EXPOsOMICs: Meet-in-the-Middle and Network Perturbation

Abstract

Systems biology has been driven by technology (the development of omics) and by statistical modelling and bioinformatics. We aim to bring biological thinking back. We suggest that three traditions of thought need to be considered: (a) causality in epidemiology, for example the “sufficient-component-cause framework”, and causality in other sciences, for example the Salmon and Dowe approach; (b) new acquisitions about disease pathogenesis, for example the “branched evolution model” in cancer, and the role of biomarkers in this process; (c) the burgeoning of omic research, with a large number of “signals” that need to be interpreted. To address the new challenges of epidemiology, the concept of the “exposome” has been proposed. We show examples from recent projects in the field, namely, new omic approaches applied to epidemiological studies; and in particular, the identification of hallmarks of cancer as intermediate steps between exposure to carcinogens and the cancer phenotype, according to the “meet-in-the-middle” concept. We use examples derived from the study of mutational spectra in tumours and benzo(a)pyrene and bisphenol A as model carcinogens. We suggest conceptualising the detection and tracing of signals in terms of information transmission.

Appendix: Exposomics: Meet-in-the-Middle and Network Perturbation

Evidence of Temporal Sequence of Hallmarks of Cancer

Colorectal Cancer

Colorectal cancer is the most frequently examined tumour with respect to its genetic alterations and branched evolution.

In 95 colorectal cancer samples, using the Hidden Conjunctive Bayesian Networks (H-CNB) model, researchers identified APC mutations as initiating events (accumulation rate (AR) 0.39/year—consistent for an early driving role), followed by KRAS (AR = 0.12/year), PIK3CA (AR = 0.009/year), and other mutations such as EVC2, FBXW7, EPHA3 and TCF7L2 (Gerstung et al. 2011). In parallel, TP53 (AR = 0.06/year) mutations could be placed either before or after APC and KRAS mutations since their presence was independent of the presence of APC and KRAS mutations (Gerstung et al. 2011).

The model places key signalling pathways in the following order: Small GTPase pathway, Apoptosis/Wnt-Notch signalling, and homophilic cell adhesion. These alterations are often found before alterations in KRAS and TGF-β signalling pathways and before alterations in G1/S phase control. DNA damage control and JNK are altered at later stages, probably through TP53 mutations. Integrin signalling and invasion are not as common, suggesting roles at later stages of colon carcinogenesis.

Translating the pathways into hallmarks, the evidence above places evasion of growth suppressors, cell proliferation and resisting cell death prior to avoiding immune destruction and invasion and metastasis. Angiogenesis occurs simultaneously with initiation of invasion, whereas immortality occurs in parallel with the hallmarks prior to invasion and metastasis.

The early presence of the APC and KRAS mutations, as well as their high mutation accumulation rates (ARs: 0.39 and 0.12 per year Gerstung et al. 2011), is consistent with an early driving role.

These findings are in agreement with the Fearon and Vogelstein model of colorectal carcinogenesis (Fearon and Vogelstein 1990) where APC, a tumour suppressor gene, is firstly inactivated, allowing the normal epithelium to become hyperplastic and eventually form an early adenoma. Activation of KRAS then contributes to sustained proliferation and the formation of an intermediate adenoma, whereas further loss of tumour suppressors, such as Smad4, contributes to the formation of a late adenoma. Loss of p-53 then allows resistance to cell death and immortality. Additional genetic alterations, such as the accumulated loss of suppressor genes on additional chromosomes, correlate with the ability of the carcinomas to metastasise and cause death. However, Fearon and Vogelstein emphasise the importance of the accumulation of these changes, rather than their temporal order which may vary.

Box

Technical Definitions:

• Mutation Accumulation Rate (AR):

  Estimated yearly accumulation rate for a particular mutation. The higher the accumulation rate, the higher the frequency of occurrence of the mutation in different tumour samples

• Driver mutations:

  Mutations involved causally in the neoplastic process, conferring a selective advantage, thus being positively selected for during tumorigenesis.

• Passenger mutations provide no positive or negative selective advantage to the tumor but are retained by chance during repeated rounds of cell division and clonal expansion.

• Passenger Mutation Rates:

  Estimated through the quantification of synonymous (silent) missense mutations, because such mutations are expected to be biologically inert and can therefore exert no positive or negative selective advantage

• Passenger Probability Scores:

  These gene-specific scores are based on a Likelihood Ratio Test (LRT) for the null hypothesis that, for the gene under consideration, the mutation rates are the same as the passenger mutation rates. Small probability scores support rejection of the null hypothesis and acceptance that mutation rates are in fact higher than the passenger mutation rates. Therefore, small passenger probability scores support a “driver” role for mutations in a particular gene.

Rosenberg et al. (2009) also identified mutations in the APC and KRAS genes as an early event in rodent colon cancer models. Even in early premalignant lesions, increased proliferative activity, growth factor signalling and K-RAS mutations were evident.

In addition, Wood et al. (2007) in their analysis of 11 colorectal tumours identified APC, KRAS and TP53 as the most frequent driving mutations, followed by mutations in PIK3CA, FBXW7, CSMD3, TNN, NAV3, SMAD4 and many more. All these genes have passenger probability scores of <0.0001.

Independently, Beerenwinkel et al. (2007) also identify APC, TP53 and KRAS as the most often mutated genes in their analysis of 78 candidate cancer genes in 35 tumour (colon adenoma and carcinoma) samples. Therefore, evasion of growth suppressors, resisting cell death and sustained proliferation are again evidenced as the first cancer hallmarks to appear.

With respect to the temporal order of angiogenesis, studies corroborate its simultaneous occurrence with the initiation of invasion. Evidence from Takahashi et al. (2003) places the angiogenic switch between mucosal and submucosal invasive cancer. Hanahan and Folkman (1996) also place angiogenesis prior to solid tumour formation but following hyperproliferation. In addition, Zhang et al. (2001) evidence angiogenesis in early premalignant stages of tumour development. Furthermore, in benign colorectal adenomas, VEGF protein and RNA levels exceed those of normal colonic mucosa (Ono and Miki 2000; Wong et al. 1999). Lastly, the presence of VEGF A and B at the adenoma stage (Hanrahan et al. 2003) further supports the order of the angiogenic switch sometime after hyperproliferation and prior to invasion.

Lastly, even though APC, TP53 and KRAS mutations were identified at a high frequency in different sections from the primary tumour of a colorectal cancer patient (Kogita et al. 2015), only one section from this tumour contained PIK3CA mutations, at a low frequency. Interestingly, the metastatic tumour from the same patient harboured PIK3CA mutations at a high frequency. This evidence supports the temporal order of replicative immortality just prior to invasion and metastasis.

Pancreatic Cancer

Using the H-CNB to model 90 cases of pancreatic cancer, mutations in KRAS (AR > 100/year, prevalence = 100%) appear to initialise progression followed by TP53 (AR = 0.34/year), CDKN2A (AR = 0.013/year) and MLL3 (AR = 0.00066). SMAD4 is also mutated independently (AR = 0.015/year) (Gerstung et al. 2011).

Extrapolating to pathway level, apoptosis, G1/S transition, Hedgehog and TGF-β signalling pathways were ubiquitous, indicating their alteration at the earliest stages of pancreatic carcinogenesis. Then, alterations in small GTPase-dependent signalling and KRAS signalling arise independently, followed by alterations in DNA damage control, JNK and Wnt/Notch signalling. These alterations also contribute to replicative immortality. Integrin signalling is altered at late stages after homophilic cell adhesion, further contributing to senescence. Lastly, invasion pathways were found unaltered in these samples, suggesting roles in even later stages of carcinogenesis.

This model indicates that in pancreatic cancer, the hallmarks of insensitivity to anti-growth signals, cell proliferation, resisting cell death and angiogenesis are followed by deregulating cellular energetics, which in turn precedes immune system evasion, cell migration, invasion and metastasis. The model does not allow temporal placing of immortality.

Interestingly, quantitative analysis of the timing of the genetic evolution of pancreatic cancer, based on cell proliferation rates of normal pancreatic tissue and pancreatic metastatic tumours, suggested that at least a decade passes between the initial driving mutation and the birth of the parental, non-metastatic founder cell and at least five more years are needed for the acquisition of metastatic ability (Yachida et al. 2010). However, these numbers are mathematical estimates based on a small number of tumour samples; therefore, the evidence for the timing is weak.

Primary Glioblastoma

In 78 primary glioblastoma tumours, TP53 (AR = 0.015/year), PTEN (AR = 0.012/year), EGFR (AR = 0.0026/year), NF1 (AR = 0.0043/year), PI3CA (AR = 0.033/year), IDH1 (AR = 0.0087/year), PIK3R1 (AR = 0.0037) and RB1(AR = 0.011) mutations were identified as the most commonly mutated genes (Gerstung et al. 2011).

The H-CNB model identified TP53 as the first mutation, with NF1, PTEN and EGFR being mutated in parallel. Mutations then proceed in the following order: PIK3CA, PIK3R1 and RB1. However, the low accumulation rates of these individual mutations indicate a low probability for a specific gene alteration in primary glioblastomas.

This mutation order highlights apoptosis and small GTPase pathways as the first and most commonly affected pathways. G1/S phase transition, Wnt/Notch signalling and KRAS signalling were also mutated early. Alterations in these pathways are followed by alterations in DNA damage control, JNK signalling, homophilic cell adhesion and integrin signalling, the latter two occurring independently of the former two pathways. No alterations in the invasion pathway were identified.

Therefore, translating the pathways into hallmarks, in primary glioblastomas, cell proliferation and resistance to cell death are the first hallmarks identified, followed by deregulated cellular energetics, evasion of growth suppressors and angiogenesis. TP53 alterations are also likely to initiate the process of replicative immortality which is completed at later stages with PIK3CA and RB1 mutations. Invasion and metastasis is the last hallmark to occur, after avoiding immune destruction.

Independently, Parsons et al. (2008) investigated the frequency of mutations in 105 samples of glioblastoma tumours. They found CDKN2A, TP53, EGFR, PTEN, NF1, CDK4, RB1, IDH1, PIK3CA and PIK3R1 as the most frequently altered genes, in this order. With the exception of PIK3CA, and PIK3R1, all other gene alterations have a passenger probability of <0.01, indicating their role as driving mutations.

Based on these genes’ functions and assuming that the most prevalent/driving alterations are necessary for initiation of carcinogenesis, whereas less frequent alterations have later roles in the carcinogenesis cascade, then in glioblastoma tumours, cancer hallmarks are expected to appear in the following order: evasion of cell death, insensitivity to anti-growth signals, sustained cell proliferation, deregulated energetics (Warburg effect), immortality, and invasion and metastasis. Unfortunately, this list does not allow estimation of the temporal position of angiogenesis and evasion of immune destruction in glioblastoma tumours.

Renal Carcinoma

Gerlinger et al. (2012) studied intratumour branched evolution in 2 renal cell carcinomas by multi-region sequencing and identified that VHL, SETD2, PTEN and KDM5C underwent multiple, distinct and spatially separated inactivating mutations.

In the first patient, VHL was identified as the first driving mutation, followed by SETD2 mutations which break the phylogenetic tree in two branches. In one branch (core biopsy samples), heterogeneity is further propagated by KDM5C and mTOR mutations. In the second branch (metastases samples), SETD2 mutations are also followed by KDM5C mutations.

VHL, the gene identified as firstly mutated, is involved in cell division and formation of new blood vessels, among other functions, identifying cell proliferation and perhaps initiation of angiogenesis as early, widespread events in renal carcinogenesis. The SETD2 gene normally trimethylates histone 3 lysine 36 at sites of active transcription. Its mutation results in silencing of transcription instead of an active chromatin conformation. This leads to altered nucleosome dynamics and DNA replication stress, as well as to failure in loading lens epithelium-derived growth factor and the Rad51 homologous recombination repair factor at DNA breaks (Kanu et al. 2015). Therefore, growth factor evasion follows cellular proliferation. KMD5C, which is mutated downstream, is involved in regulation of transcription through transcriptional repression, which most likely contributes to maintaining genetic instability. Lastly, mTOR is a target for cycle arrest, and immunosuppressive effects, and when mutated it is involved in deregulated metabolism, evasion of anti-growth signals, genetic instability and replicative immortality (Carnero et al. 2015), suggesting that immune system evasion and replicative immortality are later events.

In the second patient, VHL and PBRM1 were mutated in all tumour specimens followed by SETD2 mutations which break the phylogenetic tree in two branches. In one branch, the metastatic branch, SEDT2 mutations are accompanied by P53 mutations. In the other branch, the core tumour branch, PTEN mutations lead to further branching.

PBRM1 is necessary for ligand-dependent transcriptional activation and is involved in chromatin organisation, whereas PTEN is a tumour suppressor enzyme which regulates cell division, apoptosis, cell movement, adhesion and angiogenesis. P53 is one of the most well-established tumour suppressors associated with all cancer hallmarks, namely, increased cancer metabolism, angiogenesis, genetic instability, immune evasion, resistance to cell death, replicative immortality, sustained proliferative signalling, invasion and metastasis (Nahta et al. 2015). This branching further confirms that cell proliferation, new blood vessel formation and transcriptional activation are widespread in tumours and occur early. They are followed by repression of anti-growth signals, through chromatin inactivation (SEDT2 mutations) and anti-growth factor inactivation (PTEN and P53 mutations). Lastly this branching suggests that other hallmarks such as replicative immortality, invasion and metastasis, as mediated by P53, follow at later stages of renal carcinogenesis.

Melanoma

Multi-region sequencing in 41 multiple melanoma biopsies from 8 individual tumours (Harbst et al. 2016) found that the three melanoma driver genes, namely, BRAF, NRAS and NF1, all key components of the MAPK pathway, were ubiquitously mutated in a mutually exclusive pattern. The ubiquitous expression of these genes is consistent with an early role of this pathway (involved in cell proliferation and insensitivity to anti-growth signals) in melanoma formation. On the contrary, mutations in the PI3K pathway were heterogeneous indicating that such mutations occur later in metastatic melanoma evolution. Since the PI3K is important for many cell activities, including cell growth and division (proliferation), movement (migration) of cells and cell survival, its appearance later in carcinogenesis suggests later appearance of the hallmarks of resistance to cell death, and invasion and metastasis. IDH1, which has an essential role in glucose metabolism, is also mutated heterogeneously in some melanoma tumours, suggesting later appearance of the Warburg effect.

Breast Carcinoma

In an analysis of 11 breast cancer tumours, Wood et al. (2007) identified TP53 and PIK3CA as the most frequent, driving mutations. The two genes have passenger probability scores of <0.0001, indicating their important driving roles in carcinogenesis. Assuming that the most prevalent/driving alterations are necessary for initiation of carcinogenesis, the high mutation frequency of these genes highlights the early appearance of the resisting cell death, insensitivity to anti-growth signals and cell proliferation cancer hallmarks.

Combined Evidence on Different Tumor Types

Looking at colon, pancreatic and glioblastoma tumours together, the H-CNB model (Gerstung et al. 2011) identified the following order of pathway alterations as the most likely: Apoptosis, TGF-β signalling, small GTPase-dependent signalling (other than KRAS), Wnt/Notch signalling, control of G1/S phase transition, KRAS signalling, Hedgehog signalling, DNA damage control, JNK, homophilic cell adhesion, integrin signalling and invasion.

Based on the above, the following order of hallmarks emerges: resistance to cell death, insensitivity to anti-growth signals, sustaining proliferative signalling, deregulated cellular energetics, inducing angiogenesis, avoiding immune destruction, enabling replicative immortality, and invasion and metastasis.

Evidence for Angiogenesis in Premalignant Lesions

Raica et al. (2009) reviewed angiogenesis in premalignant lesions and concluded that tumour angiogenesis is not necessarily a characteristic of invasive tumour but may occur prior to malignancy, as defined by invasion and metastasis. In their review, they gather evidence that microvessel density (MVD) was significantly increased in a relatively large spectrum of premalignant squamous cell lesions, such as in the oral mucosa, skin, uterine cervix, vulva and anal canal. Interestingly, for a number of these lesions, MVD was found to correlate with a major pro-angiogenic factor, namely, vascular endothelial growth factor (VEGF).

Premalignant lesions of glandular epithelia, including gastric metaplasia and dysplasia, atypical adenoma of the colon, atypical hyperplasia and carcinoma in situ of the breast also exhibited VEGF overexpression.

Evidence of Temporal Sequence of Key Characteristics for Selected Carcinogens

Benzo[a]pyrene and the Hallmarks of Cancer

Benzo[a]pyrene (BaP) is a ubiquitous contaminant, belonging to the large group of organic compounds with two or more fused aromatic (benzene) rings, namely, polycyclic aromatic hydrocarbons (PAHs) (IARC 2012; WHO 2010). PAHs and Bap formed during incomplete combustion and their major sources include tobacco smoke, residential and commercial heating with wood or coal, motor-vehicle exhaust and industrial emission (IARC 2012). Occupational exposures occur in aluminum production, roofing and paving involving coal-tar pitch, coal liquefaction, coal-tar distillation, wood impregnation, chimney sweeping and power plants (IARC 2010). While no epidemiological studies on BaP alone are available, BaP produced tumours in multiple organs and tissues in all animal species tested following exposures by many different routes (IARC 2012). Mechanistic evidences from in vitro and in vivo studies, including human exposures, showed that BaP is clearly genotoxic following its metabolization into highly reactive species that form DNA adducts leading to sister chromatid exchange, chromosomal aberrations, micronuclei or DNA damage (IARC 2012). Based on these multiple experimental evidences, BaP may be considered a well-established model carcinogen. Thus, we decided to review the literature available on BaP and its capacity to affect the different hallmarks of cancer.

Hallmark 1. Sustaining Proliferative Signalling

BaP is a very well-known chemical able to immortalise human cells, such as breast epithelial cells (Gudjonsson et al. 2004). One of the earliest mechanistic evidence of BaP effects on cell proliferation was its ability to form DNA adducts and recurrent mutations were observed in the genes of the RAS superfamily both in humans and in mice (Meng et al. 2010; Hu et al. 2003; Wei et al. 1999). Beyond the mutagenic effects on genes associated to cell proliferation, Kometani et al. (2009) showed that BaP was able to promote proliferation of human lung cancer cells after 24 weeks of exposure by activating the EGFR pathway through induction of EGFR ligands, amphiregulin and epiregulin. Moreover, BaP may promote cell proliferation through the activation of nuclear receptor signalling pathways, such as aryl hydrocarbon receptor (AhR) and estrogen receptor (ER) in humans, mice and other non-mammalian organisms (Andrysik et al. 2007; Charles et al. 2000; Tian et al. 2013; Wen and Pan 2015). In conclusion, BaP is able to induce cell proliferation through both genotoxic and non-genotoxic receptor-induced mechanisms in humans and various experimental models.

Hallmark 2. Evading Growth Suppressors

BaP exposure allows cells to evade G1 arrest and induces cell abnormal proliferation (Du et al. 2006). Most studies reported an activation of the ERK pathway in different cell types (Wang et al. 2015a; Hamouchene et al. 2011; Du et al. 2006). Interestingly, a short exposure of 24 h was sufficient to induce a dose-related activation of MAPK in normal human embryo lung diploid fibroblasts (Du et al. 2006), and apparently cellular response to BaP exposure was also dependent on the growth kinetics within a target cell population (Hamouchene et al. 2011), suggesting different susceptibility based on cell state and differentiation.

Hallmark 3. Resisting Cell Death

A recent exome-wide mutation profile on immortal human mammary epithelial cells exposed to BaP showed that genes involved in various biological processes, including regulation of cell death, harbor mutations predicted to impact protein function (Severson et al. 2014). However, other in vitro reports showed that exposure to BaP or its DNA-reactive metabolite anti-benzo[a]pyrene-7,8-diol-9,10-epoxide (BPDE) induces or sensitizes human cells to either receptor-mediated or mitochondrial-mediated apoptosis (Stolpmann et al. 2012; Sang et al. 2012). Similar results were obtained using a reverse-phase protein array comparing mouse primary liver tumours induced by BaP to their normal adjacent tissues (Phillips et al. 2015). The results showed both down-regulation (e.g. cleaved caspase 7, caspase 3) and up-regulation (Bax, Bad, Bcl-xL) of some pro-apoptotic proteins in tumour tissues. In conclusion, BaP may exert either a pro-apoptotic effect or an anti-apoptotic effect. In vivo data on the temporal sequence of these effects are lacking, although it has been suggested that the carcinogenic effect of BaP or PAHs mixtures may be also mediated by chronic inflammation and cell death subsequent to the exposure (Engström et al. 2015).

Hallmark 4. Inducing Angiogenesis

It has been recently shown that 1 month exposure to different BaP concentrations increased in a dose-dependent manner the capacity of hepatoma cell lines, BEL-7404, to recruit vascular endothelial cells and promote angiogenesis through increased secretion of vascular endothelial growth factor (VEGF) (Ba et al. 2014). Similarly, low, non-cytotoxic concentration of BaP induced hypoxia-inducible factor-1a, responsible for the adaptation to hypoxic conditions and the promotion of angiogenesis (Mavrofrydi and Papazafiri 2012). Interestingly, BaP and its metabolites may have a distinct and opposite effect on VEGF expression (Li et al. 2015), suggesting that tissue-specific or genetic inter-individual differences in the CYP450 expression may play a role in determining the overall effect on angiogenesis induction by BaP and PAHs in general.

Hallmark 5. Enabling Replicative Immortality

BaP has been shown to induce efficiently immortalization of Syrian hamster normal dermal cells through its mutagenic activity causing the direct inactivation of p53 and INK4 alterations (Yasaei et al. 2013; Newbold et al. 1980). Similarly, p53 mutations in Hupki cells (embryonic murine fibroblast with human p53 gene) exposed to BaP were correlated with p53 mutations in human lung tumours, supporting the direct role of BaP in causing smokers lung tumour p53 mutations (Liu et al. 2005). In conclusion, BaP seems to enable replicative immortality mainly through its genotoxic properties.

Hallmark 6. Activating Invasion and Metastasis

There is evidence that BaP can promote cell migration, invasion and metastasis (Ochieng et al. 2015). Recently, some studies provided mechanistic clues showing that BaP may contribute to lung cancer cell invasion and metastasis by up-regulating pro-inflammatory chemokines (IL8, CCL-2, CCL-3) and one of the master regulators of the epithelial-to-mesenchymal transition, Twist (Zhang et al. 2016a; Wang et al. 2015c). Moreover, BaP was able to induce cell migration in triple-negative breast cancer MDA-MB-231 cells through a lipoxygenase- and Src-dependent pathway, notably by increasing the secretion of metalloproteinase MMP-2 and MMP-9 (Castillo-Sanchez et al. 2013). BaP-treatment was also able to increase the metastatic potential of hepatocellular carcinoma cell lines in a mouse model, likely through an activation of both angiogenesis and NF-kB pathway (Ba et al. 2014). In conclusion, BaP is able to promote cell migration and invasion, contributing to increase the metastatic potential of several epithelial cells in different experimental settings.

Hallmark 7. Deregulating Cellular Energetics

Recent studies showed that BaP is able to alter the function of mitochondria, cellular organelles having a key role in cellular energetics as well as in programmed cell death. In particular, BaP (range, 0–500 mM) was able to lower mtDNA content in TK6 cells, a human lymphoblastoid cell line (Pieters et al. 2013). Interestingly, in the same study, indoor exposure to PAHs was associated with decreased mtDNA content in the blood (Pieters et al. 2013). Similarly, BaP induced mitochondrial damage in vivo in mice cervical tissues, strongly associated with increased oxidative stress (Gao et al. 2010). Interestingly, a short-term exposure (24–48 h) to BaP induced an increased expression of several components of the mitochondrial respiratory chain (Salazar et al. 2004), suggesting an adaptative process that could cause a mitochondria-derived increased oxidative stress to the exposed cells and tissues. In conclusion, while a more metabolism-focused research on the effects of BaP is needed, some evidence exists that it could directly impact cellular metabolism and energy production, notably through an alteration of the mitochondrial function.

Hallmark 8. Avoiding Immune Destruction

Early studies in the 1980s showed BaP has immunotoxic effects in mice, by reducing antibody production and inducing DNA adducts in splenic leukocytes (Dean et al. 1983; Ginsberg et al. 1989). A more recent study showed that BaP at a dose as low as 10 mg/kg b.w. (generally considered as non-toxic) was able to induce changes in thymus weight and spleen B-cell populations (De Jong et al. 1999).

In conclusion, while no evidence was available to test the hypothesis that BaP may affect cancer cells immune recognition, BaP is able to exert immunotoxicity, thus contributing to avoiding immune destruction by cancer cells.

Enabling Characteristics: Genome Instability and Mutation and Tumour-Promoting Inflammation

The acquisition of the hallmarks of cancer is made possible by two enabling characteristics, namely, genome instability (which allows accumulation of random mutations) and the inflammatory state accompanying many premalignant and malignant lesions (Hanahan and Weinberg 2011). As a genotoxic compound, BaP and its metabolites are a well-known mutagen for both human and experimental animals. Interestingly, a recent study involving next-generation sequencing of exomes in immortal human mammary epithelial cells showed that DNA repair genes were among the genes harboring BaP-induced mutation with a potential functional impact, suggesting a potential impact on genome stability of the mutated cells (Severson et al. 2014). Moreover, BaP increased the number of oxidatively induced clustered DNA lesions in normal primary breast-derived cells which were correlated with the number of chromosomal aberrations (Sigounas et al. 2010). These lesions were associated with a decrease of antioxidant defense capacity and an increased ROS and DNA repair gene transcription (Sigounas et al. 2010), suggesting that oxidative stress and DNA damage and repair response are strictly correlated following BaP exposure.

Important cross talks between chemicals and immune system in carcinogenesis have been recently reviewed (Kravchenko et al. 2015). Interestingly, interleukin 6 (IL-6) and tumour necrosis factor alpha (TNF-a) produced by macrophages have been shown to be critical to promote malignant transformation of human bronchial epithelial cells in a bionic airway chip culture system and in an animal model (Li et al. 2015). Similarly, it was observed that TNF-a strongly augmented the formation of stable Bap diol epoxide-DNA adducts in alveolar type II epithelial cells (Umannová et al. 2011). Thus, the inflammatory response to PAHs containing BaP and the immunosuppressive ability of BaP toward acquired immunity cells previously described may synergistically cooperate in the carcinogenicity of BaP aPAHs mixtures.

Bisphenol A and the Hallmarks of Cancer

Bisphenol A (BPA) is a chemical compound, produced in large quantities for use primarily in the production of polycarbonate plastics and epoxy resins. Human exposure to BPA is widespread. More specifically, in a study conducted by the Centers for Disease Control and Prevention (CDC) found detectable levels of BPA in 93% of 2517 urine samples from people 6 years and older in the United States (“Bisphenol A (BPA)” 2016). BPA exhibits estrogen mimicking, hormone-like properties that raise health concerns about its suitability in some consumer products and food containers. However, the ability of BPA to act as a carcinogen has been a matter of great debate.

Some epidemiological studies implicate BPA exposure in cancer incidence but the direction of the association is not consistent. In a case–control study in China, urine BPA concentration was positively associated with meningioma diagnosis (Duan et al. 2013). BPA was also shown to be positively associated with established cancer risk factors such as visceral obesity, waist circumference, glucose homeostasis and concentrations of inflammatory markers. Conversely, in a case–control study in Tokyo, serum BPA was found to be lower in premenopausal women with complex endometrial hyperplasia with malignant potential compared to controls and was also significantly lower in women with endometrial cancer than in controls (Hiroi et al. 2004). Also, despite not demonstrating an association between probable occupational BPA exposure and breast cancer, a population case–control study in the USA found BPA exposure to be more common in controls rather than cases (Aschengrau et al. 1998).

In contrast to the epidemiological studies, mechanistic evidence from in vitro and in vivo studies showed that BPA impacts on several of the cancer hallmarks and may thus act as a carcinogen. Here, we review the available experimental evidence of BPA effects on the different hallmarks of cancer.

Hallmark 1. Sustaining Proliferative Signalling

Proliferative effects of BPA are supported by a wealth of in vitro and in vivo studies. Most studies evidence a direct effect of BPA on cell proliferation (Sheng et al. 2013; Sengupta et al. 2013; Park and Choi 2014; Newbold et al. 2007; Nakagawa and Suzuki 2001; Moral et al. 2008; Jung et al. 2011; Han et al. 2001; Ge et al. 2014a; Colerangle and Roy 1997; Betancourt et al. 2010a; Liu et al. 2015; Zhu et al. 2009; Pisapia et al. 2012; Lam et al. 2015; Smith et al. 2016a, b; Zhang et al. 2012; Schafer et al. 1999; Mlynarcikova et al. 2013; Recchia et al. 2004; Ibrahim et al. 2016; Park et al. 2009; Murray et al. 2007; Kang et al. 2012; Ayyanan et al. 2011; Lee et al. 2012a; Wang et al. 2015b, 2014a; Jenkins et al. 2009, 2011). Yet other studies imply proliferative effects by evidencing BPA’s ability to mimic estrogen and have estrogenic effects (Katchy et al. 2014; Hall and Korach 2013; Gould et al. 1998; Dong et al. 2011; Chun and Gorski 2000; Lee et al. 2014; Maruyama et al. 1999; Kim et al. 2003; Hwang et al. 2011; Zhang et al. 2012; Terasaka et al. 2004) and by studying the alterations in gene expression conferred by BPA exposures (Hall and Korach 2013; Hess-Wilson et al. 2006). Interestingly, BPA exposure was shown to have proliferative effects in tissues, particularly mammary tissue, of mice/rat offspring even when exposed pre- or perinatally through maternal exposure during pregnancy or lactation (Murray et al. 2007; Newbold et al. 2009; Markey et al. 2001; Wang et al. 2014b; Betancourt et al. 2010a; Mandrup et al. 2016; Ayyanan et al. 2011; Ichihara et al. 2003).

Many of these studies indicate that the proliferative effects of BPA are dose-dependent, with higher doses invoking greater rates of proliferation in the nanomolar range (Moral et al. 2008; Lin et al. 2013; Colerangle and Roy 1997; Wang et al. 2013). However, in the micromolar concentration ranges, BPA effects were shown to be reversed, causing decreased proliferation in Sertoli TM4 cells (Ge et al. 2014b).

The signal transduction pathways activated by BPA through which it exerts its effects have also been extensively investigated. BPA has been shown to bind to ERα receptors (Nakagawa and Suzuki 2001; Miyakoshi et al. 2009) but other studies show BPA effects to be G protein–coupled receptor (GPCR) dependent, with particular involvement of the GPER receptor (Pupo et al. 2012; Song et al. 2015, p. 2; Bouskine et al. 2009). BPA was also shown to confer androgen-independent stimulation of the androgen receptor AR-T877A (Wetherill et al. 2002).

BPA induces specific gene expression changes (Park and Choi 2014) targeting proteins that regulate cell cycle progression such as the p38 MAPK proteins in human mammary cancer and ovarian adenocarcinoma cell lines (Lee et al. 2014; Kang et al. 2013), CD4K and cyclin D1 proteins in human neuroblastoma cells in vitro and neuroblastoma tumours in rats in vivo (Zhu et al. 2009), PCNA, bcl2 and bax proteins in ovarian cancer cell lines (Mlynarcikova et al. 2013; Yu et al. 2004), STAT3 in a human mammary cancer cell line (Zhang et al. 2012) and p16 and cyclin E in a human mammary epithelial cell line (Qin et al. 2012). More specifically, BPA has been shown to increase expression of cyclin D1, responsible for the G1/S cycle transition, and to decrease expression of p21, a potent CDK inhibitor that arrests cell cycle in the G1 phase in human mammary and ovarian cancer cell lines (Hwang et al. 2013; Kang et al. 2012; Lee et al. 2012a). BPA also activates both cAMP-dependent protein kinase and cGMT-dependent protein kinase pathways, causing phosphorylation of cAMP responsive element-binding protein (CREB) and Rb as shown in human testicular seminoma cells (Bouskine et al. 2009). In addition, the PTEN/AKT/p53 axis was evidenced to be involved in BPA proliferative effects. Specifically, BPA induced up-regulation of oncogenic miR-19a and miR-19b and caused dysregulated expression of miR-19-related downstream proteins, including PTEN, p-AKT, p-MDM2, p53 in human mammary cancer cells (Li et al. 2015). The IGF-1 signalling pathway is also activated by BPA, resulting in increased proliferation (Klotz et al. 2000). Perhaps more importantly, induction of key genes and proteins in the PI3K-mTOR pathway—AKT1, RPS6 and 4EBP1 and a concurrent reduction in the tumour suppressor, that is phosphatase and tensin homolog gene protein—has been shown in benign breast cells following BPA exposure. Lastly, other changes conferred by BPA, such as elevated phospho-AKT, c-Raf, and phospho-ERKs1 and 2 and decreased TGFβ, have been shown to predispose to chemically induced cancer in adult rats, even when exposure was prenatal (Betancourt et al. 2010b).

On the other hand, far fewer studies suggest that BPA exposure is not sufficient to exert proliferative effects, and that co-exposures such as phytoestrogens and IGF 1 (Katchy et al. 2014; Ishido 2004), or co-activators and particular conditions such as oestrogen depletion, or selective mutation of the ER (Hess-Wilson et al. 2006) or carcinogenic exposures such as 3,2′-dimethyl-4-aminobiphenyl (DMAB) (Ichihara et al. 2003) or catechol (Oikawa 2005) are needed.

Still other studies demonstrate no proliferative effects of BPA on membrane estrogen receptor-alpha-enriched GH3/B6/F10 pituitary tumour cells (Kochukov et al. 2009), in rat liver epithelial WB-F344 cells (Dong et al. 2014), in a human breast cancer cell line MCF-7 (Diel et al. 2002), in the human endometrial cell line, ECC-1 (Bergeron et al. 1999) and in adult female prenatally exposed mice (Yoshida et al. 2004).

Lastly, few studies demonstrate decreased proliferation in lung cancer cells (Andreescu et al. 2005) or in murine-derived multipotent neural progenitor cells (NPC) (Kim et al. 2009) as exposure concentrations increased but both studies used fairly high concentrations in the micromolar range.

Overall, the majority of studies support proliferative effects of BPA in nanomolar concentrations, which are also the most relevant with regard to human environmental exposure.

Hallmark 2. Evading Growth Suppressors

As already discussed, BPA induces cell abnormal proliferation by allowing cells to evade G1 arrest. By disrupting cell cycle regulatory proteins, BPA also allows a cell to evade growth suppression. More specifically, BPA down-regulates p16 and p21 (Hwang et al. 2013; Kang et al. 2012; Lee et al. 2012a; Qin et al. 2012), both of which play important roles in regulating/halting transition from G1 to S phase. In addition, by phosphorylating Rb (Bouskine et al. 2009), BPA negates Rb’s regulatory role in cell cycle progression and thus its tumour suppressive properties. Furthermore, BPA deregulates PTEN and p53 (Li et al. 2015) and decreases TGFβ (Betancourt et al. 2010b), thus inhibiting them from exerting their tumour suppressive properties and allowing cells to evade their growth regulatory roles.

Lastly, BPA reduces expression of the tumour suppressor, phosphatase and tensin homolog gene protein, in human high-risk donor breast epithelial cells (Goodson et al. 2011).

Hallmark 3. Resisting Cell Death

Apoptotic effects of BPA seem to be contradictory.

BPA at 10⁻⁷ M in epithelial stromal co-cultures derived from the contralateral tissue of patients with breast cancer induced gene expression patterns that facilitate apoptosis evasion (Dairkee et al. 2008). In addition, in a human MCF-7 breast cancer cell line, BPA reduced cell apoptosis at all concentrations higher than 10⁻⁸ M, with the lowest rate of apoptosis observed at a dose of 10⁻⁶ M (Diel et al. 2002). In ovarian cancer cells, BPA co-exposure with leptin was shown to have anti-apoptotic properties by inhibiting caspase 3 expression and activity by modulating STAT3 and ERK1/2 signalling pathways (Ptak et al. 2013). Similarly, in ovarian cancer cells, BPA suppressed expression of pro-apoptotic genes, increased expression of pro-survival genes and decreased caspase 3 activity (Ptak et al. 2011). In vivo, BPA decreased apoptosis in neonatal/prepubertal rats exposed to BPA through lactation and this decrease was associated with changes in PR-A, SRC1–3, erbB3 and Akt activity (Jenkins et al. 2009). Lastly, in cells treated with tamoxifen, pretreatment with BPA evidenced dose-dependent apoptotic evasion (Goodson et al. 2011).

Yet other studies indicate that BPA associated cytotoxicity can decrease cell viability and induce apoptosis in a dose-dependent manner (S. Terasaka et al. 2004; Jenkins et al. 2011). In ER-positive breast cancer cells (MCF-7: WS8) and in insulin-secreting cell lines (INS-1), BPA at higher concentrations (2x10⁻⁸ to 10⁻⁶ M) promotes apoptosis (Sengupta et al. 2013; Lin et al. 2013) through the release of cytochrome C. In Sertoli cells, apoptosis was also induced but by the transmission of apoptotic signals to the mitochondria via the CAM-CAMKII-ERK1/2 pathways (Qian et al. 2014). Expression changes such as up-regulation of PCNA and bcl-2 mRNA expression and down-regulation of the bax mRNA expression, following exposure to BPA (32 × 10⁻⁷M), suggest involvement of this pathway in apoptosis in ovarian cancer cells (Yu et al. 2004). BPA-induced caspase 3 increase was also implied to mediate pro-apoptotic effects in the mammary gland of adult albino rats (Ibrahim et al. 2016). Caspase activation, phosphorylation of ERK and AKT and involvement of both intrinsic and extrinsic apoptotic pathways were also evidenced in leukemia cells (Bontempo et al. 2009).

Hallmark 4. Inducing Angiogenesis

A few studies evidenced BPA to have pro-angiogenic effects by investigating mechanisms involved in this hallmark. BPA increased microvessel density and VEGF expression in neuroblastoma tumours from mice (BPA: 200 mg/kg per day), and VEGF expression was also increased in cultured neuroblastoma cells in vitro (2 μg/mL) (Zhu et al. 2009). In breast cancer cell lines, BPA (1⁻¹⁰ μM) was shown to regulate VEGF expression by an ER-dependent mechanism, and suppression of VEGF effects via use of inhibitors suggested involvement of MEK, p38 kinase and PI3K pathways in BPA’s pro-angiogenic effects (Buteau-Lozano et al. 2008). Yet, in bovine aortic endothelial cells (BAECs), BPA’s (100 nM) pro-angiogenic effects were exerted by up-regulation of XIAP expression via GPER activation and not ERα or ERβ activation (Liu et al. 2015). Lastly, BPA (1, 40 and 100 nM) stimulated VEGF-R2 expression in both cancerous and non-cancerous cell lines and also stimulated VEGF-A only in non-cancerous cells (Ptak and Gregoraszczuk 2015). Given VEGF-A’s role in angiogenesis, these studies imply that BPA can have pro-angiogenic results.

Hallmark 5. Enabling Replicative Immortality

BPA was shown to successfully increase the levels of both p16 and cyclin E and to induce cellular senescence in human mammary epithelial cells, as assessed by the number of human heterochromatin protein-1γ positive cells (Qin et al. 2012). Even though BPA was not directly shown to induce senescence in other studies, its exposure effects such as phosphorylation of Rb (Bouskine et al. 2009), deregulation of PTEN and p53 (Li et al. 2015), decrease in TGFβ (Betancourt et al. 2010b) and other tumour suppressors, as well as induction of key genes and proteins in the PI3K-mTOR pathway (Li et al. 2015) can contribute toward replicative immortality.

Hallmark 6. Activating Invasion and Metastasis

Several studies evidence that BPA can promote cell migration, invasion and metastasis (Zhang et al. 2016b; Kim et al. 2015a; Chen et al. 2015; Liu et al. 2015), and epithelial-to-mesenchymal transition (EMT) was shown to play an important role in the acquisition of this hallmark (Chen et al. 2015; Wang et al. 2015b).

Treatment of SW840 colorectal cancer cells promoted metastasis via induction of EMT characterised by acquiring mesenchymal spindle-like morphology and increase in N-cadherin. In parallel, E-cadherin was decreased and Snail transcription factor was increased (Chen et al. 2015). BPA was also shown to stimulate migration through the up-regulation of migration-related factors MMP2, MMP3 and MMP9 in addition to N-cadherin in neuroblastoma and ovarian cancer cells (Ptak et al. 2014; Zhu et al. 2010). These effects were shown to be dose sensitive and dependent on the MAPK and PI3K/Akt signalling pathways (Ptak et al. 2014). Up-regulation of MMPs following BPA expression was also evident in lung cancer cells but in these cells BPA treatment induced the up-regulation through rapid activation of ERK1/2 via GPER/EGFR (Zhang et al. 2014). Similarly, in MDA-MB-231 breast cancer cells, BPA induced GPER-dependent activation of migration, and an invasion process involving AP-1/NFκB-DNA binding activity through an Src- and ERK2-dependent pathway (Castillo Sanchez et al. 2016). The IKK-β/NF-κΒ signalling pathway was also elsewhere shown to be implicated in BPA-induced cervical cancer cell migration and invasion (Ma et al. 2015). Another mechanism implicated in BPA-induced cell migration was the Erβ-mediated integrin B1/MMP9 pathway (Shi et al. 2016; Kim et al. 2015b).

Despite the implicated mechanisms, all studies above evidenced BPA effects on migration. However, it is important to note that one study evidenced that BPA (0.1 and 1 μM concentration) reduced cell motility of rat liver epithelial WB-F344 cells (Dong et al. 2014).

Hallmark 7. Deregulating Cellular Energetics

One study investigating primarily the effects of BPA exposure on apoptosis of INS-1 cells evidenced that BPA-induced apoptosis was associated with mitochondrial defects, including depletion of ATP, release of cytochrome c, loss of mitochondrial mass and membrane potential, and alterations in expression of genes involved in mitochondrial function and metabolism. These mitochondrial defects may confer dysregulation of cellular energetics. However, the direct effects of BPA on reprogramming cells to favour glycolysis (Warburg effect) have not been investigated.

Hallmark 8. Avoiding Immune Destruction

Even though no studies have evidenced a direct capacity of BPA to affect recognition of cancer cells by immune cells, or immunosuppressive factor secretion from cancer cells, several studies suggest adverse effects on the functioning of the immune system that result from exposure to BPA (immunotoxicity).

Firstly, BPA exposure-related gene expression changes indicated an enrichment of genes involved in immune system response and regulation (Fic et al. 2015). BPA was also shown to affect the regulation of the immune system by reducing NO and TNFα production, through inhibition of NF-κβ mediated by ER (Kim and Jeong 2003). NF-κB is a first responder to harmful cellular stimuli such as reactive oxygen species (ROS), tumour necrosis factor alpha (TNFα) and interleukin 1-beta (IL-1β). By inhibiting NF-κβ, the organism’s ability to regulate immune response is compromised and this might enable immune destruction evasion of cancer cells.

Enabling Characteristics: Genome Instability and Mutation and Tumour-Promoting Inflammation

In an update of their original publication on the hallmarks of cancer (Hanahan and Weinberg 2000), Hanahan and Weinberg recently identified two enabling characteristics which promote malignancy and the establishment of the cancer hallmarks. These characteristics are genomic instability (which facilitates accumulation of mutations) and tumour-promoting inflammation (Hanahan and Weinberg 2011).

Several publications support a genotoxic role of BPA in vivo and in vitro. BPA’s cytotoxicity was evidenced in rat INS-1 cells where it increased DNA strand breaks, expression of DNA damage associated genes (p53 and CHK2) and ROS concentration (Xin et al. 2014). Increased ROS production and compromised mitochondrial function were also evidenced following BPA exposure in sperm and brain tissue from mice (Ooe et al. 2005). In addition, at high doses (≥100 μM) BPA induces numerical chromosomal changes almost in the diploid range and associated cellular transformation (Tsutsui et al. 1998). Improper chromosome segregation in response to BPA exposure can also cause micronuclei formation (Kabil et al. 2008) and BPA can also cause induction of DNA fragmentation (H. Terasaka et al. 2005; Ptak et al. 2011). BPA’s cytotoxic and genotoxic effects are supported by studies showing BPA’s pro-apoptotic effects, especially at larger exposures including p53 activation (H. Terasaka et al. 2005; Sengupta et al. 2013; Lin et al. 2013; Ptak et al. 2011).

Furthermore, BPA-treated cells were evidenced to have increased expressions of genes involved in DNA repair in order to overcome the DNA damage caused by the chemical (Fernandez et al. 2012).

In addition to BPA itself, BPA metabolites were also shown to have genotoxic effects. BPA-Q, when not properly detoxified, can cause depurination of DNA (Qiu et al. 2004) and many other BPA metabolites can cause formation of DNA adducts in vivo (liver and mammary cells of female CD-1 mice, male Sprague-Dawley rats and Swiss ICR (CD-1) mice) (Izzotti et al. 2009, 2010; De Flora et al. 2011).

BPA can contribute to genomic instability also through epigenetic and transcriptional changes. In adult mice exposed in utero, epigenetic effects of BPA included increased histone-3-trimethylation which is usually associated with repressed gene expression (Doherty et al. 2010; Dhimolea et al. 2014; Wong et al. 2015) and in breast cancer cell lines, BPA was shown to alter the miRNA profiles, including expression of miR-21 (Tilghman et al. 2012). Furthermore, BPA exposure increased transcription and chromatin modification factor recruitment in key promoters such as HOXB9 (Deb et al. 2016), HOXC6 (Hussain et al. 2015) and HOTAIR long non-coding RNA (Bhan et al. 2014). Lastly, BPA can induce altered methylation in several regions of the genome, including imprinting regions, in the embryo and placenta after maternal exposure (Hanna et al. 2012; Susiarjo et al. 2013; Jorgensen et al. 2016).

These epigenetic changes can explain the aberrant transcriptomic profiles observed after BPA exposure in breast neoplasms and endometriosis (Roy et al. 2015), human MCF-7 breast cancer cells and T47D carcinoma cells (Tilghman et al. 2012; Vivacqua et al. 2003; Buterin et al. 2006), human ovarian cancer cell lines (Hayes et al. 2016), endometrial cancer cells including estrogen receptor (ER)-proficient Ishikawa plus and ER-deficient Ishikawa minus endometrial cancer cells (Boehme et al. 2009; Gertz et al. 2012), a mouse Sertoli cell line (TTE3 cells) (Tabuchi et al. 2002), rodent prostate cancer (PCa) models (Ho et al. 2015) and female Sprague-Dawley offspring born to mothers exposed to BPA during gestation (Grassi et al. 2016). Genes aberrantly expressed following BPA exposure include but are not limited to CYP19A1, EGFR, ESR2, FOS and IGF1 (Roy et al. 2015), and many growth- and development-related genes such as HOXC1 and C6, Wnt5A, Frizzled, TGFbeta-2 and STAT inhibitor 2 (Singleton et al. 2006).

In turn, transcriptome changes confer aberrant proteomic profiles involving several key cellular processes such as cell proliferation, inhibition of apoptosis, tissue remodeling, inflammation, stress response and glutathione synthesis in vitro (primary organotypic cultures of the mouse mammary gland and non-cancerous human high-risk donor breast epithelial cell (HRBEC) cultures) (Williams et al. 2016; Dairkee et al. 2013) and in vivo (rats exposed prenatally or prepubertally and Swiss ICR (CD-1) mice) (Betancourt et al. 2012, 2014; Izzotti et al. 2010; Lee et al. 2012b; Tang et al. 2012). Proteins whose expression is altered in response to BPA exposure have important roles in protein metabolism, signal transduction, development and cell cycle regulation (Lamartiniere et al. 2011) and can therefore make cells more susceptible to carcinogenesis.

In addition to genetic and epigenetic effects, BPA can also confer instability by antagonistically binding to cellular regulator receptors (Moriyama et al. 2002; Greathouse et al. 2012). Lastly, BPA directly affects DNA repair through impairment of the double-strand break repair machinery (Allard and Colaiácovo 2010) and through binding of key players of DNA repair such as DNA-PKcs (Ito et al. 2008). On the other hand, it might also prevent initiation of repair of oxidised base lesions by the base excision repair pathway (Gassman et al. 2015).

In addition, to genomic instability, BPA also contributes to tumour promoting inflammation, especially in the tumour microenvironment. In adult noble rats, BPA induced marked infiltration of CD4+ and CD8+ T cells into the prostatic intraepithelial neoplasia (PIN) (Lam et al. 2015). This immune cell infiltration is an integral part of the tumour microenvironment and can be used to a tumour’s advantage since it may be “utilized” in sustaining angiogenesis, stimulating proliferation, facilitating tissue invasion and supporting metastatic dissemination (Hanahan and Weinberg 2011).

In addition, treatment of THP1 macrophages and primary human macrophages with BPA increased pro-inflammation cytokines tumour necrosis factor-α (TNF-α) and interleukin-6 (IL-6) production but decreased anti-inflammation cytokines interleukin-10 (IL-10) and transforming growth factor-β (TGF-β) production (Liu et al. 2014). BPA alters expression of both anti- and pro-inflammatory modulators in mammary tissues after in-utero exposure of CD1 mice, creating a supporting microenvironment for tumour formation. Some of these changes include decreased expression of members of the chemokine CXC family, interleukin 1 (Il1) gene family (Il1β and Il1rn), interleukin 2 gene family (Il7 receptor) and interferon gene family (interferon regulatory factor 9 (Irf9)), as well as immune response gene 1 (Irg1) (Fischer et al. 2016).