Abstract

Background

   Oral lichen planus (OLP) is one of the most frequent oral mucosal
   diseases associated with chronic inflammation, despite a profoundly
   limited understanding of its underlying pathogenic mechanisms.

Results

   The microbiome analysis was conducted on buccal and lip mucosae, tongue
   dorsum, and saliva in nonerosive/erosive OLP patients and healthy
   individuals. Significant variations were observed in the oral
   microbiome of OLP patients, particularly in the buccal mucosa. Network,
   random forest, and NetShift analysis collectively indicated that
   Parvimonas micra (P. micra) emerged as a crucial bacterium in OLP. In
   vivo analysis further demonstrated that P. micra was abundant at the
   junction of epithelial and connective tissue layers in OLP lesions.
   Single-cell RNA sequencing data implicated fibroblasts as potential
   targets, characterized by upregulation of the NF-κB pathway linked to
   TNF-α. Co-culturing of P. micra or its extracellular vesicles (EVs)
   with fibroblasts showed that P. micra and EVs could activate the NF-κB
   signaling pathway and suppress autophagy in buccal mucosal fibroblasts.
   Among the pathogenic effectors, DnaK from P. micra EVs was identified
   to interact with BAG3 in fibroblasts. The interaction of DnaK with BAG3
   subsequently activated the NF-κB pathway and decreased autophagy flux.
   Additionally, we identified that IKK-γ was the key downstream protein
   that could bind with DnaK-BAG3, thereby inhibiting autophagy and
   promoting TNF-α secretion.

Conclusions

   We initially revealed that P. micra was a crucial pathogen in the
   development of OLP and demonstrated that P. micra’s EVs induce the
   inhibition of autophagy and enhanced TNF-α secretion in OLP fibroblasts
   via the DnaK-BAG3-IKK-γ axis. This study offers novel insights into the
   pathogenic mechanisms underlying OLP.
   Your browser is not supporting the HTML5 <video> element.
   You may download that video as a file and play it with a player of you
   choice.

   [42]Download video stream.
   [43]Download video file^ (37.2MB, mp4)
   [44]Open in a new tab

   Video Abstract

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s40168-025-02151-5.

   Keywords: Oral lichen planus, DnaK, BAG3, Outer membrane vesicles,
   NF-κB

Background

   Oral lichen planus (OLP) is one of the most common oral mucosal
   diseases associated with chronic inflammation, affecting up to 0.89% of
   the general population [[45]1, [46]2]. It most commonly leads to
   significant roughness, discomfort, and pain, which rarely undergo
   spontaneous remission with a high tendency for malignant transformation
   [[47]3, [48]4]. The initiation and progression of OLP have no
   definitive and complex interaction between psychological factors,
   immunological factors, endocrine disorders, and genetic factors
   [[49]5]. Emerging evidence suggests a potentially crucial role of the
   oral microbiome in the etiology of OLP [[50]6].

   Next-generation sequencing-based microbiome studies revealed distinct
   differences in the microbial composition of saliva and lesions from
   patients with OLP, compared to that of healthy individuals [[51]7]. For
   instance, the abundance of Porphyromonas, Solobacterium,
   Capnocytophaga, and Gemella exhibited a significant increase in OLP
   patients compared to healthy individuals, whereas Haemophilus,
   Corynebacterium, Cellulosimicrobium, and Campylobacter demonstrated a
   dramatic decrease in OLP patients [[52]8–[53]11]. Regarding the
   contribution of the oral microbiome to OLP development, studies have
   demonstrated that oral pathogenic bacteria can infiltrate OLP tissues
   [[54]12], destroy the integrity of the oral mucosal epithelium barrier,
   and facilitate the infiltration of additional bacteria [[55]8],
   followed by their aggregation in the lamina propria and the epithelial
   basal layer [[56]6]. However, research on the mechanisms through which
   oral microbiota facilitate the progression of OLP remains exceedingly
   scarce.

   Fibroblasts, constituting a primary cellular component of the lamina
   propria, are crucial for tissue maintenance and integrity [[57]13].
   Beyond their roles in producing and remodeling the extracellular
   matrix, fibroblasts function as sentinel cells capable of recognizing
   and responding to pathogens through the production of inflammatory
   mediators [[58]14]. Excessive activation of fibroblasts contributes to
   chronic inflammation and tissue destruction, sustained by microbial
   dysbiosis [[59]15]. Mechanistic studies have demonstrated that in
   response to bacterial virulence factors, such as Porphyromonas
   gingivalis lipopolysaccharide, fibroblasts can release cytokines and
   play a pivotal role in the pathogenesis of OLP [[60]16, [61]17]. In
   vivo experiments have shown that OLP fibroblasts, in response to oral
   bacteria, promote the proliferation and migration of CD4^+ T-helper
   cells and exacerbate the disease process of OLP by expressing the
   chemokine CCL5 [[62]18]. Presently, the interaction between fibroblasts
   and pathogenic oral microbes in promoting the development of OLP
   remains underappreciated.

   To identify key oral pathogenic bacteria enriched in OLP disease and
   elucidate their mechanisms for promoting OLP progression, we compared
   the microbial profiles of buccal and lip mucosa, tongue dorsum, and
   saliva between patients with nonerosive/erosive oral lichen planus
   (NE/E-OLP) and healthy controls. A variety of potential oral bacteria
   were differently distributed in the buccal mucosa of OLP patients. Of
   these, Parvimonas micra (P. micra), the gram-positive anaerobic coccus
   [[63]19], was prominently enriched in the buccal mucosa of OLP patients
   and correlated with the severity of OLP. P. micra is a frequently
   detected opportunistic pathogen in the oral cavity, which plays an
   important role in the occurrence of various oral diseases such as
   endodontic abscesses, periodontitis, and orofacial odontogenic
   infections [[64]20, [65]21]. Our recent study indicated that P. micra
   could commonly be found in oral squamous cell carcinoma tissues, where
   it interacteds with the cell receptor CKAP4 to trigger the activation
   of HIF-1α and autophagy, thereby promoting the metastasis of oral
   squamous cell carcinoma through the TmpC-CKAP4 axis [[66]22]. Herein,
   we conducted the mechanism study between P. micra and oral lichen
   planus fibroblasts (OLPFs), which identified a primary pathogenic
   effector of P. micra and elucidated the cellular receptors, downstream
   regulators, and pathways in fibroblasts. This study provides a deep
   insight into the role of oral microbiota in the progression of OLP.

Materials and methods

Sample collection

   The study was approved by the Ethics Committee of the School and
   Hospital of Stomatology, Shandong University (No. [67]GR201702). The
   inclusion criteria for the patients were as follows: (1) Patients aged
   18–60; (2) patients did not suffer from severe periodontal disease,
   untreated caries, or endodontic infections; (3) patients were free of
   systemic diseases; (4) patients had no smoking habits; (5) patients had
   not taken antibiotics, antiviral drugs, or glucocorticoids in the past
   6 months; (6) patients satisfied the histopathological and clinical
   diagnostic criteria of OLP [[68]23, [69]24]; (7) patients had lesions
   only located on the buccal mucosa of both cheeks; and (8) patients
   signed informed consent to participate. No water or food was allowed to
   be consumed at least 2 h before the collection of oral samples between
   8:00 and 11:00 a.m. Sterile swabs were used to collect mouth swabs from
   the buccal mucosa, lip mucosa, tongue dorsum, and saliva. Two hours
   postprandially, saliva was collected by rinsing the mouth five times
   with sterile double-distilled water, followed by the collection of
   naturally secreted saliva into an Eppendorf tube. All samples were
   collected in sterile Eppendorf tubes and stored at − 80 °C for 20 min.
   VAS scores were obtained before treatment. The scale is divided from 0
   to 10, with 0 being the minimum level of pain and 10 being the maximum
   level of pain [[70]25]. CD3^+ T cells, CD4^+T cells, and CD8^+T cells
   were sorted with flow cytometry (FCM) from the blood of OLP patients.

Generation and analysis of microbiome data

   Samples from 80 healthy individuals and 91 OLP patients (including 50
   nonerosive (NE)-OLP patients and 41 erosive (E)-OLP patients) of buccal
   mucosae, tongue dorsum, lip mucosae, and saliva were collected. The
   hexadecyltrimethylammonium bromide/sodium dodecyl sulfate (CTAB/SDS)
   method was employed to extract deoxyribonucleic acid (DNA) from
   samples. After the genomic DNA of all the samples was extracted,
   polymerase chain reaction (PCR) amplifications were performed. The
   V3–V4 hypervariable regions of the 16S ribosomal ribonucleic acid
   (rRNA) gene were amplified using specific primers (341 F:
   CCTAYGGGRBGCASCAG, 806R: GGACTACHVGGGTWTCTAAT). All PCRs were performed
   in 30 μL reactions with 15 μL of Phusion® High-Fidelity PCR Master Mix
   (Thermo, Waltham, Massachusetts (MA), United States of America (USA)).
   A GeneJET Gel Extraction Kit (Thermo Scientific) was used for purifying
   PCR products that were analyzed by electrophoresis on 2% agarose gels.
   Samples exhibiting strong and distinct bands were selected for further
   in-depth analysis, followed by the sequencing of the library on an
   Illumina HiSeq 2500 platform. Paired-end sequencing reads were
   amalgamated with sequence tags following the overlap relationship
   between reads and passed the quality control (QC) test with UPARSE
   [[71]26]. The USEARCH1 pipeline was used to filter out poor-quality
   reads and chimeras. After dereplication, clean reads were clustered
   into multiple operational taxonomy units (OTUs) with similarities above
   97%. The representative reads of OTUs were aligned to the Ribosomal
   Database Project (RDP, release 18) to obtain a clear taxonomy of the
   microbiome.

   To quantify the core microbiome, we employed stringent metrics,
   including an abundance cutoff of 0.001 and a minimum occupancy
   threshold of 90%. The adonis2 method from the “vegan” package was used
   to probe into the association between host OLP grouping and the
   microbiome of four oral niches. The α- and beta (β)-diversity distance
   matrices were computed using USEARCH11, with a minimum depth threshold
   of 36,000 reads. Principal coordinates analysis (PCoA) was conducted to
   assess differences in complexity based on Bray–Curtis distance. We
   tested the differences between the healthy individuals and OLP patients
   using multivariate association with linear models (MaAsLin2) analysis.
   Effects of age and gender were adjusted. The remaining species or
   genera were chosen if the p-value < 0.05. The intersection of
   differences between the healthy individuals and the other two groups
   was selected for follow-up analysis. K-means clustering analysis was
   carried out for the grouping of differently distributed microbes at the
   buccal mucosa. The differently distributed microbial species and genera
   at the buccal mucosa were clustered into three clusters by the
   similarity of their abundant variation patterns and named clusters 1 to
   3. Spearman correlation analysis was utilized to determine the
   interrelationships between clusters in different groups (|Spearman
   correlation|≥ 0.45, p-value < 0.05). Random forest analysis (version
   4.7–1.1) was used to test the classification effect of these microbes.
   Spearman correlation analysis was performed to determine the
   interrelationships between P. micra and other differently distributed
   microbes. The NetShift method [[72]27], available at
   [73]https://web.rniapps.net/netshift, was used for identifying
   the “driver” nodes in the network of microbial relationships between
   healthy and diseased groups. Cytoscape (version 3.7.1) and Gephi
   (version 0.9.2) were applied to draw the network diagram. Additionally,
   “ggplot2” (version 3.4.0) and “ComplexHeatmap” (version 2.16.0) were
   adopted to generate additional figures.

Fluorescence in situ hybridization (FISH)

   The FISH was done on the 5-μm formalin-fixed paraffin-embedded tissue
   sections of normal and OLP buccal mucosae. The sections were incubated
   for 30 min in a pre-heated oven at 60 °C and then 20 min in a xylene
   solution. Subsequently, they were immersed in gradient alcohol (100%,
   95%, 90%, 80%, and 70%; 1 min each) and washed four times with
   phosphate-buffered solution (PBS). The FISH procedure was performed
   following the instructions of the manufacturer (GenePharma, Suzhou,
   China). The sequence of P. micra for cyanine 3 (cy3)-labeled FISH
   probes was CTGAGCGTCAGTAAAAGTCC [[74]28]. The slides were rinsed with
   4′,6-diamidino-2-phenylindole (DAPI) and visualized under confocal
   laser scanning microscopes (CLSMs). The images were captured using
   Image-Pro Plus 6.0 (Media Cybernetics, Inc., Rockville, MD, USA).

Single-cell RNA-sequencing analysis

   The 10 × single-cell RNA-sequencing (scRNA-seq) data containing one
   normal mucosa sample and five OLP samples were acquired from the Gene
   Expression Omnibus (GEO) database ([75]GSE211630). They were
   transformed into Seurat objects by use of the R software “Seurat”
   package (version 5.0.1). The doublets were removed with the
   “DoubletFinder” package (version 2.0.3). Counts were controlled via the
   exclusion of low-quality cells based on mitochondrial gene percentages
   and cycle gene scores. The “FindVariableFeatures” function was used to
   screen the first 2000 highly variable genes. Downscaling and cluster
   identification were completed using principal component analysis (PCA)
   based on 2000 genes and uniform manifold approximation and projection
   (UMAP). All the samples were integrated using the “RunHarmony”
   function. The “FindAllMarkers” function was used for identifying the
   significant marker genes within various clusters by setting log2 FC to
   1. Cluster annotation analysis was conducted based on the literature
   review of cluster markers. According to the gene family clustering
   results of fibroblasts, differential genes and Kyoto Encyclopedia of
   Genes and Genomes (KEGG) pathways were analyzed between normal and OLP
   tissues with p.adjust < 0.01 and logFC > 2 (FindMarkers and pathview
   version 1.38.0 clusterProfiler version 4.7.1.001). The cellular
   communication of cell subsets was conducted using CellChat (version
   1.6.1).

Bacterial strains and growth conditions

   P. micra strain ATCC 33270 was obtained from the American Type Culture
   Collection (Manassas, VA, USA) and cultured in heart (5 g/L) and brain
   extracts (12.5 g/L), proteose peptone (10.0 g/L), glucose (2.0 g/L),
   sodium chloride (NaCl, 2.0 g/L), and disodium hydrogen phosphate
   (Na[2]HPO[4], 2.5 g/L) at 37 °C under anaerobic conditions with a gas
   mixture of 90% nitrogen (N2), 5% hydrogen (H2), and 5% carbon dioxide
   (CO2).

Isolation of normal and OLP fibroblasts

   Buccal mucosal tissues were acquired from 16 donors (8 were
   pathologically confirmed as OLP, and 8 were normal tissues) at the
   School and Hospital of Stomatology, Shandong University. Tissues were
   treated with dispase (2 mg/mL, Sigma-Aldrich, Darmstadt, Germany) and
   rinsed in PBS. Then, they were cut into pieces with a size of 1 mm^3
   and cultured in a T25 culture flask with α-minimum essential medium
   (α-MEM; Gibco, WA, USA) containing 10% fetal bovine serum (FBS; Gibco,
   WA, USA) and 1% penicillin–streptomycin solution at 37 °C in a
   humidified atmosphere with 5% CO[2]. The purpose was to obtain
   fibroblasts (cells from passages 3–5 were used for subsequent
   experiments). All the patients signed an informed consent form. The
   present study was approved by the Ethics Committee of the School and
   Hospital of Stomatology, Shandong University (No. [76]GR201702).

Isolation and identification of extracellular vesicles (EVs)

   The isolation and identification methods of EVs refer to the previous
   literature [[77]26] with minor modifications. Briefly, P. micra EVs
   were isolated from the above cultures with an optical density at 600 nm
   (OD600) of 0.1 at 37 °C for 48 h. The cultures were pelleted with a
   centrifuge (7800 × g, 4 °C, 20 min) and filtered with a 0.22-μm filter
   membrane. The filtered supernatant was concentrated using 100-kDa
   ultrafiltration tubes and collected by a Beckman Optima XPN-100
   ultracentrifuge (292,700 g, 4 °C, 1 h). After the removal of the
   supernatant, EVs were resuspended in 200-μL sterile PBS and normalized
   to an equivalent protein concentration with a bicinchoninic acid (BCA)
   assay kit (CWBiotech, Beijing, China). The scanning electron microscope
   image of EVs was taken by scanning electron microscopy (SEM). The
   diameters of EVs were precisely measured using a NanoSight LM10
   instrument (Malvern, Westborough, MA, USA). The composition of EVs was
   analyzed using high-performance liquid chromatography-mass spectrometry
   (HPLC–MS, UltiMate 3000).

Screening potential effector proteins of P. micra

   The FASTA protein sequences of P. micra ATCC33270 were downloaded from
   the National Center for Biotechnology Information (NCBI,
   [78]https://www.ncbi.nlm.nih.gov/). Then, they were aligned using the
   virulence factors of pathogenic bacteria (VFDB,
   [79]http://www.mgc.ac.cn/VFs/) and pathogen-host interactions (PHI,
   [80]http://www.phibase.org/index.jsp) databases. The BUSCA tool
   ([81]http://busca.biocomp.unibo.it) was used to predict subcellular
   localization.

Cloning of recombinant DnaK in Escherichia coli

   The genomic DNA of P. micra was used to clone the DnaK gene into
   Escherichia coli (E. coli) expression vector pET-28a. The primers are
   5′-GTGCCGCGCGGCAGCCATATGATGTCAAAAATTATAGGTATTGATTTAGGTAC-3′ and
   3′-ACGGAGCTCGAATTCGGATCCCTATTTATTTTCATCTTCGTCAACTACTTC-5′. Recombinant
   DnaK was generated and purified as a histidine (His)-tag protein in the
   pET-28a vector. pET-28a containing DnaK was expressed in E. coli cells
   with 0.4-mM isopropyl-β-D-thiogalactoside at 37 °C for 4 h until OD600
   reached 1. The protein was purified using a hydroxymethyl (Tris)–NaCl
   buffer system, followed by overnight dialysis against the dialysis
   buffer (10-mM Tris–hydrogen chloride (HCl) pH 8.0, 150-mM NaCl, 10%
   glycerol, 0.001% sarcosyl). The purified protein was concentrated using
   30-kDa ultrafiltration tubes and checked with SDS–polyacrylamide gel
   electrophoresis (PAGE).

   The location of DnaK in EVs was analyzed by proteinase K or Triton
   X-100 according to a reported protocol [[82]29]. There were five groups
   (DnaK, DnaK + proteinase K, EVs, EVs + proteinase K, and EVs
   + proteinase K + Triton X-100 group) involved in this assay. DnaK or
   EVs were processed by 0.05 μg/μL proteinase K (CWBiotech, Beijing,
   China) with/without 0.1% Triton X-100 (CWBiotech, Beijing, China),
   incubated at 3 °C for 10 min, and then added with 5-mM
   phenylmethanesulfonyl fluoride (PMSF) (CWBiotech, Beijing, China) at
   room temperature for 10 min. Afterwards, each sample was heated at 70
   °C for 5 min to suppress proteinase K.

Enzyme-linked immunosorbent assay (ELISA)

   The ELISA assay was carried out between the normal fibroblasts (NFs)
   group and the OLPFs group; between the P. micra-treated NFs (P.
   micra/NFs was 50:1) group, the P. micra conditional medium-treated NFs
   group, and the P. micra-originated EVs (10 μg/uL)-treated NFs group;
   between the NFs group and the NF + lentivirus-GFP-DnaK (lv-GFP-DnaK)
   group; between the NF + lv-GFP-DnaK group and the NF
   + lv-GFP-DnaK + BAG3 siRNA (si-BAG3) group; and between the NF
   + lv-GFP-DnaK group and the NF + lv-GFP-DnaK + IKK-γ siRNA (si-IKK-γ)
   group. The concentration of TNF-α was determined using an ELISA kit
   (BioLegend, CA, USA) following the manufacturer’s instructions.

   The lentivirus-GFP-DnaK (lv-GFP-DnaK) or lentivirus-Flag-DnaK (lv-DnaK)
   and lentivirus control in PCDNA 3.1 were synthesized by Wzbio (Jinan,
   China). BAG3-targeting siRNA (S: CAGCAACCUUGAAGCAGAUTT; AS:
   AUCUGCUUCAAGGUUGCUGTT) and IKK-γ-targeting siRNA (S:
   UGGAGAAGCUCGAUCUGAATT; AS: UUCAGAUCGAGCUUCUCCATT) were synthesized by
   GenePharma (Suzhou, China).

Western blot assay

   For each group, equal amounts of proteins were obtained and resolved by
   10% SDS-PAGE [[83]30] and immunoblotted with primary antibodies against
   matched antigens overnight. The antigens included
   microtubule-associated protein 1 light chain 3 beta (LC3B),
   phosphorylated inhibitory subunit of NF kappa B alpha (p-Iκβα),
   total-Iκβα (CST, CA, USA), DnaK, p-p65, total-p65, IKK-γ (Abcam,
   Cambridge, England), BAG3, sequestosome 1 (SQSTM1, Proteintech, Wuhan,
   China), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH, Affinity,
   OH, USA). These proteins were incubated with the horseradish peroxidase
   (HRP)-conjugated secondary antibodies (Affinity, OH, USA).
   Immunoreactive bands were analyzed by enhanced chemiluminescence
   reaction (ECL, Amersham Pharmacia Biotech, USA) and normalized by GAPDH
   with Quantity One.

mCherry-GFP-LC3B

   NFs, OLPFs, EVs-treated NFs, lv-DnaK, si-BAG3, and si-IKK-γ groups were
   cultured to 50% confluence in each well of 12-well plates for 24 h.
   mCherry-GFP-LC3B adenovirus (WZBio, Jinan, China) was mixed with α-MEM
   and added to every well and cultured at 37 °C for 24 h. The cell nuclei
   were counterstained using DAPI. Images were captured for autophagosomes
   and autolysosomes, which showed yellow and red fluorescence, processed
   with a confocal microscope (TCS-SP8, Leica, Germany). Fluorescence
   signals were analyzed using Image-Pro Plus 6.0 software.

Immunofluorescence histochemistry

   NFs, NFs treated with EVs, lv-DnaK, or stably expressing enhanced GFP
   (EGFP)-LC3B were seeded in 12-well plates with sections for one night
   and fixed with 4% paraformaldehyde. The sections were processed with
   3.0% hydrogen peroxide, followed by incubation in 5.0% bovine serum
   albumin. Besides, they underwent overnight incubation at 4 °C with
   antibodies against pkh67 (Maokangbio, Shanghai, China), β-tubulin (CST,
   CA, USA), BAG3, and IKK-γ and then a 1-h incubation with Alexa Fluor
   488/594-labeled goat anti-mouse/rabbit immunoglobulin G (IgG) (Abbkine,
   CA, USA) at 37 °C in the dark. The images were captured with a laser
   scanning confocal microscope. Quantification of LC3B puncta was
   performed using the ImageJ software package.

Protein pull-down assay

   Protein pull-down assay was performed using the Pierce™ Pull-Down
   PolyHis Protein: Protein Interaction Kit (Thermo Fisher Scientific, New
   York, USA) according to the manufacturer’s instructions [[84]31].
   Briefly, protein lysate was prepared from OLPF, and His-tagged DnaK was
   incubated in PBS for 1 h at 4 °C. After complete binding, the bound
   proteins were immunoprecipitated and analyzed using HPLC–MS (UltiMate
   3000).

Co-immunoprecipitation (Co-IP)

   Co-IP were performed as described with minor modifications [[85]32]. In
   brief, the lysates of lv-Flag-DnaK, si-BAG3, and si-IKK-γ were
   processed by Nonidet P-40 (NP-40) buffer. The lysates underwent
   overnight incubation at 4 °C with DnaK, BAG3, IKK-γ, and Flag
   (Proteintech, Wuhan, China) and normal IgG (Santa, CA, USA) with
   rotation. Protein A/G PLUS agarose beads (Santa, CA, USA) underwent 1 h
   of incubation and rotation at 4 °C. The complexes were washed and
   eluted from the beads by boiling in the SDS-PAGE sample. The
   immunoprecipitates were analyzed by Western blotting with DnaK or BAG3,
   IKK-γ, and Flag.

Statistical analysis

   Data were presented as the mean ± standard deviation (SD) and repeated
   three times. Student’s t-test or one- or two-way analysis of variance
   (ANOVA) was used to analyze inter-group comparisons. In addition, p ≤
   0.05 was considered to show statistical significance, which was
   analyzed using the Statistical Package for the Social Sciences (SPSS)
   software version 17.0.

Results

The overall characteristics of the oral microbiome in patients with NE/E-OLP

   To reveal the variation of the oral microbiome in OLP patients, 704
   samples of buccal mucosa, tongue dorsum, lip mucosa, and saliva from 80
   healthy individuals and 91 OLP patients (including 50 NE-OLP and 41
   E-OLP patients) in Shandong province, China (Table S1), were included.
   The clinical indicators and blood-derived CD3^+, CD4^+, and CD8^+ T
   cells for each OLP patient were collected (Table S1). The 16S rRNA
   amplicon sequencing (CRA 003224) obtained 58.56 million clean reads,
   which were classified into 44 phyla, 976 genera, and 1138 species. The
   saturation of microbiome data at each oral site was depicted in the
   rarefaction curves (Fig. S1 A). At the phylum level, Firmicutes,
   Proteobacteria, and Bacteroidetes were the most enriched microorganisms
   across all the oral sites in the population of this study, but the
   ranking of their abundance varied across oral sites (Fig. [86]1A). In
   healthy individuals, for example, Firmicutes had the highest abundance
   in buccal and lip mucosa, while Bacteroidetes and Proteobacteria were
   mostly enriched in the tongue dorsum and saliva, respectively
   (Fig. [87]1A). However, their abundance was significantly influenced by
   NE-OLP and E-OLP diseases (Fig. [88]1A). At the genus level,
   Streptococcus, Haemophilus, Neisseria, and Rothia were in high
   abundance at each oral site, but the ranking of their abundance was
   also affected by oral sites and OLP disease (Fig. [89]1B). The adonis
   analysis of Bray–Curtis and Jaccard distances simultaneously showed
   that the microbiome of buccal mucosa (p-value < 0.001), tongue dorsum
   (p-value < 0.001), lip mucosa (p-value < 0.001), and saliva (p-value
   < 0.01) was significantly correlated with OLP disease (Fig. [90]1C).
   Among them, the buccal mucosa was the site most correlated with OLP.
   Shannon and Simpson indices indicated that NE-OLP and E-OLP patients
   exhibited significantly greater microbial diversity in the buccal
   mucosa than healthy individuals (p-value < 0.0001, Fig. [91]1D),
   respectively. The microbiome at the lip mucosa and saliva showed the
   same trend as the buccal mucosa (Fig. S1B, C), while the tongue dorsum
   increased in abundance despite being insignificant (Fig. S1D).

Fig. 1.

   [92]Fig. 1
   [93]Open in a new tab

   Characteristics of the microbiome presented in the buccal and lip
   mucosa, tongue dorsum, and saliva of healthy individuals and NE/E-OLP
   patients. The abundance exhibited variability at both the phylum (A)
   and genus (B) levels among four oral sites of healthy individuals and
   NE/E-OLP patients. C Adonis analysis of Bray–Curtis and Jaccard
   distances showed the correlation of the microbiome and OLP among four
   oral sites. D Shannon and Simpson indices among NE-OLP, E-OLP patients,
   and healthy individuals. E Core microbiome of buccal mucosa at the
   species level in healthy individuals, NE-OLP, and E-OLP patients. F
   β-Diversity of the microbial composition of buccal mucosa, measured by
   Bray–Curtis and Jaccard distance. *p-value < 0.05, **p-value < 0.01,
   ***p-value < 0.001, ****p-value < 0.0001

   The core microbiome of the buccal mucosa at the genus level with
   occurrence rate ≥ 90% and relative abundance ≥ 0.001 was analyzed, and
   18 core microorganisms were found in all individuals. Meanwhile, it was
   found that seven microbes, including Parvimonas, Tannerella, and
   Treponema, were enriched in OLP patients with high frequency and
   abundance (Fig. S1E). At the species level, 22 microbes exhibited high
   abundance in the whole population (Fig. [94]1E). Besides, we identified
   an additional eight microbes as the core bacteria in OLP patients,
   encompassing the periodontal pathogen Tannerella forsythia, as well as
   P. micra (Fig. [95]1E). β-Diversity analysis showed that the microbial
   composition of the buccal mucosa of NE- and E-OLP patients greatly
   differed from that of healthy individuals, with the first principal
   component accounting for 20.2% of the variation according to
   Bray–Curtis distance and 14.1% according to Jaccard distance
   (Fig. [96]1F). The β-diversity of the lip mucosa and saliva exhibited
   no significant difference at the first principal component between NE-
   and E-OLP patients and healthy individuals (Fig. S1 F, G). The first
   principal component of the dorsal tongue showed a significant
   difference between NE- and E-OLP patients and healthy individuals
   (p-value < 0.01, Fig. S1H). These results suggest that oral
   microorganisms, especially those in the buccal mucosa site, are closely
   associated with NE/E-OLP.

P. micra in the buccal mucosa exhibited a close correlation with NE/E-OLP

   At the species level, MaAsLin analysis was performed to compare the
   buccal mucosa microbiota between NE/E-OLP patients and healthy
   individuals. The results showed that there were significant differences
   observed in 75 and 72 microbes (p-value < 0.05) in patients with NE-
   and E-OLP, respectively, compared to healthy individuals (Table S2). A
   heatmap of 31 microbes overlapped in NE-OLP and E-OLP patients was
   presented in Fig. [97]2A. These microbes were categorized into three
   clusters, with those in groups 2 and 3 being notably more abundant in
   patients with NE-OLP and E-OLP. Among them, P. micra,
   Peptostreptococcus stomatis, Streptococcus oralis, and Fusobacterium
   nucleatum exhibited the greatest divergence from those found in healthy
   individuals (Fig. [98]2B). Spearman correlation showed that the
   microbes within clusters 2 and 3 were closely and positively correlated
   with each other, while cluster 1 was seldom associated with clusters 2
   or 3 (Fig. [99]2C). MaAsLin analysis at the genus level revealed that
   Parvimonas, Peptostreptococcus, and Streptococcus were notably the
   microbes exhibiting the greatest divergence between healthy individuals
   and patients with NE/E-OLP (Fig. S2 A). These microbes may be potential
   pathogenic bacteria that contribute to the development of NE/E-OLP.

Fig. 2.

   [100]Fig. 2
   [101]Open in a new tab

   Identification of the key pathogenic bacteria correlated with the
   pathogenesis of OLP. A Heatmap of the differently distributed microbes
   in both NE-OLP and E-OLP patients. B Comparison of the abundance of P.
   micra, P. stomatis, S. oralis, and F. nucleatum among healthy
   individuals, NE-OLP and E-OLP patients. C Spearman correlation at the
   species level among healthy individuals, NE- OLP and E-OLP patients. D
   The random forest analysis utilized to differentiate patients with
   NE/E-OLP from healthy individuals. E MeanDecreaseAccuracy revealed the
   contribution of microbes distinguishing NE-OLP and E-OLP patients from
   healthy individuals. F Spearman correlation among P. micra and other
   most contributing microbes. G Regression analysis between P. micra and
   the clinical indices in NE-OLP and E-OLP patients. *p-value < 0.05,
   **p-value < 0.01, ***p-value < 0.001, ****p-value < 0.0001

   The random forest analysis was conducted on 31 differential bacteria
   present in the buccal mucosa to ascertain their classificatory efficacy
   in distinguishing NE/E-OLP patients from healthy individuals. The
   results showed that the area under the curve (AUC) was 0.785 for
   distinguishing NE-OLP patients from healthy individuals (Fig. [102]2D)
   and an AUC of 0.849 for differentiating E-OLP patients from healthy
   individuals (Fig. [103]2D). Next, the MeanDecreaseAccuracy of these
   bacteria was computed, revealing that P. micra made the greatest
   contribution to distinguishing NE- and E-OLP patients from healthy
   individuals (Fig. [104]2E), respectively. Random forest analysis
   conducted at the genus level also showed that Parvimonas was the
   primary microbe distinguishing NE-OLP and E-OLP patients from healthy
   individuals (Fig. S2 C). Spearman correlation showed a positive
   correlation between P. micra and other high contributing microbes,
   including F. nucleatum and T. forsythia (Fig. [105]2F). NetShift
   analysis showed that P. micra served as a pivotal bacterium,
   facilitating the transition from healthy individuals to E-OLP patients,
   as well as from NE-OLP to E-OLP patients (Fig. S2D). The abundance of
   P. micra was analyzed in relation to clinical indices in patients with
   NE-OLP and E-OLP. P. micra exhibited a significant correlation with VAS
   in the NE-OLP patients (p-value = 0.009), as well as with CD8^+ T cells
   (p-value = 0.024) and CD4^+/CD8^+ T cells (p-value = 0.001) in the
   E-OLP patients (Fig. [106]2G). These results suggested that P. micra
   may play an important role in the progress of OLP.

OLP increased the inflammatory activity and blocked autophagy in fibroblasts

   The presence of P. micra in the buccal mucosa of OLP patients was
   validated by the FISH assay, which showed that P. micra was
   significantly enriched in the junction of epithelial and connective
   tissue layers of OLP patients compared to those of healthy individuals
   (p-value < 0.05) (Fig. [107]3A). As a proinflammatory bacterium in the
   oral cavity, P. micra can cause a series of inflammatory responses by
   invading host tissues [[108]19]. The inflammatory characteristics in
   the buccal mucosa of OLP patients were investigated by scRNA-seq
   between normal and OLP patients ([109]GSE211630). Ten clusters of cells
   were annotated based on the enrichment of cluster-defining gene markers
   (Fig. [110]3B). Cell component analysis showed that the proportion of
   fibroblasts dramatically decreased, while the proportion of immune
   cells increased, especially NK/T cells (Fig. [111]3C). Intercellular
   and molecular communication showed that communications between CD4^+ T
   cells, CD8^+ T cells, and fibroblasts were increased. Also, fibroblasts
   could regulate a majority of other cell types in the epithelial and
   stromal tissues of OLP patients (Fig. [112]3D). KEGG pathway enrichment
   analysis indicated that a variety of inflammatory-related pathways such
   as NF-κB, TNF-α, JAK-STAT, and PI3 K-AKT were enriched in OLPFs
   (Fig. [113]3E). TNF-α was linked to the activation of T cells,
   Langerhans’ cells, mononuclear cells, and epithelial keratinocytes in
   OLP [[114]33]. The secreting cytokines in the NF-κB pathway showed that
   TNF-α, CCL21, and CCL19 were significantly upregulated in OLPFs
   compared with NFs (Fig. [115]3F). To confirm the activity of TNF-α in
   fibroblasts within the buccal mucosa, fibroblasts were isolated and
   cultured from both healthy individuals (NFs) and OLP patients (OLPFs).
   Western blot showed that p-p65 and p-Ikβα were significantly
   overexpressed in OLPFs compared with NFs (p-value < 0.05)
   (Fig. [116]3G). ELISA showed that OLPFs exhibited a higher
   concentration of TNF-α in comparison to NFs (p-value < 0.05)
   (Fig. [117]3H). These results indicate that the inflammatory response
   occurring within OLPFs represents a crucial and fundamental
   pathological characteristic of OLP disease.

Fig. 3.

   [118]Fig. 3
   [119]Open in a new tab

   The increase of inflammatory activity and the blockade of autophagy in
   OLPFs. A The FISH image of P. micra between healthy individuals (n =
   16) and OLP (n = 15) patients. The white arrow represents P. micra. B
   The clustering results of buccal mucosa cells between healthy
   individuals and OLP patients. C Comparison of cell components between
   healthy individuals and OLP patients. D Intercellular and molecular
   communication between CD4^+ T cells, CD8^+ T cells, and fibroblasts. E
   KEGG pathway enrichment analysis of inflammatory-related pathways
   between healthy individuals and OLP patients. F The comparison of
   secreted cytokines in the fibroblasts between healthy individuals and
   OLP patients. G Western blot of the NF-κB pathway in NFs and OLPFs. H
   ELISA of TNF-α between NFs and OLPFs. I Western blot of autophagic
   proteins in NFs and OLPFs. J Western blot of LC3B-I and LC3B-II in NFs
   and OLPFs treated with/without BafA1. K Confocal images of NFs and
   OLPFs, labeled with Ad-mCherry-GFP-LC3B, were utilized to distinguish
   autolysosomes (exhibiting GFP-negative, mCherry-positive vesicles) from
   autophagosomes (showing both GFP- and mCherry-positive vesicles). Scale
   bar: 10 µm. LC3B puncta signal intensities of images for three
   independent experiments are shown. *p-value < 0.05. L ELISA of TNF-α
   between NFs and OLPFs treated with/without BafA1. *p-value < 0.05

   Autophagy serves as a crucial host cell defense strategy against
   microbial stimulation [[120]34, [121]35]. We examined the autophagic
   status in NFs and OLPFs and found a significant elevation of SQSTM1
   (p-value < 0.05) and LC3B-II (p-value < 0.05) in OLPFs relative to NFs
   (Fig. [122]3L). Using the inhibitors BafA1, the autophagic flux
   analysis of OLPFs showed an increase in LC3B-II in both NFs and OLPFs
   upon the inhibition of lysosomal-dependent LC3B-II degradation
   (Fig. [123]3M). The ratio of mCherry (red) (autolysosomes) to yellow
   puncta (autophagosomes) in NFs and OLPFs, as indicated by
   mCherry-GFP-LC3, showed that OLP elevated the count of autophagosomes
   in OLPFs while simultaneously reducing the count of autolysosomes in
   NFs (Fig. [124]3N). Next, we investigated whether the blockade of
   autophagy was involved in TNF-α secretion. ELISA assays conducted on
   NFs and OLPFs, in the presence and absence of BafA1 treatment, showed
   that the addition of BafA1 significantly increased the level of TNF-α
   in OLPFs (p-value < 0.05, Fig. [125]3O). These results suggest that the
   autophagic activity is notably disrupted within OLPFs.

P. micra promoted the release of TNF-α and blocked autophagy in NFs

   We evaluated whether P. micra could induce inflammatory activity in
   NFs. ELISA was used to compare TNF-α levels in NFs and NFs treated with
   P. micra, revealing a significant increase of TNF-α upon treatment with
   P. micra (p-value < 0.05) (Fig. [126]4A). To verify whether P. micra or
   P. micra-released components promoted the inflammatory effect, culture
   medium (CM) from P. micra was obtained and treated NFs. The results
   showed that incorporation of CM into NFs led to a substantial elevation
   of TNF-α (p-value < 0.05) (Fig. [127]4A). EVs are one of the key
   components secreted by microbes in the pathogenesis of oral disease
   [[128]36]. Therefore, EVs were isolated and characterized from the CM
   of P. micra (Fig. S3 A, B). The abundance of TNF-α between NFs and NFs
   + P. micra-derived EVs was compared and showed that the EVs of P. micra
   could significantly increase the secretion of TNF-α (p-value < 0.05)
   (Fig. [129]4A). Next, whether P. micra, CM, and EVs of P. micra could
   activate the NF-κB pathway was examined by Western blot. The results
   showed a significant elevation in the phosphorylation of p65 and Iκβα
   in P. micra, CM, and EVs groups when compared to the NF group,
   respectively (p-value < 0.05, Fig. [130]4B). These observations imply
   that P. micra-derived EV could stimulate the inflammatory response of
   NFs.

Fig. 4.

   [131]Fig. 4
   [132]Open in a new tab

   P. micra and P. micra-derived EVs exacerbated the TNF-α secretion and
   blocked autophagic flux in NFs. A ELISA of TNF-α in NFs treated with P.
   micra (Pm), CM, or EVs of P. micra. B Western blot of the autophagy
   markers, LC3B-I and LC3B-II, in NFs treated with P. micra (Pm), CM, and
   EVs of P. micra. C Western blot of p-p65, total-p65, p-Iκβα, and
   total-Iκβα in NFs treated with P. micra (Pm), CM, and EVs of P. micra.
   D Western blot of LC3B-I and LC3B-II in NFs treated with EVs of P.
   micra, with/without BafA1. E Confocal images of mCherry (red), GFP-LC3
   (green), and DAPI (blue) fluorescence in NFs treated with EVs of P.
   micra. Scale bar: 25 µm. Images were representative of three
   independent experiments. *p-value < 0.05

   Next, the analysis of the autophagic state in NFs upon the stimulation
   of P. micra, CM, or EVs showed that both SQSTM1 and LC3B-II were
   over-expressed by P. micra (p-value < 0.05), CM of P. micra (p-value
   < 0.05), and EVs (p-value < 0.05), respectively (Fig. [133]4C). In
   addition, the expression of LC3B-II was not changed between NFs and
   NFs-EVs treated with BafA1 (Fig. [134]4D). mCherry-GFP-LC3 tests showed
   that the autophagic status in NFs was significantly blocked with the
   incorporation of EVs (p-value < 0.05) (Fig. [135]4E). These results
   suggested that EVs are important pathogenic components of P. micra to
   block the autophagic flux in fibroblasts.

DnaK in P. micra-derived EVs promoted TNF-α release and blocked autophagy

   To reveal how P. micra-derived EVs promote inflammation and block
   autophagy in NFs, we labeled EVs with pkh67 and labeled NFs’
   cytoskeletal proteins with β-tubulin. Immunofluorescence (IF) showed
   that EVs could transfer into the cytoplasm of NFs (Fig. [136]5A). EVs
   may exert their toxic effects by releasing pathogenic proteins or small
   RNAs within host cells [[137]36–[138]38]. Herein, the proteome of P.
   micra-derived EVs was profiled by HPLC–MS, which identified 12
   candidate effectors (Table S3). Among them, DnaK was selected for the
   following study. Western blot of DnaK in OLPFs showed that it was
   significantly enriched in OLPFs in comparison to NFs (p-value < 0.05,
   Fig. [139]5B). Similarly, a higher abundance of DnaK could be detected
   in NFs stimulated by P. micra (p-value < 0.05), CM of P. micra (p-value
   < 0.05), and P. micra-derived EVs (p-value < 0.05) in comparison to NFs
   (Fig. [140]5C), respectively. These results suggest that DnaK might
   play an important role in the pathogenicity of P. micra-derived EVs.

Fig. 5.

   [141]Fig. 5
   [142]Open in a new tab

   DnaK in P. micra-derived EVs promoted inflammatory activation and
   blocked autophagy. A Co-localization of EVs-labeled pkh67 and β-tubulin
   in NFs treated with EVs of P. micra. B Western blot of DnaK in NFs and
   OLPFs. C Western blot of DnaK in NFs treated with P. micra, CM, and EVs
   of P. micra. D Western blot of DnaK in EVs treated with proteinase K
   (K), Triton X-100 (T), and T + K. E ELISA of TNF-α in NFs and NFs with
   lv-DnaK. F Western blot of SQSTM1, LC3B-I, and LC3B-II in NFs and NFs
   with lv-DnaK. G Western blot of SQSTM1, LC3B-I, and LC3B-II in lv-DnaK.
   H Western blot of LC3B-I and LC3B-II in lv-DnaK, with/without BafA1. I
   Confocal images of mCherry, GFP-LC3, and DAPI fluorescence in NFs and
   lv-DnaK. Scale bar: 25 µm. Images were representative of three
   independent experiments. *p-value < 0.05

   To analyze the location of DnaK in EVs, EVs were treated with
   proteinase K (which degrades protein but not membranous structures)
   alone or with both proteinase K and Triton X-100, compared with DnaK or
   both DnaK and proteinase K. As shown in Fig. [143]5D, the abundance of
   DnaK reduced in the EVs treated with proteinase K compared with the EVs
   group, and DnaK in the EVs treated with proteinase K and Triton X-100
   was less abundant than in EVs treated with proteinase K. These results
   suggested that DnaK was situated both within the outer membrane and the
   cytoplasm of EVs.

   We tested the pathogenic effect of DnaK on NFs. ELISA showed that the
   transfection of DnaK (lv-DnaK) into NFs could facilitate a significant
   increase of TNF-α (p-value < 0.05, Fig. [144]5E). The activation of the
   NF-κB pathway and autophagy under the stimulation of lv-DnaK was
   examined. The results showed that the phosphorylation of Iκβα and p65
   exhibited a marked increase (Fig. [145]5F), and the expression of
   SQSTM1 and LC3B-II was also remarkably increased (p-value < 0.05,
   Fig. [146]5G) in comparison to NFs. Western blot showed that BafA1
   treatment did not significantly influence the expression of LC3BII in
   lv-DnaK (Fig. [147]5H). mCherry-GFP showed that lv-DnaK significantly
   inhibited the count of autolysosomes (p-value < 0.05) while
   significantly increasing the count of autophagosomes (p-value < 0.05,
   Fig. [148]5I). These results suggested that DnaK is an important
   pathogenic effector of P. micra-derived EVs.

DnaK interacted with BAG3 to activate the NF-κB pathway and block autophagy
flux

   GST/His pull-down assay and HPLC–MS analysis identified a series of
   candidate receptors of DnaK in fibroblasts (Table S4). Among them, BAG3
   was reported as an autophagy-related protein [[149]39]. Co-IP assay
   showed DnaK could bind to BAG3 (Fig. [150]6A). IF assay showed DnaK
   could co-localize with BAG3 in the nucleus and cytoplasm of lv-GFP-DnaK
   (Fig. [151]6B). The abundance of DnaK and BAG3 in NFs and OLPFs was
   compared and showed that in comparison to NF, BAG3 showed no
   significant increase in OLPFs (Fig. [152]6C).

Fig. 6.

   [153]Fig. 6
   [154]Open in a new tab

   BAG3 in fibroblasts interacted with DnaK to induce the subsequent
   pathogenic effect. A Co-IP of BAG3 and DnaK. B Co-localization of BAG3
   and DnaK in lv-DnaK. C Western blot of BAG3 in NFs and OLPFs. D ELISA
   of TNF-α of lv-DnaK and si-BAG3. Western blot of NF-κB proteins (E) and
   SQSTM1, LC3B-I, and LC3B-II (F) in lv-DnaK and si-BAG3. BafA1 treatment
   (G) and mCherry-GFP-LC3B (H) tests detected autophagy flux. Images were
   representative of three independent experiments. *p-value < 0.05

   To examine the role of BAG3, a transfection assay was performed to
   knock down BAG3 by small-interference RNA in lv-DnaK (si-BAG3). ELISA
   analysis showed that si-BAG3 could dramatically suppress the abundance
   of TNF-α compared with lv-DnaK (p-value < 0.05) (Fig. [155]6D). The
   activation of the NF-κB pathway in si-BAG3 was examined and showed that
   p-Iκβα and p-p65 were significantly decreased in si-BAG3 (p-value
   < 0.05, Fig. [156]6E). Moreover, the role of BAG3 in autophagy was
   tested and showed that si-BAG3 remarkably decreased the expression of
   SQSTM1 and LC3B-II (p-value < 0.05, Fig. [157]6F). With the treatment
   of BafA1, LC3B-II was significantly increased in si-BAG3 (p-value
   < 0.05, Fig. [158]6G). mCherry-GFP showed si-BAG3 significantly
   increased the count of autolysosomes and autophagosomes (p-value
   < 0.05, Fig. [159]6H). These results suggest that BAG3 is an important
   intracellular receptor mediating the pathogenic effect of DnaK.

DnaK-BAG3 interacted with IKK-γ to activate the NF-κB pathway and block
autophagy flux

   BAG3 is reported to modulate the NF-κB pathway via IKK-γ [[160]40].
   Western blot of IKK-γ in OLPFs showed that IKK-γ was significantly
   enriched in OLPFs in comparison with NFs (p-value < 0.05,
   Fig. [161]7A). The interaction of IKK-γ-DnaK-BAG3 was verified by Co-IP
   in lv-DnaK and showed that IKK-γ could combine with DnaK-BAG3
   (Fig. [162]7B). The knockdown of BAG3 significantly decreased the
   abundance of IKK-γ in lv-DnaK (p-value < 0.05) (Fig. [163]7C). In
   addition, the combination of DnaK and IKK-γ was also significantly
   reduced by the knockdown of BAG3, compared with lv-DnaK (p-value
   < 0.05, Fig. [164]7D).

Fig. 7.

   [165]Fig. 7
   [166]Open in a new tab

   Interaction of DnaK-BAG3 with IKK-γ activated inflammatory response and
   blocked autophagy. A Western blot of IKK-γ in NFs and OLPFs. B Co-IP of
   BAG3, Flag, and IKK-γ in lv-DnaK and si-BAG3. C Western blot of IKK-γ
   in lv-DnaK and si-BAG3. D The combination of DnaK and IKK-γ in si-BAG3
   and lv-DnaK. E Co-localization of IKK-γ and EGFP-LC3B in lv-DnaK. F
   Western blot of IKK-γ, autophagic proteins, and NF-κB proteins. G
   Western blot of SQSTM1, LC3B-I, and LC3B-II in lv-DnaK and si-IKK-γ
   with/without the treatment of BafA1. H Confocal images of mCherry,
   GFP-LC3, and DAPI fluorescence in lv-DnaK and si-IKK-γ. Images are
   representative of three independent experiments. I ELISA of TNF-α in
   lv-DnaK and si-BAG3. J Simplified schematic diagrams summarizing the
   key mechanisms. *p-value < 0.05

   IKK-γ mediates the effect of autophagy [[167]41] and activates the
   NF-κB pathway [[168]42]. IKK-γ and EGFP marked LC3B puncta were
   co-localized in lv-DnaK (Fig. [169]7E). We knocked down IKK-γ using
   siRNA in lv-DnaK (si-IKK-γ) and demonstrated that si-IKK-γ
   significantly reduced the abundance of SQSTM1 and LC3B-II compared to
   lv-DnaK (Fig. [170]7F). A study showed that IKK-γ can interact with
   SQSTM1 to regulate LC3 proteins [[171]43]. The knockdown of IKK-γ could
   reduce the expression of SQSTM1 with or without BafA1 treatment
   (Fig. [172]7G). mCherry-GFP transfection tests showed that si-IKK-γ
   significantly increased the count of autolysosomes (p-value < 0.05)
   while significantly decreasing the count of autophagosomes in lv-DnaK
   (p-value < 0.05, Fig. [173]7H). Furthermore, si-IKK-γ decreased the
   level of TNF-α in lv-DnaK (Fig. [174]7I). These results suggest that
   IKK-γ mediated the DnaK-BAG3-induced TNF-α expression and blockage of
   autophagy (Fig. [175]7J).

Discussion

   The oral microbiome of OLP patients differed from that of healthy
   individuals, which suggested that the oral microbiota may correlate
   with the onset and progression of OLP [[176]8–[177]11]. Nevertheless,
   how oral microbes contribute to OLP progression is still unknown. In
   the current study, the correlation between OLP and a rarely reported
   oral opportunistic pathogen, P. micra, was first established, but the
   other possible bacteria cannot be excluded. Meanwhile, the first
   pathogenic study of P. micra in OLP was conducted, and the new
   mechanism by P. micra-derived EV which promotes an inflammatory
   response and blocks autophagy was discovered in this study
   (Fig. [178]8).

Fig. 8.

   [179]Fig. 8
   [180]Open in a new tab

   Mechanic diagram of P. micra facilitated the development of OLP. P.
   micra accumulated within the lesional mucosa of the buccal region,
   permeating through epithelial tissue to infiltrate the connective
   tissue. P. micra released EVs into fibroblasts, which constitute a
   primary cellular component. The DnaK protein derived from these EVs
   interacted with BAG3 in fibroblasts, thereby influencing autophagy and
   modulating the NF-κB signaling pathway via IKK-γ, ultimately affecting
   the expression of TNF-α

   To systematically reveal the association of the oral microbiome with
   OLP disease, a comprehensive microbiome-wide association study was
   first conducted between NE/E-OLP patients and healthy individuals at
   the buccal region, the lip mucosa, the tongue dorsum, and saliva.
   Consistent with previous studies [[181]8, [182]10, [183]44], OLP
   patients and healthy individuals showed significant differences,
   especially at the oral site of the buccal mucosa. Notably, P. micra
   exhibited the most pronounced differential distribution, which is known
   as an oral commensal bacterium closely related to chronic
   periodontitis, oral cancer, and colorectal cancer [[184]19, [185]22,
   [186]45]. FISH assay validated that P. micra was enriched in the
   junction of epithelial and connective tissue layers with high abundance
   in OLP patients.

   scRNA-seq data showed that the proportion of T cells increased, and
   communications between CD4^+ T cells, CD8^+ T cells, and fibroblasts
   were increased in OLP patients. It is crucial to acknowledge that a
   hallmark histopathological feature of OLP is the presence of a
   band-like infiltrate of T cells within the superficial connective
   tissue of the subepithelial layer [[187]46]. The quantity of CD4^+ and
   CD8^+ T cells is closely correlated with the severity, progression, and
   clinical classification of OLP [[188]47]. Therefore, we speculate that
   fibroblasts possess the ability to enhance the recruitment and
   migration of T cells into connective tissue, similar to the viewpoint
   proposed by Xu [[189]48]. As a result, we directed our research efforts
   on fibroblasts, the main cellular component of the connective tissues
   of the buccal mucosa, which remain stationary in a stable environment
   [[190]14]. In response to tissue damage and inflammatory reactions,
   fibroblasts can be activated to exert multiple immune effects.
   Single-cell analysis found that the TNF-α of the fibroblasts-related
   NF-κB pathway was activated and communicated with multiple cells in the
   tissues of OLP patients. Consistently, it was proved that fibroblast
   from OLP patients could overexpress TNF-α and activate the NF-κB
   pathway. Thus, how the inflammatory response of fibroblasts is
   triggered is an important question in OLP disease.

   The FISH assay showed the significant enrichment of P. micra in the
   junction of epithelial and connective tissue layers in OLP patients,
   which is a highly enriched region of fibroblasts. Hence, the pathogenic
   study was carried out between P. micra and fibroblasts. It was
   demonstrated that P. micra could promote TNF-α secretion and block
   autophagy. To explore the underlying mechanism, CM and EVs of P. micra
   were treated to fibroblasts, and obtained similar results. As a result,
   EVs of P. micra was one of the important ways to exert
   its pathogenicity have firstly received our attention [[191]49].

   Furthermore, we isolated and demonstrated the potential toxic protein
   DnaK in P. micra-derived EVs. Studies showed that DnaK (70-kDa heat
   shock proteins, Hsp70 s) could mimic the chaperone protein of
   eukaryotic cells to induce protein misfolding in the cellular process
   and play crucial roles in diverse physiological processes and
   pathological conditions [[192]50]. For example, DnaK of Mycoplasma and
   Pseudomonas aeruginosa can be identified by host receptors, thereby
   initiating inflammation via the activity of proteins like p53 and NF-κB
   [[193]51, [194]52]. To explore the interacted protein of DnaK in
   fibroblasts, 12 candidate host proteins were captured. Among them, BAG3
   was reported to interact with Hsp70 s and regulate the NF-κB pathway
   [[195]40, [196]53]. So, we selected BAG3 for the following study. In
   addition, as a sensor of Hsp70 s, BAG3 can bind to numerous protein
   domains that support different activities by affecting the upstream
   regulators of the signaling pathway and the process of autophagy flux
   [[197]39]. BAG3 expression was silenced in lv-DnaK (si-BAG3). It was
   found that si-BAG3 could secrete less TNF-α, inhibit the NF-κB pathway,
   and activate autophagy. Next, Co-IP assays were conducted. It was
   proved that IKK-γ directly interacted with DnaK and BAG3; the
   expression of IKK-γ decreased with the decrease of BAG3 in si-BAG3;
   si-IKK-γ could secrete less TNF-α, inhibit NF-κB pathway, and activate
   autophagy. Additionally, IKK-γ was mediated by LC3 proteins binding to
   LC3-interacting region (LIR) motifs present in p62 [[198]43]. IKK-γ
   lowered the threshold of p62 in lv-DnaK and si-IKK-γ with and without
   BafA1 treatment.

   Our prior research has demonstrated that the composition and
   proportions of microbes in the human microbiome, such as the gut
   microbiome, were significantly influenced by geographical factors,
   lifestyle patterns, and dietary habits [[199]54]. It is noteworthy that
   the microbiome analysis of OLP patients in this study primarily drew
   from a single region (Shandong, China), potentially introducing a
   regional bias in the identification of microbiota and pathogenic
   bacteria. This limitation may hinder our comprehensive understanding of
   the pathogenic mechanism of OLP. In this study, we focused on the
   pathogenicity of P. micra, and there must be other oral microorganisms
   that play a very important role in the progression of OLP.
   Additionally, in the exploration of the pathogenic mechanism of P.
   micra, we focused on elucidating the mechanism of the EVs protein DnaK.
   Given the diversity of pathogenic components within the EVs secreted by
   microorganisms, it implies that P. micra’s EVs may encompass numerous
   distinct pathogenic elements and modes of action, necessitating further
   in-depth investigation [[200]55]. Meanwhile, in vitro fibroblast models
   cannot fully reflect the changing processes in vivo.

   In conclusion, we present the first suggestion that EVs of P. micra
   may play a vital role in the progress of OLP disease. DnaK of P. micra
   EVs interacted with the BAG3 of fibroblasts and modulated the
   expression of IKK-γ, which regulated the secretion of TNF-α by
   activating the NF-κB pathway and inhibiting autophagy (Fig. [201]8).
   This study provides a novel perspective on understanding the
   pathological mechanisms of OLP. The microorganisms enriched in OLP,
   identified through this research, can function as biomarkers for
   predicting the risk and progression of the disease. The elucidation of
   the pathogenic mechanisms of these newly discovered microorganisms will
   further drive advancements in the prevention and treatment of OLP.

Supplementary Information

   [202]40168_2025_2151_MOESM1_ESM.zip^ (18.2MB, zip)

   Supplementary Material 1: Supplementary figures: Fig. S1.
   Characteristics of the oral microbiome in patients with NE/E-OLP. A.
   The rarefaction curves illustrate the saturation of microbiome data at
   each oral site. B, C, and D. The α diversity of the microbial
   composition at the lip mucosa, tongue dorsum, and saliva is assessed
   using the Shannon and Simpson indices. E. The core microbiome at the
   genus level in healthy individuals, NE-OLP patients, and E-OLP
   patients. F, G, and H. The β diversity analysis of the microbial
   composition at the lip mucosa, saliva, and tongue dorsum is conducted
   using Bray-Curtis and Jaccard distances. * p-value < 0.05, **p-value <
   0.01, *** p-value < 0.001, **** p-value < 0.0001. Fig. S2. Analysis of
   the oral bacteria closely associated with NE/E-OLP at the genus level.
   A Heatmap of microbes overlapping in NE-OLP and E-OLP groups at the
   genus level. B Random forest predictions for the healthy and NE/E-OLP
   microbiota of oral sites at the genus level, and their contribution
   calculated by MeanDecreaseAccuracy (C). D NetShift network showing the
   network driver bacterium from healthy individual to NE-OLP patients,
   from healthy individual to E-OLP patients, and from NE-OLP patients to
   E-OLP patients. * p-value ˂ 0.05, ** p-value ˂ 0.01,*** p-value ˂
   0.001, **** p-value ˂ 0.0001. Fig. S3. Characterizing OMV of P. micra
   by scanning electron microscopy (A) and nanoparticle tracking (B).
   [203]40168_2025_2151_MOESM2_ESM.zip^ (76.4KB, zip)

   Supplementary Material 2: Supplementary tables: Table S1. Information
   on healthy individuals and NE,E-OLP patients and their average clinical
   indices. Table S2. The results of MaAsLin2 analysis of different
   distributed microbial genera and species among healthy, NE-OLP, E-OLP
   groups. Table S3. The putative effector proteins of P. micra-derived
   EVs by HPLC-MS. Table S4. Mass spectrometry analysis of DnaK-binding
   proteins.
   [204]Supplementary Material 3.^ (7.4MB, pdf)

Acknowledgements