The Guidelines of Peking University Institutional Review Board for the care and use of laboratory animals were followed for the animal experiments. The Laboratory Animal Center of Peking Medical University provided the female BALB/c and non-obese diabetic (NOD) mice (both 6–8 weeks old). Mice were housed in a specific pathogen-free environment and had ad libitum access to standard rodent chow and water.

The institutional review board of Peking University School of Stomatology approved the collection of human peripheral blood and human labial glands for research purposes (PKUSSIRB-201631139). Samples were collected after obtaining informed written consent from the patients.

Samples from five donors who required surgery to remove their lip mucoceles, and were free from any systemic diseases, were used to isolate LGMSCs. The adjacent normal labial glands (LGs) were collected after enucleation of lip mucoceles. After isolation, phosphate-buffered saline (PBS) (Gibco, Waltham, MA, USA) was used to wash the LGs and their surrounding fibrous tissues and blood vessels were carefully wiped away. Then, the remaining tissues were cut into small parts and cultured in cell culture flasks containing alpha-Minimum Essential Medium (α-MEM) (Gibco) with 20% fetal bovine serum (Gibco) and 1% penicillin/streptomycin (Beyotime, Shanghai, China) in a humidified atmosphere of 5% CO₂ at 37°C. When the cells reached 80%–90% confluence, 0.25% trypsin (MP Biomedicals, USA) was used to digest the adherent cells, which were passaged in vitro. All the experiments used MSCs at the third and fourth passages.

LGMSCs were identified using previously described methods (21), including surface marker assessment using flow cytometry, osteogenic differentiation, and adipogenic differentiation.

LGMSCs were grown in alpha-Minimum Essential Medium containing 10% exosome-depleted fetal bovine serum for 48 h, after which we collected the culture supernatants. The supernatants were centrifuged sequentially at 300 × g for 10 min, 2000 × g for 10 min, and 10,000 × g for 30 min. Lastly, the supernatants were subjected to ultracentrifugation at 100,000 × g at 4°C for 60 min using an Optima L-100XP ultracentrifuge (Beckman Coulter, Placentia, CA, USA), followed by washing using PBS and centrifugation at 100,000 × g for another 1 h. A BCA Protein Assay Kit (Thermo Fisher, Waltham, MA, USA) was used to determine the protein concentration to quantify the LGMSC-Exos. Nanoparticle tracking analysis (NTA), transmission electron microscopy (TEM) observation, and western blotting were used to characterize the purified LGMSC-Exos.

For western blotting, we used anti-CD63 and CD81 antibodies (1:1000 dilution; (Abcam, Cambridge, UK) (Abcam). For NTA, exosomes were diluted to a suitable concentration with PBS and analyzed using a ZetaView PMX 110 instrument (Particle Metrix, Munich, Germany). For TEM observation, the concentration of the exosomes was adjusted to roughly 10¹¹ vesicles per mL using PBS and 20 μL of the diluted exosomes were deposited on copper coated 200-mesh formvar grids and dried for 5 min at room temperature. Samples were then stained with 10 μL phosphotungstic acid for 2 min at room temperature. Filter paper was used to blot off the excess solution, and the samples were dried at room temperature, observed under a transmission electron microscope, and photographed.

Peripheral blood from patients with pSS or age- and sex- matched donors was collected into EDTA Vacutainers. Ficoll density-gradient centrifugation (Haoyang Biotech, Tianjin, China) was used to isolate PBMCs. In co-culture system, the LGMSCs and PBMCs were added at a ratio of 1:10, and exosomes (30 μg/mL) were added into complete Roswell Park Memorial Institute 1640 medium for PBMCs treatment. PBMCs were collected for subsequent experiments at 72 h after co-culture. To evaluate whether effect of LGMSCs on B cells is mediated by their exosomes, GW4869, a commonly used pharmacological agent that inhibits exosome generation was used. LGMSCs and PBMCs were cultured in regular culture medium with 10 μM GW4869. To evaluate PRDM1/Blimp1 expression in B cells after treated with LGMSC-Exos, PBMCs were collected at 72 h after co-culture and then MagniSortTM Human CD19 Positive Selection Kit (Thermo Scientific, MA, USA) was used to isolate CD19⁺ B cells from PBMCs.

All mice used were female. NOD mice transplanted with LGMSCs or LGMSC-Exos comprised the treatment groups. NOD mice intragastrically administrated with hydroxychloroquine (HCQ) served as the positive controls and mice infused with PBS comprised the negative controls. For transplantation, LGMSCs (1 × 10⁶ diluted in 200 μL PBS/mouse) or LGMSC-Exos (50 μg diluted in 200 μL PBS/mouse) were injected into NOD mice via their tail veins on alternate days for 14 days. Body weight, serum glucose levels, and the salivary flow rate of the mice were measured every other week until mice were sacrificed at week 16.

PBMCs co-cultured with LGMSCs or LGMSC-Exos and splenic lymphocytes collected after the mice were sacrificed were subjected to flow cytometry analysis for B cell subsets. For PBMCs, they were incubated with anti-human CD19, CD20, IgD, CD38, CD27, and CD24 monoclonal antibodies (5 μL per million cells in a 100 μL staining volume) (BioLegend, San Diego, CA, USA) for 30 mins. For mouse splenic lymphocytes, they were incubated with anti-mouse CD19, CD27, and CD138 monoclonal antibodies (5 μL per million cells in a 100 μL staining volume) (BioLegend) for 30 mins. Thereafter, the cells were washed two times before being analyzed on a CytoFLEX flow cytometer (Beckman Coulter, Brea, CA, USA).

The Trizol Reagent (Takara, Tokyo, Japan) was used to extract total RNA from cells according to the supplier’s instructions. A PrimeScript™ RT Reagent Kit (Takara) was used to synthesize cDNA from the RNA. cDNA representing miRNA was generated using a miRNA First Strand cDNA Synthesis Kit (Sangon Biotech, Shanghai, China). The quantitative real-time PCR step of the qRT-PCR protocol was performed in a reaction volume of 10 μL using SYBR^® Premix Ex Taq™ by Roche LightCycler^® 480II (Roche, Mannheim, Germany). The relative expression levels of genes were calculated using the 2^–ΔΔCT method (22), and the primer sequences are listed in Table 1 .

The interaction between PRDM1 (encoding PR domain zinc finger protein 1) and hsa-miR-125b-5p (miR-125b) was predicted using TargetScan software (version 7.1) (23). Position 166–172 of the PRDM1 3’ UTR was predicted to bind to miR-125b. The PRDM1 fragment containing the miR-125b binding sites was synthesized to generate wild-type (PRDM1–WT) or mutant-type (PRDM1–MUT). The reporter of fluorescence in the vector was encoded by the Renilla Luciferase gene (Rluc), and PRDM1–WT and PRDM1–MUT were cloned downstream of Rluc the gene, separately. Then, the luciferase reporter plasmids PRDM1–WT and PRDM1–MUT (RiboBio, Guangzhou, China) were co-transfected with miR-125b mimic or control mimics (RiboBio) into HEK-293 T cells. Forty-eight hours after transfection, the Glo^® Luciferase Assay System (Promega, Madison, WI, USA) was used to lyse the cells and a GLomax20/20 Luminometer (Promega) was used to determine the luciferase activity.

MiR-125b in exosomes was depleted and upregulated by transfection of miR-125b mimics and inhibitors (RiboBio), respectively, into LGMSC-Exos using a Fect™ Exosome Transfection Kit (System Biosciences, Palo Alto, CA, USA).

Data are presented as the mean ± standard deviation. The statistical analyses were carried out using GraphPad Prism version 6.01 (GraphPad Software, La Jolla, CA, USA) or IBM SPSS Statistics version 26 (IBM, Armonk, NY, USA). Differences between groups were analyzed using the Mann–Whitney U-test or Kruskal–Wallis comparison test.

LGMSCs exhibited a typical spindle-shaped morphology ( Figure 1A ) and expressed the MSC markers CD29, CD73, and CD90 (99.9%), but did not express the endothelial cell marker CD31 or the hematopoietic marker CD45 ( Figure 1B ). Moreover, costimulatory molecule expression (HLA-DR, CD86, and CD80) was negative ( Figure 1C ). Their multiple differentiation potential was demonstrated after induction for 21 days using alizarin red staining (for osteogenesis) and Oil Red O staining (for adipogenesis) ( Figure 1D ).

LGMSC-Exos were isolated and characterized by NTA, TEM observation, and western blotting analysis. As shown in Figures 1E, F , morphologically, LGMSC-Exos were spherical and cup-shaped, and NTA particle size analysis showed that they had the expected size of exosomes. Upon western blotting, we observed that isolated exosomes contained CD9, CD63 and Psg101, but had no calnexin ( Figure 1G ). Thus, the isolated vesicles exhibited characteristic morphologies and phenotypes of exosomes, confirming their successful isolation from the supernatants of cultured LGMSCs.

LGMSC-Exos were injected into NOD mice via their tail veins. The therapeutic potential of LGMSC-Exos was evaluated using saliva flow rates and lymphocytic infiltration of submandibular glands of the NOD mice ( Figure 2A ). Remarkably, treatment with LGMSC-Exos improved the reduced saliva flow rates in the NOD mice significantly ( Figure 2B ). Furthermore, histopathological assessment of the LGMSC-Exo-treated NOD mice showed a considerably reduced area and number of lymphocyte infiltration foci compared with that in the PBS-treated NOD mice (p < 0.05, Figures 2C, D ). The saliva flow rates and H&E staining both showed similar therapeutic effects of LGMSC-Exos compared with those of LGMSCs.

Splenic lymphocytes were collected and CD19⁺ B cells, CD19⁺CD27⁺ memory B cells, and CD19^-CD138⁺ plasma cells were analyzed by flow cytometry. Compared with those in the PBS-treated group, the proportions of CD19^-CD138⁺ plasma cells were significantly reduced in spleens of LGMSC-Exo-treated NOD mice ( Figures 3A, B ). However, the numbers of total B cells and memory B cells were not altered ( Figures 3C, D ).

The LGMSC-Exos had a similar treatment effect in NOD mice compared with their parental cells. To investigate whether the effect on plasma cells was mediated, or at least partially mediated, by exosomes, we evaluated the effect in vitro. As shown in Figure 4A , when LGMSCs were treated with the exosome inhibitor GW4869, their effect on plasma cells was partially abolished. Moreover, when PBMCs from patients with pSS were cocultured with LGMSC-Exos, we observed a markedly decreased in the percentage of CD19⁺CD20^-CD24⁺CD38⁺ plasma cells ( Figure 4B ), showing similar effects to their parental cells.

To identify the likely target via which LGMSC-Exos mediate their inhibitory effect on plasma cells, mRNA profiling of B-lymphocytes activated in the presence or absence of exosomes was searched in the literature (24). PRDM1, also known as B lymphocyte induced maturation protein (Blimp1), is a regulator of B cell differentiation. According to mRNA profiling data from Khare et al., PRDM1 was one of the differentially expressed genes. Thus, we investigated whether the expression of PRDM1 in PBMCs from patients with pSS was regulated by LGMSC-Exos. Firstly, we detected the mRNA and protein levels of PRDM1/Blimp1 in B cells from pSS patients. As shown in Figures 5A, B , PRDM1 expression was increased in B cells from patients with pSS compared with that in B cells from healthy controls (HCs). Moreover, immunohistochemical staining showed increased levels of PRDM1 in labial glands from patients with pSS ( Figure 5C ). Furthermore, 72 hours after coculture, PBMCs were collected and CD19⁺ B cells were purified for PCR and western blotting analysis. The results showed that both the mRNA level and protein levels of PRDM1 were reduced after co-culture with LGMSC-Exos ( Figures 5D, E ).

To identify exosomal miRNAs that might exert the therapeutic effect of LGMSC-Exos, we furthered examined the miRNAs carried by LGMSC-Exos using miRNA sequencing. The top 30 most abundant miRNAs in LGMSC-Exos are shown in Figure 6A . Among those miRNAs, has-miR-125b-5p (miR-125b) was chosen for further analysis after a literature review (25). Overall, LGMSC-Exos carry approximately 462 miRNAs, including 65 novel miRNAs ( Supplementary Table S1 ).

MiR-125b was overexpressed (OE) or knocked down (KD) in LGMSC-Exos by transfection of miR-125b mimics or inhibitors, respectively. As shown in Figure 6B , the transfection efficiency was demonstrated by qRT-PCR. PBMCs from patients with pSS were co-cultured with OE-miR125b-Exos or KD-miR125b-Exos, followed by flow cytometry detection of B cell subset percentages in the collected PBMCs, accompanied by western blotting and qRT-PCR detection of PRDM1 expression. The results showed that OE-miR125b-Exos treatment slightly reduced the mRNA levels of PRDM1 whereas the mRNA levels of PRDM1 was significantly enhanced after KD-miR125b-Exos treatment ( Figure 6C ). These results were confirmed by western blotting ( Figure 6D ). Furthermore, the percentage of CD19⁺CD20^-CD24⁺CD38⁺ plasma cells was significantly reduced in OE-miR125b-Exos treatment group, whereas KD-miR125b-Exos treatment had the opposite effect ( Figure 6E ).

To determine whether miR-125b targets potential sites in PRDM1, dual luciferase reporter assays were performed ( Figure 7A ), which demonstrated a significant reduction in the relative luciferase activity from the PRDM1–WT plasmid compared with that from the PRDM1–MUT vector in the presence of miR-125b mimics ( Figure 7B ). Thus, miR-125b targets the 3′-UTR of PRDM1 to restrain its expression.