The effect of sex steroid hormones on the ecology of in vitro oral biofilms

Sex steroid hormones (SSH) such as oestrogen, progesterone and testosterone are cholesterol derived molecules that regulate various physiological processes. They are present in both blood and saliva, where they come in contact with oral tissues and oral microorganisms. Several studies have confirmed the effect of these hormones on different periodontal-disease-associated bacteria, using single-species models. Bacteria can metabolize SSH, use them as alternative for vitamin K and also use them to induce the expression of virulence factors. However, it is still unclear what the effects of SSH are on the oral microbiome. In this study, we investigated the effects of four SSH on commensal in vitro oral biofilms. Saliva-derived oral biofilms were grown in Mc Bain medium without serum or menadione using the Amsterdam Active-Attachment model. After initial attachment in absence of SSH, the biofilms were grown in medium containing either oestradiol, oestriol, progesterone or testosterone at a 100-fold physiological concentration. Menadione or ethanol were included as positive control and negative control, respectively. After 12 days with daily medium refreshments, biofilm formation, biofilm red fluorescence and microbial composition were determined. The supernatants were tested for proteolytic activity using the Fluorescence Resonance Energy Transfer Analysis (FRET). No significant differences were found in biofilm formation, red fluorescence or microbial composition in any of the tested groups. Samples grown in presence of progesterone and oestradiol showed proteolytic activity comparable to biofilms supplemented with menadione. In contrast, testosterone and oestriol showed a decreased proteolytic activity compared to biofilms grown in presence of menadione. None of the tested SSH had large effects on the ecology of in vitro oral biofilms, therefore a direct translation of our results into in vivo effects is not possible. Future experiments should include other host factors such as oral tissues, immune cells and combinations of SSH as present in saliva, in order to have a more accurate picture of the phenomena taking place in both males and females.


Introduction
Sex steroid hormones (SSH) are cholesterol-derived molecules, mostly secreted by the testes and the ovaries, and are responsible for a variety of physiological processes in the human body [1,2]. These hormones are known for their role in sexual development and reproduction. However, they are extremely versatile in their functions and are involved in a wide range of non-reproductive processes. For example, they are capable of influencing bone and tissue metabolism, cardiovascular and brain function, the immune response, amongst others [3][4][5][6][7].
After being secreted into the bloodstream, SSH reach different target tissues, including the oral tissues and the periodontium. SSH diffuse into the salivary glands' acinar cells to then reach the oral cavity [8].
During periods of hormonal fluctuations such as puberty and pregnancy SSH trigger various responses, including inflammation and proliferation of cells [9,10]. These responses result in gingival inflammation, also known as gingivitis [11,12]. However, not only target tissues are influenced by the presence of SSH. Bacteria and other microorganisms living in and on the body also encounter SSH in their habitat and the oral cavity is no exception [9,13,14].
Several in vitro studies have shown that mainly -but not exclusively-periodontal-disease-associated bacteria can respond to the presence of SSH. In particular, three groups of SSH and their derivates, namely: progestogens, oestrogens and testosterone [13] can modify bacterial metabolism, growth and expression of virulence factors [15][16][17].
A characteristic virulence factor of several periodontal-disease-associated bacteria is their ability to produce protein-degrading enzymes (proteases) [18] which play a crucial role in the development of periodontal disease [19].
The effects of SSH on bacteria and yeast have been studied in several body niches, such as the gut [20] vagina [21] and respiratory tract [22], both in vitro and in vivo.
In the mouth, the effects of SSH on bacteria have been studied mostly in single-species models on periodontal-disease-associated bacteria. The reason behind this is based on the premise that periodontal-diseaseassociated bacteria can use SSH to replace a vital component to their growth: vitamin K [16][17][18][19][20][21][22][23] observed that oestrogen and progesterone could compensate for the absence of vitamin K in the medium and thus support the viability and growth of the periodontal pathogens Porphyromonas gingivalis and Prevotella melaninogenica, amongst other periodontal-disease-associated bacteria. Since then, many in vitro studies have assessed different aspects of the interactions between SSH and oral bacteria [13].
In vivo fluctuations of SSH, both intrinsic (puberty, menstrual cycle and pregnancy) and extrinsic (use of artificial hormones), can lead to clinical and microbiological changes in the oral cavity [9,[24][25][26][27][28]. Results from studies focusing on microbiological changes of specific bacteria and studies using broader approaches such as 16S rRNA gene amplicon sequencing suggest that hormonal shifts are associated with changes in the oral tissues and the oral microbiome. Interestingly, these changes are not limited to periodontal-disease-associated bacteria. Several studies have also reported interactions between SSH and caries associated bacteria, health associated bacteria and oral yeasts [13].
The effects of SSH on oral health result from the complex interaction between the host and its microbiome. In the present study we focus on the effects of SSH on the oral microbiota using a multi-species in vitro approach. This could help better understand the pathophysiology of non-dental plaque induced gingival disease. The aim of this in vitro study was to test the individual effects of oestradiol, oestriol, progesterone and testosterone on the growth, composition and metabolism of commensal oral microcosm biofilms.

Saliva collection and storage
In analogy with previous studies [29], saliva was collected from two self-reported healthy donors (1 female, 1 male), following the unstimulated saliva spitting method protocol [30]. Donors refrained from dental hygiene for at least 24 h and did not eat or drink for 2 h before the collection. The collected saliva was pooled, mixed 1:1 with 60% sterile glycerol and aliquoted. The study was approved by the ACTA Ethical Committee (protocol number: 2021-40362).

Red fluorescence of biofilms
Prior to harvesting, red fluorescence of the biofilms was determined [29]. The lid containing the biofilms was inverted and photographed in a dark room using a QLF-D camera system (Inspektor Research Systems BV, Amsterdam, the Netherlands) via image capture software (C31. 25 Inspektor Research Systems BV, Amsterdam, the Netherlands). Red fluorescence was assessed qualitatively as presence or absence. It is a typical characteristic of anaerobic mature oral biofilms to fluoresce red [38]. This is due to their ability to accumulate porphyrins [39].

Harvest of the biofilms
To harvest the biofilms, all the glass coverslips were aseptically removed from the lid and transferred to sterile tubes containing 2 ml of phosphate buffered saline (8.0 g/l NaCl, 0.2 g/l KCl, 1.0 g/l Na 2 HPO 4 , 0.2 g/l KH 2 PO 4 ) set at 7.4 pH. Biofilms were dislodged from the glass by sonication using a Vibracell VCX130 sonicator with a maximum of 130 W and 20 kHz for 45 s, with a pulse rate of 50% and pulses of 1 s at a vibration amplitude of 40% (Sonics & Materials, Newtown, USA). Samples were kept on ice to prevent heating.

Biofilm viability
To quantify biofilm viability, the number of anaerobic Colony Forming Units (CFU) of each biofilm was determined. Serial dilutions of the dispersed biofilms were plated on tryptic soy blood agar (TSBA) plates using an Eddy jet spiral plater (Neu-tec Group Inc., Farmingdale, NY). Plates were incubated anaerobically at 37 • C for 7 days and CFUs were counted. The remaining undiluted dispersed biofilms were centrifuged for 10 min at 21,300×g at 4 • C. Supernatants were discarded and the pellets were stored at − 80 • C for DNA isolation and 16S rRNA gene amplicon sequencing.

Spent medium pH assessment
The pH of the medium after growth was determined using an iridium microelectrode (Beetrode MEPH-1, WPI Instruments, New Haven, Conn, USA) connected to a daily calibrated Orion SA720 pH/ISE Meter (Orion Research, Boston, Mass, USA).

Protease activity using a fluorescence resonance energy transfer analysis (FRET)
Protease activity in the spent medium was determined using the FRET analysis as described previously [40]. Two probes were used: (1) PEK-054 to detect unspecific increased proteolytic activity and (2) BikKam-15 [40] to detect specific P. gingivalis increased proteolytic activity. For this, 100 μl of Tris Buffered Saline (TBS: 50 mM Tris and 150 mM NaCl, pH 7.6) were added to each well of a Blackwell clear-bottom 96-well plate (Corning, Lowell, MA). Subsequently, 100 μl of the filtered culture supernatants were added in duplicates. Lastly, 4 μl of the selected FRET-probe (800 μM) were added to each well and the fluorescence was assessed immediately. Fluorescence intensity (excitation: 485 nm, emission: 530 nm) was measured every 5 min for 120 min at 37 • C using a SpectraMax i3x microplate reader (Molecular Devices, San Jose, CA). The protease activity was calculated from the initial linear slope and defined in Relative Fluorescence Units (RFU) per minute (RFU/min). All the values were then transformed to values relative to the carrier control, which was set at 100%.

DNA isolation
DNA isolation from dislodged biofilms was performed as described previously [41]. Briefly, biofilm pellets were thawed, resuspended with 150 μl Tris EDTA buffer (10 mM Tris-Cl (pH 8.0), 1 mM EDTA (pH 8.0)) and transferred to a well in a 96 deep well plate (Axygen Scientific Inc., CA, USA). As controls, the original inoculum and sterile McBain medium without additions from each experiment were included.
To each well, the following compounds were added: 250 μl of 0.1mm diameter Zirconia beads (BioSpec Products, Bartlesville, OK, USA), 200 μl of phenol (Rotiphenol, Carl Roth GMBH&Co. KG, Germany) and 200 μl of lysis buffer (MagMini DNA isolation kit, LGC Genomics Ltd, UK). The plate was then sealed and placed in a Mini-BeadBeater-96 (BioSpec Products, Bartlesville, OK, USA) for 2 min at 2.100 oscillations/min. When this process was completed, DNA was extracted and purified with the MagMini DNA Isolation Kit (MagMini DNA isolation kit, LGC Genomics Ltd, UK). Bacterial DNA concentration was determined by qPCR as described elsewhere [42].

16S rRNA gene amplicon sequencing and data processing
To assess the bacterial composition of the biofilms exposed to the different hormones, DNA was further processed for sequencing as described above. For this, the V4 hypervariable region of the 16S rRNA gene was used and the equimolar mix was sequenced using the Illumina MiSeq platform (Core Facility Genomics, AmsterdamUMC, The Netherlands). The paired-end reads were quality-filtered, merged and clustered into operational taxonomic units (OTUs) at 97% similarity as described by Koopman et al. [43]. The most abundant sequence of each OTU was assigned a taxonomy using the Ribosomal Database Project (RDP) classifier [44] implemented in QIIME 1.9.1 and the SILVA database (version 132, [45]). The OTU table was subsampled at a depth of 13,950 reads per sample to allow comparisons among the samples.

Statistical analyses
One-way ANOVA and the Bonferroni post-hoc test were performed to analyse the total cultivatable cell counts. Results derived from the FRET analysis were assessed using ANOVA and no post-hoc correction was used [46]. Maximal activity, represented by RFU/min was analysed. Differences were considered statistically significant if p < 0.05.
The Shannon diversity index (unbiased version) and observed OTUs (richness), calculated using PAST v4.10 [47] were used to characterize the α-diversity. Next, the Kruskal-Wallis test and Dunn's test for multiple comparisons were performed using GraphPad v9.4.1 for Windows (GraphPad Software, San Diego, California USA, www.graphpad.com). The log2-transformed sequencing data (OTU table) was ordinated using a Principal-Component Analysis (PCA) to compare two groups of experimental runs. Statistical differences in the microbial profiles were analysed with one-way Permutational Multivariate Analysis of Variance (PERMANOVA), using the Bray-Curtis distance and 9999 permutations, to identify possible differences in microbiological profiles between the different treatment conditions (Menadione, Ethanol (carrier), Oestradiol, Oestriol, Progesterone and Testosterone). The latter analyses were performed using R version 4.1.3 [48] and RStudio (RStudio: Integrated Development for R. RStudio, PBC, Boston, MA). Differences were considered statistically significant if p < 0.05.
The metagenomic biomarker discovery tool Linear discriminant analysis Effect Size (LefSe) [49] was used to identify microbial taxa that significantly differed between the treatment group (a hormone) versus ethanol (carrier control). This was done using a filtered OTU table where OTUs with ≤100 reads were not considered in the analysis. Default LEfSe settings were used. The online Galaxy platform (http://huttenh ower.sph.harvard.edu/galaxy) was used to perform the LEfSe analyses.

Red fluorescence, biofilm formation and pH
In our experiment, no differences in the red fluorescence of the biofilms were observed (Fig. 1). The total cultivatable cell counts did not show any differences between the groups (Fig. 2). Similarly, the pH of the spent medium did not show any hormone-dependent differences (7.09 ± 0.25).

Effect of SSH on proteolytic activity
Differences between the groups were observed when assessing specific and unspecific proteolytic activity. Samples from days 4, 8 and 12 were tested. Specific P. gingivalis activity was not observed in any of the samples (data not shown). Unspecific proteolytic activity using the probe PEK-054 showed a separation in two groups on days 4, 8 and 12. The highest proteolytic activity was detected on the samples with added menadione, oestrogen and progesterone. The lowest proteolytic activity group was detected on the samples with added oestriol and testosterone (Fig. 3). This trend was the most distinct at day 12.

α-diversity
The average relative abundance of the 15 most abundant bacterial genera for each condition are shown in Fig. 4. All conditions presented a similar composition despite the addition of different compounds. All samples were composed of approximately 20% Megasphaera, 20% Veillonella, 10% Prevotella and 50% of a variety of taxa including Fusobacterium, Parvimonas, Leptotrichia amongst others. A slight, but not significant difference, in the relative abundance of Parvimonas and Leptotrichia was observed in the samples with the lowest proteolytic and Shannon diversity were calculated to assess α-diversity (Fig. 5). The number of OTUs in the inoculum was 106 (data not shown). The number of OTUs in all samples was 54 ± 10. Comparing the number of OTUs between the different conditions, oestriol (p = 0.0145) and testosterone (p = 0.045) exhibited a significantly higher diversity as compared to the negative control (carrier) (Fig. 5a). The Shannon diversity index did not show any significant differences between the groups (Fig. 5b).

Microbial composition of the microcosm biofilms
To investigate possible microbial composition differences in the biofilms induced by SSH, bacterial composition of the biofilms was determined using 16S rRNA gene sequencing. The microbial profiles were plotted using a Principal Component Analysis (PCA) (Fig. 6). No obvious clustering of samples for each treatment was detected. Permutational Multivariate Analysis of Variance analysis (PERMANOVA) showed significant differences in composition only between the biofilms exposed to ethanol (carrier) and testosterone (p = 0.026). The differences between the biofilms exposed to menadione and oestriol approached the cut-off value for statistical significance (p = 0.055). Accordingly, linear discriminant analysis effect size (LEfSe) did not show any significant differences in specific bacteria between the tested conditions and ethanol (carrier control).

Reproducibility of the experiments
To evaluate the inter and intra experiment reproducibility, the Bray-Curtis similarity indices were averaged within a batch and between the two experimental batches, for each condition. Both experiments showed an intra experiment similarity (range: 0.79-0.86, 0.83-0.86 resp.), and high inter experiment similarity (range: 0.82-0.85) indicating that the experimental approach is highly reproducible.

Discussion
In the present study, we assessed the response of an in vitro oral microcosm to the presence of oestradiol, oestriol, progesterone and testosterone. The response of the biofilms to these compounds did not differ significantly on most of the tested parameters. However, biofilms grown in the presence of oestradiol and testosterone showed a lower proteolytic activity and a higher microbial diversity (OTUs) than the other SSH. All things considered, none of the tested SSH had large effects on either the growth or composition of in vitro oral biofilms, therefore a direct translation of our results into in vivo effects is not possible.
The aim of the study was to assess possible changes in a healthy commensal microcosm, similar to the study by Janus et al. [32]. For this reason, saliva was selected as the inoculum. Saliva can be collected in an easy and non-invasive manner. Also, saliva is regarded as a fair representation of several niches in the oral cavity [50].
Several in vitro studies have assessed the effects of SSH on singlespecies models, with positive results [13]. Mostly periodontal-disease-associated bacteria have been studied with several reports documenting their ability to use SSH to support different biological processes [15][16][17]. Therefore, it is remarkable that in the present study: (1) a very limited effect of SSH could be measured in commensal oral microcosm biofilms and; (2) there was no clear selection of periodontal-disease-associated bacteria as determined by 16S rDNA sequencing.
It is unexpected that both menadione and hormones did not show large differences in our model system. The experimental design was aimed to test the exclusive effect of SSH on commensal oral biofilms. For this reason, menadione was used as a positive control as it is known to be vital for the growth of certain periodontal-disease-associated bacteria [51]. In our experiment, menadione did not specifically support the growth of these bacterial species (Fig. 4). This suggests that the addition of menadione does not significantly influence the composition of a multi-species commensal oral biofilm as would be expected based on single-species models. This could be explained by previous studies that have evaluated the synthesis of vitamin K by specific microbes. Ibrahim et al. [52] reported that oral Prevotella spp. can synthesize vitamin K from chorismite, with the help of MenA-MenG enzymes (classical menaquinone pathway). These enzymes have also been identified in Megasphaera spp. [53], V. parvula [54], F. nucleatum [55], Actinomyces [56], and Porphyromonas [57][58][59]. Also, C. rectus is capable of producing vitamin K from chorismite via de Futalosine pathway by action of MqnABCD enzymes [60]. The aforementioned genera were present in our samples (Fig. 4). Several of these genera have been observed in another study using a similar methodology [61]. The absence of difference in composition between biofilms supplemented with menadione and the carrier control suggests that bacteria present in oral microcosm biofilms are indeed capable of producing vitamin K.
While we did not detect differences in microbial composition of oral biofilms grown in the presence of different SSH, significant differences in proteolytic activity of these biofilms were observed. When measuring proteolytic activity, samples supplemented with a known analogue for vitamin K (oestrogen and progesterone) showed a level of proteolytic activity comparable to the one exhibited by biofilms supplemented with menadione. In contrast, addition of oestriol and testosterone resulted in a significantly lower proteolytic activity. This suggests that oestriol and testosterone may not function as substitutes for vitamin K.
The observed difference in proteolytic activity could not be explained based on the microbial composition. However, differences in enzyme activity might be related to differences in expression. For instance, Parvimonas micra has been reported to stimulate the expression Biofilm formation was estimated based on total viable cultivatable cell counts (CFU). Positive control: menadione (medium with added 0.01% menadione); Negative control: carrier (medium with added 0.01% absolute ethanol). Total cultivatable cell counts were ~10 8 , with no differences between the groups. Two independent experiments were performed with four samples per condition.
of the virulence factor gingipains of P. gingivalis [62] In our study, we did not evaluate the effects of SSH on gene expression in biofilms. For now, we can only conclude that the presence of oestradiol and progesterone induced a small yet significant increase of proteolytic enzymes similar to menadione. On the contrary, oestriol and testosterone induced a decrease in the presence of these enzymes.
Periodontal diseases such as gingivitis and periodontitis are often characterised by the presence of dysbiotic biofilms [63][64][65]. Also, periodontal-disease-associated bacteria are known to be more proteolytic and commonly produce proteases [66,67]. These proteases have been recently clinically measured using a FRET-based assay [68]. Bikker et al. reported an increase of salivary proteolytic activity on subjects taking part in an induced-gingivitis clinical study. This increased proteolytic activity could be of key importance to assess a patient's risk of developing periodontal disease. Although the in vitro differences were statistically significant, we should be careful to directly translate these results to an in vivo situation.
There is consensus that sex steroid hormones can cause morphological and metabolic changes of the gingival tissues during periods of hormone surges [9,69]. Direct changes in oral biofilms by hormonal changes though, are understudied and remain controversial. Changes in periodontal-disease-associated bacteria have been reported during puberty [28], the menstrual cycle [25], pregnancy [11,24,70,71], on women with hormonal disorders [72] and by users of synthetic SSH [73]. These reports all have something in common: changes in periodontal-disease-associated bacteria occur during periods of hormonal fluctuations. Taking into consideration that there might be an adaptation of the oral microbiome to the constant presence of SSH in vivo, distortion of the regular balance could explain compositional changes. To test this in an in vitro model, a longer exposure time would be needed along with changes in the concentration of added SSH to mimic hormonal fluctuations.
Also, several studies report a sex-preference in the incidence of periodontitis, with men being more susceptible than women [74]. However, based on our in vitro study this difference cannot be associated to distinct sex-specific-SSH effects on the oral microbiota. The absence of SSH-induced effects suggests that SSH alone are not responsible for changes in the oral microbiome in vivo. SSH play multiple roles including in the physiology of the host by modulating the immune system [75][76][77]. It is certainly possible that this host response to SSH modulates the microbiome to a larger extent than the direct modulation of the microbiome by SSH.
The complex crosstalk between the oral microbiome, SSH and oral tissues include many different factors that influence the microbiome and host's response in ways that have yet to be explained. In our in vitro model, SSH differentially modulated the proteolytic activity of oral biofilms. To our knowledge, there are no other studies that have tested this on an in vitro oral microcosm. Future experiments should include other host factors such as oral tissues, immune cells and combinations of SSH as present in saliva, in order to have a more accurate picture of the phenomena taking place in both males and females.
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Fig. 3. Unspecific proteolytic activity of the spent medium expressed in percentage relative Fluorescence Units relative to carrier (0.01% added absolute ethanol). Measurements were performed using FRET assay and PEK-054 probe. A. Proteolytic activity measured on biofilm's supernatant after 4 days B. Proteolytic activity measured on biofilm's supernatant after 8 days C. Proteolytic activity measured on biofilm's supernatant after 12 days. Significant differences (*p ≤ 0.05) can be seen on days 4 and 12 between the carrier and SSH. Results from day 8 were not statistically significant. Only comparisons between carrier and SSH are shown. Two independent experiments were performed with four samples per condition.

Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Corresponding author (Bastiaan P. Krom) is serving in an editorial capacity for the journal we are submitting to

Data availability
Data will be made available on request.