Nisin Probiotic Prevents Periodontal Disease and Inammation while Promoting Periodontal Regeneration and a Shift Toward a Healthy Microbiome/Virome

Background Dysbiosis of the oral microbiome mediates chronic periodontal disease, including its characteristic bone loss and host inammatory response. Realignment of this microbial dysbiosis towards health may prevent disease. Treatment with antibiotics and probiotics can modulate the microbial, immunological, and clinical landscape of periodontal disease with some success. Antibacterial peptides or bacteriocins, such as nisin, and nisin-producing probiotics, Lactococcus lactis, have not been examined in this context, yet warrant further examination because of their biomedical benets in eradicating biolms and oral pathogenic bacteria, and modulation of immune mechanisms. The goal of this study was to examine the potential for nisin and a nisin-producing probiotic to abrogate periodontal bone loss and related inammatory landscape while modulating the composition of the oral microbiome. A polymicrobial mouse model of periodontal disease was employed for this purpose. In a disease context, nisin and the nisin-producing Lactococcus lactis probiotic signicantly decreased the levels of periodontal pathogens, alveolar bone loss, oral inammatory host response, and host-antibody response to these pathogens. Surprisingly, nisin and/or the nisin-producing L. lactis probiotic also enhanced the number of gingival broblasts, periodontal ligament cells, and bone lining cells in response to the polymicrobial infection. Nisin and probiotic treatment signicantly shifted the oral bacteriome and virome towards the healthy control state. This shift was characterized by a unique signature; health was associated with a Proteobacteria (Marinobacter sp. B9-2), whereas 3 retroviruses (Golden Hamster Intracisternal A-particle H18, Bat gammaretrovirus, and Porcine type C oncovirus) were associated with disease. Specic disease-associated microbial species were highly correlated with IL-6 levels. of probiotics to suppress periodontal pathogens or anaerobic bacteria in human and animal studies have shown some benets. Human studies exploring probiotics as monotherapy or adjunctive therapy have shown some benet or neutral effects in reducing periodontal pathogens or anaerobes with the probiotics Lactobacillus salivarius WB21[44], L. reuteri[18, 45, 46, 47], bacillus[48], L. plantarum[33], L. rhamnosus SP1[49], B. lactis[31] and various Streptococci[24, 32, 50]. Animal studies also showed some benets. When Lactobacillus brevis or Bidobacterium lactis were applied in a murine model of periodontitis, there was a signicant decrease in the counts of anaerobic bacteria relative to aerobic bacteria[51, 52]. Use of Lactobacillus rhamnosus GG showed no antimicrobial activity against P. gingivalis and F. nucleatum[53]. The current investigation demonstrated that the nisin-producing probiotic and nisin itself reduced the oral levels of all the Red complex bacteria, P. gingivalis, T. forsythia, and T. denticola, and the orange complex pathogen, F. nucleatum, indicating nisin/nisin probiotic’s ecacy in consistently removing these pathogens from oral surfaces. In summary, this study highlights an approach to realign the oral microbial dysbiosis of periodontal disease and its related sequalae (bone loss, host immune response) towards health. Treatment with antibiotics and probiotics have been used to modulate the microbial, immunological, and clinical landscape of periodontal disease with some success. Antibacterial peptides or bacteriocins, such as nisin, and nisin-producing probiotics, such as Lactococcus lactis, have not been examined in this context. However, they warrant examination because of their well characterized biomedical benets in eradicating biolms and oral pathogenic bacteria, while also modulating immune mechanisms. This study demonstrates that nisin and nisin probiotic treatment inhibit periodontal disease-related bone loss and host immune responses while signicantly shifting the oral bacteriome and virome towards the healthy control state. This shift was characterized by a unique signature where health was associated with a Proteobacteria (Marinobacter sp. B9-2), whereas 3 retroviruses (Golden Hamster Intracisternal A-particle H18, Bat gammaretrovirus, and Porcine type C oncovirus) were associated with disease. The ability to shift the oral microbiome towards health may be a useful approach to treating periodontal disease in vivo. Further, the novel discovery that nisin and a nisin probiotic promote the numbers of host reparative cells reveals a potentially new biomedical application for nisin in regenerative medicine. Nisin’s ability to shift the oral microbiome towards health, mitigate oral disease, and promote a regenerative periodontal phenotype may benet the regenerative potential of the periodontium and negate systemic effects associated with periodontal disease and its pathogens.

induced by oral infection with P. gingivalis, T. denticola, T. forsythia, and F. nucleatum, and employed to examine the effects of nisin and the nisin-producing probiotic L. lactis in abrogating periodontal bone loss and modulating the composition of the oral microbiome and in ammatory landscape.
For the oral polymicrobial infection, P. gingivalis was mixed with an equal volume of T. denticola for 5 min. Subsequently, T. forsythia was added to the culture tubes containing P. gingivalis and T. denticola, and the bacteria were mixed gently for 1 min and allowed to interact for an additional 5 min. Lastly, F. nucleatum was added and mixed well with P. gingivalis, T. denticola, and T. forsythia. After 5 min, the four bacterial consortium was mixed thoroughly with an equal volume of sterile 4% (w/v) carboxymethyl cellulose (CMC; Sigma-Aldrich) in PBS, and this mixture was used for the oral gavage [12,83].
Lactococcus lactis growth conditions Two L. lactis strains were used in this study; nisin-producing L. lactis (ATCC 11454) was obtained from ATCC and non-nisin producing L. lactis (NZ9800) was kindly provided by Dr. Paul Cotter, Head of the Food Biosciences Department in the Teagasc Food Research Center, Cork Institute of Technology, Ireland. L. lactis strains were grown in Brain Heart Infusion (BHI, Sigma-Aldrich) media overnight in a 37 o C shaking incubator. The L. lactis strains were then pelleted by centrifugation, resuspended in PBS to a concentration of 1 ×1010 CFU/ ml, and mixed with an equal volume of sterile 4% CMC. This mixture was used for oral inoculation .

Nisin preparation
An ultra-pure (>95%) food grade form of nisin Z ( NisinZ ® P) also referred to as nisin ZP was purchased from Handary (S.A., Brussels, Belgium), a primary manufacturer of nisin in the food industry. From here forward, nisin ZP will be referred to as nisin. The stock solution was prepared at a concentration of 600 or 200 μg/ml in sterile water, lter sterilized, and stored at 4•C for a maximum of 5 days for use in experiments. For oral treatment of mice, the nisin solution was mixed with an equal volume of sterile 4% CMC to reach the nal concentration (300 or 100μg/ml).

Infection and treatment of mice
A total of 60 eight-week old BALB/cByJ female mice (The Jackson Laboratories, Bar Harbor, ME) were housed in microisolator plastic cages and randomly distributed into 10 groups (6 mice per group). The description of the experimental groups and infection and treatment protocols are shown in ( Figure 1A and B). The experimental protocols were approved by the Institutional Animal Care and Use Committee of the University of California, San Francisco (IACUC APPROVAL NUMBER: AN171564-01B). All the mice were given trimethoprim (0.17 mg per ml) and sulfamethoxazole (0.87 mg per mL) daily for 7 days in the drinking water and their oral cavity was rinsed with 0.12% chlorhexidine gluconate (Peridex) mouth rinse to inhibit the native oral microbiota [12,14]. The polymicrobial inoculum (5×109 combined bacteria per ml; 1×109 cells in 0.2 ml per mouse; 2.5×108 P. gingivalis, 2.5×108 T. denticola, 2.5×108 T. forsythia and 2.5×108 F. nucleatum) was administered topically in the morning for 4 consecutive days every week for a total of 8 weeks. Nisin (100 or 300 μg/ml, 0.2 ml per mouse) and L. lactis (5×109 bacteria per ml; 1×109 cells in 0.2 ml per mouse) were administered every day in the evening every week for a total of 8 weeks. A sterile 2% CMC solution was administered as the control treatment.
Following 8 weeks of polymicrobial infection, oral swab samples were collected to evaluate the microbial status and to examine the effect of nisin on periodontal pathogens. The samples were collected from the oral cavity of the mice using a sterile micro sized cotton swab. The teeth and surrounding gingival tissue were swabbed and the cotton tip was immersed in 10:1 Tris-EDTA buffer immediately and stored at -80°C until further processing for DNA isolation. Then mice were euthanized and the blood was collected for analysis of antibody response to the periodontal pathogens. The maxillae and mandibles were resected from each mouse for morphometric, histologic, immunologic, and sequencing analysis.
DNA isolation from oral swabs, ethanol precipitation, and real time PCR to con rm bacterial infection DNA isolated from oral swabs was used to evaluate and con rm infection in the mice using methods described in our previous study [12]. In brief, DNA was isolated from the swabs and puri ed using the QIAamp® DNA Mini kit (Qiagen, Germantown, MD, USA). Ethanol precipitation of DNA was then performed to prepare the samples for subsequent real time polymerase chain reaction (PCR). Lastly, standard real-time PCR was used to quantify the periodontal pathogens in the oral swab samples.
Morphometric analysis of periodontal alveolar bone loss overnight and then stained with 1% methylene blue. Digital images of both buccal and lingual/palatal root surfaces of all molar teeth were captured under a stereo dissecting microscope (SMZ1000, Nikon) at the magni cation shown in the images, then the line tool of Image J software (NIH Image) was used to measure the alveolar bone loss from the cementoenamel junction (CEJ) to alveolar bone crest (ABC). For bone loss measurements, the distance between CEJ and ABC were measured from a total of 28 sites on the buccal and lingual/palatal surfaces of the molars (3 sites on the rst molar, 2 sites on the second molar, and another 2 sites on the third molar) [12,84,85]. Two blinded examiners (experienced periodontists) performed all measurements twice at separate times. Both horizontal bone loss and intrabony defects were detected under the stereo dissecting microscope. The intrabony defects were marked as present or absent [12,14].
Histopathological evaluation of periodontal in ammation and cellular content The right maxilla was resected from each mouse and immediately xed in 4% paraformaldehyde for 24h, then decalci ed with diethyl pyrocarbonate-treated 0.5M ethylenediaminetetraacetic acid (pH 8) for 28 days at room temperature. The decalci ed specimens were then dehydrated and embedded in para n using a fully-enclosed tissue processor (ASP300S, Leica Biosystems, Buffalo Grove, IL, USA). The tissue blocks were cut into serial sections (4 μm) parallel to the mesiodistal plane using a microtome, then sections were stained with Mayer's hematoxylin (Sigma-Aldrich, St. Louis, MO, USA) and eosin Y solution (Sigma-Aldrich) for assessment of in ammation. The sections were examined with a stereomicroscope.
The number of in ammatory cells (round-shaped nuclei) and gingival broblast (spindle-shaped nuclei) within a square eld (100 × 100 μm) in connective tissue adjacent to the gingival epithelium between rst and second molars were morphologically evaluated and counted in three tissue sections per mouse specimen (n = 3 per group). Similarly, the number of periodontal ligament (PDL) cells (spindle-shaped nuclei in the PDL space) and alveolar bone lining cells (cell nuclei on bone surface) were counted. All cell counts were averaged for each group, and data were expressed as the mean number of cells per 1.0 mm 2 of connective tissue in the maxillary specimens.
PCR evaluation of immune cytokine pro les from gingival tissues The gingival tissue was treated overnight at 4 ºC with RNA stabilization solution (RNAlater, Invitrogen) after tissue harvesting. Samples were powdered with a mortar and pestle under continuous liquid nitrogen, and total RNA was then isolated from each sample using the RNeasy mini Kit (QIAGEN). The purity and quantity of the RNA were evaluated using the NanoVue Plus spectrophotometer (Biochrom Ltd.). Subsequently, total RNA was synthesized into cDNA using the SuperScript VILO Master Mix (11755050; Invitrogen).
To assess the immune cytokine pro les in gingival tissues, relative gene expression was evaluated by real-time PCR as in our previous study[86] using the following TaqMan primers and probes (TaqMan Gene Expression Assays; Applied Biosystems): interleukin-1β (IL-1β; Mm00434228_m1), IL-6 (Mm00446190_m1), tumor necrosis factor-α (TNF-α; Mm00443258_m1), interferon gamma (IFN-γ; Mm01168134_m1), C-C Motif Chemokine Ligand 2 (CCL2; Mm00441242_m1), C-X-C Motif Chemokine Ligand 2 (CXCL2; Mm00436450_m1), and transforming growth factor beta 1 (TGF-β1; Mm01178820_m1). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH; Mm99999915_g1) was used as a housekeeping gene to normalize the amount of mRNA present in each reaction. PCR was performed in 20 μl reaction mixtures containing the TaqMan Fast Advanced Master Mix, cDNA template (20 ng/well), primers, and probes using a QuantStudio 3 Real Time PCR system (Thermo Fisher Scienti c). The optimized thermal cycling conditions were as follows: 20 min at 95°C, followed by 40 cycles per 1 min at 95°C, and 20 min at 60°C. To compare the expression levels among different samples, the relative expression level of the genes was calculated by the comparative CT (ΔΔCT) method using QuantStudioTM Design & Analysis Software.

Serum antibody analysis
Serum from all 60 mice was collected on the day of euthanasia and used to determine the host response in the form of immunoglobulins (IgG) against P. gingivalis, T. denticola, T. forsythia, and F. nucleatum by an enzyme-linked immunosorbent assay (ELISA) [12,14]. For positive controls for the ELISA, each of these pathogens was grown to a cell density of 7×108 cells/mL and harvested by centrifugation (7000 x g, 30 min, 4 °C). After washing, the cells were treated overnight with 0.5% formalin in buffered saline, then diluted to 0.3 (OD600nm) in 0.05 M carbonate-bicarbonate buffer and used as the coating antigen. The amounts of speci c IgG antibodies present were determined by using a mouse IgG ELISA quanti cation kit (Bethyl Laboratories, USA). A pilot test was rst performed with diluted (1:100) mouse serum to con rm reactivity with the bacterial antigens after 1 h at room temperature. Then, after coating with the formalin-xed bacterial cells, wells were incubated with diluted mouse serum (1:100) for 1 h at room temperature. Wells were then washed with PBS containing 0.05% Tween-20 (PBST), and alkaline phosphatase-conjugated goat anti-mouse IgG and the 3,3′,5,5′-tetramethylbenzidine chromogenic substrate reagent were added for detection and measurement of antibody response. Samples were assayed in duplicate and the puri ed mouse IgG was used to establish a standard curve on each plate for speci c IgG quanti cation.
DNA isolation from gingival tissue for next generation shotgun sequencing DNA was extracted from the mandibular gingival tissue of all mice (6 mice/group) using the QIAamp® DNA Mini kit (Qiagen, Germantown, MD, USA) as follow. The gingival tissue was ground in liquid nitrogen with a mortar and pestle and 180 μl of Buffer ATL was mixed with 25 mg of tissue powder by vortexing. Then, 20 μl of QIAGEN proteinase K was applied to each sample and samples were incubated at 56°C for 3 hours in a shaking water bath. Subsequently, 20 μl of the RNase reagent (20 mg/ml) was added to the samples followed by incubation for 2 min at room temperature. After adding 200 μl of Buffer AL, the samples were incubated at 70°C for 10 min. In addition, 200 μl of pure ethanol was mixed with each sample. This entire mixture was then applied into the QIAamp Mini spin column and centrifuged at 6000 x g for 1 min. Next, 500 μl of Buffer AW1 were added to the spin column and samples centrifuged at 6000 x g for 1 min. Then, 500 μl of Buffer AW2 was added and samples were centrifuged at full speed (20,000 x g) for 3 min, followed by centrifugation (20,000 x g) for 1 min again to eliminate the chance of possible Buffer AW2 carryover. Lastly, samples were incubated with 200 μl of Buffer AE in the spin column, which was placed in a clean 1.5 ml microcentrifuge tube at room temperature for 5 min, then DNA were eluted by centrifugation at 6000 x g for 1 min.
The purity and quantity of the DNA were evaluated using the NanoDrop™ OneC Microvolume UV-Vis Spectrophotometer (Thermo Scienti c), which met quality control measures for subsequent shotgun sequencing analysis.
Metagenome shotgun sequencing and microbiome data production and analyses Shotgun metagenomic sequencing library preparation was performed by Novogen, Inc. The libraries were prepared according to a standard protocol from Illumina, and at least 1 Gb of 150 bp pair-end reads per sample were sequenced on the Illumina Hiseq4000 machines. FASTQ les were generated from the sequencing machines and used for the analyses of the bacteriome/microbiome and virome as described below.

Data Processing
The following criteria were used for processing and cleaning up the raw data. Low quality bases (Q-value ≤ 38), which exceeded a certain threshold (40 bp by default) were trimmed. Reads which contained N nucleotides over a certain threshold (10 bp by default) were trimmed. Reads which overlapped with adapter over a certain threshold (15 bp by default) were trimmed.

Metagenome assembly
We utilized de novo assembly for each sample as follows. Samples passing quality control were assembled initially using SOAPdenovo (http://soap.genomics.org.cn/soapdenovo.html). The Scaffolds were cut off at "N" to get fragments without "N", called Scaftigs. Clean data for all samples were then mapped to assembled Scaftigs using SoapAligner (http://soap.genomics.org.cn/soapaligner.html) and unutilized paired-end reads were collected. Mixed assembly was conducted on the unutilized reads with the same assembly parameter. The scaftigs of each sample and mixed assembly, which were less than 500 bp, were trimmed.

Taxonomy annotation
The following taxonomy annotation scheme was used. We aligned unigenes to the NCBI nonredundant database with DIAMOND to taxonomically annotate each metagenomic homolog (MEGAN). According to the abundance table of each taxonomic level, various analyses were performed using custom scripts by R and Python.
Statistical analysis SPSS 21.0 statistical software (IBM, Chicago, IL, USA) was used for statistical analysis of the non-sequencing data. Student's t-test was used to compare two independent groups. For comparison of intrabony defects, data were expressed as frequency and percentage, and a chi-square test was used for analysis. Further, analyses of the PCR data from the oral swabs and quanti cation of in ammatory cells were performed using an ANOVA followed by a Tukey's test.
Data were presented as means ± standard deviations (SD). Values of p < 0.05 were considered signi cant.
For the microbiome/virome analyses, we normalized the data to have 1 million reads per sample (reads per million, RPM). We ltered out taxa with average read counts less than 1 RPM. We removed 5 samples, namely Infection 1, Infection 6, Non-nisin L. lactis + Infection 4, Non-nisin L. lactis + Infection 6 and Nisin + Infection 3, that have low sequencing coverage. We used the Shannon diversity index to quantify bacterial and viral diversity across different groups. In order to compare the difference of bacterial contents, viral contents, and Shannon diversities between different groups, we computed the p-values using a twosample t-test assuming equal variance of samples from the two groups. For the Principal Coordinates Analysis (PCoA), we further restricted to species with RPM<500 to avoid the result being dominated by commonly present species. Three species, namely Mouse Intracisternal A-particle, Chlamydia abortus, Chlamydia trachomatis, were ltered out under this criterion. We used the Bray Curtis dissimilarity to quantify the difference between microbiome composition of different samples. The 95% con dence ellipses were computed assuming that the data in each group followed a two-dimensional normal distribution. For the differential abundance analysis, we performed a log transformation (log10 (RPM + 0.1)) for the bacterial and viral read counts and used a two-sample ttest to compute the p-values, assuming equal variance in the two groups. We further used the Benjamini-Hochberg procedure[87] to correct for multiple comparisons. We reported the corresponding false discovery rate (FDR) for conducting pair-wise comparisons (e.g., Infection versus Control), and the multiplicity is the total number of taxa. For correlating microbial species with the immune markers, we considered the data in log space for both read counts and immune marker measurements (log10(x+0.1)). We considered only microbial species that are signi cant in at least one differential abundance comparison (comparison v.s. control or v.s. infection). We computed Pearson's correlation with a p-value based on t-test. We performed the Benjamini-Hochberg procedure 88 for multiple testing for each immune marker (across all microbial species) separately.

Results
Polymicrobial oral infection is reduced by nisin probiotic A PCR-based approach was used to evaluate the ability of nisin and a nisin-producing probiotic L. lactis to modulate oral infection by periodontal pathogens in a polymicrobial infection mouse model (Fig. 1A and B). Oral swab results indicated that all four bacteria were detectable at 8 weeks post infection ( Fig. 2A). In the infection group, P. gingivalis, T. forsythia, and F. nucleatum were present at signi cantly higher levels than the control group (p < 0.001). Similarly, T. denticola showed a trend toward higher levels in the infection group, but this was not signi cantly different from the control group. Treatment with nisin or the nisin-producing L. lactis probiotic markedly decreased the number of P. gingivalis, T. forsythia, and F. nucleatum compared to the infection group (p < 0.01). In contrast, the non-nisin-producing probiotic group showed that T. forsythia and F. nucleatum didn't recover back to control levels; and these levels were signi cantly higher compared to the nisin-producing probiotic group (p < 0.05).
In addition, the frequency of mice exhibiting infection differed depending on the pathogens (Fig. 2B). P. gingivalis as well as T. forsythia and F. nucleatum were present in all mice in all groups, whereas T. denticola was present in much fewer mice across all groups.
Alveolar bone loss parameters were signi cantly inhibited in mice treated with nisin or nisin probiotic A polymicrobial infection mouse model of periodontal disease was used to evaluate the ability of nisin and a nisin-producing probiotic L. lactis to modulate periodontal bone loss. After 8 weeks of inoculation/infection with periodontal pathogens (P. gingivalis, T. denticola, T. forsythia and F. nucleatum), mice treated simultaneously with nisin or the nisin-producing probiotic L. lactis exhibited signi cantly less bone loss compared to the infection group ( Fig. 3A and B). Treatment with either high or low concentrations of nisin both showed signi cant rescue effects and signi cantly diminished bone loss in the presence of infection. The non-nisin producing L. lactis probiotic was unable to prevent the bone loss in the infected group.
The presence of alveolar intrabony defects were also evaluated following treatment. Nineteen percent of control uninfected sites showed a baseline level of intrabony defects compared to 58% of infected sites (Fig. 3C). Nisin (low or high concentrations) and the nisin-producing probiotic signi cantly decreased the number of sites that exhibited intrabony defects; 31%, 22%, and 33%, respectively. The non-nisin producing probiotic was unable to signi cantly prevent the development of intrabony defects; however the percentage of sites (49%) that exhibited defects was lower than that of control infected sites (58%). Although, the comparison between the Infection + L. lactis group (33%) and Infection plus non-nisin L. lactis group (49%) showed no signi cant difference (P > 0.05), the Infection plus non-nisin L. lactis group exhibited higher numbers. However, the Infection plus L. lactis group was signi cantly different from the infection control group, but the Infection plus non-nisin L. lactis group was not.
Host antibody response against periodontal pathogens is attenuated with nisin or nisin probiotic To evaluate the host response to the polymicrobial infection, serum antibody levels to the 4 periodontal pathogens were evaluated using an ELISA. Control infected mice showed a signi cant antibody response to all 4 periodontal pathogens compared to the uninfected control mice (Fig. 4). Nisin (low or high concentrations) and the nisin-producing probiotic signi cantly decreased the antibody response in the infected mice. The non-nisin producing L. lactis was also able to decrease the antibody response to the periodontal pathogens, however the effect was not as signi cant as that compared to the nisin-producing probiotic.

Nisin or nisin probiotic prevent an in ux of in ammatory cells into the periodontal complex upon polymicrobial infection
To evaluate nisin's ability to alter the host in ammatory response in the context of periodontal disease, we evaluated the in ammatory cell in ltrate and morphologic changes within the periodontal tissues using hematoxylin and eosin staining of sagittal sections (Fig. 5A). In the control group, few in ammatory cells were observed in the gingival connective tissue just below the thin junctional epithelium. In contrast, the gingival tissues from the polymicrobial infection group exhibited an in ltration of numerous in ammatory cells (p < 0.001; Fig. 5B) and deep periodontal pocket formation with epithelial hyperplasia and rete ridge elongation. Treatment with nisin and the nisin-producing probiotic L. lactis signi cantly decreased the in ammatory cell in ltrate in the infection group (p < 0.001). However, treatment with the non-nisin-producing L. lactis did not signi cantly decrease the in ammation compared to the infection group.
Nisin or nisin probiotic activate a proliferative phenotype in cells of the periodontium Surprisingly, nisin and/or the nisin-producing L. lactis probiotic also markedly increased the number of broblast-like and osteoblast cells (gingival broblasts, periodontal ligament cells, alveolar bone lining cells) compared to the control and/or infection groups (p < 0.05; Fig. 5B). In contrast, application of the nonnisin-producing L. lactis did not signi cantly increase the number of gingival broblasts or alveolar bone lining cells. This is the rst time that nisin or a probiotic has been shown to promote the number of oral cells responsible for the regenerative potential of the periodontium.
Nisin or nisin probiotic abrogates the host in ammatory cytokine response to the periodontal pathogens To further examine the effect of nisin or nisin probiotic on the host in ammatory response in the context of periodontal disease, the relative gene expression of in ammatory cytokines was assessed in gingival tissues by real-time PCR (Fig. 6). The infection group showed a signi cant upregulation of IL-1β, IL-6, and CXCL2; the latter is homologues to IL-8 in mice (p < 0.05). However, treatment with nisin or the nisin-producing L. lactis probiotic signi cantly reduced their expression in the infected mice. The non-nisin producing L. lactis also suppressed the in ammatory response similar to the nisin-producing L. lactis, indicating that the L. lactis itself modulated the host in ammatory response. Other cytokines, namely TNF-α, IFN-γ, CCL2, and TGF-β, showed no signi cant changes following the polymicrobial infection or nisin treatment, although the anti-in ammatory cytokine TGF-β showed a trend toward higher levels with nisin treatment.
Nisin and the nisin probiotic promote a shift from a disease-associated microbiome toward a "healthy control" oral bacteriome and virome In order to assess how nisin and the nisin-producing probiotic modify the oral bacteriome and virome, and how it compares across infection and healthy groups, we conducted metagenome shotgun sequencing analysis of these different conditions. We compared the bacterial (Fig. 7A) and viral content (Fig. 7B) of groups treated with nisin, nisin-producing L. lactis probiotic, and non-nisin producing L. lactis with and without infection and compared against the control group and against the infection group. We observed signi cant differences in viral content across groups. However, we observed only minor differences in bacterial content. With regards to viral content for different groups, the infection group had signi cantly higher viral content than the control group (nominal pvalue 0.041), nisin group (nominal p-value 0.032), infection plus nisin group (nominal p-value 0.0029), and infection plus L. lactis group (nominal p-value 0.0020). In concordance with the bacterial content, bacterial Shannon diversity for different groups showed no signi cant differences across groups (Fig. 7C).
However, the viral diversity score was different across some groups. Speci cally, the infection group was higher but not signi cantly different in diversity than the control group (nominal p-value 0.18) and L. lactis group (nominal p-value 0.16) (Fig. 7D).
To assess the overall change in the oral bacteriome and virome composition, we further performed Principal Coordinates Analysis (PCoA). As shown in Fig. 5A, we found that PC3 and PC4 separate the control group from the infection group (explained variance of 9.6% and 8.3%, respectively. See also Supplementary Fig. 1 for the rst 5 PCs). To investigate if the microbiome compositions of other groups were more similar to the control group or the infection group, we further overlaid each of the other groups on top of the control group and the infection group. Importantly, we found that among infected animals, those treated with nisin (Fig. 8B) and the nisin-producing L. lactis probiotic (Fig. 8C) were similar to the control group, indicating that nisin and L. lactis drive the microbiome composition toward the healthy state. In contrast, those treated with the non-nisin producing L. lactis (Fig. 8D) were in between the control and infection group, indicating that non-nisin producing L. lactis is less effective as a treatment in shifting the oral microbiome toward the healthy control. Other non-infection groups were more similar to the control group (nisin in Fig. 8E and L. lactis in Fig. 8F), except the non-nisin producing L. lactis group (Fig. 8G), which had a high variance.
Furthermore, we identi ed bacteria and viruses at the genus and species level that showed differences in abundance across groups. In this regard, in order to identify speci c differences, we performed two different analyses; the rst analysis was based on using the healthy control group as the reference group (Fig. 9A, B) and a second analysis was based on using the infection group as the reference group (Fig. 9C, D).
At the genus level and looking at differences relative to the control group, we observed that the genus Enterococcus was in lower abundance across different groups when compared against the control group (Fig. 9A). This was speci cally observed in the infection plus L. lactis (FDR < 0.1) and non-nisin L. lactis groups (FDR < 0.3) (Fig. 9A). The infection plus nisin group also showed a lower abundance, although this did not reach statistical signi cance. In addition, the genus Marinobacter showed a reduced abundance in the infection group (FDR < 0.3) compared to the control. Moreover, the genus Pasteurella showed a reduced abundance in the infection plus L. lactis group (FDR < 0.1) compared to the control. Also, the genus Pseudomonas and genus Enterobacter showed an increased abundance compared to the control, speci cally in the infection plus nisin (FDR < 0.2) and infection plus non-nisin L. lactis (FDR < 0.3) groups, respectively (Fig. 9A). The groups with the least change relative to the control group were the nisn and L. lactis groups.
At the species level and looking at differences relative to the control group (Fig. 9B), we observed that the Golden Hamster Intracisternal A-particle H18, Bat gammaretrovirus and Porcine type C oncovirus showed increased abundance (FDR < 0.2) upon infection compared to the control group, suggesting their role in the disease process. However, Marinobacter sp. B9-2 showed a reduced abundance (FDR < 0.3) in the infection group compared to the control group, suggesting its role in health. In addition, both Enterococcus faecium and Pasteurella multocida showed a decreased abundance (FDR < 0.1) in the infection plus L. lactis group compared to the control group. The non-nisin producing L. lactis group showed a decreased abundance of Enterococcus faecium and an increased abundance of Golden Hamster Intracisternal A-particle H18 (FDR < 0.3) compared to the control group.
At the genus level and looking at differences relative to the infection group (Fig. 9C), we observed that the infection plus L. lactis group showed decreased abundances at the genius level for Salmonella (FDR < 0. At the species level and looking at differences relative to the infection group (Fig. 9D), we also observed some microbes in higher abundance across different groups compared to the infection group (e.g. Marinobacter sp. B9-2 and Mouse mammary tumor virus); suggesting their potential involvement in maintaining health. In contrast, Mouse intracisternal A-particle, Bat-gammaretrovirus, Golden hamster intracisternal A-particle H18, Salmonella enterica and Porcine type-C oncovirus were in lower abundance across different groups compared to the infection group, suggesting their potential involvement in the transition to disease (Fig. 9D).

Speci c microbial species (bacteria and viruses) are correlated with IL-6 levels
We next examined potential correlations between microbial changes and cytokine levels across all groups. We identi ed signi cant correlations between speci c microbial species (bacteria and viruses) and IL-6 levels ( Fig. 9E and F). Speci cally, s_Golden hamster intracisternal A-particle H18 (FDR < 0.1) exhibited the highest level of positive correlation with IL-6 levels. The following species also exhibited a positive correlation with IL-6 but at a decreased level of signi cance (FDR < 0.3): s_Bat gammaretrovirus, s_Salmonella enterica, and s_Porcine type-C oncovirus. The following, s_Marinobacter sp.B9-2, was the only species showing a moderate negative correlation (FDR < 0.3) with IL-6 levels.

Discussion
Studies exploring the potential of probiotics to suppress periodontal pathogens or anaerobic bacteria in human and animal studies have shown some bene ts. Human studies exploring probiotics as monotherapy or adjunctive therapy have shown some bene t or neutral effects in reducing periodontal pathogens or anaerobes with the probiotics Lactobacillus salivarius WB21 [44], L. reuteri [18,45,46,47], bacillus[48], L. plantarum [33], L. rhamnosus SP1 [49], B. lactis [31] and various Streptococci [24,32,50]. Animal studies also showed some bene ts. When Lactobacillus brevis or Bi dobacterium lactis were applied in a murine model of periodontitis, there was a signi cant decrease in the counts of anaerobic bacteria relative to aerobic bacteria [51,52]. Use of Lactobacillus rhamnosus GG showed no antimicrobial activity against P. gingivalis and F. nucleatum [53]. The current investigation demonstrated that the nisin-producing probiotic and nisin itself reduced the oral levels of all the Red complex bacteria, P. gingivalis, T. forsythia, and T. denticola, and the orange complex pathogen, F. nucleatum, indicating nisin/nisin probiotic's e cacy in consistently removing these pathogens from oral surfaces. bene cial effects with use of various Streptococci species, Bacillus subtilis, Lactobacillus brevis, Saccharomyces cerevisiae, Bacillus subtilis, Bacillus licheniformis, Bi dobacterium animalis subsp. lactis, Bi dobacterium lactis, and Lactobacillus rhamnosus GG [24,[51][52][53][54][55][56][57][58][59]. In the current study, signi cant decreases in alveolar bone loss and intrabony defect formation were observed with the use of the nisin-producing Lactococcus lactis probiotic or nisin itself. Furthermore, low and high concentrations of nisin were equally effective at reducing bone loss.
Probiotics have not been examined for their potential to reduce the host systemic antibody response to periodontal pathogens in a periodontal setting in humans or animals [24]. The current investigation revealed that Lactococcus lactis probiotic or nisin itself can signi cantly reduce the systemic antibody response to all periodontal pathogens. This suggests that this nisin probiotic and nisin have signi cant potential for blocking the negative downstream systemic effects associated with these periodontal pathogens. It is noteworthy that the non-nisin producing L. lactis also mediated some bene cial effects. Some of the partial effects mediated by the non-nisin producing L. lactis control may be due to it is inherent properties as a lactic acid bacteria (low pH and enzymatic activity); which may contribute to its effects [34,60]. For example, the non-nisin producing probiotic was also able to decrease the antibody response to the periodontal pathogens, however the effect was not as signi cant as the nisin-producing L. lactis (Fig. 2D).
The novel discovery that nisin and the nisin-producing probiotic L. lactis promoted increases in the number of regenerative cells of the periodontium is surprising. The potential for a probiotic or bacteriocin to promote the regenerative potential of host reparative cells has not been previously documented.
These ndings have implications for the clinical sequelae of periodontal disease. Namely, in addition, to the aforementioned bene cial effects of nisin in mitigating periodontal disease bone loss and the host in ammatory response, while resetting the oral microbiome towards control levels, this additional nding suggests that nisin and a nisin probiotic may promote a regenerative potential and tissue restitution following disease.
Limited studies have examined a probiotic's ability to shift a disease-associated oral microbiome. In humans, one study found that lozenges containing L. rhamnosus GG and Bi dobacterium animalis mediated no change in the microbial composition of saliva using a focused oral microbe microarray[67]. One study in rats, using Bi dobacterium animalis subspecies lactis showed an increase in the levels of Actinomyces and Streptococci-like species while decreasing the levels of Veillonella parvula, Capnocytophaga sputigena, Eikenella corrodens, and Prevotella intermedia-like species [59]. Importantly, the current study revealed that the probioltic L. lactis and its bacteriocin nisin can shift a disease-associated oral bacteriome and virome back towards a healthier state ( Fig. 5B and C). This agrees with our recent in vitro ndings in oral bio lms that nisin and a nisin probiotic shift periodontal pathogen-spiked oral bio lms back towards a control/healthy state [42]. Maintaining or promoting a healthy microbiome in the course of treatment with probiotics is being recognized as an important parameter that should be evaluated [68,69,70]. In this study, we used the approach of identifying the complex microbial signature of periodontal health as a baseline for comparison to evaluate and con rm a restitution of "health" following antimicrobial treatment for periodontal disease.
A long-standing premise in the pathogenesis of periodontal disease has been its association with pathogenic bacteria, especially members of the so called Red Complex. The current study and others highlight the importance of new and emerging microbes, both bacteria and viruses, in periodontal disease pathogenesis [12,[71][72][73][74][75][76] and their potential shift with treatment [72,77]. These microbes may be important signatures useful in monitoring treatment and to determine shifts that signify health. We observed that the species Marinobacter sp. B9-2 was in higher abundance in the healthy control group compared to the infection group. However, 3 viruses, Golden Hamster Intracisternal A-particle H18, Bat gammaretrovirus, and Porcine type C oncovirus showed increased abundance (FDR < 0.2) in the infection group compared to the control group and also relative to other treatment groups (Fig. 6B). Thus, periodontal health was associated with Marinobacter sp. B9-2, whereas the 3 viruses, Golden Hamster Intracisternal A-particle H18, Bat gammaretrovirus, and Porcine type C oncovirus, were associated with periodontal disease. These ndings are consistent with our earlier observations showing that these three viral infectionassociated microbes were also associated with bone loss, whereas Marinobacter decreased with bone loss [12]. Treatment generally shifted microbes towards the healthy control. The signi cance of these speci c microbes and their role in health and disease and response to treatment has not been previously described. Marinobacter is a genus of Proteobacteria found in sea water and a number of strains and species can degrade hydrocarbons [78]. Intracisternal type A particles are defective retroviruses in rodent genomes [79]. Bat gammaretrovirus are retroviruses that can cause malignancies and immune de ciencies in mammals, reptiles and birds [80]. Porcine type C oncovirus is a type of gammaretrovirus that lives in extreme environments and can be found in the human microbiome [81]. Further study is warranted to determine the relevance of these microbes in human oral health and disease.
Several of these microbial species were also signi cantly correlated with the cytokine host immune response. Namely, s_Golden hamster intracisternal Aparticle H18 (highest correlation), s_Bat gammaretrovirus, s_Salmonella enterica, and s_Porcine type-C oncovirus exhibited a signi cant correlation with IL-6 levels. However, s_Marinobacter sp.B9-2 was signi cantly negatively correlated with IL-6 levels. These ndings further highlight the tight relationship between the microbiome and the host immune response; an interaction well known in conditions of health and disease[82].

Conclusions
In summary, this study highlights an approach to realign the oral microbial dysbiosis of periodontal disease and its related sequalae (bone loss, host immune response) towards health. Treatment with antibiotics and probiotics have been used to modulate the microbial, immunological, and clinical landscape of periodontal disease with some success. Antibacterial peptides or bacteriocins, such as nisin, and nisin-producing probiotics, such as Lactococcus lactis, have not been examined in this context. However, they warrant examination because of their well characterized biomedical bene ts in eradicating bio lms and oral pathogenic bacteria, while also modulating immune mechanisms. This study demonstrates that nisin and nisin probiotic treatment inhibit periodontal disease-related bone loss and host immune responses while signi cantly shifting the oral bacteriome and virome towards the healthy control state. This shift was characterized by a unique signature where health was associated with a Proteobacteria (Marinobacter sp. B9-2), whereas 3 retroviruses (Golden Hamster Intracisternal A-particle H18, Bat gammaretrovirus, and Porcine type C oncovirus) were associated with disease. The ability to shift the oral microbiome towards health may be a useful approach to treating periodontal disease in vivo. Further, the novel discovery that nisin and a nisin probiotic promote the numbers of host reparative cells reveals a potentially new biomedical application for nisin in regenerative medicine. Nisin's ability to shift the oral microbiome towards health, mitigate oral disease, and promote a regenerative periodontal phenotype may bene t the regenerative potential of the periodontium and negate systemic effects associated with periodontal disease and its pathogens.
Declarations ETHICS APPROVAL STATEMENT   Mouse treatment procedure and oral sample collection timeline. Polymicrobial infections were carried out in the morning for 4 consecutive days once per week from the 3rd to the 10th week. Nisin and L. lactis were administered every day in the evening from the 3rd to the 10th week. Oral swab samples were collected at 8 weeks following the initial infection. Blood and tissue specimen collection was performed at euthanasia following 8 weeks of infection.

Figure 2
Polymicrobial oral infection is reduced by nisin or nisin-producing probiotic treatment. Oral swab samples were collected at eight weeks after polymicrobial infection. DNA was isolated and puri ed from the swab samples of eight groups (Control, Infection, Nisin (H), L. lactis, Non-nisin L. lactis, Infection + nisin (H), Infection + L. lactis and Infection + Non-nisin L. lactis). The total bacteria were quanti ed by standard real-time PCR using primers corresponding to 16S ribosomal RNA. A. The data are shown as a percentage of each pathogen (P. gingivalis, T. denticola, T. forsythia, or F. nucleatum) among total bacteria. a, the difference in percentage of the pathogen was signi cant (p < 0.001) compared to the Control group. b, the difference in percentage of the pathogen was signi cant (p < 0.01) compared to the Infection group. *, the difference in percentage of the pathogen between the two groups was signi cant (p < 0.05). B. The table demonstrates the number of detected bacteria and detection frequency (%) of periodontal pathogens in each swab from each mouse relative to the number of collected samples.

Figure 3
Alveolar bone loss is signi cantly abrogated with nisin or nisin-producing probiotic treatment. A. Representative images of alveolar bone loss on the palatal surfaces of maxillary molars in six groups (Control, Infection, Infection + nisin (L) ,Infection + nisin (H), Infection + L. lactis and Infection + Non-nisin L. lactis).
Scale bar represents 0.2mm. B. The graph represents alveolar bone loss in all ten groups. Data represent the means ± standard deviation from 6 mice per group. For each mouse, alveolar bone loss was calculated as the average from 28 sites (3 sites on the rst molar, 2 sites on the second molar, and 2 sites on the third molar, on both sides of the left maxilla and mandible). a, the difference in alveolar bone loss was signi cant (p < 0.05) compared to the Control group. b, the difference in alveolar bone loss was signi cant (p < 0.05) compared to the Infection group. *, the difference in alveolar bone loss between the two groups was signi cant (p < 0.05). C. The percentage of intrabony defects was calculated as the number of tooth surfaces containing periodontal intrabony defects out of total tooth surfaces. For each group, there were a total of 72 tooth surfaces (6 mice, 36 molars, 72 sides (buccal, palatal/lingual)). a, the difference in the percentage of intrabony defect was signi cant (p < 0.05) compared to the Control group. b, the difference in the percentage of intrabony defect was signi cant (p < 0.05) compared to the Infection group. c, there was no signi cant difference in the percentage of intrabony defect between the Infection + L. lactis group and Infection + non-nisin L. lactis group (p > 0.05).

Figure 4
Host antibody response against periodontal pathogens is signi cantly abrogated with nisin or nisin-producing probiotic treatment. Serum IgG antibody levels to P. gingivalis, T. denticola, T. forsythia , and F. nucleatum in all ten groups. Data represent the means ± standard deviation from 6 mice per group. a, the difference in serum IgG antibody levels was signi cant (p < 0.05) compared to the Control group. b, the difference in serum IgG antibody levels was signi cant (p < 0.05) compared to the Infection group. *, the difference in serum IgG antibody levels between the two groups was signi cant (p < 0.05).

Figure 5
Nisin or nisin-producing probiotic prevent an in ux of in ammatory cells into the periodontal complex, and promote increases in host regenerative periodontal cells. Histological examination of periodontal in ammation in the interproximal area between the rst and second maxillary molars was performed in 5 groups (Control, Infection, Infection + nisin (H), Infection + L. lactis and Infection + Non-nisin L. lactis). A. Representative histological images of morphologic changes within the periodontal tissues using HE staining of sagittal sections. B. The bar graphs demonstrate the number of in ammatory cells and host in ammatory cells, gingival broblasts in connective tissues adjacent to the gingival epithelium, number of periodontal ligament cells, and alveolar bone lining cells were counted within a square eld (100 ×100 μm) between rst and second molars. Data represent the means ± standard deviation from 3 mice per group. a, signi cantly different compared to the control group (p<0.05); b, signi cantly different compared to the infection group (p <0.05).

Figure 6
Nisin or nisin-producing probiotic abrogate the host in ammatory cytokine response in gingival tissues. To evaluate the immune cytokine pro les in gingival tissues, mRNA expression of IL-1β, IL-6, TNF-α, IFN-γ, CCL2, CXCL2, and TGF-β1 were measured by real-time PCR. The amount of mRNA in each reaction was normalized to GAPDH, which is a housekeeping gene. Data are shown as means ± standard deviation from 6 mice per group. a, p < 0.05 compared with the Control group. b, p < 0.05 compared with the Infection group.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.