Mobile elements and viral integrations prompt considerations for bacterial DNA integration as a novel carcinogen

Abstract

Insertional mutagenesis has been repeatedly demonstrated in cancer genomes and has a role in oncogenesis. Mobile genetic elements can induce cancer development by random insertion into cancer related genes or by inducing translocations. L1s are typically implicated in cancers of an epithelial cell origin, while Alu elements have been implicated in leukemia as well as epithelial cell cancers. Likewise, viral infections have a significant role in cancer development predominantly through integration into the human genome and mutating or deregulating cancer related genes. Human papilloma virus is the best-known example of viral integrations contributing to carcinogenesis. However, hepatitis B virus, Epstein-Barr virus, and Merkel cell polyomavirus also integrate into the human genome and disrupt cancer related genes. Thus far, the role of microbes in cancer has primarily been attributed to mutations induced through chronic inflammation or toxins, as is the case with Helicobacter pylori and enterotoxigenic Bacteroides fragilis. We hypothesize that like mobile elements and viral DNA, bacterial and parasitic DNA may also integrate into the human somatic genome and be oncogenic. Until recently it was believed that bacterial DNA could not integrate into the human genome, but new evidence demonstrates that bacterial insertional mutagenesis may occur in cancer cells. Although this work does not show causation between bacterial insertions and cancer, it prompts more research in this area. Promising new sequencing technologies may reduce the risk of artifactual chimeric sequences, thus diminishing some of the challenges of identifying novel insertions in the somatic human genome.

1. Introduction

In 2013, 1.6 million people were estimated to develop cancer in the US, while 49% of men and 38% of women will develop cancer in their lifetime. Genetic changes, environmental factors, and infectious agents can all cause increased cell proliferation leading to cancer through abnormal epigenetic alterations, point mutations, translocations, and other modifications. Infectious agents and mobile elements that integrate into the human genome, in whole or in part, are an example of events that inflict such DNA damage [122]. Here, we explore whether bacterial integrations may also occur in tumors.

2. Role of mobile elements in carcinogenesis

2.1 Mobile elements basics

Mobile elements are possibly the best-studied insertional mutagens of the human genome. While currently inactive in the human genome [92], DNA-based transposons account for ~45% of the human genome [40]. RNA-based retrotransposons can be divided into classes based on the presence/absence of long terminal repeats (LTR). The LTR-retrotransposons, such as endogenous retroviruses, are not currently active in humans [79]. However, the non-LTR retrotransposons actively jump throughout the human genome acting as mutagens [40, 54, 79] (Figure 1A). These non-LTR retrotransposons consist of long interspersed elements 1 (LINE-1s or L1s), Alu elements (a type of short interspersed element, or SINE), and SVAs (a combination of SINE-R, variable number of tandem repeats, and Alu-like sequences). L1s are the only non-LTR retrotransposons that encode enzymes for retrotransposition [54], and as such, they mobilize themselves as well as Alu elements and SVAs [54]. Of the >500,000 L1s in the human genome, only a few are highly active and responsible for the majority of new insertions [13]. In every generation, L1s and Alu elements move to new genomic locations through germline retrotransposition [54]. Many disease-causing transpositions have been identified thus far including ones in breast cancer [76, 128], Apert Syndrome [91], hemophilia [47, 85], and colon cancer [77, 123]. While L1s and Alu elements tend to be inactive in somatic tissues, their-reactivation could aid in tumorigenesis [54].

Figure 1. Carcinogenesis by mobile element movement, viral integrations, and microbial infections.

(A) Mobile genetic elements, such as retrotransposons, can play a role in cancer mutagenesis. They can either move to a new location in the genome or recombine with similar mobile elements and cause genomic instability. This vulnerability often contributes to cancer progression. (B) Viruses use three broad mechanisms to promote carcinogenesis. Viruses can integrate into the human genome inducing genomic instability (purple arrow on the left). The deregulation of signaling pathways induced from these chromosomal disruptions transform normal cells into cells with cancerous phenotypes. Regardless of life cycle stage, viruses can express proteins that alter normal cell cycle patterns by increasing proliferation and decreasing apoptosis (purple arrow on the right). Viruses can also alter inflammatory responses that cause DNA damage and contribute to cancer development (gray arrow), a mechanism that is not discussed in this review. (C) Microbial infections can lead to carcinogenesis through multiple different pathways. It is important to note that these pathways can occur in conjunction with each other or individually. Microbes can first cause irritation and inflammation that leads to DNA damage and thus cancer (gray arrow). Or bacteria can secrete microbial proteins that alter host-signaling pathways, leading to increased cell proliferation and decreased apoptosis, the classic characteristic of cancer (blue arrow on the right). While multiple examples of viral integrations that progress to cancer have been described, evidence for microbial integration into the human genome has only been reported recently.

2.2 L1 insertions and cancer

L1s have been implicated in tumorigenesis of various epithelial cell cancers [14], including lung cancer [40, 46]. A genomic comparison of 20 lung cancer samples with normal matched samples revealed 9 L1 insertions in only the tumor samples [46]. Six of these samples had 1–3 new L1 insertions/tumor and also showed an increase in hypomethylated DNA when compared to the matched normal samples [46], meaning the DNA in these samples had decreased methylation. Since host cells reduce transposition through increased methylation, hypomethylated tumor cells are more permissive to L1 transposons [46].

Colorectal cancer samples also have an increased rate of L1 insertions, some of which have disrupted genes with known cancer driver functions [123], like the APC gene [77]. More tests are necessary to resolve the exact effect of L1 insertions since they may be passenger mutations or may be directly related to tumor formation [46, 123]. It has recently been reported that the methylation status of L1 insertions in rectal cancer predicts the clinical outcome of the patient, with individuals with L1 hypomethylation having shorter survival times and higher incidences of tumor recurrence [8].

2.3 Alu elements and cancer

Alu elements are the most successful type of transposon with >1 million copies in the human genome [60]. Alu insertions have been associated with familial adenomatous polyposis, breast cancer, acute myeloid leukemia, and hereditary non-polyposis colorectal cancer syndrome [40].

2.3.1 Alu elements can disrupt DNA repair

Alu elements have been found to mutate tumor suppressor genes that aid in DNA repair, like BRCA1 and BRCA2, which are important for homologous recombination in DNA double strand break repair. Recombination between two similar Alu elements led to the loss of a 3-kbp region in exon 17 of the BRCA1 gene and subsequent inactivation of BRCA1 causing breast cancer [82]. Other breast cancer studies have demonstrated multiple Alu-mediated mutations in the BRCA1 and BRCA2 genes [104]. Likewise, hereditary non-polyposis colorectal cancer syndrome can occur when mutations arise in genes involved in the mismatch repair system, such as MLH1 or MLH2 [80]. Both of these genes have increased Alu concentrations in their introns with about 75% of the rearrangements in the MLH2 gene caused by Alu-mediated non-allelic homologous recombination [65].

2.3.2 Alu elements and leukemia

In acute myeloid leukemia, Alu-mediated partial duplications frequently occur in the coding regions of the MLL1 gene, most commonly resulting in duplicate exons 5 through 11 or 12 to fuse upstream of the original exon 5 [112]. MLL1 is part of the mixed lineage leukemia gene family, which have histone methylase capabilities and are involved in HOX gene regulation [2]. Some experiments suggest that the wild-type allele can be suppressed, allowing for expression of the MLL1 copy containing the Alu-mediated partial duplication. This in turn induces a leukemic phenotype [134] through altered protein structure and inactivation of inherent MLL1 function [111]. Overall, transposable elements could play a major role in somatic and germline mutagenesis and more focused research in this area is necessary before rates of L1 and Alu element involvement in cancer development and progression can be estimated.

3. Viral integrations in cancer

3.1 Viruses involved in carcinogenesis

In 2002, viral infections were estimated to cause 12.1% of cancers [93]. Known viral carcinogens include EBV, hepatitis B virus (HBV), hepatitis C virus (HCV), human papilloma virus (HPV), Kaposi’s sarcoma-associated herpesvirus (KSHV), and human T-cell lymphotrophic virus type 1 [108]. Human immunodeficiency virus type-1 [108, 116], Merkel cell polyomavirus (MCV) [30, 108], and BK polyomavirus [1, 74] are also proposed to be associated with cancer. While both DNA and RNA viruses have been linked to carcinogenesis, only the DNA viruses discussed in this review are known to integrate into the host genome and contribute to carcinogenesis through insertional mutagenesis.

3.2 Human papilloma virus

The insertion of HPV DNA into human chromosomes is the single most important event leading to tumorigenesis in cervical cancer (Figure 1B). Clonal integration of HPV-16 or HPV-18, two of the “high-risk” strains, [130], occurs in 80–100% of cervical carcinoma tumors [18, 19, 75]. Normally, the HPV E2 protein regulates E6 and E7 oncogene expression [6] and while the virus replicates, the host cell maintains control of proliferation [133]. However, when HPV integrates into the host nuclear genome it frequently does so in a manner that leads to disruption of E2 [133]. The integration of E6 and E7 without a functional E2 deregulates E6 and E7 [103]. Uncontrolled expression of E6 leads to down-regulation of the TP53 pathway and thus reduction of apoptosis contributing to cellular transformation to a malignant phenotype [110]. Deregulation of E7 leads to increased cell proliferation through many mechanisms, including degradation of the retinoblastoma protein [11], which controls the duration of the G1 phase of the cell cycle [43].

3.3 Hepatitis B virus

HBV DNA often integrates into hepatocellular carcinoma cells as infected hepatocytes regenerate [3, 12, 16, 25, 29, 70], but its necessity for carcinogenesis is uncertain [101]. Nonetheless, the number of HBV integrations in tumor cells is predictive of time of survival [126]. Individuals with >3 integrations typically have a decreased time of survival when compared to that of individuals with <3 integrations [126]. In addition, individuals with >3 integrations have an earlier age of onset of cancer and a lower incidence of cirrhosis, the key risk factor for development of hepatocellular carcinoma [126]. Therefore, the increase in HBV integrations may play a key role in hepatocellular carcinogenesis in younger individuals without cirrhosis [126].

HBV has been shown to integrate recurrently into seven human oncogenes and tumor suppressor genes: TERT [35, 86, 94, 106, 126], ITPR1 (also known as IP3R1) [86, 94], IRAK2 [94], MAPK1 [86, 94], MLL2 [86], MLL4 [86, 106, 126], and CCNE1 [126]. About 40% of the integrations are from a region of the HBV genome that contains the viral enhancer, the X gene, and the core gene [126]. Expression of the integrated HBV X protein and X antigen can transform liver cells to a malignant phenotype in culture [42, 137] and give rise to hepatocellular carcinoma in mice [49, 53, 137], in addition to altering host gene expression, thus increasing cell growth and survival [28, 29, 41, 98]. Some researchers speculate that integration of the viral enhancer could facilitate human-HBV fusion proteins, disruption of tumor suppressor genes, and dysregulation of other downstream genes [126].

3.4 Epstein-Barr virus

Burkitt’s lymphoma is a type of B-cell derived non-Hodgkin lymphoma that is highly aggressive and fast growing [81] and was linked to EBV in 1964 [83]. Burkitt’s lymphomas are classified as: endemic, sporadic, and immunodeficiency-related. Endemic Burkitt’s lymphoma is associated with endemic malaria in Africa [81, 87] where EBV is found integrated into 90% of endemic Burkitt’s lymphoma cases [97]. All types of Burkitt’s lymphoma, regardless of EBV infection, have a translocation involving the MYC oncogene, most commonly rearranged with an immunoglobulin chain gene [51, 127]. The MYC-immunoglobulin translocation constitutively activates cellular proliferation by putting an immunoglobulin enhancer element in control of MYC expression, which is responsible for cell cycle control and contributes to carcinogenesis [127]. Sporadic cases of Burkitt’s lymphoma typically have different translocation breakpoints compared to endemic cases of Burkitt’s lymphoma, suggesting that EBV integration may play a role in generating the MYC translocation in endemic Burkitt’s lymphoma [81].

3.5 Merkel cell polyomavirus

MCV is a common childhood infection that after extreme sun exposure can progress to Merkel cell carcinoma, a type of aggressive skin cancer [78]. MCV is clonally integrated into 80% of Merkel cell carcinomas at random genomic locations [30, 109]. MCV has a T-antigen protein complex known to target tumor suppressor proteins, such as retinoblastoma [121], and is solely expressed in tumor cells [120]. It has been suggested that in order for MCV to contribute to carcinogenesis there must be immunosuppression [83], then the virus must integrate into the host genome, and finally mutations must occur to the large T-antigen of the T-antigen protein complex, while maintaining an intact binding domain for retinoblastoma [83, 121]. This hypothesis has been backed by the knowledge that integrated MCV typically has a truncated C-terminal domain of the large T-antigen gene [64, 121]. The full length large T-antigen activates DNA damage responses leading to upregulation of TP53 signaling, while the truncated large T-antigen does not have this capability instead triggering increased cell proliferation [64]. The large T-antigen truncation also causes loss of MCV-induced viral replication from the loss of C-terminal ATPase/helicase domain, which may allow the virus to evade the host immune system and contribute to carcinogenesis [121].

4. Microbial involvement in carcinogenesis

An estimated 15–20% of cancers worldwide are linked to infectious agents like viruses, parasites, and bacteria, resulting in 1.5 million cancer deaths worldwide in 2008 [20]. Given that there are 10X more bacterial cells in the human body than human cells [68], bacteria have a large potential to effect human health. As they relate to cancer, it is often difficult to establish causation. Microbial infections may contribute to an illness, or they may merely colonize as a result of microenvironment changes caused by the onset of disease. Yet clearly some bacteria are associated with cancer and some subset of those bacteria are thought to contribute to carcinogenesis.

Helicobacter pylori has been strongly associated with gastric carcinoma and gastric mucosa-associated lymphoid tissue (MALT) lymphoma [95]. Schistosoma haematobium has been associated with bladder carcinoma [31], Salmonella typhi with gallbladder cancer [24], Chlamydia pneumonia with lung cancer [66], and Bacteroides fragilis, Streptococcus bovis, Escherichia coli and Fusobacterium spp. with colorectal cancer [10, 15, 55, 56, 72, 105, 119]. Most of the bacteria described here have been shown to contribute to carcinogenesis through DNA damage initiated by increased inflammation [62] (Figure 1C). For example, Helicobacter pylori activates cyclooxygenase enzymes necessary for the formation of inflammatory prostaglandins. This increase in inflammation causes DNA damage by the release of mutagens like inducible nitric oxide synthase [95].

In addition, bacteria can secrete microbial proteins that affect cell signaling (Figure 1C). For example, enterotoxigenic Bacteroides fragilis secretes the B. fragilis toxin (BFT), a zinc-dependent metalloprotease toxin, which is highly related to diphtheria toxin [118]. BFT triggers cleavage of E-cadherin [135], which in turn activates β-catenin [136] and thus Wnt signaling, increasing colonic proliferation [71]. The NF-κB pathway is also activated through interaction with BFT, causing the secretion of pro-inflammatory cytokines by colonic epithelial cells [118]. Therefore, like viruses, bacteria can contribute to carcinogenesis by expression of bacterial proteins and by causing increased inflammation [27, 96, 117] (Figure 1). The role of bacterial DNA integration in oncogenesis is less well studied.

5. Bacterial DNA integration as a putative carcinogen, challenges and future directions

5.1 Background

Viral integration is one example of lateral gene transfer (LGT), which describes the transfer of DNA between diverse organisms and is synonymous with horizontal gene transfer (HGT). LGT is most frequently described as occurring between different bacteria, but examples of LGT from bacteria to animals have been increasingly reported [21]. Bacteria-animal transfers frequently occur between endosymbionts and their host [21] and to specific nematodes and insects that parasitize plants [22]. The predominant limiting factor in the occurrence of LGT is the proximity between both organisms [7, 21], and in animals this means proximity to the germ cells [21]. One example of this are Wolbachia endosymbionts, germ line intracellular endosymbionts, which frequently transfer large portions of their genomes into their arthropod and nematode hosts [23]. However, genomic analysis of the arthropod mealybug genome reveals that multiple other germ line endosymbionts may also participate in such interactions [44].

5.2 Bacteria-human lateral gene transfer

If bacterial DNA can integrate into the germ cell genome of invertebrates and be inherited, as is the case with Wolbachia DNA, it seems feasible that bacterial DNA can become integrated into vertebrate somatic genomes. The bacterial type four secretion system of Bartonella henselae has been shown to transfer DNA into human cells in culture [115] as has the analogous system in Agrobacterium tumefaciens [58]. However, integration of microbial DNA into primary cancer cells has only been described recently [99]. It is known that human cells are protected against bacterial integration by both the innate immune response and the separation of germline cells from somatic cells [100]. For example, the human immune system recognizes bacterial DNA [37, 107] and mRNA [107] as pathogen-associated molecular patterns (PAMPs) and mounts an immune response. However, components of the immune system have not been identified that recognize all bacterial rRNA [90, 107]. Absence of such components would provide a gateway for integration of unrecognized bacterial rRNA into host genomes. In addition, immune reactions, while efficient, are also imperfect meaning that some nucleic acids may go undetected. Any integrations that result could then act as potential mutagens.

A recent study using sequencing data from The Cancer Genome Atlas reported evidence for integrations of Pseudomonas-like DNA into known tumor suppressor and proto-oncogenes in stomach adenocarcinoma samples [99]. The same report also described putative Acinetobacter-like DNA integrations into either the mitochondrial genome or nuclear mitochondrial transfers of acute myeloid leukemia samples [99]. In both cases, the putative integrations have homology to bacterial rRNA [99]. However, it does not demonstrate whether these integrations may be causative or if the cancer cells may merely be more permissive for such integrations. Previously, other interactions between microbes and human cells, such as chronic inflammation, have been implicated in bacteria-associated cancers, such as stomach adenocarcinoma [62]. However, the recent demonstration of integrations of Pseudomonas-like DNA into proto-oncogenes in a subset of TCGA stomach adenocarcinoma samples may indicate direct carcinogenic potential [99].

The integrations of Pseudomonas-like DNA into stomach adenocarcinoma samples were predominantly into five genes. Four of the five genes are proto-oncogenes, including TMSB10, CEACAM5, CEACAM6, and CD74 [99]. All four of these genes are known to be up-regulated in gastric cancer [32, 39, 57, 89]. CEACAM5 and CEACAM6 are both carcinoembryonic antigen-related cell adhesion molecules, which are a type of immunoglobulin-related glycoproteins with varied functions including cell-cell recognition and regulation of cellular functions [57]. Specifically these molecules have been known to supervise tissue architecture [45] and serve as receptors for specific bacteria and viruses [57]. The CEA family of antigens is also one the most widely used tumor markers [9, 33]. Thymosin β10 (TMSB10) is an actin monomer-sequestering protein that is involved in cell motility and migration through its ability to bind G-actin [138]. It can also alter angiogenesis, apoptosis, cell proliferation and many other cellular processes involved in cancer progression [124]. CD74 is an important modulator of the immune response as it is a transmembrane glycoprotein that controls antigen presentation and is known to associate with major histocompatibility complex (MHC) class II [67]. CD74 is also involved in various signaling pathways such as NF-κB and AKT, which can both contribute to carcinogenesis [73, 125]. In the stomach adenocarcinoma samples containing Pseudomonas-like DNA insertions, the afore-mentioned genes possessed integrations of bacterial DNA in their 5’-untranslated region, which prompts speculation that the overexpression of these genes may have been triggered by the insertional mutagenesis of a repressor binding site [99] or a site governing transcript stability (Figure 2).

Figure 2. Illustration of stomach adenocarcinoma bacterial integration sites in relation to other gene features.

Four of the five genes implicated in Pseudomonas-like DNA integrations in stomach adenocarcinoma samples are depicted. Their bacterial integration sites are shown as well as the exons, introns, direction of translation, and untranslated regions.

If bacterial integration is an important mechanism of oncogenesis, we expect that further studies in this area will reveal integrations analogous to those from mobile elements and viruses. Such integrations would include insertional inactivation of coding genes, and transfer of oncogenic proteins and peptides (e.g. the B. fragilis metalloprotease). Some phenonmena found with mobile elements, like translocations, require multiple copies in the genome. Therefore, phenonmena like translocations are not likely to involve bacterial integrations unless there are multiple integrations or duplications of the same piece of bacterial DNA.

Should bacterial LGT be an additional mechanism used by specific pathogenic bacteria to promote carcinogenesis, specific vaccines could be developed to prevent cancer development, similar to that of HPV. The HPV vaccine is believed to decrease HPV-associated cancers by 70% [93] and serves as a great example of how a vaccine can be used to prevent cancer development caused by microbial insertional mutagenesis.

Future experiments examining bacteria-human LGT in cancer are necessary to validate these transfers and demonstrate their oncogenic potential. Effective validation methods include techniques used to validate mobile element and HPV integrations. For example, fluorescence in situ hybridization (FISH) is frequently used to visualize HPV integrations in human chromosomes [84, 129]. PCR amplification of the junction followed by sequencing is the standard experiment used to validate L1 insertions [46]. Southern blots are also frequently used to study mobile element break points [38, 77] and HPV integrations [26, 113] as they allow for detection of specific pieces of DNA. To evaluate the oncogenic potential of bacterial LGT in cancer, experimental approaches from the study of HPV could again be used. For example, expression constructs demonstrated that the presence of E6/E7 or the entire HPV genome in human cells could result in a cancerous phenotype [36, 48]. Finally, the recently developed CRISPR-Cas9 system, which enables efficient genome editing to modify specific genes in human cells and conduct in situ genetic screens [69], could be used to evaluate the impact of bacteria-human LGT.

5.3 Challenges of studying lateral gene transfer

Despite the evidence supporting the existence of LGT in animals, including humans, as well as the clinical benefits of identifying such an event, evaluation of such events has been challenging thus far. Most techniques used to identify foreign genetic material in a sample involve sequencing of PCR products, which carries the risk of contamination and detection of artifactual chimeric sequences. These chimeras would be formed in the ligation and/or PCR amplification during library construction [131, 132]. Unfortunately, it is hard to differentiate true LGT from chimerism, as sequence validation using PCR has the same risk of chimera formation [131, 132].

Another challenge in identifying LGT comes from assembling genome sequences. When genome assemblies are created they are often purged of reads that resemble bacterial sequences—either intentionally through foreign sequence screens or unintentionally through analysis of k-mers. In this situation it is possible that LGT from bacteria is present but never reflected in the final genome assembly [21]. Furthermore, modern mapping-based techniques that use BWA [63] and/or Bowtie [61] will miss bacterial reads if not included in the reference genome. Tumor heterogeneity is also a challenge for identifying LGT in cancer samples. The significance of sequencing data is based on the number of sequencing reads supporting that element, referred to as the sequence coverage [5]. If LGT does not occur early during clonal expansion it could be at such low levels that there may be no reads supporting it, or the coverage of the LGT may be so low that its validity is doubted.

5.4 Promising techniques to study lateral gene transfer

New sequencing technologies, such as Oxford Nanopore, will be extremely beneficial for investigation of LGT. As advertised, this new technology does not rely on amplifying DNA or on ligation of adapters [4, 114], which will significantly reduce the risk of generating artifactual chimeric sequences in sequencing data [4]. Another benefit of this technology is that the starting sample is conserved during the steps leading to sequencing [4]. This improvement allows for verification of initial findings by validating them on the same starting sample, an important aspect for confirming rare variants in a heterogeneous sample [4]. When this technology becomes available it will allow for a more thorough examination of LGT in animals, potentially leading to the identification of disease-causing integrations in humans.

6. Concluding remarks

HPV, HBV, and EBV integrations into the human genome are a major part of carcinogenesis in their associated cancers [83]. H. pylori and B. fragilis also contribute considerably to their respective cancers, but through different mechanisms [31, 50, 119]. Microbial LGT in cancer genomes has not been studied extensively enough to definitively determine the frequency that such integrations occur and cause functionally significant mutations. With this in mind, it is important to consider all viruses and microbes present in tumors and thoroughly investigate all potential mechanisms for mutagenesis. More work needs to be done using up-and-coming sequencing technologies in order to critically evaluate the role of LGT from carcinogenic infectious agents. Only when LGT between bacteria and humans is given more careful consideration can we establish its relevance, or lack thereof, in human carcinogenesis.

Mobile elements can cause cancer by inserting into cancer-related genes.

Viruses contribute to carcinogenesis by mutating and deregulating human genes.

Microbes induce chronic inflammation as well as toxins that are linked to cancer.

Viral and mobile element integration allude to microbial integration in the cancer genome.

New technologies may facilitate detecting microbial integrations in the human genome.