Adult female New Zealand white rabbits (3.5-4 kg) were obtained from Vital River Laboratory Animal Technology (Beijing, China). All procedures with animals were performed in accordance with the guidelines of the Laboratory Animal Care and Use Committee of Tianjin Medical University Eye Hospital and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research (Permit Number: TJYY20181217001). The collection and use of human blood samples and human umbilical cords were approved by the Ethics Committee of Tianjin Medical University Eye Hospital (Permit Number: 2018KY(L)-41).

Fresh human umbilical cords were obtained and digested to isolate hUC-MSCs as previously described in our work (6). Briefly, the umbilical cords were washed twice, cut into around 1 mm³ pieces and then digested with 0.1% collagenase type II (Gibco, USA) for 1h on shaker at 37°C. After that, the cell suspension was filtered through meshes and then centrifuged. The harvested pellets were resuspended in complete culture medium and subsequently incubated at 37°C. The medium with 10% fetal bovine serum (FBS, cat# 16000-044, Gibco, USA) was refreshed every 2-3 days. The phenotype of hUC-MSCs was determined by flow cytometry using specific antibodies as follows: CD29-FITC (clone TS2/16), CD34-FITC (clone 4H11), CD44-FITC (clone IM7), CD45-FITC (clone HI30) and CD105-PE (clone SN6) (eBioscience™, USA); CD73-FITC (clone AD2), CD90-FITC (clone 5E10), CD11b-FITC (clone M1/70) and HLA-DR-PE (clone LN3) (Biolegend, USA). The abilities of hUC-MSCs to differentiate into adipocytes, osteoblasts and chondrocytes were evaluated as described previously (23).

Passage 3-5 hUC-MSCs (2-2.5×10⁶ cells) were cultured with 25ml culture medium containing 10% sEVs-free FBS (prepared by overnight ultracentrifugation at 110,000 g at 4°C) in 175 cm² cell culture flask for 48h, and at least 400ml culture medium was collected for one batch of cultures. sEVs were isolated using differential centrifugation based on previously described methods (24). In brief, 400ml supernatants collected from cultured MSCs were centrifuged at 2000 ×g for 10 min followed by 10000 ×g for 30 min to remove cellular debris and dead cells, and then at 110000 ×g for 70 min twice via ultracentrifugation (Type 45Ti rotor, Beckman Coulter, USA), all at 4°C. Then the pellets were resuspended in 400-500μl PBS, sterilized by filtration through a prerinsed 0.22-μm filter and stored at -80°C until use. BCA protein assay reagent kit (Solarbio, China) was used to detect the protein concentration of the collected sEVs, and nanoparticle tracking analysis (NTA) (NanoSight NS300, UK) was performed to determine the diameter and particle concentration of sEVs. The mean protein and particle contents of sEVs were 0.55 ± 0.05 μg/μl and 5.72×10⁸ ± 3.12×10⁷ particles/μl. Visualization of sEVs was assessed by HT7700 transmission electron microscope (TEM) (HITACHI, Japan). 20 μg sEVs were denatured in SDS loading buffer at 95°C for 5 min, and then were used for western blot assays to detect the typical sEV markers including CD9, CD63, CD81 and tumor susceptibility 101 (TSG101), and the non-sEV marker Calnexin.

Autoimmune dacryoadenitis was induced as described previously (25). Briefly, the epithelial cells purified from normal rabbit LGs (pLGECs) were irradiated and then co-cultured with autologous peripheral blood lymphocytes (PBLs) at the ratio of 1:1. After 5 days, the activated PBLs from mixed cell reactions were harvested and adoptively injected back into rabbits via ear margin vein to induce autoimmune dacryoadenitis.

To explore the preventive and therapeutic effects of hUC-MSC-sEVs, two administration schedules were arranged: 1) Model rabbits in the preventive experiments were subconjunctivally injected with hUC-MSC-sEVs (30 μg) (hUC-MSC-sEVs group, n=9) or PBS (untreated group, n=9) at the early stage of disease after adoptive transfer of activated PBMCs (days 1, 3, 5, 7 and 9). Normal rabbits were used as controls (normal group, n=9). 2) Model rabbits in the therapeutic experiments were treated at the developed stage (2 weeks after transfer) with every-two-day injection of hUC-MSC-sEVs (30 μg) (hUC-MSC-sEVs group, n=6) or PBS (untreated group, n=6) for 5 doses. Based on our preliminary experiments, this administration protocol had been confirmed to be most effective for rabbit autoimmune dacryoadenitis.

The clinical assessments were conducted every 2 weeks after the first hUC-MSC-sEVs administration. The tear break-up time (BUT), the severity of the corneal fluorescein staining and the tear production were detected and recorded as described previously (26). Histopathological evaluation of LGs and conjunctivas was performed by hematoxylin and eosin (H&E) staining at the end of the experiment (8 weeks after transfer), and the number of focus (aggregate of >50 lymphocytes) per 4 mm² in LGs was recorded as previously reported (26).

The human monocytic cell line THP-1 was obtained from Stem Cell Bank, Chinese Academy of Sciences, and was differentiated into macrophages by treatment with 160 ng/ml phorbol 12-myristate 13-acetate (PMA, Sigma Aldrich, USA) for 24h. Then the differentiated macrophages (5×10⁵ cells/ml) were stimulated with 100 ng/ml lipopolysaccharides (LPS) from E. coli K12 (ultrapure, cat# tlrl-peklps, Invivogen, USA) and 50 ng/ml recombinant human interferon-γ (IFN-γ, R&D Systems, USA) for an additional 24 hours. After that, the adherent cells were incubated with fresh medium containing hUC-MSC-sEVs (5 μg/ml) or PBS for 48 h. The treated macrophages were finally collected for the following co-culture experiment or RNA and protein harvest.

miR-100-5p mimics and negative-control (NC) mimics, as well as miR-100-5p inhibitor and NC inhibitor, were designed and synthesized by Gene Pharma (Suzhou, China). LPS+IFN-γ-stimulated macrophages were transfected with miR-100-5p mimics (100 nM) or NC mimics using Lipofectamine 2000 Reagent (Thermo Fisher Scientific, USA) according to the manufacturer’s instructions. After 48h transfection, the cells were collected for subsequent experiments. hUC-MSCs were transfected with 100 nM of miR-100-5p inhibitor, NC inhibitor, miR-100-5p mimics and NC mimics for 6h. Then the culture medium was replaced with that containing 10% sEVs-free FBS for sEVs isolation. The effect of transfection was confirmed by Q-PCR.

To identify the internalization of hUC-MSC-sEVs into macrophages, hUC-MSC-sEVs were stained with PKH26 fluorescent dye (Sigma-Aldrich, USA) according to the manufacturer’s instructions and centrifuged again to remove contaminating dye. Then, the PKH26-labeled hUC-MSC-sEVs (10 μg/ml) were co-cultured with PMA-treated macrophages. After 6h, the macrophages were washed with PBS and fixed in 4% paraformaldehyde for 15min. Thereafter, the nucleus was stained with DAPI (Solarbio, China). Finally, the cells were observed and photographed with the laser scanning confocal microscope (LSM800, Zeiss, Germany).

CD4+ T cells were isolated from the healthy human peripheral blood by magnetic activated cell sorting (MACS) (negative selection), using the CD4+ isolation Kit II (Miltenyi Biotec, USA) following Ficoll density gradient centrifugation, and then were seeded in 24-well plate (1×10⁶ cells/well) or 96-well plate (1.2×10⁵ cells/well) with the stimulation of ImmunoCult™ Human CD3/CD28 T Cell Activator (STEMCELL, Canada). Then CD4+ T cells were co-cultured with hUC-MSC-sEVs-treated macrophages at a ratio of 40:1 (CD4+ T cells: macrophages). After 5 days, CD4+ T cells were prepared for flow cytometric analysis or Bromodeoxyuridine (BrdU) assay.

Activated human CD4+ T cells treated with hUC-MSC-sEVs-educated macrophages were incubated with BrdU labelling reagent (10 μM) for the last 24h. Cell proliferation was quantitatively determined based on the measurement of BrdU incorporation, using a cell proliferation ELISA kit (Roche Diagnostics GmbH, Germany) following manufacturer’s instructions.

The lymphocytes were isolated from LGs as described previously (26). The lymphocytes in LGs, rabbit PBMCs treated by hUC-MSC-sEVs or human CD4+ T cell co-cultured with macrophages were fixed, permeabilized and stained with mouse anti-human Foxp3-PE (clone 206D), rat anti-human CD4-FITC (clone A161A1), mouse anti-human CD25-APC (clone BC96) (BioLegend, USA), and mouse anti-rabbit CD4-FITC (clone KEN-4, BIO-RAD, USA). Stained cells were analyzed with FACS Calibur flow cytometer (BD Biosciences, USA).

Total RNA from LGs or cells was extracted using EZ-press RNA Purification Kit (EZBioscience, USA) following manufacturer’s instructions. Quantitative real-time PCR (Q-PCR) was performed using Sybr Green Master Mix and an ABI 7900 HT Sequence Detection System (both from Thermo Fisher Scientific, USA). The relative expression was calculated using the following equation: relative gene expression= 2^{[ΔCt(control)–ΔCt(target)]}.

Total RNA of hUC-MSC-sEVs was isolated using Exosome RNA Purification Kit (EZBioscience, USA). Additionally, the synthetic miRNA Caenorhabditis elegans miR-39 (cel-miR-39, 200fmol/μl, Sequence: UCACCGGGUGUAAAUCAGCUUG), was added to the isolated total RNA, and was used as an exogenous control (27). The steps for real-time amplification were the same as above.

The sequences of the primers are listed in Table S1 - S3 .

After dewaxed and rehydrated, LG tissue paraffin sections were subjected to citric acid antigen retrieval. Then sections were incubated with CD68 antibody (clone KP1, 1:50, Abcam, UK) overnight at 4°C after blocking with 1% goat serum, followed by incubating with anti-mouse IgG H&L (Alexa Fluor 488) secondary antibody (1: 500, Abcam, UK) at room temperature for 2 h away from the light. Nuclear DNA was labelled with DAPI. Images were observed with laser scanning confocal microscope (LSM800, Zeiss, Germany).

Proteins isolated from cells, tissues or sEVs were loaded on 8-10% SDS-PAGE and transferred onto PVDF membrane. The membranes were blocked with 5% skimmed milk at RT for 1 h and further incubated with mouse antibodies against β-actin (clone OTI1, 1:2000, ZSGB-BIO, China), iNOS/NOS2 (clone C-11, 1:500, Santa Cruz, USA), Arg1 (clone C1, 1:500), CD9 (clone MEM-61, 1:500), CD63 (clone TS63, 1:500), CD81(clone M38, 1:500), TSG101 (clone 4A10, 1:500), or Calnexin (clone 6F12BE10, 1:500) (Abcam, UK) overnight at 4°C. Detection was carried out using anti-mouse IgG, HRP-linked secondary antibody (1:1000, CST, USA) for 2 h at room temperature. The bands were visualized using Tanon 4800 Multispectral Imaging System (Tanon Science & Technology, China).

Data from at least three independent experiments were presented as mean ± SD. The Shapiro-Wilk test was used to test the normality of the data. Comparisons between groups were made using the Student’s t test and one-way ANOVA, or Mann-Whitney U test and Kruskal-Wallis test according to their distribution. All analyses were performed using SPSS 25.0 software (IBM Corporation, USA), and a value of P ≤ 0.05 was considered statistically significant.

The hUC-MSCs we obtained were positive for CD29, CD44, CD105, CD73, CD90 and negative for CD45, CD34, CD11b and MHC Class II antigen (HLA-DR) ( Figure 1A ), and were able to differentiate into adipocytes, osteocytes and chondrocytes under in vitro culture ( Figure 1B ). hUC-MSC-sEVs were purified from the conditioned medium of hUC-MSCs by serial differential centrifugation plus ultracentrifugation. Examination of TEM revealed that hUC-MSC-sEVs were round or ellipsoid, cup-shaped with double layer membrane structure ( Figure 1C ). NTA demonstrated that the diameters of these particles mainly ranged from 50-150 nm with a peak at around 113 nm ( Figure 1D ). Using western blot, we further confirmed that these nanovesicles were positive for specific markers: CD81, CD9, CD63 and TSG101, and were negative for the non-sEV marker Calnexin ( Figure 1E ).

The rabbit autoimmune dry eye model usually developed obvious clinical symptoms on the second week after adoptive transfer of activated PBLs (25). To investigate whether hUC-MSC-sEVs have preventive effects on autoimmune dacryoadenitis in rabbits, we subconjunctivally injected hUC-MSC-sEVs or PBS into rabbits during the early stage after adoptive transfer (days 1, 3, 5, 7 and 9) ( Figure 2A ), and observed the clinical signs until the eighth week. The hUC-MSC-sEVs administration attenuated the disease severity as early as 2 weeks after transfer. Compared with the untreated group, the hUC-MSC-sEVs group displayed milder corneal epithelial damage, more tear secretion and longer tear BUT at all the time points observed ( Figures 2B-E ). To further validate the effect of hUC-MSC-sEVs, the histological examination was performed on week 8 after transfer. As shown in Figures 2F-G , less infiltration of inflammatory cells and more minor tissue damage were observed in LGs and conjunctivas of rabbits in the hUC-MSC-sEVs group when compared to those without hUC-MSC-sEVs administration. All together, these findings suggested that hUC-MSC-sEVs administered in the early stage of disease were effective in attenuating the clinical severity and reducing lacrimal and conjunctival inflammation in induced rabbit dry eye model.

Macrophages are important regulators in inflammation of LGs (5, 28). To clarify its underlying mechanism of how hUC-MSC-sEVs prevented the rabbit autoimmune dacryoadenitis, we investigated the effects of hUC-MSC-sEVs on macrophages accumulated in inflamed LGs. Immunofluorescence assay showed that the expression of CD68, a typical pan macrophage marker, was lower in perivenular/periductal area of LGs from model rabbits treated with hUC-MSC-sEVs than those from untreated groups ( Figure 3A ). It is well recognized that macrophages exhibit the ability to switch between M1 and M2 in response to microenvironmental changes (29). We therefore asked whether hUC-MSC-sEVs could regulate macrophage polarization. LGs collected from rabbits with or without hUC-MSC-sEVs administration were applied for Q-PCR and western blot. As shown in Figure 3B , in comparison with those of the untreated group, the gene expression of M1 markers nitric oxide synthase 2 (NOS2), interferon-regulatory factor 5 (IRF5), TNF-α, IL-1β and IL-6 was significantly reduced, whereas the gene expression of M2 markers, such as arginase-1 (Arg1), CD206, krueppel-like factor 4 (KLF4), IL-10 and TGF-β, was significantly increased in LGs of hUC-MSC-sEVs group. Consistently, the western blot analysis showed that hUC-MSC-sEVs administration led a dramatic increase in protein level of Arg-1 and a significantly reduced NOS2 protein expression in LGs ( Figure 3C ). Collectively, these findings suggested that hUC-MSC-sEVs could convert macrophages into the anti-inflammatory M2 phenotype in response to the chronic inflammation in LGs. These M2-like macrophages may have beneficial effects on alleviating autoimmune dacryoadenitis through secreting anti-inflammatory cytokines and promoting tissue repair.

Considering the critical role of Tregs in preventing lacrimal gland autoimmunity (14), we next examined the effect of hUC-MSC-sEVs on Tregs. Tregs comprise a distinct anti-inflammatory lineage specified by the X-linked transcription factor Foxp3, which can be directly regulated by the orphan nuclear receptor, Nurr1 (30). We observed that the mRNA expression of Foxp3 and Nurr1, and the proportion of CD4+Foxp3+Tregs in inflamed LGs were significantly elevated after hUC-MSC-sEVs administration ( Figures 3D, E ). Together, these data indicated that hUC-MSC-sEVs administration promoted Treg induction in rabbit autoimmune dacryoadenitis.

In order to observe the therapeutic effects of hUC-MSC-sEVs in vivo, we started treating model rabbits with every-two-day injection of hUC-MSC-sEVs or PBS at the developed stage of the disease (2 weeks after transfer) for 5 doses ( Figure 4A ). As shown in Figures 4B-E , the relief of clinical signs was observed 2 weeks after first hUC-MSC-sEVs injection. Treatment with hUC-MSC-sEVs resulted in significantly lower fluorescein staining scores and obviously increased BUT from week 4 to week 8, and remarkably increased aqueous tear volume at week 6 and 8 in rabbits compared with those in the untreated group. Histologically, immune cell infiltrates were diminished in LGs and conjunctivas of model rabbits by hUC-MSC-sEVs administration ( Figures 4F, G ). As expected, these results indicated that hUC-MSC-sEVs administered at the developed stage of diseases alleviated the severity of rabbit autoimmune dacryoadenitis.

We next investigated the modulatory effects of hUC-MSC-sEVs on M2 macrophage polarization and Treg cell differentiation in vitro. PBMCs from model rabbits stimulated with irradiated pLGECs were treated with or without hUC-MSC-sEVs (5 μg/ml). Firstly, we observed that the transcript levels of M1 markers (NOS2, TNF-α, IL-1β and IL-6) were obviously reduced, whereas the mRNA levels of M2 markers (Arg1, CD206, KLF4, IL-10 and TGF-β) were significantly increased in the hUC-MSC-sEVs group compared with the control group, which was consistent with the protein expression of NOS2 and Arg1 ( Figures 5A, B ). Then we evaluated the effect of hUC-MSC-sEVs on Tregs. The Q-PCR analysis revealed the significantly higher expression of Foxp3 and Nurr1 in the hUC-MSC-sEVs group than that of the control group ( Figure 5C ). Flow cytometry analysis of PBMCs collected after 5-day co-culture showed that hUC-MSC-sEVs treatment significantly increased the percentage of CD4+Foxp3+Tregs ( Figure 5D ). Taken together, these data demonstrated that hUC-MSC-sEVs could induce M2 polarization and promote Treg generation in stimulated PBMCs.

Given the essential role of macrophages in innate and adaptive immunity, we next focus on the influence of hUC-MSC-sEVs on macrophages. Under confocal laser-scanning microscope, we observed that the PKH26 (red)-labeled hUC-MSC-sEVs were internalized into PMA-treated THP-1 cells (called as Mac) after six-hour co-incubation and localized in the cytoplasm ( Figure 6A ). To further validate the role of hUC-MSC-sEVs on macrophage phenotype, LPS+IFN-γ-stimulated Mac was co-cultured with or without hUC-MSC-sEVs for 48h. The levels of M1 and M2 markers were detected by Q-PCR and western blot analysis. The results showed that hUC-MSC-sEVs could effectively elicit human M1 macrophage to an anti-inflammatory M2 state in vitro ( Figures 6B, C ).

Previous studies have shown that MSC-sEVs required other mediator cells such as monocytes to promote Treg expansion (31, 32), and anti-inflammatory macrophages could induce Tregs (33). Hence, we sought to examine whether the macrophages educated by hUC-MSC-sEVs influenced T cell proliferation and Treg differentiation. After incubated with hUC-MSC-sEVs as aforementioned, Mac and LPS+IFN-γ -stimulated Mac were isolated from each group and co-cultured with human peripheral blood-derived CD4+ T cells in the presence of anti-CD3/-CD28 antibodies for 5 days. The BrdU proliferation assay revealed that CD4+ T cell proliferation was significantly repressed by hUC-MSC-sEVs-educated macrophages, especially those pre-stimulated with LPS+IFN-γ ( Figure 6D ). In addition, the flow cytometric analysis showed that hUC-MSC-sEVs-educated macrophages (with LPS+IFN-γ pre-stimulation) significantly elevated the proportion of CD4+Foxp3+Tregs compared with other groups ( Figure 6E ). These combined findings showed that hUC-MSC-sEVs-educated macrophages with M2-like phenotype could suppress CD4+ T proliferation and augment Treg generation.

Increasing evidence have shown that miRNAs are the important functional cargos in sEVs (34, 35). Therefore, we determined to investigate whether hUC-MSC-sEVs contained any pivotal miRNAs that contributed to their immunomodulatory effects on macrophages. We analyzed the miRNA sequencing outcomes of hUC-MSC-sEVs reported in three published studies (36–38), and aligned them to a public available database of exosomal miRNA profiles (ExoCarta). The top 5 miRNAs from each of the sequencing data were shown in Table S4 . Among them, miR-100-5p attracted our attention: First, it was one of most highly enriched miRNAs in hUC-MSC-sEVs in comparison with both sEVs derived from human embryonic kidney 293T cells (293T-sEVs) and human dermal fibroblasts (HDF-sEVs). Second, it was previously reported to be crucial in maintaining tumor-associated macrophage M2 phenotype (39). Thus, we hypothesized that sEVs-miR-100-5p might play a role in sEVs-mediated M2 polarization. To test this, we performed Q-PCR assay and confirmed that miR-100-5p was significantly highly expressed in hUC-MSC-sEVs we isolated ( Figure 7A ). In addition, we found that miR-100-5p in LGs of model rabbits was significantly upregulated after treatment with hUC-MSC-sEVs ( Figure S1 ), suggesting that miR-100-5p may be helpful for improved rabbit autoimmune dacryoadenitis. Subsequently, in order to explore the direct effects of miR-100-5p on macrophage polarization in vitro, LPS+IFN-γ-stimulated Mac were transfected with miR-100-5p mimics or NC mimics. The results showed that miR-100-5p overexpression significantly facilitated the polarization of macrophages from M1 to M2 phenotype ( Figure S2 ). Collectively, these data indicated that miR-100-5p might be the key components contributing to the regulation of hUC-MSC-sEVs on M2 macrophage polarization.

To further confirm the role of miR-100-5p in hUC-MSC-sEVs, sEVs purified from the conditioned medium of hUC-MSCs transfected with miR-100-5p inhibitor, NC inhibitor, miR-100-5p mimics or NC mimics (called as miR-100-5p inhibitor-sEVs, NC inhibitor-sEVs, miR-100-5p mimics-sEVs or NC mimics-sEVs, respectively), were co-cultured with macrophages as mentioned before. The transfection efficiency was confirmed by Q-PCR analysis ( Figure S3 ). We observed that the FAM (green)-labeled miR-100-5p inhibitor was contained in PKH26 (red)-labeled sEVs, which were internalized into macrophages after 6 hours of co-incubation ( Figure 7B ). Further analysis revealed that, compared with NC inhibitor-sEVs, miR-100-5p inhibitor-sEVs were less effective in suppressing M1 marker expression and enhancing M2 marker expression ( Figures 7C-E ), indicating that inhibiting miR-100-5p in hUC-MSC-sEVs blunted their effects on M2 macrophage polarization. Furthermore, miR-100-5p mimics-sEVs markedly upregulated M2 markers and downregulated M1 markers compared to those NC mimics-sEVs ( Figures S4A, B ). These results together support the notion that miR-100-5p was involved in the hUC-MSC-sEVs mediated M2 macrophage polarization.

Next, we explored whether miR-100-5p in hUC-MSC-sEVs influenced the function of hUC-MSC-sEVs-educated macrophages on T cells. To this end, THP-1 derived macrophages pretreated with miR-100-5p inhibitor-sEVs or NC inhibitor-sEVs were collected and then co-cultured with human CD4+ T cells. The BrdU proliferation assay showed that CD4+ T cell proliferation was markedly inhibited by NC inhibitor-sEVs-educated macrophages, not by miR-100-5p inhibitor-sEVs-educated macrophages ( Figure 7F ). Likewise, the percentage of the CD25+Foxp3+Treg populations was significantly lower in CD4+ T cells co-cultured with miR-100-5p inhibitor-sEVs-educated macrophages than in those co-cultured with NC inhibitor-sEVs-educated macrophages ( Figure 7G ), whereas the percentage of the CD4+Foxp3+Treg populations was obviously upregulated by miR-100-5p mimics-sEVs-educated macrophages ( Figure S4C ). Hence, these findings indicated that miR-100-5p in hUC-MSC-sEVs was, at least, partly responsible for the T cell suppression and Treg cell expansion induced by hUC-MSC-sEVs-educated macrophages.