The potential role of infectious agents and pelvic inflammatory disease in ovarian carcinogenesis

Ovarian cancer (OC) is a major threat to female health, with a global prevalence of 239,000 cases in 2012 [1]. Due to the lack of symptoms in early stage OC, advanced disease is often present at the time of diagnosis [2].

Almost 90% of OC originates from epithelial cells, which have been thought to arise from the surface mesothelial lining of the ovaries. However, recent research has indicated that a proportion of serous epithelial OC could originate in precancerous lesions called “serous tubal intraepithelial carcinomas” (STICs) located in the fimbriated end of the fallopian tubes [3]. This is an important finding because previous theories of ovarian carcinogenesis have been unable to explain the fact that precancerous lesions have never been identified in the ovaries [3].

The female internal genitalia and the peritoneal cavity are accessible to outside pathogens through the genital tract. The mesothelial lining of the peritoneal cavity is identical in both sexes. However, primary serous peritoneal cancer, which may represent a continuum with serous OC [3], occurs almost exclusively in female patients [4]. This suggests that extrinsic factors could play a crucial role in ovarian carcinogenesis. Furthermore, epidemiological studies have demonstrated that patients with tubal factor infertility have a higher risk of OC [5]. Since the fallopian tubes are often affected by pelvic inflammatory disease (PID), it is therefore highly relevant to consider the potential role of infectious agents in OC. Previous reviews have focused on human papillomavirus (HPV) and OC, but this review is the first to include studies on all bacterial and viral infections as well as studies on pelvic inflammatory disease (PID).

To review the existing literature on OC and infectious agents, we performed a systematic literature search by means of a database search for English language articles published in the period from 1980 to 2016. Using the Pubmed research tool, the keyword: “ovarian neoplasms” was by means of the Boolean logical “AND”, combined with the following keywords: “Chlamydia infections” (MeSH), “Neisseria gonorrhoeae” (MeSH), “Mycoplasma genitalium” (MeSH), “Papillomaviridae” (MeSH), “pelvic inflammatory disease” (MeSH). Correspondent searches were performed in EMBASE and Web of Science. The initial search identified 935 articles. After removal of 130 duplicates, 805 studies remained for further analysis. Contents were outside the scope of the present review in 740 articles that were excluded. The remaining 65 articles were subjected to full-text screening. Original studies, examining the association between infectious agents and ovarian cancer by use of tissue-based or serologic methods were included. Epidemiological studies examining the association between pelvic inflammatory disease and ovarian cancer were also included. Case-reports and reviews were excluded, leaving a total of 40 articles to be included in the review. More detailed information on the search strategy is given in the “Additional file 1” to this article.

Table 1 summarises the viruses investigated in relation to OC, and their potential oncogenic mechanisms.

In relation to OC, previous research has focused on the potential association between Chlamydia trachomatis (C. trachomatis), Mycoplasma genitalium (M. genitalium), and Neisseria gonorrhoeae (N. gonorrhoeae) [13, 18, 57‐61]. An overview of the potential carcinogenic mechanisms of these bacteria is provided in Table 3.

Microorganisms that cannot directly induce cellular transformation may still play a role in OC oncogenesis due to the paradoxical effect of the host inflammatory response. It has long been recognised that the cells involved in innate and adaptive immune responses are recruited to the site of tumorigenesis. Here, they can release a multitude of tumour promoting inflammatory cytokines, chemokines, and ROS that can potentially facilitate tumour cell migration, metastasis, and angiogenesis [65]. Cancer has aptly been described as “a wound that does not heal [71]”. Several conditions associated with pelvic inflammation such as perineal talc use and endometriosis have been demonstrated to increase the risk of OC [72]. Likewise, ovulation is speculated to induce focal damage and inflammation on the ovarian surface epithelium, and factors that reduce the number of lifetime ovulatory cycles have been shown to reduce the risk of OC [73]. Consequently, it has been speculated that PID could play a role in ovarian carcinogenesis, and the potential association between PID and OC have been investigated in seven case-control studies (Table 4). The majority of studies relied on patient-reported outcomes and identified cases of PID through interviews or questionnaires. Risch et al. reported a statistically significant association between PID and OC (odds ratio (OR) 1.53; 95% CI 1.10–2.13) [74]. Moreover, they found a dose-response effect in relation to repeated episodes of PID, implying a higher risk of OC with increasing episodes of PID (OR 1.88; 95% CI 1.13–3.12). A prospective, population-based study from Taiwan verified the diagnosis of PID by the International Statistical Classification of Diseases and Related Health Problems codes registered in a national database [75]. After 3 years of follow-up, a higher risk of OC was observed in the case group (hazard ratio (HR) 1.92; 95% CI 1.27–2.92)). In this study, the association was also stronger when analysis was restricted to cases exposed to more episodes of PID (HR 2.46; 95% CI 1.48–4.09) [75]. However, a statistically significant association between PID and OC could not be confirmed in any of the remaining studies [72, 74, 76‐79].

The reviewed studies were very heterogeneous in terms of study population, study design, and the analysis methods used.

However, based on the reviewed studies, it is fair to conclude that high-risk HPV is unlikely to be associated with OC in Western countries. Interestingly, a higher prevalence was consistently reported in Asia and the Middle East. This finding is supported by previous reviews focusing on HPV and ovarian cancer [80, 81] and a potential association may exist in these regions. The geographic variation could be caused by varying potency of HPV strains. Thus, it has been demonstrated that latent infection of the cervix uteri with the non- European variant of HPV 16 is associated with a 2- to 9-fold higher risk of cancer or high- grade cancer precursors, compared to infection with the European variant [82]. Differing genetic predispositions between ethnical groups may also be a factor because polymorphisms of the TP 53 gene or the promoter of the tumour necrosis factor alpha gene have been associated with increased vulnerability to HPV oncogenes [83, 84]. Lastly, environmental or life-style factors could play a role. For instance, the global incidence of EBV infection among adults is estimated to be over 90%, and there is strong evidence linking EBV infection to the development of NPC [47]. Despite the ubiquity of infection, the incidence of NPC is very low in most regions. However, dramatically elevated rates are observed in certain parts of Asia, and dietary factors, such as the intake of salt-preserved fish, are thought to play a role [85]. This underlines that, even though infection with oncogenic viruses is common, cancer is a rare event that can arise from a combination of genetic, environmental, and lifestyle factors.

Most of the epidemiological studies found an association between OC and PID, but only two studies reported a statistically significant association [74, 75]. It is noteworthy, however, that these studies also identified a dose- response effect. This is one of the key points in Hill’s criteria for causation that state that a biological gradient must be present, with more exposure to the suspected agent leading to a larger effect [86]. These findings are supported by studies that found evaluated markers of inflammation such as CRP and interleukins to be significantly associated with an increased risk of OC [87]. A possible limitation to the reviewed studies is that they relied primarily on patient-reported outcomes, which may be subject to recall bias. Furthermore, due to the potential subclinical course of PID, patients can be unaware of previous episodes. This is supported by studies that have demonstrated a relatively high seroprevalence of chlamydia antibody titres in women that reported no history of PID, and this issue would tend to produce bias towards the null [88, 89].

Interestingly, CMV was reported in 50% of cases in a single study, but this finding requires confirmation in future and larger studies. No other viral agent was detected.

The only detected bacteria were C. trachomatis and M. genitalium. The results of this review do not support an association between C. trachomatis and OC, the majority of studies reporting negative results. The association between M. genitalium and OC was investigated and found in only a single study and more research is needed.

Several factors may account for the conflicting results. First of all, there may be no connection between the investigated infectious agents and OC. Secondly, the heterogeneity of analysis methods, study populations, and study designs may play a part. However, other explanations must be discussed. Importantly, it is unclear to what extent signs of microbiological agents are present in a tumour that arises decades after infection. For some oncogenic viruses, like high-risk HPV, expression of viral oncogenes is obligate for the maintenance of the malignant phenotype [90]. In these cases, viral genes will remain integrated and detectable in tumour cell DNA. But another potential mechanism is debated in the “hit and run” hypothesis. Here, integration of viral transforming genes and potential epigenetic reprogramming of host cells initiate tumorigenesis, but the viral gene expression is subsequently lost during neoplastic development [91]. Accordingly, it has been demonstrated that CMV oncogenes IE1 and IE2 can cooperate with adenovirus E1A gene to transform rat kidney cells. However, expression of the transforming oncogenes was absent in the clonal cell lines from the transformed foci [92].

Finally, other oncogenic agents act more indirectly by inducing chronic inflammation that may function as an initiator or promoter of carcinogenesis [93]. Consequently, the involved pathogen may no longer be present when cancer is diagnosed. For instance, despite the established role of HCV in hepatocellular carcinoma, only a minority of transformed hepatocytes contain viral RNA [94].

These examples highlight the fact that the abscense of microbiological agents in tumour tissue does not rule out their potential role in the transformation of host cells. These issues seem to strengthen the need for serologic studies. However, serologic methods are limited by the fact that antibody levels are likely to decline over time. In a study from Finland, 43% of women had declining C. trachomatis antibody titres 6 years after the diagnosis of PID [95]. Moreover, not all patients with prior chlamydial infection will have detectable antibodies [96].

These obstacles call for a revised strategy in the attempt to uncover the potential role of infectious agents in OC. It is fair to assume that signs of microbiological presence will be easier to detect when the interval between primary infection and investigation is short. Therefore, it seems more relevant to investigate precursor lesions for the presence of potential microbiological agents [91]. Until recently, this has been impossible due to the lack of a clearly defined premalignant lesion to OC. However, this has changed because increasing evidence indicates that STIC lesions in the distal fallopian tubes are precursor lesions to serous epithelial OC, which constitutes the majority of OC cases [3]. Yet, no explanation has been formulated as to why STIC lesions arise. Since the fallopian tubes are often affected and damaged by chronic PID, it is biological plausible that pathogens with transforming capacities, or the chronic inflammation they induce, could lead to subsequent neoplastic transformation. We therefore recommend that future studies focus on the detection of bacterial or viral agents in fallopian tube tissue samples with STIC lesions verified through the “Sectioning and Extensively Examining of the Fimbriated end” protocol.

The reviewed articles indicate a potential association between PID and OC. An association between OC and high-risk HPV genotypes may exist in Asia, whereas an association in Western countries seems unlikely due to the low reported prevalence. Two studies reported an association between bacterial agents and OC, but the majority of bacterial studies reported negative findings. However, the available literature was scarce and more studies are warranted.