TGF-β-induced CD4+ FoxP3+ regulatory T cell-derived extracellular vesicles modulate Notch1 signaling through miR-449a and prevent collagen-induced arthritis in a murine model

Abstract

CD4⁺FOXP3⁺ Treg cells are central to the maintenance of self-tolerance and can be defective in autoimmunity. In autoimmune rheumatic diseases, dysfunctional self-tolerance, is to a large extent, caused by insufficient Treg-cell activity. Although nTregs have therapeutic effects in vivo, their relative scarcity and slow rate of in vitro expansion hinder the application of nTreg therapy. It was previously reported that EVs contribute significantly to the suppressive function of FOXP3⁺ Treg cells. Considering that the stability and plasticity of nTregs remain major challenges in vivo, we established EVs derived from in vitro TGF-β-induced Treg cells (iTreg-EVs) and assessed their functions in a murine model of autoimmune arthritis. The results demonstrated that iTreg-EVs preferentially homed to the pathological joint and efficiently prevented the imbalance in Th17/Treg cells in arthritic mice. Furthermore, we found that miR-449a-5p mediated Notch1 expression modulation and that miR-449a-5p knockdown abolished the effects of iTreg-EVs on effector T cells and regulatory T cells in vitro and in vivo. Taken together, our results show that iTreg-EVs control the inflammatory responses of recipient T cells through miR-449a-5p-dependent modulation of Notch1 and ameliorate the development and severity of arthritis, which may provide a potential cell-free strategy based on manipulating iTreg-EVs to prevent autoimmune arthritis.

Introduction

CD4+CD25+FoxP3+ regulatory T cells fall into distinct subpopulations according to origin: thymus-derived Treg (tTreg) cells and peripherally derived Treg (pTreg) cells; FoxP3+ Treg cells generated in vitro from conventional T cells are called in vitro-induced Treg (iTreg) cells, whereas those differentiated in vivo from conventional T cells in the periphery are called pTreg cells [1]. The Treg-cell program and subsequent suppressive function in the periphery are imparted by the transcription factor FoxP3 [2]. Continued expression of FoxP3 is associated with control of immune responses and maintenance of immunological tolerance. Treg cells primarily regulate immune responses via immunosuppressive cytokine-dependent and antigen-presenting cell (APC)-dependent mechanisms, enabling suppression of bystander cells [3–5].

Extracellular vesicles (EVs) are generally divided into exosomes, microvesicles, and apoptotic bodies based on their size, origin, biogenesis, and cargo [6–8]. Exosomes are small secreted membrane vesicles (~40–160 nm in diameter) that are formed through inward budding by endosomal membranes [9, 10]. As a consensus on specific markers for EV subtypes has not yet emerged, it is difficult to distinguish exosomes from microvesicles; therefore, exosomes and microvesicles are referred to as small EVs, following the classic references [11–13]. EVs participate in important biological functions and are involved in numerous physiological processes. Immune and nonimmune cell-derived EVs can mediate immune homeostasis and drive inflammation, autoimmunity, and infectious disease pathology. Multiple kinds of cells, such as CD4+ and CD8+ T cells, B cells and dendritic cells (DCs), produce EVs. EVs from these cells have been shown to mediate immune stimulation or immune modulation [14–19]. Recently, EVs released by murine Treg cells following T-cell receptor (TCR) activation were identified, first by Smyth et al. [20] and later by Okoye et al. [21]. EV production by murine Treg cells appears to be quantitatively greater than that by other T cells, including naïve CD4+ and CD8+ T cells, Th1 cells, and Th17 cells, and is regulated by changes in intracellular calcium, hypoxia, and sphingolipid ceramide synthesis, as well as the presence of IL-2 [21]. EVs contribute significantly to the function of murine CD4+ CD25+ FoxP3+ Treg cells. Rab27-DKO Treg cells that fail to release EVs are compromised in regard to disease prevention, resulting in colon shortening, weight loss, and increased IFN-γ expression [22].

Regulation of the immune system is vital for the prevention of many autoimmune diseases, such as rheumatoid arthritis (RA). Numerous reports show that FoxP3+ T cells are abundant in inflamed joints [23]. The presence of abundant inflammatory cytokines in inflamed joints might cause pathogenic effector T cells to compromise Treg cell-mediated suppression. Although synovial Treg cells are maintained, they are not resistant to IL-6- and TNF-α-induced inflammation [24]. The aim of treatments for autoimmune diseases is not complete elimination of pathogenic autoreactive T cells but alteration of the T-cell imbalance by numerical expansion of antigen-specific nTreg or iTreg cells, together with a numerical reduction in effector T cells and attenuation of their effector cell activity. Importantly, we first reported that TGF-β has the ability to induce CD4+ CD25− cells to become CD4+ CD25+ Treg cells in vitro [25]. The similarities and disparities of nTreg and iTreg cells have also been discussed [26, 27]. While the proinflammatory cytokine IL-6 can convert nTreg cells into IL-17-producing cells, iTreg cells are resistant to the effects of this compartment [28–31]. The more stable properties of iTreg cells during expansion for sufficient EV preparation offer an advantage for the clinical application of iTreg cells as an immunotherapy. Thus, the therapeutic strategies that tip the balance toward maintaining and augmenting immunosuppressive activity raise the question of whether iTreg-EVs maintain suppressive function and prevent immune imbalance in RA. Our study demonstrated that iTreg-EVs displayed immunomodulatory properties in vitro and in vivo, which were attributed to miR-449a-5p production and this miRNA targeting the Notch1 signaling pathway. Our data highlight an additional and novel mechanism by which iTreg-EVs mediate immunoregulation and provide an innovative cell-free approach for preventing RA and other autoimmune diseases.

Materials and methods

Mice

DBA/1J mice and C57BL/6J mice were purchased from Charles River Laboratories (Beijing, China) and The Jackson Laboratory (Bar Harbor, ME). The animal study was performed at the animal facilities of Sun Yat-sen University and approved by the Institutional Animal Care and Use Committees (Approval Number: 160520). All institutional and national guidelines for the care and use of laboratory animals were followed, and mice aged 6–12 weeks were used in the experiments.

TGF-induced Treg-cell preparation and EV isolation and identification

Naïve CD4+ CD62L+ T cells were purified from the spleens of C57BL/6J mice using magnetic isolation (Miltenyi Biotec). In general, 15 mice were used for the isolation of naïve CD4+ CD62L+ T cells (naive CD4). Then, 50 million naïve CD4 were cultured in culture medium containing EV-free 10% FBS at a concentration of 2 million cells/well in a 48-well plate and stimulated with recombinant IL-2 (2 ng/mL, R&D) and TGF-β (2 ng/mL, R&D) in the presence of 1:5 mouse T-Activator CD3/CD28 Dynabeads (Life Technologies, Thermo Fisher Scientific, USA). Finally, the collected volume of 7.5-mL cultures was used for EV preparation. Naive CD4+ CD62L+ T cells under the same conditions but without TGF-β stimulation served as the control. EV-free FBS was prepared by centrifugation at 300 × g for 10 min, 3000 × g for 10 min, 10,000 × g for 30 min, and 110,000 × g for 70 min, followed by filtration using a 0.22-μm filter [32, 33]. After 3 days, differentiated cells were evaluated by flow cytometry. The culture supernatant was collected and prepared for EV isolation with Umibio^® EV isolation kits (Umibio Biotechnology, Cat. No: UR52121, China) according to previous reports or the manufacturer’s instructions [34, 35]. In brief, an initial spin was performed with a sequential centrifugation procedure at 3000 × g for 10 min to remove dead cells or pellets and at 10,000 × g for 60 min to remove cell debris. Then, corresponding amounts of EV concentration solution were added according to the manufacturer’s instructions. The mixtures were vortexed and incubated for up to 2 h and then centrifuged at 10,000 × g for 60 min to precipitate the EV pellets. Finally, the pellets were resuspended in 1× PBS and purified with an Umibio EV Purification Filter (EPF column) at 3000 × g for 10 min. The EVs were stored at −80 °C immediately after isolation until further analysis. All centrifugation steps were carried out at 4 °C. In this study, one separate experiment was used with one separate set of EVs.

A transmission electron microscope (TEM) was utilized to observe the morphology of isolated EVs. The particle size distribution of EVs was analyzed using a NanoSight NS300 (Malvern, UK).

Purification of nTreg and iTreg cells

In this experiment, CD4+ FoxP3+ cells sorted from FoxP3^GFP transgenic C57BL/6J mice were cultured in EV-free culture medium at a concentration of 2 million cells/well in a 48-well plate in the presence of TGF-β, IL-2, and 1:5 mouse T-Activator CD3/CD28 Dynabeads for 2 days, and then the supernatant was collected for nTreg-EV isolation. For iTreg cell production, naive CD4+ CD62L+ T cells isolated from FoxP3^GFP transgenic C57BL/6J mice were cultured in T-cell complete medium at a concentration of 2 million cells/well in a 48-well plate and stimulated with IL-2 (2 ng/mL) and TGF-β (2 ng/mL) in the presence of 1:5 mouse T-Activator CD3/CD28 Dynabeads for 3 days to produce iTreg cells. Then, CD4+ FoxP3+ cells were purified from the iTreg cells by fluorescence-activated cell sorting, cultured in EV-free culture medium at a concentration of 2 million cells/well in a 48-well plate and stimulated with TGF-β, IL-2, and 1:5 mouse T-Activator CD3/CD28 Dynabeads for an additional 2 days. Then, the supernatant was collected for iTreg-EV isolation. Thus, with these approaches, nTreg-EVs and iTreg-EVs were successfully acquired for further analysis.

Fluorescent labeling of EVs

For uptake studies, purified EVs were labeled with a PKH67 (green) kit (Thermo Fisher Scientific) according to previously reported protocols [36]. Briefly, EVs diluted in 100 μL PBS were mixed with 0.4 ml Diluent C. In parallel, PKH67 dye was added to a 0.5-ml mixture solution and incubated for 4 min at room temperature. After staining was complete, 0.5 mL 0.5% bovine serum albumin/PBS was added to bind any excess PKH67 dye. The labeled EVs were washed at 110,000 g for 1 h, and the PKH67-labeled EV pellet was resuspended in PBS and used for uptake experiments. In uptake experiments, 20 μg PKH67-labeled EVs were applied to cultured splenic T cells for 24 h. Then, the cells were fixed and stained with CM-DiI (Thermo Fisher Scientific, red fluorescence) and DAPI (Thermo Fisher Scientific, blue fluorescence). Imaging was performed with a confocal microscope (Zeiss LSM800, Germany).

For in vivo tracking of EVs, purified EVs were labeled with a lipophilic dye 1,1-dioctadecyl-3,3,3,3-tetramethylindotricarbocyanine iodide (DiR, Red) kit (Thermo Fisher Scientific) according to the kit protocols. Briefly, EVs were incubated in a DiR solution (5 µM) for 15 min at room temperature in the dark and then washed with PBS at 110,000 g for 1 h three times. The final EVs were resuspended in PBS and stored at −80 °C for an in vivo optical imaging (OI) assay.

Apoptosis assessment

The influence of iTreg-EVs on cell apoptosis was quantified using the Annexin-V Alexa Fluor^® 647 Apoptosis Detection Kit (EZbioscience, USA). Following a 24 h incubation with iTreg-EVs under stimulation with soluble anti-CD3 (1 μg/mL) and soluble anti-CD28 (1 μg/mL), splenic T cells were collected, washed twice with PBS and then labeled with propidium iodide (PI) and Annexin-V Alexa Fluor^® 647. Cell apoptosis was measured using a flow cytometer (BD Biosciences, San Jose, California, USA). All experiments were performed three times. Data from one of three independent experiments are shown.

In vitro suppression assay

Mouse splenic T cells isolated from C57BL/6J mice using nylon wool were labeled with carboxyfluorescein diacetate succinimidyl ester (CFSE; 1 μM, Thermo Fisher Scientific). Then, the CFSE-labeled T cells were stimulated with an anti-CD3 antibody (0.025 μg/mL) in the presence of mitomycin C-treated antigen-presenting cells (APCs, 1:1). The cells were cocultured with 20 μg/mL EVs for 3 days, and CFSE dilution by CD8+ and CD4+ T cells was tested by flow cytometry (BD LSRFortessa™, BD Biosciences, San Jose, California, USA).

For detection of inflammatory and anti-inflammatory phenotypes, splenic T cells isolated from C57BL/6J mice were cultured in T-cell complete medium at a concentration of 2 million cells/well in a 48-well plate and stimulated with soluble anti-CD3 (1 μg/mL) and anti-CD28 (1 μg/mL) Abs in the presence of recombinant IL-12 and IL-6. After 3 days, the cells were collected to detect the relative expression of IL-1A, TNF-α, IFN-γ, IL-17A, IL-6, Stat1, Stat3, IL-21, IL-23, Stat5, FoxP3, and CTLA-4 by qPCR.

Mouse naive CD4+ T-cell differentiation in vitro

Naïve CD4+ CD62L+ T cells were purified from the spleens of C57BL/6J mice using magnetic isolation (Miltenyi Biotec). The naive CD4+ CD62L+ T cells were cultured in T-cell complete medium at a concentration of 0.3 million cells/well in a 96-well plate under polarization conditions for Th1 (IL-12, 10 ng/mL; anti-IL-4, 5 μg/mL), Th17 (IL-6, 20 ng/mL; TGF-β, 2 ng/mL; anti-IFN-γ, 5 μg/mL; anti-IL-12, 5 μg/mL; anti-IL-4, 5 μg/mL) or Treg (TGF-β, 2 ng/mL; IL-2, 30–50 U/mL) cells in the presence of mitomycin C-treated APCs (1:1) and soluble anti-CD3 (1 μg/mL) and anti-CD28 (1 μg/mL) antibodies (Abs) for 3 days. The levels of CD4+ IFN-γ+, CD4+ IL-17A+, and CD4+ FoxP3+ cells were detected by flow cytometry. The recombinant cytokines IL-12, IL-6, IL-2, and TGF-β were purchased from R&D; anti-IFN-γ, anti-IL-12 and anti-IL-4 were purchased from BioLegend.

Establishment of an animal model of collagen-induced rheumatoid arthritis (CIA)

DBA-1J mice were immunized via intradermal injection at the base of the tail with 100 μL type-II bovine collagen (Chondrex, Redmond, WA, USA) mixed with 100 μL complete Freund’s adjuvant supplemented with Mycobacterium tuberculosis (3 mg/mL, Biolead, Beijing, China). Synovial hyperplasia, irreversible destruction of the articular cartilage, chondrocyte death, cartilage degradation, and fibrosis-like structures were observed after 3–4 weeks. Finally, the presence of a chronic inflammatory response indicated that CIA was successfully induced, and the arthritis incidence of CIA mice ranged from 80 to 100%. In this study, we used the CIA animal model to examine the effects of C57BL/6J mouse-derived iTreg-EVs on allogeneic cell activity, cytokine production and bone erosion. Then, one mouse received EVs in 100 μL PBS at a dose of 1 μg/μL via intravenous injection on Days 0, 15, and 30 (n = 5 mice). Each paw was evaluated and scored individually for arthritis severity as described previously [37–39]. The sum of the scores for the four limbs was used to determine arthritis incidence. The clinical scores for arthritis features were evaluated, the features were scored individually using a 0–4 scoring system, and the paw scores were summed to yield an individual mouse score, with a maximum score of 16. The thickness of paw swelling was measured and scored by examiners blinded to the group conditions using a 0 to 4 scoring system as well as the representative gross appearance of arthritic joints. Arthritis incidence, clinical scores, and paw thickness were evaluated every 2–3 days by investigators who were blinded to the experimental conditions.

On the 60th day, the mice were sacrificed by CO2 asphyxiation and cervical dislocation. The hind limbs containing the knee joint and toe joint were fixed in 10% formalin and dehydrated with different concentrations of ethanol. Tissues were cut into 4- to 7-μm slices, placed in a 65 °C constant-temperature oven for 30 min, and soaked in xylene I for 15 min and then in xylene II for 15 min. The slices were soaked in 100% alcohol, 95% alcohol, 85% alcohol, and 75% alcohol for 5 min each and washed with running water for 10 min. The slices were stained with hematoxylin and eosin (H&E). All slides were evaluated by investigators who were blinded to the experimental conditions. The extent of synovitis and pannus formation was determined using a graded scale as follows: 0 = no signs of inflammation, 1 = mild inflammation with hyperplasia of the synovial lining without cartilage destruction, and 2–4 = increasing degrees of inflammatory cell infiltration and cartilage/bone destruction. Paws were subjected to micro-CT analyses as described previously [40]. Micro-CT imaging was performed with a Siemens Inveon CT Scanner and Inveon Acquisition workplace software. The dataset was loaded into Amira and viewed using the Voltex display and the VolrenRed Psewdo-color Scale. For micro-CT scoring, three volumes of interest were set with ±1-mm length in the distal and proximal directions from the center of each metatarsophalangeal joint. The bone volumes of the three metatarsophalangeal joints were calculated.

In vivo optical imaging (OI)

Mice were anesthetized with 2.5% isoflurane (Merial, Lyon, France), and images were acquired in the prone and supine positions. Fluorescence imaging was performed using an IVIS 200 small-animal imaging system (PerkinElmer, Waltham, MA, USA) using an Ex filter at 700 nm and an Em filter at 780 nm. Background fluorescence was measured and subtracted by setting up a background measurement (Ex filter, 530 nm). The fluorescence Em was normalized to photons per second per centimeter squared per steradian (p/sec/cm2/sr). Color images represent the spatial distribution of fluorescence within an animal overlaid on black and white photographs of the mouse, collected at the same time. Images were acquired and analyzed using Living Image 4.0 software (PerkinElmer). The relative mean fluorescence intensity of each designation of regions of interest (ROIs) was obtained by subtracting the mean fluorescence intensity of the corresponding ROI of the control mouse from the measured mean fluorescence intensity. Data are expressed as the average radiance ± SD. After these images were acquired, the mice were sacrificed, and dissected tissues (the joint, kidneys, lymph nodes, spleen, liver, lungs and heart) were imaged immediately as described above.

Osteoclastogenesis

Mouse CD11b+ cells were isolated from the bone marrow with a biotinylated anti-mouse CD11b antibody (BioLegend) and anti-biotin MicroBeads (Miltenyi Biotec) using an AutoMACS platform; purity was confirmed to be >95%. Purified cells were cultured with minimum essential medium (MEM) (containing EV-free 10% FBS) in the presence of macrophage colony-stimulating factor (M-CSF, 50 ng/mL) for 3 days, followed by stimulation with mouse RANKL (50 ng/mL) (R&D Systems) and M-CSF (50 ng/mL) (R&D Systems) for an additional 6 days to induce osteoclast formation. To evaluate the level of osteoclast formation, cells were stained with a TRAP kit (Sigma–Aldrich, 387 A) according to the manufacturer’s instructions, and TRAP+ positive cells were imaged and counted by microscopy.

Small RNA sequencing and bioinformatic analysis

Total RNA was isolated from Med- and iTreg-EVs via TRIzol (Invitrogen, Carlsbad, CA, USA). RNA concentration and integrity were measured with an Agilent 2100 Bioanalyzer (Applied Biosystems, Carlsbad, CA). Small RNA libraries were prepared using the Illumina TruSeq RNA Sample Preparation Kit (Illumina, San Diego, CA, USA) following the manufacturer’s recommendations. The libraries were quantified using a qPCR-based KAPA Biosystems Library Quantification kit (Kapa Biosystems, Inc., Woburn, MA). All six libraries were diluted as required and sequenced on an Illumina MiSeq or Illumina NextSeq 500 sequencing platform using a MiSeq reagent kit v3 or NextSeq 500/550 High Output kit v2 (51 cycles using a 75-cycle kit) at Norgen Biotek Corp. (ON, Canada). Raw reads were quality-controlled using Fast-QC. To identify known small RNAs, clean reads were mapped to the miRBase database (http://www.mirbase.org/) using BWA. The RNA-seq data were deposited in the NCBI Gene Expression Omnibus (GEO) database under the accession code SRP271989. p-Adj < 0.05 and log^{fold change} ≥ 1 were used as the threshold values for significance to define upregulated or downregulated miRNAs. Heat maps representing differentially regulated genes were generated using Cluster 3.0 software. The DIANA-microT-CDS (http://diana.imis.athena-innovation.gr), TargetScan (http://www.targetscan.org/), and miRanda (www.microrna.org) platforms were used to predict miRNA target genes of miRNAs. The differentially expressed miRNAs had putative target genes according to all three programs. We used the Kyoto Encyclopedia of Genes and Genomes (KEGG) database to analyze the combinatorial effect of the target genes. The results consisted of selected KEGG pathways. The P value threshold was set at 0.05. Network interactome maps for target genes of miR-449a-5p were constructed with DIANA-mirPath3.0 (http://snf-515788.vm.okeanos.grnet.gr/).

Transfection of a miR-449a-5p mimic and inhibitor

Some authors previously proposed a transfection-based method to modify EVs [41, 42]. Cells were transiently transfected with a miRNA negative control (NC), miR-449a-5p inhibitor or miR-449a-5p mimic using Lipofectamine 3000 (Thermo, MA, USA) according to the manufacturer’s instructions. The transfection concentration of the miRNA constructs was 30 nM. The NC, miR-449a-5p inhibitor and miR-449a-5p mimic were purchased from GenePharma (Shanghai, China).

Dual-luciferase reporter assay

A biological database (http://www.targetscan.org) was used to predict the putative sites for binding between miR-449a-5p and Notch1, which were then verified by Renilla luciferase and firefly luciferase dual-luciferase reporter assays. Wild-type Notch1 (Notch1-WT) and Notch1-(MUT) plasmids were constructed, and the two reporter plasmids were cotransfected into human embryonic kidney (HEK) 293T cells along with a miR-449a-5p mimic or negative control (NC) plasmid. Following 48 h of transfection, the cells were lysed, and the collected supernatant was used to detect luciferase activity using a Dual-Luciferase^® Reporter Assay System (Promega, Madison, WI). Luciferase activities were normalized to the activity of Renilla luciferase.

Flow cytometry and intracellular staining

Surface staining was performed with appropriate fluorophore-labeled antibodies specific for CD4 or CD8. For FoxP3 staining, cells were fixed and permeabilized using a FoxP3 staining buffer set (eBioscience, San Diego, CA) according to the manufacturer’s protocol. For intracellular IFN-γ and IL-17A staining, cells were stimulated with phorbol 12-myristate 13-acetate (PMA, 50 ng/mL) and ionomycin (500 ng/mL) in the presence of brefeldin A for 5 h. Then, the collected cells were fixed and permeabilized using an intracellular staining kit (BioLegend, California, USA) according to the manufacturer’s protocol. Flow cytometry data were acquired with a BD LSRFortessa™ (BD Biosciences, San Jose, California, USA). FlowJo 10 software (TreeStar, Ashland, OR) was used to analyze the flow cytometry data.

Quantitative RT–PCR

Total RNA was extracted from tissues and cells using a quick RNA extraction kit (EZbioscience, USA) according to the manufacturer’s instructions. Reverse transcription of the extracted RNA was carried out according to the protocols of the Mir-X miRNA First-Strand Synthesis Kit (TaKaRa, Tokyo, Japan) and cDNA Reverse Transcriptase 5× RT Master Mix (TaKaRa, Tokyo, Japan). Diluted cDNA (50 ng/µL) was subjected to real-time PCR in accordance with the specifications of the SYBR^® Premix Ex Taq TM II (Tli RNaseH Plus) Kit (Takara, Tokyo, Japan). PCR was performed on a Quantstudio 5 (Thermo Fisher, MA, USA). Fold changes were calculated using relative quantification (the 2^−ΔΔCt method). β-actin is a major protein component in striated muscle fibers and is also a major component of muscle filaments and cell microfilaments. β-actin is widely distributed and was used as the normalization control. U6 is a class of small nuclear RNAs (snRNAs) that are transcribed by RNA polymerase III. U6 mainly forms small nuclear ribonucleoprotein particles and is stably expressed and localized in the nucleus. U6 is widely used as the normalization control in miRNA qPCR analysis. Primers were synthesized by Sangon Biotech (Shanghai, China) (Table 1).

Table 1.

Primer for Real-time RT-qPCR

	Primer sequence		Product
Target	Sense	Antisense	Size(bp)
β-actin	AGGGAAATCGTGCGTGAC	CAAGAAGGAAGGCTGGAAA	122
Notch1	CAACTGCCAGAACCTTGTGC	AGAGTGACGTCAATGCCTCG	180
IL-1A	CGCTTGAGTCGGCAAAGAAAT	TGGCAGAACTGTAGTCTTCGT	98
TNF-α	AGTCTGTATCCTTCTAAC	TTCTGAGTAGTTGTTGAA	107
IFN-γ	AGCTCTTCCTCATGGCTGTT	TTTGCCAGTTCCTCCAGATA	149
Stat1	GATCGCTTGCCCAACTCTTG	ACTGTGACATCCTTGGGCTG	198
IL17A	TCAAAGCTCAGCGTGTCCAA	CGTGGAACGGTTGAGGTAGT	120
IL-6	CTTCAGAGAGATACAGAAACTCTAAT	GCTTATCTGTTAGGAGAGCAT	88
Stat3	CAATACCATTGACCTGCCGAT	GAGCGACTCAAACTGCCC	97
IL-21	ATCTCAGCAGTGACTCCCCT	TGCTGTGTCCCAGACCTACT	191
IL-23	TGAGCCCTTAGTGCCAACAG	CTTGCCCTTCACGCAAAACA	131
Stat5	GGCCTTGAGGGGAACTCTTC	CGGTGGAGGCTGTTACTTCTA	153
Foxp3	TGACAGACACCATCCTAAT	AGTTCACGAATGTACCAAG	142
CTLA4	ACGCAGATTTATGTCATTGATCCAG	AACCCCAAGCTAACTGCGAC	83
U6	CGCTTCGGCAGCACATATAC	TTCACGAATTTGCGTGTCAT
mmu-miR-449a-5p	GCAGTGGCAGTGTATTGTTAG	3′ universe

Western blot analysis

Cells, tissue samples or EV samples were lysed in RIPA buffer (Beyotime Biotechnology, Jiangsu, China) supplemented with protease inhibitors (Roche, Basel, Switzerland), and the total protein concentration was determined using a BCA protein assay kit (TIANGEN, Beijing, China). Equal amounts of protein (30 μg) were separated by 10% SDS–PAGE and transferred to polyvinylidene difluoride membranes (Millipore, MA), which were blocked with 5% BSA in TBS with Tween 20 for 1 h at room temperature. Then, the membranes were incubated with primary Abs overnight at 4 °C, followed by incubation with corresponding HRP-conjugated secondary Abs for 1 h at room temperature. Abs against CD63, CD81, CD9, TSG101, Notch1, NICD and β-actin were purchased from Abcam (Cambridge, UK). Bands were visualized using the SuperSignalTM West Chemiluminescent Substrate Kit (Thermo Fisher, MA, USA) following the manufacturer’s instructions. To test for equal loading, blotted membranes were stripped and reprobed with primary and secondary antibodies. Autoradiographic films were scanned and quantitated using Quantity One software (Bio–Rad, Hercules, CA).

Enzyme-linked immunosorbent assay

For detection of the secretion of the cytokines TNF-α, IFN-γ, IL-17A, and IL-10, splenic T cells isolated from C57BL/6J mice were cultured in T-cell complete medium at a concentration of 2 million cells/well in a 48-well plate and stimulated with soluble anti-CD3 (1 μg/mL) and anti-CD28 (1 μg/mL) Abs in the presence of recombinant IL-12 and IL-6. After 3 days of culturing, PMA (50 ng/mL) and ionomycin (500 ng/mL) were added for an additional 5 h. Finally, the culture medium was collected and evaluated by ELISA (ELISA kit, Bioo Scientific, USA) according to the manufacturer’s instructions. The relevant cytokines, including TNF-α, IFN-γ, IL-17A, IL-6, and IL-10, and IgG, IgG1, IgG2a, and IgG2b antibodies in mouse serum were analyzed using ELISA (ELISA kit, Bioo Scientific, USA). Samples were detected in triplicate relative to standards supplied by the manufacturer and analyzed for significant differences among different groups.

Bioinformatic analysis of differential miRNAs between nTreg-EVs and iTreg-EVs

Total RNA acquired from nTreg-EVs or iTreg-EVs was used for small RNA sequencing as previously described. The RNA-seq data have been deposited in GEO under the accession code SRP272038. We identified significant differentially expressed miRNAs through volcano plot filtering and calculated a P value using Student’s t test. MiRNAs were selected using the criteria of p < 0.05 and fold change ≥ 1.5, with nTreg-EVs as the control. Heat maps representing differentially regulated genes were generated using Cluster 3.0 software. The target genes of selected microRNAs were predicted using the DIANA-microT-CDS, TargetScan, and miRanda platforms. The gene set described above was obtained by biological pathway enrichment analysis based on KEGG and gene ontology (GO) analyses.

Statistical analysis

Results are expressed as the mean ± SD and were statistically analyzed using one-factor analysis of variance (ANOVA). When ANOVA showed significant differences, pairwise comparisons between groups were performed with Dunnett’s test. Student’s t test was used to assess statistical significance between two groups, and nonparametric tests were used to assess statistical significance among multiple groups. The chi-square test was used to compare arthritis incidence and clinical scores between groups. All statistical analyses were performed with GraphPad Prism Software (version 8.00), and p values below 0.05, 0.01, or 0.001 were considered statistically significant.

Results

Isolation, purification, and functional characterization of iTreg-EVs

Previous research has indicated that strong and sustained TCR stimulation can activate the mTOR/Akt signaling pathway, which facilitates T effector (Teff)-cell production and suppresses Treg-cell generation [43]. We previously found that suboptimal TCR activation is most conducive to FoxP3 expression and established a unique protocol for the production of induced Treg cells [44]. As shown in Fig. 1a, CD4+ CD25+ FoxP3+ Treg cells, called iTreg cells, were successfully induced in vitro, with non-TGF-β-stimulated cells as the control cells. EVs were acquired from the culture medium according to the EV isolation kit procedure described exhaustively in the Methods. Accordingly, we identified EV phenotypic characteristics. As expected, analysis with a TEM presented a typical morphology of EVs, with a complete enveloped structure containing low-density substances (Fig. 1b). To examine the size and distribution of EVs, we used a nanoparticle tracking analyzer and observed that most particles ranged from 40 to 160 nm, while iTreg-EVs had a significantly different size distribution, which was consistent with the EV sources of the heterogeneous iTreg population used for EV isolation (Fig. 1c). For further verification, we performed Western blot analysis to evaluate the special protein markers of EVs. Our results demonstrated that CD9, CD63, CD81, and TSG101 were expressed in these EVs (Fig. 1d). Thus, we successfully acquired EVs derived from in vitro TGF-β-induced Treg cells (iTreg-EVs) and EVs derived from non-TGF-β-stimulated cells (Med-EVs).

Fig. 1.

Isolation, identification, and functional characterization of iTreg-EVs. a Identification of in vitro TGF-β-induced Treg cells by flow cytometry. b Transmission electron microscopy analysis of iTreg-EVs. The image shows small vesicles with a size of ~100 nm in diameter. Scale bar, 200 nm. c Nanoparticle tracking analysis of the size distribution of particles. d The protein markers CD9, CD63, CD81, and Alix were detected by Western blot analysis. e In vitro iTreg-EV uptake by splenic T cells. f, g Staining for apoptotic markers in splenic T cells. Splenic T cells were incubated with iTreg-EVs under stimulation with soluble anti-CD3 (1 μg/mL) and soluble anti-CD28 (1 μg/mL), and then the splenic T cells were collected and stained with Annexin-V and PI. The percentage of apoptotic cells (Annexin-V+PI+) was detected by flow cytometry. h In vitro suppression assay evaluating CD3+ T-cell proliferation. I, j In vitro suppression assay evaluating Th1 and Th17 cell differentiation. k In vitro Treg-cell differentiation assay. *p < 0.05; **p < 0.01. Data are shown as the mean ± SD; n = 4, from one of three independent experiments

Having established successful isolation of EVs from cell cultures, the possibility that these particles modulated T-cell responses in vitro was investigated. We initially used splenic T cells isolated from C57BL/6J mice and cocultured them with PKH67-labeled iTreg-EVs (green). As shown in Fig. 1e, iTreg-EVs were present in the cytoplasm of cultured T cells, which indicated EV uptake by the T cells. Before examining the in vitro immunoregulatory effects of iTreg-EVs on T cells, we assessed the toxicity of iTreg-EVs to splenic T cells in vitro. We observed no overt proapoptotic effects induced by iTreg-EV treatment at various concentrations ranging from 10 to 500 μg/mL (Fig. 1f) or at various time points (Fig. 1g). Collectively, these results demonstrated that iTreg-EVs could be taken up by T cells and had no toxic effect on T cells, suggesting that they are suitable for T cell-targeted treatment. Moreover, we utilized an in vitro suppression assay to test whether iTreg-EVs have the capacity to suppress T-cell proliferation, with Med-EVs used as the control. We found that the proliferation of CD3+ T cells was reduced by iTreg-EVs (Fig. 1h). To assess whether iTreg-EVs control the differentiation of CD4+ T-cell subsets, such as Th1 and Th17 cells, mouse naive CD4+ T cells were isolated and cultured under Th1- or Th17-polarizing conditions. As shown in Fig. 1i, j, the differentiation of Th1 or Th17 cells was markedly suppressed by iTreg-EVs.

FoxP3 is a critical transcription factor in the differentiation of Treg cells, and effects on its expression were examined. Mouse naive CD4+ T cells were cultured for 3 days under Treg differentiation conditions in the presence of iTreg-EVs to examine whether iTreg-EVs modulate FoxP3 expression. Curiously, iTreg-EVs showed higher FoxP3 expression than control EVs at 3 days. For up to 10 days under this induction condition, iTreg-EVs controlled FoxP3 expression, whereas control EVs did not change FoxP3 expression (Fig. S1a). We further tested whether iTreg-EVs had the same effect when naïve CD4+ cells were cultured under Treg differentiation conditions without TGF-β stimulation. As shown in Fig. S1b, iTreg-EVs promoted FoxP3 expression. These data indicated the roles of iTreg-EVs in maintaining and augmenting the FoxP3 phenotype. In summary, we demonstrated that iTreg-EVs suppressed immune responses by inhibiting inflammatory T-cell proliferation and differentiation while maintaining or augmenting the suppressive activity and function of Treg cells.

Functional characterization of nTreg-EVs and iTreg-EVs

It is now well accepted that Treg cells are heterogeneous. The optimal Treg subset required to induce immune tolerance remains unknown. To test the functional difference between nTreg-EVs and iTreg-EVs, we sorted CD4+ FoxP3^GFP+ cells from the spleen of C57BL/6J FoxP3^GFP transgenic mice (nTregs) and CD4+ FoxP3^GFP+ cells from in vitro-induced iTregs. Then, these cells were cultured in EV-free complete medium for 2 days in the presence of TGF-β, IL-2, and TCR stimulation to prepare individual EVs (Fig. 2a). Thus, in vitro suppression assays assessing T-cell proliferation were used to analyze the immunoregulatory function of nTreg-EVs and iTreg-EVs. We found that both nTreg-EVs and iTreg-EVs suppressed CD8+ cell proliferation in vitro but there was no significant difference between nTreg-EVs and iTreg-EVs (Fig. 2b). Similarly, the suppression of Th17 cell differentiation was also not significantly different between nTreg-EV and iTreg-EV treatments (Fig. 2c). Moreover, we detected the tolerance-related phenotypic changes in activated T cells, and the results showed that both nTreg-EVs and iTreg-EVs skewed T cells toward a tolerogenic phenotype (Fig. 2d). Collectively, these data indicated that nTreg-EVs and iTreg-EVs had similar suppressive functions in vitro.

Fig. 2.

Functional characteristics of nTreg-EVs and iTreg-EVs. a Cell sorting of nTreg and iTreg cells for EV preparation. b In vitro suppression assay evaluating CD8+ T-cell proliferation. c In vitro suppression assay evaluating Th17 cell differentiation. d Splenic T cells were stimulated with soluble anti-CD3 and anti-CD28 Abs in the presence of recombinant IL-12 and IL-6. After 3 days, the relative expression of inflammatory phenotypic markers was detected by qPCR. In vivo imaging of transferred DiR-labeled nTreg-EVs and iTreg-EVs (100 μg) in CIA mice 15 days after immunization. The fluorescence percentage of the joint was evaluated 24 h after injection (e) and 15 days after injection (f). g Mouse CD11b+ cells were isolated and stimulated under osteoclast-generating conditions. Representative images of osteoclast generation are presented; scale bar, 500 μm. The average TRAP-positive osteoclast numbers per area (×4 magnification under a microscope) under different conditions were quantified. h CIA mice were treated with 100 μg EVs at 0, 15, and 30 days after immunization, and the arthritis severity scores of CIA mice were determined at various time points. i Knee joint sections were stained with H&E on Day 60 (experiment termination), and severity scores were evaluated; scale bar, 500 μm. j Toe joint sections were imaged by micro-CT analysis on Day 60, and bone volumes of the metatarsophalangeal joints were evaluated. k, l Intracellular staining for IL-17A and FoxP3 in LN cells on the 60th day was detected by flow cytometry. m Serum was obtained from the blood of CIA mice on Day 60, and then the protein levels of anti-collagen II antibodies (IgG, IgG1, IgG2a, and IgG2b) were detected by ELISAs. In vitro data are shown as the mean ± SD; n = 4, from one of three independent experiments. For the in vivo assay, data are shown as the mean ± SD; n = 5 mice from one of three independent experiments. Representative in vivo tracking images were from three separate experiments; n = 5 mice

Our previous research suggested that iTreg cell infusion significantly ameliorated the severity of collagen-induced arthritis (CIA) [39]. As the effects of Treg-EVs on an arthritis animal model have not been reported, we initially utilized CIA mice to determine the preventive effects of iTreg-EVs on arthritis induction, disease progression, and inflammatory responses in this arthritis animal model; nTreg-EVs were used as the control. First, we utilized an in vivo small-animal imaging device to investigate EV distribution in CIA mice. The lipophilic membrane dye DiR was used to label EVs, and DiR-labeled nTreg-EVs and iTreg-EVs were intravenously transplanted into CIA mice for 24 h. We observed that there was no significant difference in the fluorescence distribution in the joints or other organs between nTreg-EV- and iTreg-EV-infused mice (Fig. 2e). Thus, we prolonged the circulation time to 15 days, but there was still no significant difference (Fig. 2f). It is well recognized that the activation of osteoclasts through antibodies, either alone or incorporated in immune complexes, ultimately leads to bone erosion. We isolated CD11b+ cells from the bone marrow to induce osteoclast formation in vitro and found that the osteoclast-inducing capacity of monocytes was not significantly different between the groups (Fig. 2g).

The CIA mouse model shares most pathological features of human RA, including synovial hyperplasia, joint swelling, and bone and cartilage destruction [37, 45]. Accordingly, significantly lower average clinical scores were observed in nTreg-EV- and iTreg-EV-infused mice (Fig. 2h). Moreover, we found that both nTreg-EVs and iTreg-EVs significantly reduced the degrees of inflammatory infiltration, synovial hyperplasia, pannus formation and bone destruction (Fig. 2i, j). It is universally acknowledged that predominant Th17 cell expansion and insufficient Treg function are implicated in arthritis disease progression, with both having central roles. We observed that both nTreg-EVs and iTreg-EVs significantly lowered the CD4+ IL-17A+ (Th17)-cell frequency and reversed the bias toward the Th17 cell subset to favor Treg cells in CIA mice (Fig. 2k, l). Additionally, the presence of autoantibodies is a critical factor in RA initiation. We found that both nTreg-EVs and iTreg-EVs significantly suppressed the levels of autoantibodies (IgG, IgG1, IgG2a, and IgG2b), but there was no significant difference between nTreg-EVs and iTreg-EVs (Fig. 2m). Therefore, the collective data suggest that nTreg-EVs and iTreg-EVs have similar functional characteristics related to immunomodulation.

miR-449a-5p is responsible for the suppressive function of iTreg-EVs

EVs function as cargo carriers to encapsulate parental cell-derived proteins, lipids, mRNAs, and regulatory miRNAs and deliver them to target cells to modulate their activities [46]. The role of Treg-EV miRNAs in modulating target cells and the associated mechanisms remain unknown. We initially utilized small RNA sequencing to determine the miRNA profile of iTreg-EVs. Hierarchical clustering analyses showed distinguishable miRNA expression profiles. We observed that 17 upregulated and 13 downregulated miRNAs were significantly differentially expressed in iTreg-EVs compared with Med-EVs (p-Adj < 0.05, log^{fold change} ≥ 1) (Fig. 3a, b). The miRNAs that were obviously enriched in iTreg-EVs included miR-5112, miR-7213-5p, miR-181d-3p, miR-99b-3p, miR-449a-5p, miR-92-5p, miR-676-3p, miR-24-1-5p, miR-1954, miR-449a-3p, miR-8117, miR-7001-3p, miR-6900-5p, miR-196a-1-3p, miR-let-7c-1-3p, miR-3473e, and miR-7235-3p. The KEGG database was used to analyze the significantly enriched signaling pathways of the target genes of the identified miRNAs, and the pathways included the Notch signaling pathway (Fig. 3c). Among the miRNAs that were highly enriched in iTreg-EVs, miR-449a-5p was previously reported to inhibit the expression of the inflammatory gene Notch1 [47, 48]. We also found that miR-449a-5p was significantly more highly expressed in iTreg-EVs than in Med-EVs (Fig. 3d).

Fig. 3.

MiR-449a-5p is highly expressed in iTreg-EVs, and knockdown of miR-449a-5p reverses iTreg-EV suppression of T-cell proliferation and differentiation in vitro. a Cluster heat map of miRNAs in iTreg-EVs. Med-EVs were used as the control. b Volcano plot showing differentially expressed miRNAs. MiRNAs that met the thresholds p-Adj < 0.05 and log^{fold change} ≥ 1 were considered statistically significant. c The signaling pathways identified as enriched by KEGG analysis. The x-axis represents -log10(p value), the y-axis represents the KEGG term, and p < 0.05 was considered statistically significant. d Identification of miR-449a-5p expression in iTreg-EVs using RT–qPCR. e In vitro suppression assay evaluating CD8+ and CD4+ T-cell proliferation. We knocked down miR-449a-5p expression in Treg-EVs (i-EV), and miRNA negative control-transfected iTreg-EVs (NC-EV) were used as the control. f In vitro suppression assay evaluating Th1 and Th17 cell differentiation. g In vitro Treg-cell differentiation assay. *p < 0.05; **p < 0.01. Data are shown as the mean ± SD; n = 4, from one of three independent experiments

We used in vitro T-cell proliferation and differentiation assays to investigate the possible modulatory mechanism of iTreg-EVs and assess whether miR-449a-5p was responsible for the suppressive function of iTreg-EVs. First, we knocked down miR-449a-5p expression in Treg-EVs (miR-449a-5p-i-EVs or i-EVs) by transfection of a devised miR-449a-5p inhibitor into iTreg cells, and miRNA negative control-transfected iTreg-derived EVs (NC-EVs) were used as the control (Fig. S2a–c). We found that the proliferation of CD8+ and CD4+ cells was reduced by NC-EVs, but i-EVs exhibited a poor suppressive capacity against CD4+ and CD8+ T-cell proliferation (Fig. 3e). To assess whether miR-449a-5p mediates the suppression of the differentiation of CD4+ T-cell subsets, such as Th1 and Th17 cells, mouse naive CD4+ T cells were isolated and cultured under Th1- or Th17-polarizing conditions. As shown in Fig. 3f, i-EVs did not suppress Th1 or Th17 cell differentiation. More importantly, i-EVs abolished the elevation in the CD4+ FoxP3+ cell percentage (Fig. 3g). To determine whether miR-449a-5p in iTreg-EVs is responsible for the functional maintenance of T cells toward a tolerogenic phenotype, we conducted a qPCR assay to analyze the mRNA levels of transcription factors for relevant cytokines. We found that i-EVs did not suppress IL-1A, TNF-α, IFN-γ, IL-17A, IL-6, Stat3, or IL-23 and failed to promote Stat5-, FoxP3- and CTLA-4-expressing phenotypes (Fig. S3a). In addition, proinflammatory and anti-inflammatory cytokine levels reflect the state of the immune response, so we observed the effects of iTreg-EVs on the secreted levels of TNF-α, IL-17A, IFN-γ, and IL-10. As shown in Fig. S3b, the production of TNF-α, IL-17A and IFN-γ was reduced, but the production of the anti-inflammatory cytokine IL-10 was enhanced by NC-EVs. Conversely, i-EVs restored the above effects. Collectively, these data suggest that miR-449a is required for the regulatory function of iTreg-EVs involved in modulating inflammatory T-cell responses.

Blockade of miR-449a-5p on iTreg-EVs restores histopathological improvement and inflammatory responses in arthritis mice

We utilized CIA mice to determine the joint-targeting property of iTreg-EVs in this arthritis animal model, with Med-EVs used as the control. DiR-labeled EVs were intravenously transplanted into CIA mice for 24 h. The fluorescence signal in iTreg-EV-transferred mice was obvious, but the fluorescence signal in control EV-transferred mice was hardly detectable (Fig. 4a). After images were acquired, the mice were sacrificed, and the joints, lymph nodes, spleen, liver, lungs, kidneys, and heart were dissected and subjected to imaging. We found that most fluorescence signals accumulated in the liver in both Med-EV- and iTreg-EV-infused mice, but a higher intensity of fluorescence signals in the joint was specifically detected in iTreg-EV-infused mice compared with Med-EV-infused mice (Fig. S4a, b). We found that i-EVs failed to delay arthritis onset and lost the abilities to reduce clinical scores and paw swelling (Fig. 4b–d). Moreover, i-EVs poorly protected against disease severity and bone erosion in the joint (Fig. 4e, f).

Fig. 4.

Blockade of miR-449a-5p in iTreg-EVs reverses the protective effects of iTreg-EVs on CIA mice. a The dynamic distribution of iTreg-EVs in CIA mice. In vivo imaging of transferred DiR-labeled EVs (100 μg) in CIA mice 15 days after immunization. The fluorescence percentage of the joint was evaluated 24 h after injection. CIA mice were treated with 100 μg EVs at 0, 15 and 30 days after immunization. The incidence of arthritis (b), arthritis severity scores (c) and paw thickness (d) of CIA mice were determined at various time points. e Knee joint sections were stained with H&E on Day 60 (experiment termination), and severity scores were evaluated; scale bar, 500 μm. f Toe joint sections were imaged by micro-CT analysis on Day 60, and bone volumes of the metatarsophalangeal joints were evaluated. g, h Intracellular staining for IL-17A and FoxP3 in LN cells on the 60th day was detected by flow cytometry, as indicated by the T-distributed random neighbor embedding (tSNE) plot. Serum was obtained from the blood of CIA mice on Day 60, and the protein levels of TNF-α, IFN-γ, IL-17A, IL-6, and IL-10 (i) and anti-collagen II antibodies (j) were detected by ELISAs. *p < 0.05; **p < 0.01; ***p < 0.001. Data are shown as the mean ± SD; n = 5 mice from one of three independent experiments. Representative in vivo tracking images were acquired during three separate experiments; n = 5 mice

The findings of the present study suggest that infiltrating effector cells and a dysfunctional immune microenvironment may ultimately contribute to joint destruction in CIA mice. We observed that NC-EV treatment significantly lowered the CD4+ IL-17A+ (Th17) cell frequency and reversed the bias toward Th17 cells to favor Treg cells in CIA mice (Fig. 4g, h). NC-EV treatment resulted in significantly lower levels of the signature Th17 cytokine IL-17A (Fig. 4i) and TNF-α, IFN-γ, and IL-6 (Fig. 4i), the main signature cytokines of effector T cells, and could aggravate the progression and severity of RA diseases, which had an essential role in all phases of RA pathophysiology in mediating arthritis damage. For the important regulatory cytokine IL-10, we found that NC-EVs induced an approximately twofold elevation in IL-10 (Fig. 4i). Additionally, the presence of autoantibodies is a critical factor in RA initiation. We also showed that NC-EVs significantly suppressed the levels of autoantibodies (IgG, IgG1, IgG2a, and IgG2b) (Fig. 4j). More importantly, we observed that blockade of miR-449a-5p on iTreg-EVs restored the above histopathological improvement and inflammatory responses in CIA mice (Fig. 4e–j). Overall, we present systematic data demonstrating that iTreg-EVs preferentially home to the inflamed joint and that iTreg-EVs depend on the transfer of miR-449a-5p to prevent disease progression and inflammatory responses in an arthritis animal model.

iTreg-EVs modulate the expression of Notch1, which is a direct target of miR-449a-5p

To determine whether Notch1 is a direct target of miR-449a-5p affected through binding to Notch1 mRNA transcripts and whether iTreg-EVs modulate Notch1 signaling through transfer of miR-449a-5p in the prevention of murine arthritis, a fragment of the 3′ UTR of Notch1 containing the normal sequence or a mutant sequence for the potential miR-449a-5p binding site was initially cloned into a vector with a firefly luciferase reporter gene (Fig. 5a). We found that luciferase activity was significantly reduced by a miR-449a-5p mimic, but this was not observed with the mutant reporter (Fig. 5b). Furthermore, we transiently transfected the miR-449a-5p mimic for 48 h and then detected the mRNA and protein expression of Notch1 or NICD in HEK-293T cells. The Notch1 intracellular segment, NICD, enters the nucleus to bind with the transcription factor CSL to form the NICD/CSL transcriptional activation complex. As expected, transfection of the miR-449a-5p mimic significantly inhibited Notch1 mRNA expression (Fig. 5c). Similarly, the protein levels of Notch1 and NICD were suppressed by miR-449a-5p (Fig. 5d). Next, we used a mouse splenic T-cell in vitro culture system to assess the modulatory function of iTreg-EVs in Notch1 signaling. As shown in Fig. 5e, NC-EVs significantly inhibited Notch1 mRNA expression in splenic T cells in vitro, but miR-449a-5p-i-EVs restored Notch1 mRNA expression and Notch1 and NICD protein expression (Fig. 5f). More importantly, we verified that the iTreg-EV-mediated transfer of miR-449a-5p protected against arthritis progression through modulation of Notch1. The protein levels of Notch1 and NICD in spleen tissues from CIA mice were investigated, and the results showed that relatively high protein expression levels of Notch1 and NICD were detected in miR-449a-5p-i-EV-treated mice (Fig. 5g). Taken together, the results demonstrate that iTreg-EVs protect against arthritis in mice through miR-449a-mediated suppression of Notch1 expression.

Fig. 5.

iTreg-EVs modulate Notch1 expression and are dependent on miR-449a. a Sequence alignment of miR-449a-5p and its putative target sites in the 3′-UTR of the Notch1 mRNA transcript. Mutations were generated in the complementary sites for the seed region of miR-449a-5p, as indicated. b Identification of the Notch1 region targeted by miR-449a-5p using luciferase activity analysis in HEK-293T cells. c The mRNA expression of Notch1 in miRNA-449a-5p mimic-transfected HEK-293T cells. d The protein expression of Notch1 and NICD in miRNA-449a-5p mimic-transfected HEK-293T cells. e The mRNA expression of Notch1 in i-EV-treated splenic T cells. f The protein expression of Notch1 and NICD in i-EV-treated splenic T cells. g The protein levels of Notch1 and NICD in spleen tissues from CIA mice on Day 60 were detected by Western blot analysis, and two typical spleen tissues for individual groups are presented. *p < 0.05; **p < 0.01. Data are shown as the mean ± SD; n = 4, from one of three independent experiments. For the in vivo assay, data are shown as the mean ± SD; n = 5 mice

EV miRNA profile differences between nTregs and iTregs

It is generally acknowledged that distinct gene modifications and phenotypes impart functional plasticity to Treg cells. Previous reports have indicated that miR-let-7d-containing nTreg-EVs mediate the suppression of Th1 cells to prevent systemic disease [21]. However, the elaborate miRNA-based mechanisms of Treg-EVs remain unknown. We investigated the individual miRNA profiles of nTreg and iTreg cells using small RNA sequencing. Hierarchical clustering analyses showed distinguishable miRNA expression patterns (Fig. 6a). Through hierarchical clustering analysis, we found that samples with similar miRNA expression patterns were clustered together, and the volcano map visualized the miRNA expression distribution. We observed that 147 upregulated and 179 downregulated miRNAs were significantly differentially expressed in iTreg-EVs compared with nTreg-EVs (p < 0.05, fold change ≥ 1.5) (Fig. 6b). Interestingly, we found that iTreg-EVs carried similar miR-let-7d levels and higher miR-155 levels than nTreg-EVs.

Fig. 6.

Bioinformatic analysis of the miRNA profiles of nTreg-EVs and iTreg-EVs. a Heat map showing differentially expressed miRNAs of nTreg-EVs and iTreg-EVs. b Volcano plot showing differentially expressed miRNAs between nTreg-EVs and iTreg-EVs. The red dots represent upregulated miRNAs, the green dots represent downregulated miRNAs, and the gray dots indicate no significant change. The x-axis represents log2(fold change), the y-axis represents −log10 (p value), and p < 0.05 was considered statistically significant. c The distribution map of gene ontology (GO) analysis results revealed the possible functions of miRNA-related genes. The y-axis is the enriched gene ratio, the x-axis represents the GO term, and p < 0.05 was considered statistically significant. d The classification map for KEGG pathways. The x-axis represents the overlapping gene count, and the y-axis represents the KEGG term. The column color represents enrichment significance, and p < 0.05 was considered statistically significant

Many factors contribute to Treg-cell heterogeneity, such as differentiation stimuli, transcriptional modulation and antigen priming [4]. However, nTreg cells exhibit high turnover and fine sensitivity to a range of signals in the inflammatory milieu, whereas iTreg cells do not. In this study, we proposed to examine whether EV-mediated cellular communication partially modifies the functional characteristics of nTreg and iTreg cells and whether these EVs are associated with nTreg and iTreg cell plasticity under inflammatory conditions. GO analysis of each sample was used to detect broad classes of protein functionality, namely, molecular functions, biological processes and cellular components. Based on the target genes predicted using the DIANA-microT-CDS, TargetScan, and miRanda platforms, the distribution map of the GO analysis results revealed the possible function of the target genes (Fig. 6c). Furthermore, based on the target genes predicted using the DIANA-microT-CDS, TargetScan, and miRanda platforms, KEGG pathway analysis revealed that the target genes were enriched in certain signaling pathways, including cytokine-cytokine receptor interaction, mTOR signaling pathway, Wnt signaling pathway, focal adhesion, prolactin signaling pathway, EGFR tyrosine kinase inhibitor resistance and IL-17 signaling pathway (Fig. 6d). In these investigations, we particularly focused on cytokine-cytokine receptor interactions, the mTOR signaling pathway, the IL-17 signaling pathway and the FoxO signaling pathway. These terms were characterized by high comprehensive database scores and exhibited a linear association with potential miRNAs (miR-342-5p, miR-297a-3p, miR-297b-3p, miR-297c-3p, and miR-466g). Finally, we proposed that iTreg-EV-mediated transfer of these miRNAs into recipient cells alters inflammatory signaling pathway-associated gene expression, which contributes to the weaker inflammatory challenge and more stable tolerance phenotypes of iTreg cells. The above analysis elucidated the possible mechanism of EV miRNA-mediated immune homeostasis and might provide an explanation for the functional heterogeneity between nTreg and iTreg cells.

Discussion

Treg cells have broad immunosuppressive roles, affecting processes ranging from the triggering of innate immune cells to adaptive cell-mediated responses. Treg cells are important for maintaining immunological self-tolerance and homeostasis and preventing a variety of autoimmune diseases, including rheumatic diseases, such as RA and systemic lupus erythematosus (SLE). Treg cells have been regarded as immunotherapeutic targets. Manipulation of the number and/or suppressive function of Treg cells has been shown to be impactful in the treatment and prevention of organ-specific and systemic autoimmune diseases, including RA and SLE, even if the primary causative abnormality is not in Treg cells. It is well recognized that EVs make a significant contribution to Treg-cell function, as inhibiting the release of EVs can reverse the suppressive capabilities of Treg cells. Thus, investigating the therapeutic efficacy of Treg-EVs and associated mechanism facilitates an understanding of the local immune regulation and regulatory mechanisms of Treg cells involved in immune tolerance and homeostasis.

Regulatory T-cell homeostasis is governed by a number of factors. Recently, it has been demonstrated that the immunomodulatory function of Treg-EVs is impaired in patients with relapsing-remitting multiple sclerosis (MS) [49]. MS patient-derived EVs cause decreased expression of insulin-like growth factor 1 receptor (IGF1R) and transforming growth factor β receptor 1, leading to inhibition of Treg-cell differentiation [50]. A distinct immune challenge might drive context‑specific Treg-cell function and restoration of immune homeostasis. Accordingly, diseases can occur when Treg-cell homeostasis is impaired. Yu et al. observed that adoptive transfer of autologous rat Treg-cell-EVs postponed allograft rejection and prolonged the survival of transplanted kidneys by inhibiting T-cell proliferation [51]. Notably, human Treg-derived EVs inhibit alloimmune-mediated skin tissue damage and immune cell infiltration in a humanized mouse skin transplant model, as reported by Tung et al. [52]. These observations suggest that Treg-cell-EVs are crucial to the homeostasis of Treg cells and could be considered an exciting new therapy for the induction of tolerance based on strategies that maintain homeostasis by enhancing immunosuppressive activity.

Previous research has shown that EVs derived from anergic rat T cells inhibit inflammatory cell activity following coculture with B cells and DCs in vitro [53]. What happens to Treg-EVs, which cells acquire them and whether this acquisition is receptor driven are poorly understood. Tung et al. proposed that Treg-cell-derived EVs inhibit immune reactions by modifying DCs to a tolerogenic phenotype [54]. Our in vitro uptake assay revealed the uptake of iTreg cell-EVs by T cells. Recently, nTreg cell-EVs were found to exert immunomodulatory function and reduce CD4+ T-cell proliferation and cytokine (IL-2 and IFN-γ) production in vitro [20]. In parallel, iTreg-EVs were found to exert suppressive effects on CD8+ and CD4+ T-cell proliferation. Treg cells have been shown to possess a broad spectrum of immunomodulatory capabilities, such as regulating the function of professional APCs and influencing the differentiation and associated cytokine secretion profile of T-cell subsets [3]. In this study, mouse naive CD4+ cells were cocultured with mitomycin C-treated APCs under Th1-, Th17- or Treg-polarizing conditions, and we found that iTreg-EVs impaired Th1 and Th17 cell differentiation while promoting Treg differentiation. Our results demonstrated that iTreg-EVs controlled immune homeostasis by affecting the function of recipient T cells.

To study autoimmune disease, an experimental model of RA (collagen-induced RA, CIA) was established in this study and used to assess the immunotherapeutic efficacy of iTreg-EVs. Our results first demonstrated that iTreg-EVs could delay arthritis incidence and ameliorate clinical scores. Histopathological images showed that iTreg-EVs attenuated synovial hyperplasia and irreversible destruction of the articular cartilage and prevented chondrocyte death, cartilage degradation, and fibrosis-like structure formation. The findings of our present study supported the concept that the Th17 cell-predominant state of immune responses initially contributes to bone erosion and cartilage destruction in arthritis. We also found that adoptively transferred iTreg-EVs controlled effector cell activation and differentiation and enhanced Treg-cell numbers. Additionally, the current data suggest that dysfunction in the microenvironment in joint lesions contributes to the insufficient numbers and suppressive function loss of Treg cells in this compartment. Moreover, the production of cytokines such as TNF-α, IFN-γ, IL-17A, and IL-6 was remarkably suppressed by iTreg-EVs in CIA mice. Overall, our collected data suggest that the therapeutic efficacy of iTreg-EVs is realized through reversing the balance to favor inhibiting inflammatory responses and maintaining immunosuppressive activity to prevent arthritis incidence.

EVs function as a cargo delivery system dependent on encapsulated proteins, lipids, mRNAs, and regulatory miRNAs, which regulate the activity of target cells through diverse mechanisms [46]. One recent study investigated whether the suppressive feature of nTreg cell-EVs could be attributed to the ectoenzyme CD73. Loss of CD73 on nTreg-EVs reversed their suppressive nature [20]. Additionally, although a high level of CD25 expression was observed on Treg-EVs, this molecule may not play a role in their suppressive function [20]. Recent evidence indicates that the transfer of miRNAs contained in Treg-cell-EVs has been shown to modify the function of recipient APCs by inhibiting the translation of target mRNA molecules [54, 55]. Likewise, the transfer of miRNAs, including Let-7d, miR-155, and Let-7b, to Teff cells through the acquisition of Treg-cell-EVs has been shown, suggesting that miRNAs may play a critical role in the suppressive capacity of Treg-EVs [21]. MiRNAs are a class of naturally occurring, small noncoding RNA molecules that target the 3′-untranslated region (3′-UTR) of mRNA, resulting in mRNA degradation or suppression [56]. Nonetheless, iTreg-EV miRNAs modulate target cells, and the associated mechanisms remain unknown. Thus, we investigated miRNA transfer and the miRNA-mediated modulation of recipient cells. In this study, we found for the first time that miR-449a-5p was highly enriched in iTreg-EVs and identified Notch1 as a direct target gene of miR-449a-5p. Previous studies have emphasized that Notch signaling is important in several stages of T-cell development and differentiation. Inhibition of Notch signaling reduces Th1 and Th17 responses [57–59]. Inactivation of Notch signaling reverses the Th17/Treg-cell imbalance in patients with immune thrombocytopenia [60]. We found that blockade of miR-449a-5p on Treg-EVs restored the effects of iTreg-EVs on effector cells and regulatory T cells in vitro and in vivo, suggesting that iTreg-EVs control inflammatory activation through miR-449a-5p-dependent modulation of Notch1. Additionally, Okoye et al. demonstrated a wide range of transcripts enriched in nTreg cell-EVs, including those of chemokines, interleukins, collagen and matrix proteins, and ephrins. The roles of mRNAs and proteins in modulating target cells remain unknown but seem likely to participate in the immunomodulatory effects of Treg-EVs [21]. Whether other molecules contribute to the suppressive function of Treg-EVs is currently unknown.

Current reports support that the intrinsic properties of EVs involved in regulating complex intracellular pathways advance the potential utility of EVs in achieving therapeutic control of many diseases, including RA. In this study, we performed an in vivo imaging assay to initially assess the distribution of intravenously transferred iTreg-EVs in CIA mice. We observed that iTreg-EVs home into the inflamed joint through the circulatory system. In addition, it is worth noting that Treg cells are plastic in the presence of a complex immune microenvironment in vivo. In contrast to parental cells, adoptively transferred Treg-EVs cannot be modified under inflammatory conditions in vivo, providing an ideal immunomodulatory agent. Therefore, we initially compared the suppressive function and in vivo distribution of nTreg-EVs and iTreg-EVs and found similar immunomodulatory characteristics. Although nTregs and nTreg-EVs have been proposed to be useful for treating systemic disease [21], their low numbers and instability create challenges. Our data extend the advantages and application of iTreg-EVs in treating RA and other autoimmune diseases.

Previous studies have emphasized that Treg cells become specialized for different environmental contexts, tailoring their functions and homeostatic properties to a wide range of tissues and immune conditions. Although both the nTreg and iTreg subsets share similar phenotypes and display comparable suppressive activity, several factors distinctly affect their development, stability and function [26]. Zhou et al. proposed that inflammatory environments could impair human FoxP3+ Treg-cell function by converting these cells into Teff cells in vivo [61]. Our previous research concluded that inflammatory conditions could convert nTreg cells into Th17 and other T-helper cell subtypes, whereas TGF-β-induced iTreg cells failed to undergo Th17 cell conversion [30, 31, 62]. Our recent report suggests that iTreg cells but not nTreg cells maintain their regulatory function after exposure to arthritic conditions, which could have implications for Treg-cell-based therapies in autoimmune conditions, such as RA [63]. These investigations suggest possibly distinct profiles of gene modification between nTreg and iTreg cells. It has been claimed that CpG methylation of the FoxP3 promoter in TGF-β-induced iTreg cells but not in nTreg cells accounts for the stability of iTregs [64, 65]. We previously observed that the methylation status of the FoxP3 gene loci does not affect FoxP3 stability but that FoxP3 histones are associated with iTreg stability [66]. Additionally, it is important to propose that EVs contribute significantly to the function of murine CD4+ CD25+ Foxp3+ Treg cells. Rab27-DKO Treg cells lose the ability to release EVs and fail to prevent disease in animals [22]. These results for Treg-cell heterogeneity raise the question of whether Treg phenotypic and functional maintenance occurs through vesicle-mediated cellular communication upon exposure to inflammatory conditions. Thus, the miRNA profiles of nTreg-EVs and iTreg-EVs were reported. Recently, the transfer of miRNAs, including Let-7d and miR-155, via nTreg cell-EVs to conventional T cells was shown [21]. Let-7d-containing nTreg cell-derived EVs contributed to the suppression of Th1 cell proliferation and IFN-γ secretion. nTreg cells transfer miR-155 to conventional T cells with concomitant upregulation of several Treg-cell-associated genes in recipient cells. However, our data showed that iTreg-EVs had no significant enrichment of miR-let-7d or miR-155 compared to control EVs (Fig. 3a). More importantly, iTreg-EVs showed a higher miR-155 level than nTreg-EVs. These discrepant results suggested that differential miRNA expression might drive the functional properties of these EVs. Therefore, we analyzed enriched KEGG pathways (nTreg-EVs as the control), and the results demonstrated that specific signaling pathways, including cytokine-cytokine receptor interactions, the IL-17 signaling pathway and the FoxO signaling pathway, were involved. IL-17 binding to its receptor (containing IL-17RA) activates the MAPK pathway and the nuclear factor kB (NF-κB) pathway, which promotes the expression of proinflammatory cytokines and chemokines, such as IL-1β, IL-6, IL-8, TNF-α, and granulocyte-macrophage colony-stimulating factor (GM-CSF), as well as the subsequent priming of Th17 cells and their maintenance [67, 68]. Moreover, the serum/glucocorticoid regulated kinase 1 (SGK1)-mediated induction of proinflammatory Th17 cells was recently shown to occur in tandem with activation of the p38/MAPK and NFAT5 pathways, leading to a FOXO1-dependent increase in the expression of the IL-23 receptor and heightened signaling through the IL-17 inflammatory cascade [69, 70]. Overall, we emphasize the use of different miRNA-mediated immunomodulatory mechanisms between nTreg-EVs and iTreg-EVs and suggest a putative role for EV miRNAs that imparts iTreg cell resistance to Th17 cell conversion upon exposure to inflammatory challenge.

Although clinical trials using Treg-cell-EVs have not been proposed, Yu et al. investigated the use of Treg-cell-EVs as a therapy in a transplantation setting [51]. Sullivan et al. reported that IL-35-containing EVs derived from Treg cells promoted lymphocyte exhaustion and exerted infectious tolerance [71]. These positive outcomes have paved the way for clinical trials using EVs isolated from TCR-stimulated Treg cells. Despite these encouraging findings, several limitations regarding the utility of Treg-cell-EVs cannot yet be addressed. First, no standardized protocol for isolating and purifying Treg-cell-EVs exists. Second, EV release by Treg cells is not constitutive and requires activation using anti-CD3/CD28 Abs, and whether these Abs contaminate Treg-cell-EV preparations remains untested. Proteins/cytokines present in the culture medium may also cause a potential problem, especially as EVs are isolated from culture supernatants. Last, the production of large amounts of EVs from Treg cells and the quantity of Treg cells needed for therapeutic purposes are currently unknown, and producing large quantities of sufficiently pure Treg-cell-EVs under GMP conditions remains a challenge. Collectively, our study results highlight the possibility that manipulation of in vitro TGF-β-induced iTreg-EVs, a cell-free therapy, will provide an innovative approach to combat RA and other autoimmune diseases.

Author contributions

Conceived and designed the experiment: SGZ Performed the experiments: JRC, XJH, YLH, RZL, JZ, FH, and JW. Processed and analyzed the data: JRC, FH, YLH, and XRL. Wrote and revised the paper: JRC, NO, and SGZ.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41423-021-00764-y.