The vaginal microbiota, human papillomavirus infection and cervical intraepithelial neoplasia: what do we know and where are we going next?

Cervical cancer, the commonest infection-associated neoplasm, and its premalignant precursor cervical intraepithelial neoplasia (CIN), are caused by strains of the human papillomavirus (HPV). Over 100 subtypes of HPV exist with 13 being identified as high risk (high-risk HPV; hrHPV) and causal of cervical cancer in 100 % of cases [1]. HPVs-16 and -18 are the most oncogenic and prevalent of these and are responsible for around 70 % of cases [2]. The peak age for infection in girls is around 20 years. It is estimated that 80 % of sexually active women will have been infected at some point by age 50 [3]. Over 90 % of HPV infections are transient, being cleared by an incompletely understood immune response within 6–18 months [4], although re-infection with the same or different HPV subtypes can occur [5]. Persistence of the virus is essential for development of high-grade CIN and cervical cancer and factors that correlate with higher persistence rates include age, immunodeficiency, smoking, oral contraceptives and Chlamydia trachomatis infection. Emerging evidence indicates that cervicovaginal microbiota plays a substantial role in the persistence or regression of the virus and subsequent disease. This review will summarise this evidence, discuss possible mechanisms linking vaginal bacteria with cervical pathology and finally consider the potential for future therapeutic strategies.

Bacteria account for 50 % of the cells of the human body, and together with archaea and lower eukaryotes are collectively termed ‘human microbiota’ [6, 7]. Due to the limitations of culture-based techniques [8], the composition of microbiota in different body compartments is increasingly studied through the use of next-generation sequencing (NGS) techniques. This may involve shotgun metagenomic sequencing of all of the DNA in a biological samples (human and bacterial) but most commonly involves amplifying, sequencing and analysing specific regions of bacterial 16S rRNA genes, although other rRNA genes (18S for eukaryotic microbes) or genomic regions (for viruses) can be used. A variety of bioinformatics tools and platforms are used to assign resulting sequences to specific microbial taxa at different taxonomic levels as well as for in-depth phylogenetic analysis of microbial community structure. There are several excellent reviews covering these topics in greater depth [6, 9, 10], which is beyond the scope of this review.

The link between health, disease and the human microbiota is a fast-moving and contentious area of research, and an appreciation of the variation in microbiota composition amongst individuals is expanding our understanding of the pathophysiology underlying a variety of diseases affecting many body systems, from colorectal cancer [11, 12] to atopic dermatitis [13]. In the majority of human body sites to be examined to date, highly diverse microbial communities are generally considered a signature of health [14, 15]. However, in the case of the female reproductive tract, health is more commonly associated with low microbial diversity and dominance by only one or a few species of Lactobacillus [16–18]. Ravel and co-workers demonstrated that this concept is broadly observed in the majority of reproductive age women; a significant proportion harbour comparatively diverse vaginal bacterial communities [17]. In their study, vaginal samples collected from 396 ethnically diverse reproductive age women were analysed using Roche 454 FLX-based sequencing of bacterial 16S rRNA genes. In total, 282 taxa were identified and, using hierarchical taxonomic clustering, the vaginal microbial profile of each woman could be classified into a total of five ‘community state types’ (CSTs), which have subsequently been used by numerous other studies [16, 19, 20]. CSTs I, II, III and V are characterised by dominance of Lactobacillus crispatus, L. gasseri, L. iners and L. jensenii respectively and tend to have low species diversity and evenness. In contrast, CST IV is typically devoid of Lactobacillus spp. and instead enriched with strict anaerobic species often associated with bacterial vaginosis (BV) including Gardnerella, Megasphera, Sneathia and Prevotella. As will be discussed below, the structure of the vaginal microbiota (VMB) is dynamic and transitioning between CSTs can occur. In the vagina, the most common transition observed is from CST III to CST IV [21], which suggests that L. iners may be less able to inhibit colonisation of strict anaerobes and pathobionts compared to other Lactobacillus spp. [21] or because L. iners appears more capable of surviving and adapting to a wide range of pH and other metabolic stress-related conditions due to the constitutive and inducible expression of genes not seen in other lactobacilli [22, 23].

BV is a polymicrobial disorder characterised by a microbial community structure consistent with CST IV; that is, by diminished levels of Lactobacillus spp. with associated overgrowth of anaerobes, particularly Gardnerella, Prevotella and Peptostreptococcus species [24]. The prevalence of BV is around 9 % in the UK [25] and up to 29 % in the US [26]. The Hay/Ison criteria used for diagnosis of BV in the UK is largely based on the presence or absence of both Lactobacillus and Gardnerella or Mobiluncus morphotypes [27]. BV has previously been correlated with higher incidence, prevalence and persistence of HPV infection and with development of CIN [28–31]. However, other studies did not find a positive correlation between CIN and BV [32–34], which may partly be explained by the lack of objective diagnostic tests for BV, the reliance on subjective scoring systems [35] and the heterogeneity of BV itself.

Recent observational cross-sectional studies support the concept that CSTs III and IV, in particular, are frequently linked with the presence of HPV infection and development of preinvasive cervical disease states [52–54]. While microbial diversity is considered to be a sign of health in many body sites, highly diverse vaginal microbiota are often considered atypical or a state of dysbiosis and are associated with disease states. However, there is a lack of investigation into how exactly the vaginal microbiota could play a role, and further mechanistic studies are warranted. Vaginal Lactobacillus spp. prevent colonisation of bacterial vaginosis-associated bacterial species through maintenance of a low pH [60–63] and bacteriocin production [64–66]. This is important for maintenance of the cervical epithelial barrier function that inhibits entry of HPV to the basal keratinocytes [67]. When BV-associated strict anaerobes are able to colonise, they produce enzymes and metabolites, which may compromise this barrier, facilitating HPV entry. They also act on several cellular pathways that can enable a persistent, productive viral infection and subsequent disease development and progression [68–72]. Indirect evidence generated by existing mechanistic studies in vivo and in vitro in complementary fields and models of BV, as well as studies of other viral genital infections, supports potential mechanisms that warrant further investigation.

Cellular targets and a role for specific bacterial species

It is currently unclear if dysbiotic vaginal microbial communities act in synergy with HPV to manipulate its cellular targets such as p53, pRB, survivin and hTERT [100] or whether this occurs independently. However, evidence points towards the likelihood that particular species have a pathological role in HPV acquisition and persistence, rather than dysbiosis as a whole. G. vaginalis is commonly found in CST IV [17] and often presents at relatively high relative abundance in the adolescent vagina [40]. The immature adolescent cervix is known to be more susceptible to HPV infection compared to older women, and this correlates positively with the rate of squamous metaplasia [101]. It is plausible that the higher levels of G. vaginalis may play a role during this period of greater susceptibility. Sneathia spp. have frequently been identified in association with HPV positivity [52] and with CIN and cervical cancer [54, 55]. Furthermore, Sneathia spp. have also been associated with adverse obstetric outcomes including miscarriage and preterm labour [102, 103]. Previously, Nawrot and colleagues used a PCR-based method to show Leptotrichia amnionii (now re-named Sneathia amnii) is associated with cervical cancer but not HPV infection or CIN [104]. The species was, however, not unique to women with cancer, suggesting that it may play a role in carcinogenesis, rather than occurring as a consequence of the disease. Sneathia spp. belong to the Fusobacterium genus, which has further been implicated in colorectal carcinogenesis [12, 105, 106] through activation of proinflammatory pathways and inhibition of immunocytotoxicity [107]. Fusobacterium spp. produce FadA, a virulence factor, which is capable of activating the WNT signalling pathway, a key cell survival and proliferation pathway which is found to be dysregulated colorectal carcinogenesis [108] and cervical cancer [71]. Fusobacteria are also implicated in the modulation of immunomodulatory pathways. For example, F. nucleatum DNA levels have been shown to be inversely proportional to CD3+-T cell counts in colonic mucosa [109], which is interesting given that aberrant CD3+-T cell signalling and function has been observed in cervical cancers [72] and relapsing disease [110] (see Fig. 1).

Probiotics are defined by the World Health Organisation as ‘live microorganisms that, when administered in adequate amounts, confer a health benefit on the host’. They have been used successfully as an adjunct to traditional antibiotics in BV to improve cure rates and prevent recurrence through their ability to replenish the depleted pool of Lactobacillus spp. [111].

Metronidazole and clindamycin are the most commonly prescribed antibiotics for BV. They target the overgrowth of anaerobes but do not appear to have a mechanism of action that would actively promote re-establishment of Lactobacillus spp. Following successful metronidazole treatment for BV, classified as improvement in Nugent and Amsel’s clinical scoring systems, L. iners is often seen to be the predominant re-colonisation species [112], which may arise through expansion of an existing population of this species. An oral preparation of Lactobacillus rhamnosus GR-1 in combination with L. reuteri RC-14 was shown to increase the prevalence of Lactobacillus-dominant vaginal microbiota, as well as improve BV cure rates when given in combination with metronidazole [113]. As neither of these are endogenous vaginal species, the study provides evidence that oral administration of Lactobacilli is capable of modulating vaginal microbial structure. These orally administrated bacteria are thought to reach the vagina through a poorly understood mechanism involving transition via the anus and the perineal and vulval skin [114]. A potential mechanism by which these species modulate community structure involves L. reuteri secretion of the bacteriocin, Gasserin [66], and L. rhamnosus produced Lactocin 160, a similar bacteriocin that is particularly active against G. vaginalis. Interestingly, this particular preparation has also been associated with increased relative abundance of L. iners as shown using NGS [115]. In a study of bacteria isolated from premenopausal women, PCR-based techniques demonstrated that L. gasseri negatively associated with L. iners and A. vaginae [116]; two species which co-associate [116] and are suggested to pose an intermediate and high risk for development of CIN respectively [56]. This observation provides further evidence that L. iners is not always associated with health, and this should be taken into consideration when designing probiotic therapies, to ensure that they do not promote dominance of this particular Lactobacillus. Probiotics have also been suggested as an intervention to promote HPV clearance, and in vitro and in vivo evidence exist to support this technique. Treatment of SiHA cells, an HPV-16-infected cervical cell line, with Bifidobacterium adolescentis significantly reduced the production of E6 and E7 mRNAs, suggesting that this species may represent a novel therapeutic of virally transformed cells [117]; however, the efficacy of this species as a probiotic is yet to be proven in humans.

As described above, L. gasseri is associated with rapid clearance of incident HPV infections [53]. This species, along with L. crispatus, has also been shown to be cytotoxic to HPV-18-infected HeLa cervical cancer cells but not to normal cervical cell lines, independent of pH or lactate concentration, suggesting a more sophisticated mechanism of action [85, 118]. Furthermore, a semi-randomised, interventional study of 54 HPV-positive women with low-grade cervical lesions showed that women treated with oral L. caseii showed greater clearance of HPV infections (29 vs 19 %) and were significantly more likely to clear their cervical lesion (60 vs 31 %), compared to an untreated cohort [119].

Prebiotics are indigestible carbohydrates, which include the fructo-oligosaccharide (FOS) and gluco-oligosaccharide (GOS) families, which promote the growth of healthy bacteria already present in the body. They have been most well studied in the gastrointestinal tract, where they have been shown to modulate microbiota composition, as well as exert immunomodulatory effects independent of the microbiota (reviewed in [120]). Several encouraging in vitro studies, and a handful of small in vivo studies, suggest proof of concept in the vagina. FOS and GOS have been shown to promote the growth of L. crispatus, L. jensenii and L. vaginalis in vitro but not Candida albicans, Escherichia coli or G. vaginalis, with the investigators using high-performance liquid chromatography (HPLC) to show that GOS and FOS could not be used as energy sources by the three latter pathobionts [121]. GOS, applied as an intravaginal gel, has been shown in a randomised controlled trial of 42 women, immediately following metronidazole treatment for BV, to result in a significant reduction in Nugent scores at 8 and 16 days of treatment [122]. Konjac glucomannan hydrolysates (GMH) have similarly been shown to promote Lactobacilli spp. colonisation in women with C. albicans infection [123]. Beyond promotion of bacterial growth through acting as a growth substrate, mannose and GMH have been shown to inhibit adhesion of E. coli to human cheek epithelial cells in vitro [124], suggesting additional mechanisms of pathobiont inhibition. When concomitantly administered with probiotics in a synbiotic preparation, they may enhance the growth of probiotic species as well as their bacteriocin production [125]. These results are encouraging and represent a very cheap, safe intervention with few side effects for a disease that cannot otherwise be treated without the risk of significant reproductive and obstetric morbidity [126–130]. While further studies are required to both understand the mechanisms by which the vaginal microbiota plays a role in the pathophysiology of cervical disease, and to identify the most protective species or strain to defend against HPV-induced dysplasia and neoplasia, and their therapeutic doses, pre- and probiotics may offer a practical intervention for the developing world, where cervical cancer is a major cause of female cancer-related mortality [131].

The ability to derive a causal link between vaginal microbiota and HPV infection and CIN/cervical cancer is limited by the cross-sectional nature of most studies undertaken in this area. This difficulty is further compounded by the slow natural history of the disease, with time from acute HPV infection to high-grade CIN taking years to decades. In addition, numerous other confounders can impact results including smoking [46] and vaginal intercourse without the use of barrier contraception [47], which have both been associated with depletion of Lactobacillus spp. The impact of other sexual practices such as oral intercourse, use of lubricants and having multiple sexual partners is poorly understood, and the information gathered in the current observational studies is very heterogeneous as highlighted in Table 1. This information may represent major confounders affecting the VM composition as well as the presence of oncogenic HPV infections and should be clearly documented in future reports.

Furthermore, many of the published studies describe the VM in relatively small cohorts with absent or limited representation of a sizeable group of normal, HPV-negative controls for the described comparisons. Oh and colleagues used samples collected from women with both low-grade and high-grade preinvasive disease and as controls grouped women with normal and ASCUS cytology irrespective of their HPV status [56], including women that may harbour underlying higher grade disease [132]. Piyathilake and colleagues only compared hrHPV-positive patients with CIN1 (non-cases) to those with CIN2 and CIN3 (cases) and lacked a healthy control population [57]. Audirac-Chalifour and colleagues defined as normal controls women with negative cytology and colposcopy, irrespective of HPV status [55]. The study by Mitra and co-workers was the only to group the compared populations according to cytology, histology and HPV status. The authors present subgroup analyses including HPV negative women with normal/ASCUS/LSIL cytology separate to those positive for hrHPV, although the samples in those subgroups were small [54]. Studies must therefore be appropriately designed to permit accurate interpretation of data and ensure any observed changes in vaginal microbial communities are directly associated with the pathology. Appropriately stored samples in historical biobanks are thus a precious commodity, which may present an opportunity to perform longitudinal studies to answer these kinds of questions. There are however limitations as these biobanks often use long interval sampling designs. Prolonged intervals between samples may fail to describe rapid changes in VMB composition occurring throughout the disease process, while important meta-data such as sexual history and smoking may be lacking.

A number of recent studies have led the scientific community to re-think the traditional conceptual model of the natural history of HPV. Previously regarded as a viral infection which simply causes transient infections or persists as a chronic infection based on studies conducted with protracted sampling intervals, it has now been shown that HPV status can fluctuate quickly over a short time period, based on studies involving very frequent testing over a short space of time [53, 133]. Whether this is due to detection and re-detection of low-level persistent infections due to wavering loss and regained immune tolerance, rather than true re-infection, is currently unclear. A small study of HPV status in 20 women has shown that HPV detection peaked at days 7–11 of the menstrual cycle [134], which are the days immediately following the time of highest VM diversity [41]. This suggests that it may be possible to correlate fluctuations in HPV status with the changes in the VM occurring as a result of the menstrual cycle to explain in role of the VM in the emerging, rapidly dynamic model of infection. However, these studies have been performed in normal, presumably healthy women, and to our knowledge the temporal dynamics of HPV status over a short period of time has not yet been studied in women with known CIN. The current theories behind the VM in cervical disease will require re-evaluation if similar HPV dynamics are observed compared to that of healthy women. Furthermore, integration of alternative HPV tests for detecting DNA, such as mRNA and E6/E7 levels, into future microbiome studies may help us begin to answer these important questions. In addition studies, focus on the interplay between the microbiota and the host immune system within these alternative conceptual models on the natural history of HPV infection is also required. It has recently been suggested that the immune system contributes as little as 20 % towards viral clearance, and that stem cell stochasticity plays the biggest role, based on the integration of epidemiological data with mathematical cellular modelling [135]. This model may also explain the concept of latency and thus fluctuating HPV status. How the VM may influence the stochastic dynamics of basal epithelial cells however is unknown.

Alternative approaches for the assessment of vaginal microbiota and its interactions with the host immune system may also offer useful means to monitor HPV acquisition, persistence and subsequent cervical dysplasia and neoplastic transformation. For example, metabonomics, defined as ‘the quantitative measurement of the dynamic multiparametric metabolic response of living systems to pathophysiological stimuli or genetic modification’ [136], is emerging as a novel way to investigate the host-microbe interaction through inspecting functional metabolic changes associated with disease phenotypes [137, 138]. Using nuclear magnetic resonance (NMR) or mass spectroscopy (MS) coupled to separation technologies, e.g. HPLC, it is possible to identify particular metabolites or pathways that are altered in association with the VMB structure. This approach will not only increase our understanding of the impact of bacteria on host biochemical and immune response, which is likely extremely complex [138], but may also present the opportunity for development of novel prognostic tests for triage of patients who are most likely to develop a high-grade or cancerous cervical lesion. The identification of such patients is one of the biggest clinical challenges in current colposcopy practice.

By definition, members of a particular bacterial species have a total nucleotide identity of >70 % across their genomes [139]. The remaining genome diversity gives rise to the existence of different strains, which may have different functional genes that induce different biological properties. Further evidence is required in order to determine whether only certain strains of a particular bacterial species are either protective or pathogenic with regards to HPV and cervical dysplasia, and a recent study by Abdelmaksoud and colleagues [140] hints this is very likely. The team compared strains of L. crispatus that colonised women with, and without BV, demonstrated considerable genomic diversity within the species and identified several genes exclusive to the presence or absence of BV [140]. These genes require further investigation to help understand the protective mechanisms exploited by certain bacteria, and other disease-associated species should also be studies using metagenomic techniques, which in turn will support development of appropriate probiotic preparations.

Afro-Caribbean women have a fourfold higher prevalence of CST IV VMBs compared to Caucasian and Asian women [17]. They have also been found to have a higher age standardised rates of cervical cancer (6.3–11.2 per 100,000 women), compared to Caucasian and Asian women (8.2–8.7 and 3.6–6.5 per 100,000, respectively) in the UK [141] and the USA [142]. Although there may be a slightly higher risk of invasive disease, this does not correspond to the increase in the prevalence of CSTIV in black women, suggesting the presence of a far more complex mechanism and interactions between the bacteria and the individual host beyond the simple presence of CST IV that promotes or not HPV persistence and cervical carcinogenesis. Future studies will help to further explore these possible associations in long-term samples and comprehend why particular species or VM CSTs may be associated with health in some but disease in others. Compliance with cervical screening and many behavioural and social factors may also explain the higher risk of invasive disease in this ethnic group.

Pre- and probiotics clearly present an enticing novel therapeutic approach to this disease, because they are cheap, easy to administer, with a low side effect profile, unlike the current gold standard treatment for high-grade CIN, which involves a surgical method that carries significant risk to future reproductive outcomes [126–130]. Furthermore, pre- and probiotic use would have an impact in other areas of women’s health, with dysbiosis being responsible for a two- to fourfold increase in risk of preterm birth [143], increased risk of miscarriage [143] and increased rates of HIV transmission [144], which highlights the importance of investing time and resources into exploring this therapeutic strategy.

The vaginal microbiota appears to play a role in the acquisition and persistence of HPV in the human vagina and in the subsequent development and progression of CIN. There is a need for further longitudinal studies to prove that these disease outcomes are influenced by VMB composition. This information may present the opportunity for development of novel therapeutic agents in the form of probiotics, to prevent HPV infection, promote its clearance in infected women and negate the risk of cervical dysplasia and future adverse reproductive outcomes that are associated with the current treatment methods [126–128]. Mechanistic studies are required to identify the most protective species. Furthermore, it is possible that only certain strains of a bacterial species are able to protect or promote disease processes.

Alongside the bacterial microbiota, the virome is now a new emerging area of interest. Although we have known for many years that HPV is the aetiological agent in precancerous and cancerous pathologies of the cervix and lower genital tract, other viral genera present in the normal vagina, alongside papillomaviridae, may be involved in disease progression [145]. Furthermore, we are aware of a symbiotic relationship between bacterial and viral communities, which requires further investigation specific to HPV and cervical pathology.