Oral bacteria and gene activity related to bipolar disorder

The design of our study was matched case–control. According to the DSM-5 criteria, the cases were those who had been diagnosed with bipolar disorder, and the controls were chosen among people who had not been diagnosed with the disorder. These subjects have been described in our previously published work [36]. Briefly, in order to compare the Microbiome Taxa and miRNAs profile of the two groups, the control group was created using block randomisation matched for sex and age (± 5 years) to ensure comparability between cases and controls. Another analysis was conducted by treating individuals who did not have a diagnosis of bipolar disorder but scored positively on the Mood Disorder Questionnaire as cases and those who scored negatively on the Mood Disorder Questionnaire as controls [85]. The inclusion criterion was patients aged between 20 and 69 years without any exclusion by sex. Participants were gathered and assessed over three days at Cagliari University Hospital. The subjects completed the MDQ, underwent a psychiatric interview, and provided a general anamnesis after being fully informed about the research study and signing the informed consent form. Block randomisation matching was used to choose the control group, guaranteeing that each case was matched with two controls who were within ±5 years of age and of the same sex. For every case, a control-matching cell that included all eligible people who had not been diagnosed with bipolar disorder was created. For every case, two controls were chosen at random from these matched cells. To preserve independence across matched sets, controls were disqualified from further eligibility after they were chosen. Overall, in this study, we evaluated a total of 25 patients (No. 10 males, mean age: 52.40; No. 15 females, mean age: 58.53) and 46 controls (No. 16 males, mean age: 52.18; No. 30 females, mean age: 63.16). This study was conceived as a pilot, exploratory matched case–control investigation designed to screen for large between-group differences rather than to deliver definitive estimates. No a priori power calculation was performed; therefore, statistical power is limited, and the generalisability of the effect estimates is restricted. Approval for the study was granted by the Ethical Committee of the Institutional Review Board of the University Hospital of Cagliari, Italy (authorisation signed on 11 July 2022, with reference number NP/2022/2893; amendments approved on 6 November 2023 and 15 November 2023). All included subjects provided written informed consent. The 1964 Helsinki Declaration's standards were followed in the conduct of this investigation. Following a thorough explanation of the study's objectives, methods, and data protection, all study participants signed a written informed consent form. They were also made aware that the study could end at any time.

For each subject recruited in this study, we collected saliva samples by using a method already described [85]. The patients rinsed their mouth with about 200 µL of commercial mineral water (Rocchetta^®, Roma, Italy) twice. After 20 min, we collected about 2 mL of saliva in a 2 mL RNase- and DNase-free Eppendorf^® tube (Merck KgaA, Darmstadt, Germany). Saliva samples for DNA extraction were used as such, while in tubes for RNA extraction, 100 µL of RNAlater^® solution was added immediately, and all samples were then frozen at −80 °C until nucleic acid extraction. DNA and RNA were extracted by using (i) the Bosphore Genomic DNA Extraction Spin Kit v2 and (ii) the Bosphore RNA Extraction Spin Kit, manufactured by Anatolia Geneworks (Istanbul, Turkey), following the kits' work instructions for each extraction type. The samples were then measured to assess the total DNA/RNA quantity and purity by using a Nanodrop Spectrophotometer (Thermo Fisher Scientific—Waltham, MA, USA).

We examined the main types of oral microbial taxa/genera by using a modified method described by Bachetti et al. [86], which included two new target bacterial groups, Fusobacterium spp. and Candida spp., as listed in Table 1. We then evaluated all oligos in silico, employing various approaches.

An in silico evaluation of the oligonucleotide sequence was performed by using the Basic Local Alignment Search Tool program (BLAST version 2.17.0), https://blast.ncbi.nlm.nih.gov/Blast.cgi↗ (accessed on 17 July 2025).

The Oligo version 7 (MedProbe, Oslo, Norway) calculator and the DNA-melt program, available at https://www.unafold.org/hybrid2.php↗ (accessed on 17 July 2025) were used to assess the theoretical melting temperatures (Tms) of all primers used in this work, as well as the formation of potential oligonucleotide dimers and self-complementarity.

The DNA folding analysis of the amplicons was assessed using a DNA fold program: https://www.unafold.org/↗ (accessed on 17 July 2025). This evaluation was performed to ensure that the primer matching zone was free of amplicon loops, which could cause thermodynamic problems, such as varying PCR efficiency for different taxa [87]. In each mentioned program, we used the same physical oligo conditions, setting [DNA] at 10⁻⁷ M, 50 mM [NaCl], and 3 mM [MgCl].

This study employed a targeted qPCR-based approach designed to quantify the relative abundance of selected bacterial taxa. As such, the data do not permit calculation of ecological diversity indices (e.g., Shannon and Simpson) or ordination analyses (PCA/PCoA), which require broader sequencing-based datasets. This methodological choice was intentional to focus on specific phyla and genera previously implicated in psychiatric conditions.

Real-time PCR reaction was performed to quantify the total bacterial taxa/genera in DNA extracts. Real-time qPCR profiling was performed using a CFX-96 apparatus (Bio-Rad Laboratories, Hercules, CA, USA) and SYBR Premix Ex Taq Kit (TaKaRa-Clontech ^®, Kusatsu, Japan), according to the manufacturers' instructions. The final volume of 0.02 mL contained Premix Ex Taq (1X), 1× SYBR Green (10,000×), 0.22 μM of each primer, and 1 to 10 ng of DNA extract. qPCR thermal profiles consisted of a denaturation step at 95 °C for 30 s, followed by 40 cycles of 5 s at 95 °C, 30 s at 60 °C, and 20 s at 80 °C. For the quantification of different phyla, universal primers were used as positive controls for each sample (Table 1). As a positive reaction control for each sample, we used the universal primers as described in Table 1. qPCR thermal profiles consisted of a denaturation step at 95 °C for 30 s, followed by 40 cycles of 5 s at 95 °C, 30 s at 60 °C, and 20 s at 80 °C. Fluorescence was detected at the end of the 80 °C segment in the PCR step. The bacterial concentration was obtained by interpolating the threshold cycle value of the sample with the standard curve, which we obtained using a series of Escherichia coli cell dilutions ranging from 5 × 10² to 1 × 10⁸ bacterial genomes/µL.

All qPCR products were validated by melting-curve analysis, which consistently showed single, sharp peaks, and by comparison with the expected amplicon sizes reported in Table 1. No-template controls remained negative in all runs, confirming amplification specificity.

For each taxonomic group, the concentration was then expressed as genomes/µL corresponding to a suspension of 100 ng DNA/mL, using the following formula: C_100µg = C_pcr × 100/[DNA] where C_100µg represents the genomes/µL for the considered bacterial group, contained in a suspension of 100 µg/µL of total DNA. G_pcr represents the genomes/ul evaluated with qPCR and contained in the sample. [DNA] = DNA concentration in µg/µL of the sample measured with the Nanodrop instrument version 8.

The percentage concentration for each phylum is calculated as follows: G% = C_100ng × 100/∑_bg where C_100ng = represents the genomes in 100 ng/mL of total DNA calculated with the previous formula. ∑_Bg = sum of the concentration of genomes/100 ng [DNA] of all bacteria in the analysed group.

The qPCRs showed an efficiency range from 0.95 to 0.98.

Among the group of BD and control salivary RNA samples, we recruited samples for molecular analysis of Pri-miRNAs gene expression. The ACTB gene related to β-actin was used as the housekeeping gene.

We designed PCR oligos using the same in silico procedures previously described for absolute quantitation of different bacterial taxa. The RT-qPCR procedure was performed following the method described by Maxia et al., 2023 [88]. Table 2 shows the main parameters of oligonucleotides used in RT-qPCR. Relative expression was performed by using the 2^−∆∆Ct method by Livak et al. [89].

For each sample, three distinct biological replicas were obtained, and quantitative data were expressed as means ± SDs. Changes in gene expression above 2 or below 0.5 were considered significant. Statistically significant differences for the exact determination among cases and control samples were determined using Fisher's exact test to substantiate a significant difference between the means of two specific groups; in this case, statistical analysis was performed by using social science statistics software (version 2025, https://www.socscistatistics.com/↗, accessed on 20 January 2025). Given the small sample and non-normal distributions, we prioritised exact and non-parametric procedures, and we report two-sided p-values with a significance threshold of p = 0.05. No formal multiplicity adjustment, which increases the risk of type I error, was applied owing to the exploratory scope of the study; this is explicitly acknowledged in the limitations. Where informative, we emphasise the direction and magnitude of effects rather than dichotomous significance. Data were collected anonymously using an ID number for each person recruited. A dedicated database was created. For continuous variables, we applied the paired t-test when the normality assumption was met and verified the results using the Shapiro–Wilk test; we used the Wilcoxon signed-rank test otherwise. For categorical variables, we chose to replace the Chi-square test with McNemar's test, which accounts for the dependence between paired observations. In addition, the Bonferroni multiple comparison test for molecular analysis was performed (α = 0.05), using GraphPad InStat software ver. 3.10 (San Diego, CA, USA).

The data obtained from samples of BD patients and controls using real-time qPCR are shown as taxa percentages in Figure 1; these represent an overview of the percentage relationships between the main bacterial phyla/genera in the oral cavity.

The taxonomic profile revealed in Figure 2 shows a substantial difference between mean and median % values. However, if we screen the results affected by extreme values according to the mean, the three main target bacterial groups that showed a significant difference were α-Proteobacteria, γ-Proteobacteria, and Actinobacteria; this aspect will be discussed in the next chapter. Because several taxa displayed right-skewed distributions with extreme values, medians provide a more robust summary than means. Importantly, the direction of the case–control differences was consistent across both central tendency measures. Statistical contrasts were conducted with non-parametric tests, which are less sensitive to outliers in small samples.

The specificity of amplification was supported by single-peak melting curves and product sizes matching the theoretical amplicons, ensuring that the observed taxa differences reflected true quantitative variation.

An elevated oral F/B ratio in BD should be interpreted as a marker of oral biotic status rather than a diagnostic signature. It reflects a relative increase in Firmicutes and/or depletion in Bacteroidetes and, in the oral niche, does not imply causality nor disorder specificity.

The F/B ratio was about 4/1 for BD subjects. Given the small sample size and skewed distributions, medians were prioritised as the most robust measure of central tendency, while means are reported descriptively. No formal adjustment for multiple testing (e.g., FDR or Bonferroni correction) was applied, consistent with the exploratory nature of this pilot study; this increases the risk of type I error and is acknowledged in the limitations.

Figure 3 demonstrates that pri-miRNAs evaluated at p > 0.05 showed no significant differences between the two patient cohorts. At the same time, the results showed a high dispersion since the SD was ± 2 for cases and ± 1.7 for controls. This behaviour may indicate that other factors, such as habits or health conditions of many subjects, are altering the equilibrium of the oral mucosal tissues. For both pri-miR-146a and pri-miR-155, melting-curve profiles confirmed single products of the expected size, with no amplification in negative controls. The lack of between-group differences for pri-miR-146a and pri-miR-155, together with the wide inter-individual variability, suggests that these transcripts may be strongly modulated by host factors (e.g., oral hygiene, smoking, medications, and diet) and pre-analytical variability. Accordingly, the negative finding should not be taken as evidence against a role of miRNAs in BD but rather as a call for more controlled, adequately powered studies.

Our aim was to examine whether oral microflora contribute to gut eubiosis or dysbiosis. Evidence suggests that oral pathobionts can alter the gut microbiota, with studies showing thatpromotes systemic and neurodegenerative disorders such as Parkinson's and Alzheimer's. Porphyromonas gingivalis

A potential link between oral bacteria, gut microbiota, and mood disorders is theoretically plausible, though current data remain inconsistent. SCFA production illustrates this duality: locally, high SCFA levels can promote dysbiosis and oral disease, whereas in the gut, SCFAs exert systemic benefits, including anti-inflammatory and metabolic effects. These divergent outcomes arise from differences in mucosal structure, with the gut epithelium being adapted to absorb and utilise SCFAs. Oral microbes may, therefore, act as direct producers, inducers, or inhibitors of gut SCFA metabolism (and). Figure 4 Figure 5

The figure, presented in schematic mode, illustrates the complexity of oral–gut interactions. Bacteria from the oral cavity can interact positively or negatively with SCFAs, leading to changes in the amount of fatty acids in the gut and blood. Considering the results shown inand, the significant variations in bacterial taxa are observed in three main microbial groups, already cited in the: α-Proteobacteria, γ-Proteobacteria, and Actinobacteria. Specifically, we can observe a significant reduction in these taxa in the oral cavity of BD patients. Figure 1 Figure 2 Section 3

These taxa contain more species of fatty acid-producing bacteria, such as those in the phylum Actinobacteria and the phylum Proteobacteria []. However, it is plausible that changes in the levels of SCFAs in the gut might be caused by certain bacteria that either work against or together with SCFA-producing bacteria [,,,,]. The production of bacteriocins against gut bacterial SCFA producers could represent an interesting antagonistic mechanism []. The reduction in these phyla in BD patients may contribute to altered short-chain fatty acid (SCFA) production in the gut, possibly through modulation of biofilm networks. Oral bacteria, via quorum sensing mechanisms, could favour biofilm communities less associated with SCFA generation [,]. ^{113 114 115 116 117 118 119 119 120}

Finally, we found that the levels of pri-miRNAs 155 and 146a were similar in both the cases and control subjects, with no significant differences between the two groups. This might suggest that these genomic regulators in saliva are not strongly linked to BD conditions and could be influenced by other habits of the host. In another, already published work that uses used the amounts of volatile sulphur compounds in the breath of BD patients, which is related to dysbiotic status, the authors found a moderate production of these bacterial metabolites in these patients which suggests that the change in the oral microbiome able to drive the association between gut microflora and mood disorders could often be due to silent dysbiotic processes [,,,]. Second, while multiple statistical contrasts were performed, no correction for multiplicity was applied, in keeping with the hypothesis-generating scope of this work. These findings should not be regarded as evidence of biomarker potential at this stage but rather as hypothesis-generating signals that warrant confirmation in larger cohorts. Mechanistic studies, ideally including direct measurement of microbial metabolites such as SCFAs, as well as longitudinal designs accounting for medication exposure, will be essential to validating and contextualising these associations. ^{36 121 122 123}

Limitations: Several limitations should be acknowledged when interpreting the present findings. First, the relatively small sample size, particularly in the bipolar disorder group, inevitably reduces statistical power and may limit the generalisability of the results, especially for highly variable biomarkers such as salivary pri-miRNAs. Second, salivary RNA is inherently subject to variability in quality and yield, which, although partially controlled through the use of a stable housekeeping gene (ACTB), may still have influenced the quantitative results and contributed to the observed heterogeneity. Furthermore, while primer design and validation were performed according to standard in silico and in vitro procedures, we did not conduct a comprehensive MIQE-compliant evaluation (e.g., the systematic reporting of amplification efficiencies, dynamic ranges, and reference gene validation). This omission should be regarded as a limitation, and future studies should implement full MIQE guidelines to maximise reproducibility and comparability across laboratories. Third, this study employed a targeted qPCR-based approach designed to quantify selected oral bacterial taxa; as such, it does not allow for the calculation of comprehensive microbial diversity indices (e.g., Shannon and Simpson) or ordination analyses (PCA/PCoA), which are standard in sequencing-based microbiome studies. Future investigations integrating 16S rRNA sequencing or metagenomic profiling will be necessary to fully characterise ecological diversity in this population. Furthermore, while multiple statistical contrasts were performed, no correction for multiplicity was applied, in keeping with the hypothesis-generating scope of this work. Therefore, the observed differences should be interpreted cautiously, and replication in larger, adequately powered samples will be essential. In addition, medication exposure was not systematically assessed. Psychotropic agents such as mood stabilisers, antipsychotics, and antidepressants, as well as antibiotics and other treatments, may substantially influence both oral microbiota and pri-miRNA expression. The absence of structured data on pharmacological profiles represents a relevant limitation of this pilot study and should be rigorously addressed in future research.

Finally, information on potential confounders such as smoking habits, oral hygiene, recent antibiotic exposure, and psychotropic medication use was not systematically available; thus, residual confounding cannot be excluded. Future studies should aim to incorporate detailed assessments of these factors and, where possible, adjust for them analytically to refine the observed associations.