Introduction
Breast cancer is the leading cause of cancer-related death among women worldwide [
1]. To this regard, several factors are involved in the initiation and promotion of breast tumors including molecular alterations at the genomic level such as mutations or copy number alterations [
2,
3]. Indeed, using functional studies, some of these genomic modifications have been clearly associated with a malignant phenotype, contributing to the oncogenesis of epithelial cells [
4,
5]. In addition to these molecular alterations, cancer cells rely on the surrounding microenvironment, where non-transformed cells and stromal components facilitate tumor growth by the secretion of autocrine signals like growth factors [
6]. Stimulation of cancer cells by paracrine-secreted factors from interstitial cells including fibroblasts, neutrophils, or endothelial cells can stimulate functions such as proliferation, survival, or migration, which are necessary to the tumor formation and dissemination [
7,
8]. However, components of the tumor stroma depend on different conditions and can differ among individuals. Of note, adipose tissue is one of the main components of the breast cancer microenvironment, and therefore, accumulation of fat tissue in the stroma can modify settings of tumor cells and influence their survival [
8]. As an example, increased presence of insulin or insulin-like growth factors can affect tumor growth but also response to treatment [
9]. In this context, breast tumors that express estrogen receptors are more dependent on stimulating factors [
10].
Besides being a risk factor for cancer, obesity has also been associated with detrimental patient outcome, especially in postmenopausal patients [
5,
4]. A number of epidemiological studies have demonstrated that how obesity is directly related to cancer mortality. In this sense, an increased body mass index (BMI) has been strongly linked with poor survival in postmenopausic patients carrying estrogen receptor positive tumors [
11]. One of the mechanisms proposed to explain how obesity increases breast cancer risk is that adipocyte-secreted hormones could be promoting tumor progression through an increase of cellular proliferation [
12]. However, little effort has been put into clarifying how the excess of adipose tissue in the tumor niche influences the molecular characteristics of the residing malignant cells.
In the present article, we aimed to evaluate biological functions that differentiate breast cancer tumors from obese and non-obese patients. To do so, we performed transcriptomic followed by protein–protein interaction network analyses to recognize relevant biological functions with druggable implications.
Discussion
In the present article, we describe biological functions and PPI networks associated with obesity in luminal A tumors, which we found generally associated with worse outcome in luminal A patients. Moreover, we uncover a druggable PPI network on luminal A obese patients which could be of utility to design potential therapeutic strategies.
Although we initially investigated all four subtypes of breast cancer, we only found significant differences between normal-weight and obese patients for the luminal A subtype.
Although this might be due to the fact that luminal A group was the most abundant, the relevant association with outcome and the strong PPI network suggests a relevant biological role. Moreover, the combined analysis of genes within each function was associated with poor outcome in luminal A patients. Of note, the subgroup analysis did not correlate with patient outcome in the other three molecular subtypes (luminal B, HER2, and basal-like). This supports the idea that deregulation of these functions has a specific role in luminal A breast cancers. Our results are in line with those obtained in another study using the same transcriptomic database (GSE 78958) [
18], although in this study, PPI analysis and druggable target opportunities were not explored. In addition, our study also identified a signature that is specific for luminal A tumors and explored the relation of this signature with patient outcome.
Although the level of deregulation between obese and non-obese patients was not highly elevated, we found significant differences using a fold changed of 1.4. We are aware that the fold change used is smaller than the one used in other studies [
19,
20]. However, this finding could be explained by the fact that obese patients included in our study lack the exact BMI information. It could be expected that the exclusion of the obese-I subtype might have been of utility to increase the difference among our study groups. In any case, even the number of deregulated genes and the level was not high, the upregulation of these transcripts was associated with an important detrimental outcome.
The most frequent functions identified in the overexpressed genes and linked with worse outcome included: cell cycle, cell differentiation, cell proliferation, and cellular response to EC stimuli. Control of cell cycle was the most frequent pathway with genes involved in the formation of the mitotic spindle and centrosome or microtubules formation like
BUB1, NUSAP, CENPF, CEP55, or the
KIF family [
21]. In addition, other genes were associated with the regulation and control of the cell cycle such as
CDK, GTSE1, CDC25C, or
CCNB [
22]. Finally,
FOXM1, a transcription factor linked with the presence of a Luminal phenotype, was found to be upregulated [
23]. Of note, our study also uncovered some interesting downregulated genes in the obese group. Notably, while
EGFR is overexpressed in around 50% TNBC and inflammatory breast cancers [
24], we found that this gene, as well as its ligand
AREG, was downregulated in the luminal A obese group. Wnt signaling has been implicated in carcinogenesis as well as in obesity promotion [
25,
26]. In this line, luminal A obese group also showed a lower expression of Wnt pathway inhibitors, such as
WIF1, BICC1, and the secreted proteins
SFRP1 and
SFRP4, together with a downregulation of the negative regulators of MYC,
TFAP2B, or
TCF7L [
27,
28].
Protein–protein interaction (PPI) networks offer information of how different proteins cooperate with others to trigger biological processes within the cell [
15,
29]. In this context, we have constructed PPI networks for the deregulated genes in luminal A obese patients. While proteins coded by deregulated genes poorly interact, we have found that exist a solid-clustering unit within the overexpressed PPI network. Remarkably, this dense cluster was comprised by proteins specifically coded by overexpressed genes that were associated with detrimental patient outcome. Thus, interference of one of its components might have an impact on several nodes, which could in turn lead to the destabilization of the network. This could open the window to new therapeutic strategies targeting this overexpressed PPI network in luminal A obese patients.
Next, we decided to search for potential drug targets within the PPI networks, linked with poor prognosis. Using Drug Interaction Database [
30], we first identified eight druggable genes: BUB1, TOP2A, BIRC5, KIF11, NEK2, RRM2, TYMS, and PBK. Then, expanding the search to other drug databases, we added eight more druggable candidates: CCNB1, CDK1, FOXM1, KIF4A, KIF20A, MELK, NEK2, and CDC25C. The PPI network built with them exhibited a high degree of interactions and, as indicated by its high clustering coefficient, might act as a cohesive functional unit.
Mitotic-related targets in this druggable network are the aim of new chemical entities with potential for preclinical or clinical translation development [
31]. For instance, a well-described target is TOP2A, where doxorubicin-like chemotherapies inhibit their effect [
32,
33]. Similarly, strategies to target BUB1 are under preclinical evaluation as this kinase has been described as associated with detrimental prognosis in breast and ovarian cancer [
20,
34]. Compounds against KIF11 are under development [
35] and some in clinical development [
36].
BIRC5 codifies for a protein which is vital for the growth and survival of cancer cells. Survivin is found to be essential for several functions linked with oncogenic transformation [
37]. It is known that normal tissues do not express survivin, and high expression in tumors is an indicative of poor prognosis and intrinsic resistance to radio- and chemotherapy [
38]. As obese patients express high levels of BIRC5, evaluation of BIRC5 inhibitors, alone or in combination, could be a potential option.
Finally, NEK2 codifies for a serine–threonine kinase with a key role in mitosis that has been found to be aberrantly overexpressed in several cancer types, among them breast cancer [
39,
40]. NEK2 expression levels are associated with tumor progression and detrimental outcome, as well as with drug resistance [
41]. Besides, preclinical studies have shown that high NEK2 levels can induce tumorigenesis by mediating chromosome instability and aneuploidy, while its downregulation can lead to cancer cells death [
42], suggesting a role for NEK2 as a potential target to treat cancer. Its elevated levels in obese patients points at NEK2 as a good candidate for targeted therapy in these patients. However, although several inhibitors have been developed for NEK2, some of them being highly specific and showing an irreversible inhibition, they have not been taken to clinical evaluation yet [
43].
Notably, NEK2 as well as BIRC5 and TOP2A were amplified in more than 12, 6, and 5% of breast cancers, respectively, reinforcing their potential role as key therapeutic targets.
In conclusion, in the present work, we describe functional pathways and protein–protein interacting networks associated with clinical outcome in luminal A tumors from obese patients. Moreover, we identify a druggable interacting map with potential for target inhibition. Although we acknowledge that this is an in silico analyses, and data should be confirmed in samples from patients, our results open new venues for further characterization and have potential for translation into the clinical setting.