P. gingivalis ATCC 33277 was cultured anaerobically in trypticase soy broth (TSB) or on TSB-blood agar plates supplemented with 1 g/L yeast extract, 5 μg/mL hemin, and 1 μg/mL menadione at 37°C. When needed, erythromycin (10 μg/mL) was added to the medium to eliminate isogenic mutants. Human umbilical vein endothelial cells (HUVECs) (EA. hy926) were maintained in Dulbecco’s modified Eagle’s medium with fetal bovine serum (10%) (Millipore Sigma, St. Louis, MO, USA) and penicillin/streptomycin (1%) at 37°C in a 5% CO₂ atmosphere (Wang et al., 2017). The PCR fusion technique and electroporation were utilized to generate the P. gingivalis BCAA ABC transporter permease (livh)-deficient strain (Δlivh) and the ABC-type BCAA transport system periplasmic component (livk)-deficient strain (Δlivk) via homologous recombination, as described in Figure S1 (Wright et al., 2014). The primers used for mutant construction and confirmation are shown in Table S1 . The mutants were confirmed via PCR and DNA sequencing ( Figure S1 ).

HUVECs were seeded in 6-well plates and cultured to 80% confluence. After washing with phosphate-buffered saline (PBS), cells were infected with P. gingivalis at a multiplicity of infection of 100 for 2 h at 37°C in 5% CO₂, unless otherwise stated. Non-adherent bacteria were removed by washing with PBS, and the cell-adherent bacteria were killed using gentamicin (300 μg/mL) and metronidazole (200 μg/mL) for 1 h at 37°C in 5% CO₂ (Moffatt et al., 2012). For RNA isolation, the cells were washed with PBS and lysed using TRIzol^® reagent (Invitrogen, Carlsbad, CA, USA).

Total RNA was extracted using TRIzol^® reagent (Invitrogen, Carlsbad, CA, USA) according to protocols provided by the manufacturer. Further, genomic DNA was removed using the commercially available DNase I system (Qiagen, Germantown, PA, USA), and RNA concentration was measured by determining the A260/A280 ratio using a Nanodrop 2000 system (ThermoFisher, Waltham, MA, USA). RNA integrity was verified via 1.5% agarose gel electrophoresis, and qualified RNAs were quantified using a Qubit 3.0 fluorometer with a QubitTM RNA Broad Range Assay Kit (ThermoFisher, Waltham, MA). Furthermore, the RNA samples were analyzed at the Huayin Genome Research Facility (Huayin Health Medical Group Co, Wuhan, China) to determine ribosomal RNA (rRNA) depletion and subsequent RNA sequencing, and rRNA was removed using a Ribo-off rRNA Depletion Kit (Vazyme Biotech Co., Nanjing, China) for humans (Catalog No. N406) and/or bacteria (catalog NO. N407). Additionally, cDNA libraries for Illumina^® sequencing were generated using the KC-Digital™ Stranded mRNA Library Prep Kit (Illumina, San Diego, CA, USA). The library products corresponding to 200–250 bps were then enriched, quantified, and sequenced using a Nova-seq 6000 sequencer (Illumina, San Diego, CA, USA) (Westermann et al., 2016; Westermann et al., 2017; Westermann and Vogel, 2018).

Trimmomatic version 0.36 was used to pre-process raw sequencing data to remove low-quality reads and trim adaptor sequences. To eliminate duplication bias introduced during library preparation and sequencing, the clean reads were first clustered according to the unique molecular identifier sequences, the reads in the same cluster were then compared via pairwise alignment to generate new sub-clusters based on sequence identity (>95%), and multiple sequence alignments were conducted to acquire one consensus sequence for each sub-cluster. De-duplicated consensus sequences were then mapped to the human or P. gingivalis ATCC 33277 genome using STAR 2.5.3a. Feature Counts (Subread-1.5.1; Bioconductor) was used for counting reads mapped to the exon regions of each gene, after which the number of reads per kilobase million was calculated. Further, the edgeR package v3.12.1 was used to identify differentially expressed genes (DEGs) between the infected group and control group. Genes with false discovery rate (FDR) P-values < 0.05 and absolute log₂FC value ≥ 1 were considered as differentially expressed. Principal-component analysis was conducted via the function rda in the Vegan package in R. Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis was conducted after the DEGs were identified using the KEGG orthology-based annotation system. Thereafter, to statistically test the results of the enrichment analysis, a hypergeometric test was performed; a cutoff P-value of 0.05 was applied for the identification of the enriched KEGG pathways (Nuss et al., 2017; Westermann and Vogel, 2018). Gene Ontology (GO) enrichment analysis for DEGs was carried out by the R/Bioconductor package clusterProfiler, and P < 0.05 was set as the threshold level to determine the significance of the GO terms (Yu et al., 2012).

For the gene expression assay, total mRNA was extracted as mentioned above and reverse transcribed into cDNA using the PrimeScript RT reagent Kit (Takara Bio, Kusatsu, Japan). qRT-PCR was then performed using the SYBR^® Premix Ex TaqTM II Kit (Takara Bio, Kusatsu, Japan) or via TaqMan Fast Universal Master Mix and TaqMan gene expression assay (ThermoFisher, Waltham, MA, USA) using an Applied Biosystems 7300 system (Foster City, CA, USA) (Liu et al., 2021). PCR amplification was performed as follows: 95°C for 5 min, and 40 cycles at 95°C for 15 s, 56°C for 15 s and 72°C for 15 s, and melting curve analysis. Relative mRNA expression was quantified through the comparative 2^-ΔΔCT method with GAPDH as the internal control (Liu et al., 2015). The mRNA expression levels of livk and livh were measured using 16S rRNA as an internal reference based on the primers listed in Table S1 .

The overnight culture of P. gingivalis was diluted to an optical density at 600 nm (OD₆₀₀) of 0.1 using fresh medium, and then anaerobically cultured at 37°C. The OD₆₀₀ of the bacterial culture was then measured at 2-h intervals for 24 h using a spectrophotometer (Multiskan GO; Thermo Scientific, Inc., Waltham, MA, USA). The experiments were each repeated three times, and growth curves were generated.

Invasion experiments were set up as mentioned in 2.1 except that HUVECs were seeded in 12-well plates (2.0 × 10⁵ cells per well) and cultured to 60% confluence. After antibiotic treatment, the cells were washed with PBS and then lysed with sterile distilled water (1 mL per well).The lysates were diluted and plated on TSB-blood agar. After incubation anaerobically at 37°C for 10 days, the number of bacterial colonies on TSB-Blood agar was counted to determine invasion efficiency, which was expressed as the percentage of the initial inoculum recovered (Moffatt et al., 2012).

The amount of EPS produced by P. gingivalis was determined using fluorescent lectins, as described previously (Liu et al., 2017a). Briefly, P. gingivalis cells were labeled with Syto-17 (Thermo Fisher, Waltham, MA, USA) and deposited on glass coverslips, while the polysaccharides were labeled with concanavalin A-fluorescein isothiocyanate isomer-I (FITC) and wheat germ agglutinin-FITC (100 μg mL^-1) for 30 min at room temperature. Thereafter, images were collected via laser scanning confocal microscopy (SP8; Leica, Germany) and analyzed using Volocity software version 6.3 (PerkinElmer, Waltham, MA, USA).

All the animal procedures applied in the present study were approved by the Ethics Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology and the Ethics Committee of West China Hospital of Stomatology, Sichuan University (WCHSIRB-D-2017-071). C57BL/6 mice (male, five-week-old) were housed in specific pathogen-free facilities, and after one week of environmental acclimation, the mice were randomly assigned into five groups of 5 mice each as follows: normal chow (NC) control, HFD, P. gingivalis WT infection with HFD (WT + HFD), P. gingivalis △livh infection with HFD (△livh + HFD), and P. gingivalis △livk infection with HFD (△livk + HFD). The mice in the HFD group and NC group were administered Research Diet No. D12492and No. D12450J, respectively ( Table S2 ). After 12 weeks, the mice were euthanized, and liver and sera samples were harvested. Further, blood samples were drawn via cardiac puncture using non-anticoagulant vacuum tubes and 23G1 needles (Tian et al., 2020).

Mice in the P. gingivalis infection groups were inoculated with 1 × 10⁹ bacteria in 100 μL PBS with 2% carboxymethyl cellulose (CMC) (Kuboniwa et al., 2017). Infection was repeated once every 3 days for 8 weeks. Conversely, mice in the NC and HFD groups were inoculated with PBS containing 2% CMC as controls. The body weights of the mice were measured at baseline and at the end of the experiment.

Total DNA was isolated from aliquots of liver samples, and the copy number of the P. gingivalis genome in the liver DNA samples was determined via qPCR using the Wizard^® Genomic DNA Purification Kit (Promega, Madison, WI, USA) and the SYBR^® Premix Ex TaqTM II Kit (Takara Bio, Kusatsu, Japan). P. gingivalis-specific primers are listed in Table S1 . Amplification reactions were performed using the following cycling parameters: 95°C for 5 min, 40 cycles at 95°C for 10 s, 60°C for 15 s, and 72°C for 15 s. The corresponding copy numbers were calculated using a standard curve ( Figure S2 ), generated as previously described (Kuboniwa et al., 2017).

The mice were fasted for 6 h prior to blood sample collection for the determination of fasting blood glucose (FBG) levels. Specifically, after the fasting, blood samples were collected from the tail vein of the mice and FBG levels were measured using a glucometer (Roche, Basel, Switzerland). Serum was collected after centrifuging the blood samples at 1,000 × g for 10 min at 4°C and stored at -80°C until further analysis. Serum AST and ALT levels were also measured using AST and ALT activity assay kits, respectively (Millipore Sigma, St. Louis, MO, USA), while serum BCAA levels were determined using BCAA assay kits (Abcam, Cambridge, MA, USA).

Statistical analysis was performed using GraphPad Prism software version 7.0 (San Diego, CA, USA). Unpaired two-tailed Student’s t-tests were used to assess differences between groups, while analysis of variance (ANOVA) with Tukey’s test was carried out for multiple comparisons to analyze datasets with more than two groups. Statistical significance was set at P < 0.05. Further, the results were presented as mean ± standard deviation (mean ± SD).

Principal component analysis (PCA) was used to determine the distances between P. gingivalis transcriptomes within endothelial cells and the free culture medium. Specifically, the first principal component (PC1), which showed the largest variance (43.8%), separated intracellular P. gingivalis and P. gingivalis in the culture medium ( Figure 1A ). Additionally, intracellular P. gingivalis showed a total of 573 DEGs compared with the control bacteria, and of these DEGs, 304 and 269 were upregulated and downregulated, respectively ( Figure 1B ). The complete list of the DEGs corresponding to intracellular P. gingivalis is shown in Table S3 . GO analysis revealed that upregulated genes were significantly enriched in the GO terms amino acid biosynthetic and metabolic processes ( Figure 1C ). Interestingly, ABC transporter genes, which are critical for the virulence of P. gingivalis, were also significantly upregulated in intracellular bacteria ( Table S3 ). Further analysis of the expression levels of PGN_RS06330 (BCAA ABC transporter permease, livh) and PGN_RS06345 (ABC-type BCAA transport system periplasmic component, livk) via qRT-PCR confirmed that the expression levels of livh and livk increased significantly in intracellular P. gingivalis ( Figure 1D ). Adaptation through horizontal gene transfer, which is mediated by tra gene homologs (Tribble et al., 2007), is regarded as an important process that is required for the survival of P. gingivalis within its host. In this regard, it was observed that P. gingivalis traA, traI, traG, traN, traJ, traO, traQ, traM, and traP genes were significantly upregulated in HUVECs compared with levels in the culture medium ( Table S3 ). Further, HcpR, which is necessary for the growth of P. gingivalis with nitrite and nitric oxide, was also upregulated ( Table S3 ).

To explore the role of livh/livk in P. gingivalis, we further assessed the effect of livh/livk deficiency on bacterial growth, invasion, and EPS production in vitro. The construction of the growth curve of P. gingivalis showed that the growth rates of either the livh or livk mutant strains were comparable to that of the parental strain ( Figure 2A ). Additionally, the invasion efficiency of the Δlivk strain was confirmed using an invasion assay ( Figure 2B ). However, no significant difference in invasion efficiency was observed between the Δlivh and WT strains ( Figure 2B ). Moreover, EPS production often plays an important role in biofilm communities developing. Therefore, to assess the role of livh/livk in EPS production, P. gingivalis was labeled with Syto-17, and EPS was stained with fluorescently labeled lectin concanavalin A and wheat germ agglutinin ( Figure 2C ). Thereafter, the level of EPS was normalized to that of P. gingivalis, and it was observed that the rates of production of EPS in Δlivh and Δlivk were not significantly different from those in the parental strain ( Figure 2D ).

To gain insight into the pathways corresponding to the endothelial cells targeted by bacteria, we analyzed the DEGs and enriched KEGG pathways corresponding to HUVECs challenged with P. gingivalis. It was observed that the transcriptional landscapes of the infected and uninfected HUVECs showed two independent clusters, indicating distinct gene expression patterns ( Figure 3A ). Additionally, a total of 838 DEGs were detected in P. gingivalis-infected HUVECs compared with their uninfected counterparts, and of these, 297 and 541 genes were upregulated and downregulated, respectively ( Figure 3B ). The complete list of DEGs corresponding to the HUVECs is shown in Table S4 . Further, we annotated the DEGs identified during the comparison of the infected and uninfected HUVECs based on the KEGG pathway. It was observed that both the upregulated and downregulated genes were enriched in pathways associated with chronic systemic diseases, including Parkinson’s disease, Huntington’s disease, NAFLD, and Alzheimer’s disease ( Figures 3C, D ). However, only the upregulated DEGs were found to be enriched in the taste transduction, metabolic, and hypoxia inducible factor 1 (HIF-1) signaling pathways ( Figure 2C ). Moreover, the downregulated DEGs were also enriched in pyrimidine metabolism, purine metabolism, RNA transport, RNA polymerase, viral carcinogenesis, transcriptional misregulation in cancer, systemic lupus erythematosus, and alcoholism pathways ( Figure 3D ). To corroborate the involvement of NAFLD-related genes in HUVECs challenged with P. gingivalis, the expression levels of MTCO2, MTCO3, UQCRB, and NDUFB10 were detected via qRT-PCR ( Table 1 ). Consistent with the dual RNA sequencing results, MTCO2 and MTCO3 were significantly upregulated, whereas UQCRB and NDUFB10 were significantly downregulated in cells infected with P. gingivalis ( Figure 3E ).

To clarify the potential effect of P. gingivalis infection on NAFLD, an HFD-fed mouse model was used to confirm the transmission of P. gingivalis in the liver, and detect the effect of bacterial infection on circulating BCAA levels, FBG levels, and liver function. As expected, at the end of the experiment, it was observed that the HFD induced an increase in the body weight of the mice, which were unaffected by infection with P. gingivalis ( Table S5 ). Additionally, the serum levels of FBG, ALT, and AST, as well as the AST/ALT ratios were significantly elevated in mice fed HFD compared with those in mice fed NC ( Figure S3 ), suggesting that HFD resulted in impaired liver function in the mice. Consistent with the results of previous studies, P. gingivalis was detected in the liver tissue of WT + HFD mice ( Figure 4A ). Further, P. gingivalis infection also induced increased serum ALT and AST levels, elevated the AST/ALT ratio, and resulted in higher FBG levels in the HFD-fed mice, confirming that P. gingivalis can aggravate HFD-induced liver injury ( Figures 4B –E ).

Circulating BCAAs participate in the regulation of liver function, and livh/livk has been annotated as a BCAA ABC transporter permease or ABC-type BCAA transport system periplasmic component of P. gingivalis. Therefore, we measured serum BCAA levels in mice infected with different strains of P. gingivalis. The results showed that P. gingivalis infection induced increases in BCAA levels in HFD-fed mice ( Figure 4F ). However, this was compromised by either Δlivh or Δlivk ( Figure 4F ). Given that the reduced invasion efficiency of the Δlivk strain was confirmed via an invasion assay ( Figure 2B ), we further investigated the effect of Δlivh or Δlivk deficiency on P. gingivalis liver transmission. We observed that the liver of HFD-fed mice contained less Δlivk than the parental strain ( Figure 4A ). Interestingly, FBG, ALT, and AST levels, as well as the AST/ALT ratios in both the Δlivh + HFD and Δlivk + HFD groups were significantly lower than those in the WT + HFD group. This notwithstanding, the Δlivh + HFD and Δlivk + HFD groups showed no significant difference ( Figures 4B –E ).