Introduction

Approximately 600,000 new cases of invasive cervical cancer (ICC) are diagnosed annually, and over 50% of these women (311,000) die within one year of diagnosis.(1) The high burden of HIV in sub-Saharan Africa and evidence that women living with HIV have 6 times greater risk of develo** ICC compared to women without HIV, calls for an urgent focus on prevention and control of cervical cancer in these high-risk populations [1]. Furthermore, epidemiologic studies have confirmed that ICC is primarily caused by human papillomaviruses (HR-HPV) [2] whose persistence and oncogenic potentials are enhanced by HIV-mediated decreases in cellular immunity with other co-factors playing roles in cervical carcinogenesis [3,4,5,6,7,8,9,10,11,12,13].

To improve cervical cancer prevention and control, the WHO launched the cervical cancer elimination initiative in 2018, recommending HPV vaccination, early detection of precancer through screening with a high-performing HPV test, and treatment of precancer and early invasive cancer [14]. The HIV and HR-HPV are sexually transmissible viruses and their attribution to cervical carcinogenesis has led to experts ascribing cervical cancer as a sexually transmissible cancer, [15] and previous study reports have shown that cervico-vaginal microbiome play a significant role in the process [10, 16,17,18]. Additionally, natural history studies of cervical carcinogenesis have expanded our understanding of the potential mechanism by which vaginal microbiome alone or in conjunction with local pro inflammatory cytokines act either directly or mediated through HR-HPV persistence in advancing progression of cellular changes in a normal cervix (NILM), low-grade precancer (LSIL) to high-grade precancer (HSIL), and eventually invasive cervical cancer stage [19,20,21,31].

Microbial community characterization

A two-stage PCR protocol was used on the extracted microbiome DNA to amplify the V3-V4 variable region of bacterial 16S rRNA genes, as described previously [32]. Briefly, gDNA was used as template for PCR amplification with the primers CS1_357wF [33] and CS2_806R [34] (ACACTGACGACATGGTTCTACACCTACGGGNGGCWGCAG and TACGGTAGCAGAGACTTGGTCTGGACTACNVGGGTWTCTAAT, respectively; linker sequences underlined). Reactions were performed in 10 µl volumes using repliQa HiFi ToughMix (QuantaBio). Cycling conditions were 2 min denaturation at 98°C, followed by 28 cycles of 98°C for 10 sec, 50°C for 1 sec, and 68°C for 1 sec. Subsequently, a second PCR amplification was performed, also in 10 microliter reactions in 96-well plates using repliQa HiFi ToughMix. Each well received a separate primer pair with a unique 10-base barcode, obtained from the Access Array Barcode Library for Illumina (Fluidigm, South San Francisco, CA; Item# 100–4876). One microliter of PCR product from the first stage amplification was used as template for the 2nd stage, without cleanup. Cycling conditions were 98°C for 2 minutes, followed by 8 cycles of 98°C for 10”, 60 °C for 1” and 68 °C for 1”. Libraries were then pooled and sequenced with a 15% phiX spike-in on an Illumina MiSeq sequencer employing V3 chemistry (2 × 300 base paired-end reads). Library preparation and sequencing were performed at the Genomics and Microbiome Core Facility (GMCF) at Rush University.

Quality control and taxonomic identification and classification into community state types (CSTs) was conducted by University of Maryland Institute for Genomic Science, as described previously [35]. [36]. CSTs were determined by the VALENCIA algorithm implemented by the University of Maryland Institute for Genomic Science, which uses a distance-based metric to classify each sample to a CST based on the similarity of the sample to the centroid of CSTs identified in a reference set. In brief, dominant taxa by CST are as follows: CST-I, L. crispatus; CST-II, L. gasseri; CST-III, L. iners; CST-IV, diverse; CST-V, L. jensenii. [36].

Data management

All data were entered in REDCap and exported as CSV file for coding and subsequent analysis on Stata version 14.1, College station, Texas, USA. The setting and use of REDCap for managing our research data has been reported in an earlier publication [37].

Stratification of high-risk HPV status

The Seegene AnyplexTM II HPV28 Detection kit detects 28 human papillomavirus types [31].

High-risk HPV (HR-HPV) types include: HPV16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 69, 73, and 82. The low-risk types HPV (LR-HPV) include: HPV6, 11, 40, 42, 43, 44, 54, 61, and 70. The HPV status were categorized as follows: category 1 as HPV negative, category 2 as low-risk risk HPV, category 3 as single high-risk HPV (any of the HR-HPV), and category 4 as multiple high risk (2 or more of the HR-HPV types). For subsequent statistical analysis we recoded HPV types into category 1 as HPV negative; category 2 as low-risk HPV positive, and category 3 as any high-risk HPV positive (single or multiple).

Statistical analysis

We compared baseline socio-demographic, sexual risk characteristics, CSTs, HPV status, and HIV status between cervical pathology categories using the Pearson’s chi squared or Fisher’s exact tests for categorical variables and ANOVA for normally distributed continuous variables. To assess the association between vaginal microbiome CSTs with grades of cervical pathology, we considered the clinical relevance of the four cervical cytology categories. We combined NILM and LSIL as one group (normal or minor grades dysplasia) while HSIL and ICC were combined as the clinically important group that requires immediate evaluation and treatment. Observations with CST-II (n = 4) were excluded (none was CST-V) from analyses as being too sparse for inference. We performed bivariable logistic regression for the primary exposure variable (CST) and other predictor variables using robust logistic regression to obtain unadjusted odd ratios, 95% confidence intervals (CI), and Wald p-values for associations between HSIL/ICC and NILM/LSIL. We used backward stepwise selection method to build final models checked by the Akaike’s Information Criteria for model selection, including age in years, HIV status, HPV status, CSTs, years of completed education, and total number of births (parity). While tobacco use has been shown to be an important co-factor in cervical cancer risk, only two participants in our sample reported ever smoking, precluding its inclusion in analysis. We then modelled the analyses stratified by high-risk HPV versus negative for HPV (the numbers for low-risk HPV were too small for stratification in the model) to assess the possible effect of CST on cervical pathology probably mediated by high-risk HPV, a known epidemiologic factor in cervical carcinogenesis [2, 38].

Results

One hundred and fifty-five (155) participants were enrolled, 151 with microbiome data for this analysis (four excluded for having CST-II). The median age of the participants was 52 years (IQR:43–58), 47.7% were HIV positive, and 58.1% had HSIL/ICC. The HR-HPV genoty** results were available in 138 of the participants (the HPV sample of 13 participants were not processed for PCR genoty** and were excluded in all analysis with HR-HPV as an outcome). Of the 138 with HPV data, 56 (40.6%) were negative for HPV, 14 (10.1%) had low-risk HPV, 37 (26.8%) had single HR-HPV, and 31 (22.5%) had multiple HR-HPV types. Of the 31 participants with multiple HR-HPV types, 16 or 18 with any other HR-HPV type were seen in 14 (45.2%), 35 and other HR-HPV types in 4 (12.9%), while other mixed infections other than 16/18/35 were 13 (41.9%). The overall prevalence of any HR-HPV type (single and multiple) was 49.3%, with a higher proportion in women with HSIL/ICC (NILM 31.6%, LSIL 46.5%, HSIL 40.8%, and 81.5% ICC; p = 0.007).

Women with high grade cervical precancer and ICC were on average older than women with NILM and LSIL (Table 1). The distribution of CST showed that 101 participants (66.9%; 95% CI: 58.9–74.0%) had CST-IV. CST-III (L. iners dominated) was found in 33 participants (21.9%; 95% CI: 15.9–29.2), while 17 participants had CST-I VMB (11.3%; 7.1–17.4). The distribution of CSTs did not differ significantly by cervical pathology (p = 0.153, Table 1), though cervical cancer diagnosis was more common among women with CST-IV (Table 2). The association of other factors with CSTs as outcome have been summarized in Table 2. HPV status did not significantly differ by CST, though CST-I was more likely among women with HIV infection, and CST-IV was more likely with increased parity.

Table 1 Characteristics of participants in comparison to cervical pathology outcomes (N = 151)
Table 2 Distribution of participant characteristics by vaginal microbiome community state types

The distribution of both low and high-risk HPV types in the study sample is shown in Fig. 1 with HR-HPV 16 being the most common. Women with HIV were also significantly more likely to have HSIL/ICC (70.2% vs. 29.3% for women without HIV, p-value = 0.001).

Fig. 1
figure 1

Frequency distribution of type-specific HPV genotypes at enrollment (HPV-16 is the commonest, then types 56,58, 18, followed by 45, 51, 52, and 53)

The unadjusted logistic regression (Table 3) showed that age in years (OR = 1.05; 95% CI: 1.01–1.08), positive HIV status (OR = 2.67; 95% CI: 1.38–5.20), and total births (parity) > 4 (OR = 5.94; 95% CI: 1.83–19.3) were significantly associated with HSIL/ICC, while completion of 7–12 years of education (OR = 0.28; 95% CI: 0.09–0.88), or more (OR = 0.29; 95% CI: 0.10–0.85) had significantly lower likelihood for HSIL/ICC. In the final multivariable logistic regression model (Table 3), adjusting for age, HR-HPV status, HIV status, CST, parity, and years of completed education, women with HIV had over three times the odds (aOR = 3.21). Women with any HR-HPV had 2.72 times greater odds of having HSIL/ICC.

Table 3 Results of Logistic Regression: Unadjusted and Adjusted odd ratios of the association of various exposure variables with high grades/invasive cervical cancer

To further understand the possible associations between CSTs and HSIL/ICC, we stratified our multivariable logistic regression models by HR-HPV status (Table 4). Among women with HR-HPV detected, the effect of HIV was strengthened with aOR = 14.1 (95% CI: 2.34–85.2). It was also observed that the association between CST-III/CST-IV and HSIL/ICC were accentuated in the presence of HR-HPV, although this finding did not reach statistical significance: CST-III (aOR = 6.69; 95% CI: 0.67–66.6) and CST-IV (aOR = 3.64; 95% CI: 0.63–20.9).

Table 4 Results of Multivariate logistic regression Stratified by High-risk HPV status

Discussion

This study aimed at understanding the predominant CSTs in Nigerian women and to estimate the associations with cervical precancer and invasive cervical cancer, HIV, and HPV infection found that CST- III and CST-IV vaginal microbiome (VMB) communities were more frequent among women with high-grade cervical precancer and ICC, especially in the presence of HR-HPV. In agreement with a previous report in Nigeria [28], we found that CST-IV was the most frequent in the study sample and did not differ significantly by HR-HPV status, although the former report did not relate the CSTs and HR-HPV to cervical pathology as in our study. Our analyses also confirmed strong associations between HIV/HR-HPV and high-grade cervical precancer or invasive cervical cancer found in several prior studies [2, 39]. Our data showed that women with HIV had 3.2 times greater odds of high-grade cervical precancer and invasive cervical cancer compared to women without HIV. Similarly, we found that the detection of any HR-HPV types increased the odds of high-grade cervical dysplasia and ICC by at least 2.7 times.

In the presence of any HR-HPV types and compared to CST-I (L. crispatus dominated), CST-III (L. iners dominated) and CST-IV (diverse, non-optimal bacterial species) vaginal microbiomes were associated with high-grade precancer or invasive cervical cancer in a large systematic review and network meta-analysis [40]. Specifically, the meta-analysis showed that vaginal microbiota dominated by non-Lactobacilli species (CST-IV) or L. iners were associated with 3–5 times higher odds of detecting any prevalent HR-HPV and cervical precancer or invasive cervical cancer compared with a L. crispatus-dominated state [40]. Also a study report among South African women assessing incident HR-HPV, vaginal dysbiosis and incident cervical intraepithelial neoplasia grade 2 or more (CIN 2+) found that the acquisition of HR-HPV changes the vaginal microbiome and that anaerobic dysbiosis risk seem to increase concurrently with development of CIN2 + suggesting that the role of the vaginal microbiome composition in CIN2 + development could be mediated by persistent HR-HPV infection [41]. Our findings are therefore in agreement with the literature implying that a better understanding of the cervico-vaginal microenvironment could be a key step for HPV viral clearance [42]. This mechanistic role of the cervico-vaginal microbiome community state types is critical and could be explored in develo** therapeutic targets for cervices already infected with the HR-HPV types but yet to advance the process of cervical carcinogenesis given the high prevalence of these common sexually transmissible viral infection in the general population. Currently, the primary HPV preventive vaccines are not readily available or affordable in most LMICs making the search for therapeutic options for cervical HPV infection a promising effort.

Furthermore, the associations between CST and the two sexually transmissible viruses - HR-HPV and HIV - have public health implications for cervical cancer prevention. In 2018, the World Health Organization launched a global initiative to eliminate cervical cancer as a public health problem by adopting highly effective evidence-based interventions 2030 [14]. These interventions includes increasing the population-level coverage for HPV vaccination of 90% of eligible young girls between 9 and 15 years; screening of 70% eligible women by 35 years and 45 years, and effective treatment of 90% women with precancer and invasive cervical cancer by 2030 [14]. As discussed earlier, most countries in LMICs including Nigeria do not have a national program on HPV vaccination [43], making interventions for therapeutics, and HIV prevention, and to maintain a healthier cervico-vaginal environment effective and plausible public health approaches for cervical cancer prevention and control efforts. Indeed, some of the known epidemiologic factors for cervical cancer [15, 24, 44,45,46,47,48,49,50,51,52,53,54] such as early sexual debut, multiple sexual partners, smoking, and unsafe sexual practices are modifiable and could be controlled using public health approaches. To further support this, the Global Burden of Cancer attributable to modifiable risk factors confirmed the contribution of behavioral risk factors such as smoking, alcohol use, and unsafe sex as modifiable risk factors attributable to a huge burden of cancer in the population [54]. Achieving population level changes in unsafe sexual practices, however, will require societal efforts in addressing the structural and system drivers of early sexual debut and multiple sex partners such as economic, legal, cultural, and religious beliefs. In the area of therapeutics, it is possible that the biological interplay of CSTs, HR-HPV, and HIV immunosuppression seen in our study could be further examined to develop point-of-care molecular testing for rapid detection and quantification of Lactobacillus and other non-Lactobacillus dominant cervico-vaginal microbiota as biomarkers for evaluation, early detection, and treatment of high-grade precancer in the population.

Our study also showed the importance of socioeconomic factors, such as more years of completed education was associated with lesser odds for high-grade cervical precancer/ICC, while age and higher number of total births (parity) were associated with greater odds for high-grade cervical precancer/ICC. Furthermore, the unadjusted analyses of age at first pregnancy showed a significant relationship favoring women whose first pregnancies were at relatively younger age (See Table 1), and these women were more likely to have a CST-IV (Table) compared to women who were older at first pregnancy. The epidemiologic findings seem to be in agreement with recent genome-wide association study which showed that women who were older at first pregnancy had a lower odds for ICC [55]. The interplay of these sociodemographic variables in association with unfavorable CSTs and in predicting high grades precancer and/or invasive cervical cancer is an interesting finding particularly in sub-Saharan Africa where educational attainment is low, and lower educational status has been shown to be associated with higher number of births [56]. As shown in Table 4, there is a greater odds for high-grade cervical precancer or ICC if parity is > 4 in women with negative HR-HPV. The epidemiologic nuances surrounding this relationship, and the mechanism by which repeated childbirths contributes to cervical carcinogenesis requires further investigation particularly in the absence of HR-HPV given emerging ideas that certain sub-types of cervical cancer are not caused by HPV [57, 58].

Our data revealed trends that need confirmatory longitudinal studies with larger sample size to further characterize biomarkers of the vaginal microbiome and HR-HPV detection in Nigerian women at various stages of cervical precancer to understand how they predict progression to high-grade precancer or invasive cervical cancer stages. We intuitively hypothesized that therapeutic approaches that modifies and maintains a more optimal cervico-vaginal microenvironment could reduce the acquisition or help the clearance of HR-HPV in cervices before the process of HPV-induced cervical carcinogenesis advances to high-grades precancer and invasive stages. This would be a highly translatable public health research effort and holds the prospect for reducing population level incidence of cervical precancer and/or invasive cervical cancer particularly in Nigeria and other LMICs where primary preventive HPV vaccines and cervical cancer screening coverage are still very low.

Our study findings have a few limitations, mainly in the relatively small sample size and cross-sectional design. We also lacked the data on duration for HPV infection and did not have data on HIV treatment with antiretrovirals, viral load, or immune suppression levels with CD4 count markers. We also did not have control on intravaginal practices such as douching with soap and water, and other agents on vaginal microbiome as previously reported in the literature [59]. Nonetheless, our findings, the first of their kind from Nigeria characterizing vaginal microbiome and relating it to HR-HPV and spectrum of normal cervix, low and high-grades precancer and invasive cervical cancer pathology, suggest a potential relationship between CST, HR-HPV, and progression to high-grade precancer and/or ICC. The findings also support the well-known evidence that HR-HPV is the principal causative factor [1, 2, 60] in cervical precancer and progression to invasive cervical cancer and add to the literature in the paucity of this information from Nigeria and sub-Saharan Africa.

Conclusions

We conclude that non-optimal CSTs, particularly when HR-HPV is detected, may assist in identifying women with high-grade cervical precancer. Characterization of CST using point-of-care molecular testing particularly in women identified with any HR-HPV may be a useful biomarker for early detection of severe precancer and cervical cancer prevention. Longitudinal data will improve our understanding of the temporal effect of non-optimal CSTs, HR-HPV, and other factors in cervical cancer development, and may lead to translatable public health interventions towards cervical cancer prevention and elimination globally.