Salivary Prevotella qPCR Signal as an Exploratory Non-Invasive Diagnostic Adjunct for Rotterdam Phenotype Stratification in Women with Polycystic Ovary Syndrome

1. Introduction

Polycystic ovary syndrome (PCOS) is a common endocrine disorder among reproductive-age women and is characterized by heterogeneous reproductive, endocrine, and metabolic manifestations. The Rotterdam criteria remain widely used for clinical classification and generate four major phenotypes based on combinations of hyperandrogenism, ovulatory dysfunction, and polycystic ovarian morphology [1,2]. This phenotypic heterogeneity is clinically important because different phenotypes may carry different reproductive, metabolic, and inflammatory implications [3,4,5,6,7,8].

Microbiome research has become increasingly relevant to PCOS because microbial ecosystems may interact with androgen exposure, insulin resistance, low-grade inflammation, bile acid metabolism, and immune regulation [9,10,11,12,13,14,15,16]. Most PCOS microbiome studies have focused on the gut or vaginal microbiome, whereas saliva remains less explored despite being easy to collect, repeatable, and acceptable in outpatient clinical settings [9,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31].

Prior salivary microbiome work in PCOS suggests that oral microbial composition may be associated with disease-related characteristics, but the evidence remains limited and methodologically heterogeneous [9,10,41]. Saliva may reflect both oral ecological conditions and systemic host factors; however, qPCR-based genus-level signals should be interpreted as molecular target signals rather than direct evidence of species-level community structure or functional microbiome activity.

Among candidate genera, Lactobacillus, Bifidobacterium, and Prevotella are biologically plausible targets because they have been implicated in microbial barrier function, host metabolic regulation, dietary pattern, immune signaling, and inflammatory tone [32,33,34,35,36,37]. Prevotella is especially context-dependent: different species and strains may be associated with either eubiotic fiber-related profiles or dysbiotic inflammatory states depending on host and environmental conditions [34,35,36,37].

Accordingly, this study analyzed qPCR-derived salivary Cq signals of Lactobacillus, Prevotella, and Bifidobacterium across Rotterdam-defined PCOS phenotypes and controls. From a diagnostic perspective, the objective was to examine whether genus-specific salivary microbial signals, particularly Prevotella, show phenotype-stratified patterns that could justify further validation as a non-invasive adjunct to conventional clinical phenotyping, rather than as a standalone diagnostic test.

2. Materials and Methods

This study used a cross-sectional design. Participants were recruited from premenopausal and infertile women who regularly attended outpatient clinical services. PCOS was diagnosed according to the Rotterdam criteria. Phenotype A was defined as hyperandrogenism, ovulatory dysfunction, and polycystic ovarian morphology; phenotype B as hyperandrogenism and ovulatory dysfunction without polycystic ovarian morphology; phenotype C as hyperandrogenism and polycystic ovarian morphology without ovulatory dysfunction; and phenotype D as ovulatory dysfunction and polycystic ovarian morphology without hyperandrogenism [1,2]. Controls were women without clinical PCOS according to the same diagnostic framework.

Eligible participants were women aged 18-42 years, clinically diagnosed with PCOS according to the Rotterdam criteria or classified as controls, willing to participate in the study, able to provide written informed consent, and willing to undergo the required research procedures, including blood and saliva sampling. Participants with incomplete key clinical or salivary qPCR data were excluded from the final analysis.

The final analysis set included 110 women in total, comprising 87 women with PCOS and 23 controls. Baseline variables recorded for analysis included age, Ferriman-Gallwey score, menstrual irregularity or ovulatory dysfunction, and polycystic ovarian morphology on transvaginal ultrasonography. Ferriman-Gallwey scoring was used as the clinical measure of hirsutism [42].

Saliva samples were collected in the morning after 1-2 hours of fasting. Participants were instructed not to brush their teeth, use mouthwash, or consume food immediately before sampling. Samples were collected into sterile containers and processed for DNA extraction. The manuscript should be interpreted with the limitation that detailed oral hygiene status, periodontal examination, recent antibiotic exposure, and dietary intake were not fully modeled as covariates.

Salivary microbial signals for Lactobacillus, Prevotella, and Bifidobacterium were assessed using genus-specific quantitative polymerase chain reaction (qPCR). DNA was extracted from saliva samples using the QIAamp DNA Stool Mini Kit (Qiagen, USA). qPCR amplification was performed using GoTaq qPCR Master Mix (Promega Corporation, USA) on a MyGo Mini S Real-Time PCR Cycler with MyGo Mini software version 3.5.6 (IT-IS Life Science Ltd., Ireland). The reaction mixture consisted of 12.5 uL of master mix, 5 uL of nuclease-free water, 1 uL of forward primer, 1 uL of reverse primer, and 5.5 uL of DNA template, for a total reaction volume of 25 uL.

The thermal profile consisted of an initial denaturation at 95 deg C for 120 s, followed by 40 amplification cycles of denaturation at 95 deg C, target-specific annealing, and extension at 72 deg C. Melting curve analysis was performed from 60 deg C to 97 deg C at 0.1 deg C/s to assess amplification specificity. Instrument run files indicated an annealing temperature of 60 deg C for Lactobacillus and Prevotella and 50 deg C for Bifidobacterium. Genus-specific standard curves were generated using serially diluted standards ranging from 1.5 x 10^8 to 1.2 x 10^9 nominal units. The standard curves were used for assay monitoring and quality assessment.

The inferential analysis was based on Cq values rather than GAPDH-normalized Delta Ct/Delta Delta Ct expression analysis. GAPDH was not used as a bacterial reference gene in the final revised analysis because it is a host housekeeping gene and is not an appropriate denominator for genus-level bacterial abundance in saliva. In addition, no universal bacterial 16S reference or fully validated absolute copy-number conversion was available for all targets. Therefore, results are reported as qPCR-derived Cq signals. Lower Cq values indicate higher target DNA signal. Standard-curve-derived quantity outputs and curve-quality parameters are presented as supplementary quality-control information, but no claim of absolute CFU/mL bacterial load is made. qPCR reporting was revised in accordance with core MIQE principles of transparent assay description and cautious interpretation [48].

Normality was assessed using the Shapiro-Wilk test [44]. Because the data were not normally distributed, continuous variables are reported as median with minimum and maximum values. Between-group comparisons were performed using the Kruskal-Wallis test [45]. Pairwise post-hoc comparisons were performed using Mann-Whitney U tests with Bonferroni correction [46]. Correlation analyses among women with PCOS were performed using Spearman rank correlation. False discovery rate correction was applied to exploratory correlation analyses using the Benjamini-Hochberg procedure [47].

Additional exploratory analyses were performed to support interpretation of the observed Prevotella-related findings. ROC analyses and age-adjusted regression models were retained as exploratory supplementary analyses only. These analyses were not used to establish diagnostic thresholds because the study was cross-sectional, single-center, and not designed for external biomarker validation.

The study protocol was reviewed and approved by the Ethics Committee of the Faculty of Medicine, Universitas Airlangga, Surabaya, Indonesia. Written informed consent was obtained from all participants before data and sample collection.

3. Results

3.1. Baseline Characteristics

A total of 110 participants were included: 23 women in phenotype A, 19 in phenotype B, 22 in phenotype C, 23 in phenotype D, and 23 controls. Baseline characteristics are summarized in Table 1.

Age did not differ significantly across groups (p = 0.811). Ferriman-Gallwey score differed markedly across groups (p < 0.001), as expected from phenotype definitions. Menstrual irregularity or ovulatory dysfunction and polycystic ovarian morphology on TVS also differed significantly across groups. In the corrected dataset, all participants in phenotype A had polycystic ovarian morphology on TVS, consistent with the Rotterdam phenotype A definition.

3.2. Salivary Microbial qPCR Signals Across PCOS Phenotypes and Controls

Salivary microbial Cq values across PCOS phenotypes and controls are shown in Table 2. These values should be read as qPCR-derived Cq signals; lower Cq values indicate higher target DNA signal.

Lactobacillus Cq values did not differ significantly across phenotypes and controls (p = 0.249). Prevotella Cq values differed significantly across groups (p < 0.001), with lower median Cq values in phenotypes A, B, and C than in phenotype D and controls. Because lower Cq values indicate higher target DNA signal, these findings suggest stronger Prevotella molecular signal in selected hyperandrogenic and/or PCOM-related PCOS phenotypes. Bifidobacterium Cq values also differed significantly across groups (p < 0.001), but the post-hoc pattern was less consistent than that observed for Prevotella.

3.3. Post-Hoc Pairwise Comparisons and Exploratory Discriminatory Analyses

Post-hoc pairwise comparisons were performed for genera with significant overall between-group differences. Significant comparisons after Bonferroni correction are summarized in Table 3; the complete pairwise comparison table is provided in the supplementary section.

For Prevotella, phenotype A, phenotype B, and phenotype C differed from phenotype D and/or controls after correction, reinforcing that the strongest phenotype-related salivary qPCR signal was observed for Prevotella. For Bifidobacterium, significant corrected differences were observed for phenotype A versus control and phenotype C versus control, but these findings were treated as secondary because Bifidobacterium assay performance was less robust in the available standard-curve files.

3.4. Exploratory Correlation Analysis Among Women with PCOS

Exploratory correlation analyses were performed among women with PCOS only (n = 87). The main findings are summarized in Table 4.

Ferriman-Gallwey score was not associated with Lactobacillus Cq values, but it correlated positively with Prevotella Cq values and negatively with Bifidobacterium Cq values. PCOM on TVS correlated negatively with Prevotella Cq values, indicating that PCOM was associated with a higher Prevotella target DNA signal. Menstrual irregularity or ovulatory dysfunction was not significantly associated with any of the three microbial Cq signals after FDR correction.

4. Discussion

This cross-sectional study examined genus-specific salivary qPCR signals of Lactobacillus, Prevotella, and Bifidobacterium across Rotterdam-defined PCOS phenotypes and controls. The revised analysis deliberately reports Cq-based microbial signals rather than absolute bacterial abundance or GAPDH-normalized expression because the available assay files did not support a uniformly robust absolute CFU/mL conversion across all targets. Within this cautious analytical frame, Prevotella showed the clearest and most interpretable phenotype-related pattern.

The main finding was that Prevotella Cq values differed significantly across PCOS phenotypes and controls. Phenotypes A, B, and C tended to show lower Prevotella Cq values than phenotype D and controls, indicating stronger Prevotella target DNA signal in selected PCOS phenotypes. This pattern is clinically plausible because these phenotypes include hyperandrogenism and/or PCOM-related features, both of which may reflect broader endocrine and inflammatory dysregulation [1,2,3,4,5,6,7,8,38,39,40].

The direction of the correlation findings requires careful interpretation. A positive correlation between Ferriman-Gallwey score and Prevotella Cq values means that higher hirsutism scores were associated with lower Prevotella target signal, because higher Cq reflects lower target DNA signal. In contrast, the negative correlation between PCOM on TVS and Prevotella Cq values suggests that PCOM was associated with stronger Prevotella target DNA signal. These apparently divergent associations reinforce the need to avoid simplistic claims that Prevotella is uniformly increased or decreased in PCOS.

Prevotella is biologically context-dependent. Depending on species, strain composition, diet, host metabolic status, and local mucosal ecology, Prevotella may be linked to fiber-rich dietary patterns, insulin sensitivity, or inflammatory dysbiosis [34,35,36,37]. Because the present study used genus-level qPCR rather than species-level sequencing, the data cannot determine whether the observed Prevotella signal reflects beneficial or pathogenic Prevotella taxa. This is a key reason why the finding should be framed as an exploratory molecular signal rather than a mechanistic conclusion.

The Bifidobacterium results were statistically significant overall but less stable interpretively. Some corrected group differences were observed, particularly against controls, yet the qPCR standard-curve files indicated weaker assay quality for Bifidobacterium than would be expected for confident absolute quantification. Therefore, Bifidobacterium should be treated as a secondary exploratory finding requiring technical revalidation.

A methodological strength of this study is the use of saliva as a non-invasive sample type, which is relevant for future translational studies in outpatient reproductive endocrinology. Another strength is explicit Rotterdam phenotype stratification rather than treating PCOS as a single homogeneous entity. This is important because phenotype-level differences may be diluted when all PCOS cases are pooled.

Several limitations remain. First, the cross-sectional design prevents causal or temporal inference. Second, the microbiological analysis was genus-specific and qPCR-based, without 16S rRNA gene sequencing, shotgun metagenomics, species-level resolution, or functional profiling. Third, the available standard-curve data showed variable efficiency and linearity; therefore, the revised manuscript avoids claims of absolute bacterial load. Fourth, BMI, insulin resistance, androgen biochemistry, diet, oral hygiene, periodontal status, and recent antibiotic exposure were not fully included as covariates. Fifth, no external validation cohort was available.

Despite these limitations, the study provides a useful exploratory signal: salivary Prevotella Cq values differ across Rotterdam-defined PCOS phenotypes and show associations with selected PCOS diagnostic features. The findings justify larger, technically strengthened studies incorporating standardized oral examination, dietary and metabolic covariates, validated qPCR efficiency, universal bacterial reference targets or absolute copy-number standards, and sequencing-based microbial profiling.

5. Conclusions

In this cross-sectional dataset, salivary Prevotella showed the clearest exploratory genus-level qPCR signal across Rotterdam-defined PCOS phenotypes. The revised interpretation is intentionally conservative for a Diagnostics submission: the study supports phenotype-associated variation in salivary microbial Cq signals as a candidate non-invasive diagnostic adjunct, but it does not establish Prevotella as a validated standalone diagnostic biomarker or quantify absolute bacterial load. Future studies should combine saliva-based microbial profiling with robust qPCR validation, species-level sequencing, metabolic and hormonal covariates, oral-health assessment, prespecified diagnostic thresholds, and independent validation cohorts.