutros at OICR.
Boutros is currently a researcher in the Human Genetics Department at the University of California, Los Angeles (UCLA). Guarischi de Sousa now works at DASA, Brazil’s largest medical diagnostics company.
The researchers plan to find applications for this signature, especially in treatment for breast cancer.
“Our next goal is to continue studying angiogenesis in cancer,” Giordano said. “We’re interested in identifying genes on this list that can be targets for the development of new drugs or new applications of existing drugs.”
From retinopathy to breast cancer
The discovery of the prognostic gene signature for breast cancer is the result of a long study supported by a research scholarship from the National Council for Scientific and Technological Development (CNPq) starting in 2010, when Giordano began studying the transcriptome (RNA expression) and proteome (protein expression) of pathological retinal angiogenesis.
“As a result of this study, our lab implemented a mouse model to research retinopathy. The murine model is important as it’s difficult to study blood vessel formation outside living tissue. This model enables us to induce retinopathy by modulating oxygen levels and study angiogenesis in the lab,” Giordano said.
The researchers took blood samples from mice to investigate the differences between physiological angiogenesis, which occurs in healthy individuals (in wound healing, ovulation and placental growth, for example) and pathological angiogenesis, which is part of disease (e.g. cancer or arthritis).
“We observed which genes were expressed by endothelial cells [the inner lining of blood vessels] in both kinds of angiogenesis, always looking for the genes that were more expressed in one kind than the other,” Giordano said.
A key element of the experiment was oxygen variation in the chambers containing newborn mouse pups.
In chambers with oxygen at 75 percent, the mice became retinopathic, whereas in ambient air (oxygen at 20 percent) the retina developed normally.
The relationship between oxygen and cells has been in the news lately.
William Kaelin, Peter Ratcliffe and Gregg Semenza won the 2019 Nobel Prize in Physiology or Medicine for discovering how cells sense and adapt to changing oxygen availability.
Gene expression-based prognostic signature
It is important to stress that clinical applications of the genetic signature for angiogenesis may differ from applications deriving from the marker gene BRCA1.
The BRCA1 gene mutation became world-famous in 2013 when US actor Angelina Jolie underwent a preventive double mastectomy after genetic testing showed that she carried the mutation and hence ran an 87 percent risk of developing breast cancer.
“BRCA1 is a genomic gene,” Setubal said. “A woman with mutations in this gene faces a higher risk of developing breast cancer but won’t necessarily do so.
The presence of mutations in this gene serves to help predict the appearance of the disease. The signature we describe in our study proved promising to predict the development of breast cancer after it actually appears.”
Breast cancer is the most common cancer among women worldwide, representing nearly a quarter (25%) of all cancers with an estimated 2.1 million new cancer cases diagnosed in 20181. Over the past two decades, there has been a rapid increase in breast cancer incidence throughout Asia, mainly South-Eastern Asia, including India2–6.
Breast cancer is the most common cancer among Indian women in a majority of urban cancer registries at Delhi, Mumbai, Bangalore, Thiruvananthapuram (AAR ranges between 33-41/100000 women) and has rapidly overtaken cervical cancer7.
In India, although age-adjusted incidence rate of breast cancer is lower (25.8 per 100 000) than the United States of America (93 per 100 000), age-wise distribution of incidence shows a higher percentage (46.7%) of breast cancer incidence among women below the age of 50 years compared to United States of America (19%)8.
An incidence rate of 45.5% has been observed in Asian countries for this age group, suggesting a higher incidence of breast cancer in the younger age group in India and other Asian countries as compared to the western population8.
To our knowledge, there is a single report describing gene expression profiles of breast cancer from Indian patients, focusing mainly on estrogen receptor (ER) positive and ER-negative tumours profiles alone11.
In the present study, we have analysed the gene signatures and molecular pathways involved in breast carcinogenesis in Indian women by transcriptome profiling.
Breast cancer incidence is increasing globally (24.2%) as well as in India (15.46%) and has become the most common cancer among Indian women7,8. To gain insight into the molecular mechanisms involved in the pathogenesis of breast cancer in Indian women, we have carried out gene expression profiling, wherein we have analysed the gene expression profiles associated with breast tumours and those associated with age and tumour stage.
So far, there has been a single report where gene expression profiles of breast tumours were studied in Indian women, however, the authors have analysed gene expression profiles of ER positive and negative tumours, where they have found four hormone-responsive genes as DEGs11. To our knowledge, this is the first study describing comprehensive gene expression profiles in Indian women and demonstrating the existence of molecular subtypes using gene expression profiles in breast tumours.
The present study identified 2413 differentially expressed genes comprising of top up-regulated genes such as COL10A1, COL11A1, MMP1, MMP13, MMP11, GJB2, CST1, KIAA1199, CEACAM6, and BUB1; top down-regulated genes comprising of PLIN1, FABP4, LIPE, AQP7, LEP, ADH1A, ADH1B, CIDEC, THRSP, GPD1, TIMP4, and KIAA1881 (Supplementary Fig. S2, Supplementary Table S2). Among the top DEGs, up-regulated expression of genes such as COL10A1, MMP1, MMP11, and BUB1; down-regulated expression of genes such as ADH1B, CIDEC, FABP4, AQP7, RBP4, CDO1, FIGF, and LPL were reported to be differentially expressed in breast and or other cancers by various authors using microarray profiling in western population26–38, showing concordance with the present study. In the present study, we found up-regulation of cell cycle genes such as BUB1, CCNA2, CCNB2, and CDC2; up-regulated expression of BUB1, CCNA2, CCNB2 and CDC2 has also been reported in breast39,40 and several other cancers41–46 using microarray and were found to be associated with poor prognosis of the disease44,47–49.
Overexpression of the above cell cycle genes may be contributing to the uncontrolled proliferation of the tumour cells and hence may serve as biomarkers and targets for therapy.
Overexpression of genes involved in DNA replication such as MCM2, MCM6, MCM10, and RAD21 was observed in the present study, increased expression of MCM2, MCM6, and MCM10 have been reported in breast and several other epithelial malignancies by transcriptome profiling, and was associated with poor prognosis50–52.
Furthermore, the focal adhesion genes such as COL1A1, COL10A1, and COL11A1 were also found to be up-regulated in the present study and are also reported to be up-regulated in various cancers including breast tumours37,53–55. In cancer cells, collagen gene expression is known to increase drug resistance by inhibiting drug penetration as well as cause an increase in apoptosis resistance, thus, in turn, promoting tumour progression37,56–58.
In the present study genes such as PLIN1, FABP4, LIPE, LEP, CIDEC, THRSP, AQP7, ADH1A, ADH1B, GPD1, and TIMP4 were found to be down-regulated, involved mainly in lipid metabolism, lipolysis, oxidoreductase activity, and PPAR pathways. In concordance with the above findings, down-regulated expression of lipid metabolism genes such as LEP, CIDEC, THRSP, PLIN1, GPD1, and FABP4 genes were also reported at the transcript level by various authors in breast54,59–64 and other cancers such as gastric65, hepatocellular66 and keratoacanthomas67.
Similar to that observed in the present study, down-regulated expression of aquaporin, AQP7 gene belonging to water channel family, TIMP4 belonging to mettalloproteinases inhibitor family member was also reported in breast and hepatocellular carcinoma at transcript level68–70.
Contrary to the down-regulation observed in the present study, up-regulation of FABP4, LEP, CIDEC genes was reported at transcript and or protein level35,71–74 in lung, thyroid, colorectal, and tongue squamous cell carcinoma; these differences may be attributed to tissue-specific differences in gene expression, differences due to the techniques employed in the studies, which need to be established by experimental validation.
Networking analysis was done to identify genes involved in regulation of gene expression in cancer cells, AURKB, CENPA, TOP2A, BUB1, CCNB2, MMP1, and SPP1 were identified as top hub genes from the up-regulated genes; suggesting these genes might be playing key regulatory role in breast carcinogenesis through deregulation of cell cycle and in invasion/metastasis. Similarly, genes such as CAV1, ACACB, NTRK2, KLF4, and MYH11 were the key down-regulated hub genes suggesting a possible role of their decreased expression in breast carcinogenesis.
Comparison of gene expression profiles of Indian patients with that of western patients led to the identification of 558 genes specifically found to be deregulated in Indian patients, suggesting some differences in the gene sets between these populations. The differences in DEGs among the two populations may be partly due to differences in platforms, experimental procedures or genetic makeup of the populations. Up-regulated expression of COL10A1, MMP11, CST1, GJB2, MMP1, MMP13, and CEACAM6; down-regulated expression of ADH1B, CIDEC, THRSP, GPD1, TIMP4, FABP4, and SCARA5 genes was common in breast cancers of the two populations.
The similarity in the DEGs between the Indian and western patients suggests a similarity in the molecular events associated with breast carcinogenesis. Further, we compared DEGs obtained in the present study with that of Lebanese population54, where several genes were found to be common between the two populations. Up-regulated expression of COL11A1, GJB2, MMP13, EPYC, CEP55, and MELK and down-regulated expression of PLIN, TIMP4, LEP, LYVE1, SDPR, FIGF, and LPL was observed in common with the Lebanese population from the top 50 genes found in the present study.
This is pointing towards a possible existence of greater similarity in the molecular pathogenesis of breast cancers amongst the Asian population compared to the western population.
Comparison of expression profiles of ET (≤40 years) and LT (≥55 years) yielded few genes that are unique between the two groups, 7 genes B4GALNT1, S100P, KLK4, HIST3H2A, DRD4, PCSK1N, and BAPX1 were significantly overexpressed in early-onset tumours compared to late-onset tumours. Overexpression B4GALNT1 causes deregulation of glycosphingolipid biosynthesis and is reported to be up-regulated in breast cancer stem cells75 at the transcript level, similarly, S100P76, KLK477,78, DRD479, and BAPX180 genes were also reported to be up-regulated in breast and other cancers at mRNA level.
Several of these genes are known to induce invasion and metastasis81–84, breast cancer in young patients is known to be aggressive85–88, the overexpression of these genes may be thus contributing to the aggressive behavior of the early-onset cancers.
Anders et al.89–91 have analysed gene expression profiles between early-onset patients and late-onset patients, where 693 DEGs were found initially, later the significance was lost when they have corrected the gene differences for subtypes and for ER and histological grades. However, such segregated analysis could not be carried out in the present study due to the small sample size.
Further, we compared gene expression profiles between lower and advanced stages of tumours; we identified 200 genes uniquely deregulated in advanced stage cancers, involving pathways such as cell adhesion (VCAN), ECM receptor interaction (COL1A2, COL3A1, ITGA11, and TNN) and pathways in cancer (JUN, and MYC) which may be contributing to increased proliferation, migration and increased angiogenesis in advanced stage of tumours.
Molecular subtyping and hierarchical clustering of gene expression profiles of breast tumours using PAM50 molecular signatures, yielded distinct clusters corresponding to each molecular subtype, showing the existence of molecular subtypes in these patients. Among these patients, 44% tumours falling into luminal subtype, 31% into basal subtype and 20% into HER2/neu overexpressing subtype, and 3% into normal-like subtype, which is more or less similar to that reported in the western population28,29.
To our knowledge this is the first study acknowledging the existence of molecular subtypes from the Indian subcontinent based on gene expression profiles; earlier, Kumar et al.92 reported the existence of molecular subtypes based on the expression of ER, PR, HER2/neu and cytokeratins at protien level, however, transcriptome-based subtyping has not been demonstrated so far.
Metallopeptidases genes were one of the functional class of genes found to have deregulated expression in breast cancers in the present study and hence we validated some of the genes by qPCR. QPCR analysis confirmed the up-regulated expression of MMP1, MMP13, and MMP11 genes and down-regulated expression of ADAMTS1 and ADAMTS5 genes in breast cancers observed in the microarray experiments. Overexpression of MMP1, MMP11, and MMP13 was also reported in breast95 and several other cancers such as gastric, oral, colorectal, oesophageal and nasopharyngeal at the transcript and or protein level30,54,96–103.
Upregulated expression of MMPs has been reported in cancer, vascular diseases and many different types of inflammatory diseases104, their overexpression results in increased invasion and metastasis in cancer cells105.
A disintegrin and metalloproteinase with thrombospondin motif (ADAMTS), superfamily genes play, an important role in ECM assembly and degradation, several of them act as tumour or metastasis suppressors by influencing cell proliferation, migration, apoptosis, and angiogenesis106.
In the present study, ADAMTS1 and ADAMTS5 genes were observed to be underexpressed, down-regulated expression and antitumour acitivity of these genes were reported in breast cancers107 and gastric carcinoma carcinoma108 respectively at both transcript and protein.
An association between overexpression of MMP1 transcripts, with loss of ER (p = 0.01), and PR (p = 0.006) was found in the present study, Nakopoulou et al.109 has also found a similar inverse association with PR expression at the protein level, supporting findings of the present study. Further, overexpression of MMP13 was found to be associated with overexpression of HER2/neu in patients (p = 0.023), Zhang et al.110 have also reported similar association in breast cancer with protein level expression. Interestingly, down-regulation of ADAMTS5 was found to be associated with late-onset tumours (≥55 years) compared to ET (≤40 years), (FC = −6.5 in LT and FC = −4.5 in ET, p = 0.013), suggesting the involvement of loss of this gene in the molecular pathogenesis with late-onset breast cancer. To our knowledge, the down-regulated expression of ADAMTS5 in breast tumours and its association with late-onset breast cancer (old age of patient) is reported for the first time in the present study.
Together, deregulated expression of these matrix remodeling factors in breast tumours may be contributing to the degradation of ECM and invasion and metastasis in breast tumours, suggesting a pivotal role played by these genes in breast tumorigenesis. However, up-regulation of MMP3 and MMP14 genes didn’t reach statistical significance, unlike found in microarray data. The discrepancy between microarray and qPCR data could be due to a different number of samples analysed by each method, to some extent, tumour heterogeneity might have also contributed to such differences.
The present study describes comprehensive gene expression profiles of breast tumours from Indian women and the presence of molecular subtypes in this population. Genes involved in cell cycle, ECM, metastasis were some of the essential pathways found to be up-regulated, on the other hand genes involved in lipid metabolism, PPAR were some of the pathways that were found down-regulated. Genes belonging to cell adhesion, cell cycle, ECM receptor interaction pathways were deregulated in early-onset breast cancers.
This study confirmed the presence of molecular subtypes in breast tumours based on gene expression profiles, for the first time from Indian patients. Comparison with western data has revealed the presence of several deregulated genes that are common between Indian and western patients suggesting a similarity in the molecular mechanisms; however, a higher similarity was with that of the Asian population.
Comparison of gene expression profiles in early- and late-onset tumours showed several common DEGs between the groups, but with differences in fold change of their gene expression. Further, significant down-regulation of ADAMTS5 in old age patients had been reported for the first time in breast cancer patients.
More information: Rodrigo Guarischi-Sousa et al, A transcriptome-based signature of pathological angiogenesis predicts breast cancer patient survival, PLOS Genetics (2019). DOI: 10.1371/journal.pgen.1008482