An initial group of 17 patients (female/male: 11/6; mean age: 48 ± 13 years) and 35 additional patients (female/male: 19/16; 60 ± 15 years) were recruited from the University Dental Clinic of the Faculty of Medicine and Dentistry at the University of Santiago de Compostela between December 2019 and March 2020, and between June 2020 and May 2022, respectively. In the first group of patients (n = 17), saliva and subgingival samples were taken to optimize biofilm generation conditions (see section 2.5). In the second group of patients (n = 35), only subgingival samples were taken, as a result of the optimization done with the first group. Oral health status was assessed according to the proceedings from the 2017 World Workshop on the Classification of Periodontal and Peri-Implant Diseases and Conditions (Papapanou et al., 2018; Tonetti et al., 2018) ( Supplementary Tables S1, S2 ). Antibiotic treatment up to one month prior to sampling was the exclusion criterium set for the study.

The investigation protocol 2009/319 for patient recruitment and sample handling, modified in July 2017, was approved by the Ethical Committee of Clinical Investigations of Galicia (Xunta de Galicia). Written informed consent was obtained from all participants in this study.

The QQ enzyme Aii20J, a lactonase enzyme obtained from the QQ marine bacterium Tenacibaculum 20J (Romero et al., 2011; Romero et al., 2014), was obtained as previously described (Mayer et al., 2018). Briefly, protein expression was induced by adding 1 mM Isopropyl-D-thiogalatopyranoside to a suspension of a recombinant E. coli BL21(DE3)pLysS containing Aii20J. The induced cells were lysed by sonication, and purification of Aii20J was achieved using the His GraviTrap affinity column protein purification kit (GE Healthcare). The remaining imidazole from the purification buffer solutions was removed by dialyzing the sample in D-tubes (MWCO 10 kDa) (Merck KGaA, Darmstadt, DE) in sterile Milli-Q water. When appropriate, purified Aii20J was added to the biofilms at a working concentration of 20 µg mL^-1, corresponding to ten times the minimum concentration of purified enzyme needed to degrade 10 µM of C₆-HSL in 24 h. The AHL-degradation capacity of purified Aii20J and remaining Aii20J activity in biofilms’ supernatants was routinely verified using a chromogenic assay (see “Quorum quenching activity solid plate assays” section below).

The AHL degradation activity of Aii20J, both purified and after total incubation time in biofilm supernatants, was tested using solid plate assays. QQ activity in saliva samples was assessed using the same methodology. Assays were carried out using the AHL biosensor strains Chromobacterium subtsugae CV026 for short-chain AHLs (McClean et al., 1997; Harrison and Soby, 2020) and C. violaceum VIR07 for long-chain AHLs (Morohoshi et al., 2008). Samples were mixed with either C₆-HSL (10 µM) or C₁₂-HSL (10 µM) (Sigma-Aldrich, Merck). These AHLs are commonly used as representatives of short- and long-chain AHLs, respectively (Muras et al., 2020b). PBS pH 6.5 with the corresponding AHL (10 μM) was used as negative control (Muras et al., 2018b). The pH of the saliva samples was checked beforehand and acidified when it exceeded pH 7 to avoid the spontaneous opening of the lactone ring of the AHLs at high pH values (Yates et al., 2002). All mixtures were incubated for 24 h at 22°C.

The presence of the AHLs was detected by adding a mixture of an overnight culture of the corresponding biosensor and soft Luria-Bertani (LB) (0.8% agar) in a 1:4 proportion on top of an LB agar plate. Wells were manufactured onto the agar, and 100 µL of the sample were placed in each well. Plates were incubated for 24 h at 30°C. The lack of violet pigmentation, as opposed to the violet halo observed around the negative controls, indicated QS inhibition. This assay allows identifying in a rather precise manner if the QS inhibition is due to an enzymatic degradation or to the presence of a QS inhibitor molecule. Enzymatic degradation is often displayed as a lack of pigmented halo or as a halo of reduced diameter respect to the negative control. On the other hand, QS inhibitors diffuse with the AHL added to the sample; the different titers of both components, with the AHL normally present in a much higher titer, result in an inner non-pigmented halo surrounded by an outer violet halo (Muras et al., 2018a). Furthermore, this assay allows differentiating between QS inhibition (observation of growth-turbid halos) and cell growth inhibition (observation of transparent halos) (Muras et al., 2018a).

The initial biofilm collection was derived from saliva and subgingival samples from seven healthy donors (codes H1 – H7) and ten patients with PD (codes P1 – P10) ( Supplementary Table S1 ) and cultured immediately. Saliva samples were collected in sterile tubes and diluted 1:100 in each culture media used (Fernandez y Mostajo et al., 2017; Muras et al., 2020a). Subgingival samples consisted of gingival crevicular fluid (GCF) collected by inserting 16 sterile paper points (4 in each mouth quadrant) in the gingival sulcus for 1 minute. In patients with PD, samples were taken from the deepest periodontal pockets. Samples were transported in 1 mL of thioglycollate broth (Merck), from which a 1:53 dilution with the corresponding medium was made (Mira et al., 2019). For this initial collection, saliva and subgingival samples were inoculated in 1) Brain Heart Infusion (BHI) (PanReac, Barcelona, ES) in aerobic conditions and 2) Schaedler medium supplemented with vitamin K1 (0.1 mg L^-1) (SCH-K1) (Conda Pronadisa, Madrid, ES) in anaerobic conditions. The Aii20J enzyme was added at the moment of the inoculation, at a final concentration of 20 µg mL^-1. Negative controls received sterile Milli-Q water. The purified Aii20J (35 kDa in size) was filtered through a Centricon with a molecular weight cut-off of 10 kDa (Merck) to remove the enzyme and use the remaining soluble fraction as an additional control ( Supplementary Figure S7 ). A heat-inactivated Aii20J (90 min at 121°C) that did not conserve AHL-degrading activity was also tested ( Supplementary Figure S7 ). Biofilms were grown for 24 h at 37°C.

In vitro oral biofilms (n = 17) were generated using the culture media and conditions described above and measured in parallel using two different methodologies. The first is a modification of the Amsterdam Active Attachment model (AAA model) (Exterkate et al., 2010; Muras et al., 2020a). Briefly, the lids of 12-well cell culture plates (VWR) were replaced by custom-made stainless-steel lids onto which 12 silicone tubes were attached, allowing for the insertion of glass coverslips (18 × 18 mm) (Menzel Gläser, Braunschweig, DE) to serve as substrata for biofilm formation. The glass coverslips fit vertically into the culture plate wells, filled with 3 mL of the corresponding adjusted inocula (Muras et al., 2020b), prompting the active attachment of bacterial cells onto their surface. Aii20J was added at a final concentration of 20 µg mL^-1. The AAA model allows the refreshment of culture media and treatments at 12 h by transferring the lid with the glass coverslips to a new plate. When required, anaerobiosis was achieved using an anaerobiosis jar (OXOID, Basingstoke, GB) with AnaeroGen 2.5L Atmosphere Generation System (OXOID). Biofilms were quantified by crystal violet (CV) assay. After incubation, coverslips were allowed to dry in a fresh plate, and biofilms were stained for 20 minutes with 2 mL of a 0.04% CV solution (v/v in distilled water) (Panreac). The excess dye was removed by washing twice with distilled water. Before releasing the CV for quantification, pictures were taken to perform visual examinations of the biofilms. When two independent observers found control and Aii20J-treated biofilms to be visually distinguishable, the biofilms were considered to have macroscopical differences ( Supplementary Figure S2 ). Finally, CV bound to the biofilms was diluted in 3 mL of a 33% acetic acid solution (v/v in distilled water) (Scharlab, Barcelona, ES). Sample absorbance was measured at 590 nm. All experiments were performed in triplicate.

The second methodology used was the xCELLigence Real Time Cell Analyzer (RTCA) System (ACEA, Biosciences Inc., San Diego, US) (Muras et al., 2018b, Muras et al. 2020b; Mira et al., 2019). E-plates 16 (ACEA, Biosciences Inc.) were inoculated with 100 μL of the corresponding culture medium and 80 μL of the saliva/subgingival samples. To lactonase-treated samples, 20 μL of Aii20J (to a final concentration of 20 µg mL^-1) were added, while 20 μL of sterile Milli-Q water were added to the control wells. An overlay of sterile paraffin was added when an anaerobic atmosphere was required. The xCELLigence measures changes in impedance as cells attach to the bottom of the E-plates, and the software computes cell index (CI) values that correspond to the amount of biofilm formed. All experiments were performed in triplicate.

A broader screening of culture conditions was then performed using subgingival samples (n = 35). Samples were collected following the previously described methodology and cultured in the AAA model. Culture conditions for four initial patients (H8, G1, P11, and P12) ( Supplementary Table S2 ) were BHI in aerobiosis, BHI in anaerobiosis, SCH-K1 in anaerobiosis, and semi-defined McBain medium (McBain et al., 2005; Janus et al., 2015) in anaerobiosis. Subsequent biofilms from ten healthy donors (H9 – H18), two patients with gingivitis (G2, G3), and 19 patients with periodontitis (P13 – P31) ( Supplementary Table S2 ) were inoculated in BHI-anaerobiosis and McBain-anaerobiosis. Biofilms were grown for 24 h at 37°C, with culture media and treatment refreshment at 12 h. Biofilm quantification was performed as previously described.

Subgingival-derived biofilms from patients P30 and P31 grown in McBain-anaerobiosis in the AAA model were further visualized using confocal-laser scanning microscopy (CLSM) to compare control and Aii20J-treated biofilms. Biofilms formed onto the glass coverslips were stained with two nucleic acid dyes; SYTO 9, which is membrane permeable and enters all bacterial cells, and propidium iodide (PI), which only enters membrane-damaged cells (LIVE/DEAD BacLight Bacterial Viability Kit, Thermo Fisher Scientific, Waltham, US). The staining solution was prepared by mixing SYTO 9 and PI in a 1:1 proportion and then diluting it 150-fold in sterile PBS (pH 7.4). Approximately 50 μL of the staining solution were poured onto each biofilm. Samples were kept for 15 minutes at room temperature and protected from light before image acquisition with a Leica Stellaris 8 FALCON confocal microscope, with the objective HC PL APO CS2 20x/0.75 Dry (Leica Microsystems GmbH, Wetzlar, DE). Biofilms were observed using the 499 nm laser excitation and 504 – 555 nm emission band for SYTO 9, and 561 nm excitation laser and 570 – 675 nm emission band for PI. Photographs were acquired at 20x magnification. To calculate the area covered by biofilms, eight fields for each condition were randomly selected and processed using the Leica Application Suite X software (LAS X v.3.7.4, Leica Microsystems). The area emitting photons from the SYTO 9 and PI channels was divided by the total field of view. The results are expressed in percentage of the area covered by biofilm structures in each condition.

Genomic DNA for sequencing analysis was obtained from biofilms grown in the AAA model. Biofilms were harvested by transferring the glass coverslips into 5 mL of sterile PBS. Then, they were sonicated for 15 minutes to disperse the biomass. Genomic DNA extraction was done using the “DNeasy PowerBiofilm Kit” (Qiagen, Germantown, US), following the manufacturer’s instructions. Briefly, biofilms were subjected to chemical and mechanical lysis. Proteins and inhibitors were removed, followed by pH-driven precipitation of large insoluble macromolecules. Total genomic DNA was captured using a silica spin filter column and finally eluted. DNA concentration was measured using a NanoDrop (Thermo Scientific).

Microbial genomic DNA (5 ng µL^-1 in 10 mM Tris pH 8.5) was used to amplify the 16S rRNA gene V3-V4 hypervariable regions. After size verification, the libraries were prepared according to the supplier’s protocol (Illumina, Inc.), and the resulting 11 pM 20% on PhiX pooled library was sequenced using a 2x300 base pairs paired-end run (MiSeq Reagent kit V3(MS-102-3001)) on a MiSeq Sequencer.

Quality assessment was performed using the prinseq-lite program (Schmieder and Edwards, 2011). The pipeline DADA2 (Callahan et al., 2016) was used to analyze and cluster the sequences into amplicon sequence variants, classified to the genus level with the SILVA database (Quast et al., 2013). Computations and statistics were carried out in R RStatistics (R Core Team, 2022) using knitr, knitcitations, markdown (Allaire et al., 2014; Boettiger, 2014; Xie, 2014), biostrings, and vegan (Oksanen et al., 2017). The differential relative abundance of the bacterial genera identified was assessed with the packages Phyloseq and DESeq2 (Love et al., 2014). Differences were filtered with a log2 fold change threshold higher than 2 and by their base mean value, discarding the lower quartile. Beta diversity was studied using a PCoA of weighted UniFrac distances using the package Phyloseq. Differences between samples, grouped according to the studied variables, were analyzed using the PERMANOVA test implemented in the adonis function of the package vegan (Oksanen et al., 2017). Alpha diversity was analyzed with the R package Phyloseq (McMurdie and Holmes, 2013), using the Chao1 and the Shannon and Simpson indexes as richness and diversity estimators. The data’s normality and homoscedasticity were assessed before choosing parametric or nonparametric tests included in the package stat (R Core Team, 2022).

The amount of total bacteria, P. gingivalis, and P. endodontalis present in 33 subgingival biofilms from patients with periodontitis and grown in McBain-anaerobiosis in the AAA model was evaluated by quantitative PCR (qPCR) in a LightCycler 480 II Thermocycler (Roche Diagnostics GmbH, Mannheim, DE). The reactions were performed in a 20 μL volume containing LightCycler 480 II Probes Master (Roche Diagnostics GmbH), specific primers and probes for each bacteria or universal primers for Eubacteria ( Supplementary Table S3 ), 5 μL of isolated DNA and PCR grade sterile water. Positive and negative controls and sample reactions were performed in duplicate. The samples underwent an initial amplification cycle of 95°C – 10 min, followed by 40 cycles of 95°C – 10 s, annealing at 60°C – 30 s and extension at 72°C – 1 s. Crossing points (C_P) were calculated with LightCycler 480 Software 1.5 (Roche Diagnostics GmbH) using the second derivative maximum method. Crossing point values were then extrapolated to colony-forming units (CFUs) using standard curves, as described in Àlvarez et al. (2013). Briefly, to obtain standard curves, first pure bacterial cultures were generated in three biological replicates. When cultures reached log phase, CFU plating and, in parallel, DNA extractions for qPCR amplification were performed, both in 10-fold serial dilutions. Then, curves correlating CFUs/mL and C_P values were generated using the qPCR software. Reproducibility of the R² values of the curves obtained from the different biological replicates was used as a quality criterium to consider the correlations reliable (Àlvarez et al., 2013).

To detect the presence of AHLs at low concentrations in untreated biofilms, the supernatants of the biofilm cultures were extracted to concentrate the polar molecules and analyze them by high-performance liquid chromatography-mass spectrometry (HPLC-MS). Glass coverslips from the AAA models were transferred, together with the spent culture media, into fresh tubes (VWR) and sonicated for 15 minutes to disperse the biomass. Samples were then acidified with HCl 1 M to induce the circularization of the lactone rings (Yates et al., 2002). Samples were extracted sequentially with two volumes of ethyl acetate (Scharlab) and two volumes of dichloromethane (Scharlab) (Muras et al., 2020a). The organic phases were evaporated in a Rotavapor-R (Büchi), and the dry samples were resuspended in 1 mL of acetonitrile (Merck) and stored at -20°C until its quantification by HPLC-MS. Fresh culture media were also extracted following the same procedures.

HPLC-MS analyses were performed at the Unit of Chromatographic Techniques and Water Analysis (University of A Coruña, ES) using an HPLC 1100 series (Agilent). The mass spectrometer (MS) used was an API 4000 triple quadrupole (Applied Biosystems). A calibration curve was obtained running standard synthetic AHLs (with and without oxo- and hydroxy-substitutions, ranging from 4 to 18 carbons in their acyl chains). The calibration curve was used to quantify AHLs in the samples. Elution times and molecular spectra of the standard synthetic AHLs were used to identify the AHLs present in the samples (Romero et al., 2010; Mayer et al., 2018).

A selection of QS and QQ-related sequences was made by manually searching NCBI’s databases (https://www.ncbi.nlm.nih.gov/) for annotated nucleotide and amino acid sequences with demonstrated activity. Selected QS sequences comprised AHL synthases (n = 36), AHL receptors (n = 28), the AI-2 synthase LuxS (n = 11), and AI-2 receptors (n = 3). PQS synthases (n = 2) and a PQS receptor (n = 1), were included as well. QS “dialect” systems were also selected, including non-AHL synthases (n = 5) and receptors (n = 10) in the analyses. Selected QQ sequences included three types of enzymes: lactonases (n = 63), acylases (n = 20), and paraoxonases (n = 3). Nucleotide sequences were translated into amino acid sequences using the software Expasy (Duvaud et al., 2021).

The initial 182 sequences were clustered into protein families using CD-HIT (Fu et al., 2012), establishing a minimum of 70% length coverage and 40% sequence identity to reduce redundancy. Even though several protein sequences have the same annotation, some were clustered into different protein families, reflecting evolutionary divergence, e.g., LuxS sequences were clustered into four groups with the representative type sequences corresponding to Vibrio harveyi, Streptococcus pyogenes, Bacillus thuringiensis, and Bacteroides heparinolyticus. The 182 sequences were clustered and grouped by function, resulting in 78 protein families. The accession numbers of the sequences in each protein family are available in Supplementary Table S4 .

The representative sequences selected for each protein family were then used to query 118 shotgun oral metagenomes from the Human Microbiome Project (HMP) (Human Microbiome Project Consortium, 2012a; Human Microbiome Project Consortium, 2012b) (https://portal.hmpdacc.org/) ( Supplementary Table S5 ) and seven oral metagenomes from a Spanish collection (Belda-Ferre et al., 2012). These 125 metagenomes corresponded to supragingival plaque samples. Due to the absence of metagenomic data from subgingival sites in the HMP, a “synthetic” subgingival metagenome was generated ( Supplementary Table S6 ). To select the information contained in this metagenome, the 16S rRNA gene sequencing data from subgingival and supragingival sites available in the HMP (Human Microbiome Project Consortium, 2012b) was investigated. First, a comparison of the relative abundances of the bacterial genera present in subgingival and supragingival niches was done to evaluate the resemblance between both sites ( Supplementary Table S7 ). Then, the bacterial genera of subgingival sites with abundance values over 1% were selected, and the genomes of the most representative species from each genera were retrieved ( Supplementary Table S6 ) to generate the synthetic subgingival metagenome.

Sequence searches were performed using DIAMOND (Buchfink et al., 2021) with a minimum subject coverage of 70% and an e-value cut-off of 1e-10. Alignment results were parsed into tables using in-home scripts (available at https://github.com/genomica-fciencias-unam/tabla_genes). Statistical analyses were done using R (R Core Team, 2022) with the phyloseq (McMurdie and Holmes, 2013), ggplot2 (Wickham, 2009), vegan (Oksanen et al., 2017), and tydiverse (Wickham et al., 2019) packages. All R procedures are available as a jupyter notebook in the Supplementary Material .

Statistical analyses were performed using Prism 8.3.0 (GraphPad, San Diego, CA, USA, www.graphpad.com) or R (R Core Team, 2022). Two-tailed Student’s t-tests (referred to in the text as t-tests) or ANOVAs were performed for normally distributed samples. Mann-Whitney tests were performed for non-normally distributed samples to compare control and Aii20J-treated biofilm mass and cell index values within the same donor. Wilcoxon matched-pairs signed-rank tests (referred to in the text as Wilcoxon tests) were performed to assess the significance of the differences between biofilms that were reduced in the presence of the enzyme and those that were increased in the presence of the enzyme, subject by subject, grouped according to their oral health status. Quantitative PCR results were analyzed using Mann-Whitney tests for independent samples. For all the statistical analyses, significant differences were determined with an α = 0.05.

To assess the potential use of Aii20J in the prevention of oral diseases, the biofilm inhibitory capacity of the enzyme was first explored using two cultivation techniques: the AAA model (Exterkate et al., 2010; Muras et al., 2020a) and the xCELLigence system (ACEA, Biosciences Inc.). The AAA model was assembled with vertically placed glass coverslips onto which bacterial cells can only attach in an active manner. The AAA model allows the refreshment of culture media, further prompting biofilm formation. Besides, the substrata can be easily manipulated at the end of the incubation time to perform quantification assays, microscopy observations, and DNA extraction, allowing a complete characterization of the biofilms that is hampered by the use of the CV assay alone (Muras et al., 2018b). On the other hand, the xCELLigence system is a real-time monitoring technique based on cell index (CI) values that correlate to the amount of biofilm formed onto the bottom of modified culture plates (Mira et al., 2019). This system has been used to monitor oral biofilm formation (Mira et al., 2019), but is not suitable to detect structural differences in biofilms. Both systems were inoculated with saliva and subgingival samples from 17 subjects ( Supplementary Table S1 ). Samples were grown aerobically in BHI medium and anaerobically in SCH-K1. Both culture media contain a high protein load, but the Schaedler medium has additional components that favor the growth of anaerobes, such as cysteine and hemin, and in addition to the commercial recipe, vitamin K1 (Murray, 1978; Marsh et al., 2016). The use of either aerobic or anaerobic atmospheres aimed to assess whether the effect of the enzyme would be different on bacterial communities grown under such conditions. We observed important variations in biofilm formation capacity among samples ( Supplementary Figures S3, S4 ) without significant differences in biomass values between healthy and PD groups grown in the AAA model (t-test) ( Supplementary Figure S3 ). In the xCELLigence system, PD samples yielded significantly higher cell index values than healthy samples for subgingival biofilms grown in BHI and saliva biofilms grown in SCH-K1 (t-test) ( Supplementary Figures S4A, D ). In addition, initial biofilm mass achieved in the control condition did not correlate with sensitivity to Aii20J in any cultivation system (Pearson r) (data not shown). Regarding the effect of Aii20J on biofilm formation, when cultivated in the AAA model, subgingival biofilms were more sensitive to the enzyme than saliva biofilms in both culture media ( Figure 1 ). Analysis of the effect of the enzyme on subgingival biofilms grouped according to the oral health status of the patients showed a significant decrease in PD-derived biofilms grown in BHI and both healthy- and PD-derived biofilms grown in SCH-K1 (Wilcoxon test) ( Figures 1A, B ). For saliva biofilms cultured in the same conditions in the AAA model, only PD samples grown in SCH-K1 presented a significant biomass decrease (Wilcoxon test) ( Figures 1C, D ). However, the xCELLigence system was much less sensitive than the AAA model in detecting the effect of Aii20J, as significant changes were only found in two saliva-derived biofilms treated with Aii20J, either in BHI or SCH-K1 (t-test) ( Supplementary Figure S2 ). Therefore, the AAA biofilm cultivation model and subgingival samples were selected for further experiments.

Additionally, we evaluated the presence of QS inhibitory activity among the initial saliva samples against short- (C₆-HSL) and long-chain (C₁₂-HSL) AHLs with standardized bioassays (Romero et al., 2011). QQ activity against C₁₂-HSL was common, with five out of seven saliva samples from healthy donors and seven out of eight saliva samples of PD origin inhibiting AHL detection by the biosensor strain VIR07 ( Supplementary Figure S5 ). Conversely, QQ against C₆-HSL was scarce, with only two healthy and one PD saliva samples decreasing violacein production in the biosensor strain CV026 ( Supplementary Figure S5 ). The small volume of the saliva samples collected did not allow assessing the presence of AHLs in these samples.

To assess which culture media and incubation conditions resulted in higher microbial diversity within the PD-derived biofilms, we chose four culture conditions: BHI-aerobiosis, BHI-anaerobiosis, SCH-K1-anaerobiosis, and in addition, McBain-anaerobiosis ( Figures 2A, B ). The use of McBain culture medium provides with a high protein load, and like the SCH-K1 medium, McBain comprises cysteine, hemin, and vitamin K1, that favor the proliferation of Gram-negative anaerobes. To achieve a higher resemblance with the oral environment, the McBain medium contains also mucin, one of the major salivary polypeptides, present in the acquired enamel pellicle. Experiments were performed using subgingival samples, as they are expected to be more representative of the periodontal microbiota than saliva samples. Regarding microbial diversity, the genus Streptococcus was predominant in all biofilms ( Figure 2B ), but McBain medium harbored the most diverse biofilms, with increased proportions of genera such as Veillonella, Granulicatella, Gemella, and Haemophilus both in control and Aii20J-treated biofilms, as compared with the other culture media. The presence of the QQ enzyme resulted in significant reductions of subgingival-derived biofilm mass even in Gram-positive predominated biofilms obtained in BHI and SCH-K1 media, without significantly changing bacterial relative abundance at the genus level ( Figures 2A, B ).

Once we established the optimal conditions for obtaining diverse oral biofilms in the AAA model, we studied a greater number of samples, assessing changes in biofilm mass and microbial diversity in the presence of Aii20J. Subgingival samples were chosen as inocula since they present higher sensitivity to Aii20J than saliva samples, as seen in previous experiments ( Figure 1 ). For these experiments, 35 patients were recruited; 11 healthy donors, three patients with gingivitis, and 21 with periodontitis ( Supplementary Table S2 ). Biofilms were grown in BHI-anaerobiosis and McBain-anaerobiosis to compare the effect of Aii20J on biofilms dominated by Gram-positive bacteria and in more diverse biofilm communities, respectively. As seen with the initial set of patients, biofilm formation abilities varied widely depending on the donor ( Supplementary Figure S6 ), with no significant differences in biomass values between healthy and PD groups being observed (t-test) ( Supplementary Figure S6 ). Biomass values of untreated biofilms did not correlate with sensitivity to Aii20J (Pearson r) (data not shown).

Biomass was reduced in the presence of Aii20J in biofilms grown in both BHI-anaerobiosis and McBain-anaerobiosis, with few exceptions ( Figure 3 ), confirming previous results obtained in Streptococcus-dominated biofilms ( Figure 1 ). When the effect of Aii20J was analyzed by grouping the biofilms according to the oral health classification of the patients, PD-derived biofilms were significantly reduced in all culture conditions, whereas healthy-derived biofilms were significantly reduced only when grown in BHI (Wilcoxon test) ( Figure 3 ). In subgingival, PD-derived biofilms, macroscopical differences ( Supplementary Figure S1 ) were detected in 71% of biofilms grown in BHI and 38% of biofilms grown in McBain in the presence of Aii20J, and biomass decrease was statistically significant in 50% and 42% of the samples, respectively (t-test) ( Figure 3 ). In subgingival, healthy-derived biofilms, macroscopical differences were found in 91% of biofilms grown in BHI and 18% of biofilms grown in McBain. Additionally, biomass reduction due to Aii20J was significant in 64% of healthy biofilms grown in BHI and 18% of healthy biofilms grown in McBain (t-test) ( Figure 3 ). Assessment of additional controls, i.e., the fraction smaller than 10 kDa resulting from the purified Aii20J solution and the heat-inactivated Aii20J, yielded no differences in biofilm mass compared with water-only negative controls in biofilms from two periodontal donors susceptible to Aii20J ( Supplementary Figure S7 ).

Subgingival biofilms from patients P30 and P31 grown in McBain were visualized under CLSM to compare biofilms that presented different responses to Aii20J in the CV assay ( Figure 3B ). Biofilms from patient P30 were significantly reduced in biomass in the CV assay and also showed a significant reduction in biofilm coverage when comparing CLSM photographs of control (11.0 ± 4.9%) and Aii20J-treated biofilms (5.8 ± 3.7%) (t-test) ( Supplementary Figure S8 ). On the other hand, biofilms from patient P31 presented no significant differences in biofilm mass in the CV assay nor in biofilm coverage between control (3.26 ± 2.57%) and Aii20J-treated biofilms (3.57 ± 1.43%) ( Supplementary Figure S8 ). However, control biofilms of patient P31 were predominantly stained with PI, whereas Aii20J-treated biofilms were predominantly stained with SYTO 9 ( Supplementary Figure S8 ).

In parallel to biofilm quantification, biofilm mass from control and Aii20J-treated samples grown in McBain was collected for DNA extraction, as we previously observed that this culture medium allowed higher microbial diversity in biofilms from subgingival samples ( Figure 2 ). Sequencing of the 16S rRNA gene confirmed the dominance of the genus Streptococcus, reaching a 45 – 61% relative abundance in the biofilms ( Figures 4A–C ). Veillonella, Haemophilus, Granulicatella, Gemella, and Aggregatibacter were also found in all biofilms within the top ten most relatively abundant genera. Beta diversity examined with Principal Coordinate Analysis (PCoA) and alpha diversity indexes revealed no significant differences in microbial composition between control and Aii20J-treated biofilms ( Figures 4D–F , Supplementary Figure S9 ). In addition to diversity indexes, a differential abundance analysis was performed grouping the biofilms according to the oral health status of the donors. In periodontitis-derived biofilms, the genus Porphyromonas, among others, significantly decreased its differential abundance in the presence of Aii20J ( Figure 5A ). Further quantification of Porphyromonas gingivalis and Porphyromonas endodontalis, both species linked to oral diseases (Hajishengallis et al., 2012; de Brito et al., 2020), was done using qPCR on periodontitis-derived biofilms (n = 19). No significant differences were found for P. gingivalis nor P. endodontalis CFUs between control and Aii20J-treated biofilms (Mann-Whitney test) ( Figure 5B ). Total 16S rRNA gene quantification for Eubacteria was performed as a normalization control.

Last, supernatants of untreated biofilms (n = 16) were investigated for the presence of the most common AHLs (from C₄-HSL to C₁₈-HSL, either with or without oxo and hydroxy substituents) by HPLC-MS analysis. None of these standard AHLs were identified in any of the samples.

The absence of standard AHLs in the supernatants of biofilm cultures and the impossibility of performing the same analytical procedures to detect AHLs on the original saliva and subgingival samples due to their reduced volume prompted the investigation of the presence of sequences related to AHL synthesis and sensing, as well as AHL enzymatic degradation in 125 metagenomes from supragingival plaque samples available in public databases (Belda-Ferre et al., 2012; Human Microbiome Project Consortium, 2012a; Human Microbiome Project Consortium, 2012b), and one synthetic subgingival metagenome that comprised the most representative bacterial species from the subgingival niche ( Supplementary Table S6 ), and that displayed very similar results to those of the supragingival metagenomes in terms of sequences found and their abundance. The analysis of these metagenomes revealed a surprisingly high abundance of proteins homologous to the AHL synthase HdtS ( Figure 6 , Supplementary Table S4 ). This protein family comprised three HdtS sequences from Pseudomonas, and it appeared in nearly every metagenome, almost as abundantly as the four protein families of LuxS, included as reference in this analysis ( Figure 6 ). Among the AHL receptors investigated, three protein families appeared in elevated abundance in both the supragingival metagenomes and the synthetic subgingival metagenome. Representative members of these families are AbaR of Acinetobacter, RhlR of Pseudomonas, and the promiscuous AHL-receptor SdiA of Burkholderia ( Figure 6 ). These protein families appeared present in a high number of metagenomes, although their abundance was lower than that observed for HdtS. It is worth mentioning that the protein family represented by RhlR also comprised several SdiA sequences, including those from Escherichia coli and Salmonella, that were divergent from those classified together with the SdiA sequences of the Burkholderia family.

In addition to AHL synthases and receptors, we investigated the supragingival and subgingival metagenomes for the presence of AHL-inactivating enzymes: lactonases, acylases, and paraxonases. Protein families of lactonase and acylase enzymes appeared in extremely high abundances, in some cases, even higher than LuxS sequences ( Figure 6 ). These in silico results align with those obtained in vitro on the high abundance of QS inhibitory activity among saliva samples ( Supplementary Figure S5 ).