MiR-1290 in natural killer cell derived extracellular vesicles: a pathogenic mediator of lupus nephritis and therapeutic target for th17 regulation

Abstract

Lupus nephritis (LN), a severe organ manifestation of systemic lupus erythematosus (SLE), is primarily driven by an imbalance between pathogenic Th17 cells and regulatory T (Treg) cells. We found that NK cell-derived extracellular vesicles (NK-EVs) from LN patients, but not healthy controls, exhibited a distinct miRNA cargo that potently drives Th17 polarization. Small RNA sequencing identified miR-1290 as the most significantly upregulated miRNA in NK-EVs derived from LN patients. This finding was validated in an independent clinical cohort, where miR-1290 levels correlated with key disease activity indices. Functional analysis revealed that miR-1290 promoted Th17 differentiation and suppressed Treg generation by downregulating NR4A2. In the MRL/lpr lupus mice, systemic delivery of NK-EVs engineered to carry a miR-1290 antagomir restored the Th17/Treg balance, alleviated renal inflammation and fibrosis. Collectively, we found miR-1290 in NK-EVs disrupted T cell homeostasis by targeting NR4A2, driving Th17/Treg imbalance. Delivering miR-1290 antagomirs via engineered NK-EVs restored this balance and alleviated renal damage in LN mice.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04212-9.

Introduction

Systemic lupus erythematosus (SLE) is a complex autoimmune disease characterized by a loss of self-tolerance, the production of pathogenic autoantibodies, and systemic inflammation that can damage multiple organs, including the kidneys, skin, joints, and hematopoietic system [1]. Lupus nephritis (LN), which develops in approximately 60% of SLE patients, represents one of the most severe disease manifestations [2]. Despite improving short-term outcomes, current immuno-suppressive strategies fail to prevent progressive renal decline in about 30% of LN patients, which leads to end-stage kidney disease within 15 years of diagnosis for many [3]. Renal damage in LN is primarily driven by a dysregulated adaptive immune response. This pathogenic process begins with the infiltration of autoreactive T cells that facilitate local autoantibody production by B cells and the secretion of pro-inflammatory cytokines (e.g., IL-17, IFN-γ, TNF-α) [4, 5]. These mediators recruit and activate additional inflammatory cells, amplify complement deposition, and directly injure renal resident cells, ultimately leading to podocyte depletion and glomerulosclerosis. Recent single-cell transcriptomic studies of LN kidney biopsies have further revealed expanded clonal T effector populations alongside a relative deficiency in regulatory T cells (Tregs), suggesting a failure of immunoregulation within the renal microenvironment [6, 7].

Beyond adaptive immune dysregulation, SLE also involves profound innate immune alterations, with NK cell biology being particularly affected. Clinical evidence reveals consistent NK cell defects in patients, characterized by diminished counts and impaired function of both CD56dim and CD56bright circulating subsets [8, 9]. Notably, diminished expression of inhibitory receptors (e.g., NKG2A, TIGIT) on NK cells correlates with disease activity, suggesting that impaired inhibitory signaling contributes to their dysregulation [10]. NK cells regulate the functions of T cells through cytokines. However, the precise impact of the altered NK cell compartment in SLE on the subsequent activation and differentiation of pathogenic T cells remains poorly understood [11]. Extracellular vesicles (EVs) have emerged as important mediators of intercellular communication in autoimmune diseases, capable of transferring proteins, lipids, and nucleic acids to modify recipient cell function [12, 13]. Therapeutically, mesenchymal stromal cell-derived EVs have shown promise in murine lupus models, delivering regulatory molecules that suppress T-cell activation and promote Treg expansion [14]. Despite these advances, the immunomodulatory capacity of EVs derived from NK cells (NK-EVs) themselves remains almost entirely unexplored in the context of autoimmunity.

Herein, we identified a distinct molecular profile in NK-EVs from LN patients. Through integrated analysis, miR-1290 was pinpointed as a key disease-associated miRNA. We demonstrated that miR-1290 in NK-EVs aggravated pro-inflammatory T cell differentiation by directly targeting the transcription factor NR4A2. Furthermore, engineered NK-EVs loaded with miR-1290 antagomirs exerted renal protective effects in a murine LN model. Collectively, our findings identified NK-EV miR-1290 as a pathogenic mediator and highlighted the miR-1290/NR4A2 axis as a promising therapeutic target for LN.

Materials and methods

Acquisition and isolation of NK-EVs

Ethical approval

For the use of clinical blood samples was obtained from both the Tenth Affiliated Hospital of Southern Medical University (MCSC2025-002-1) and the First Affiliated Hospital of Jinan University (KY-2026-034). NK cells were isolated from PBMCs using a sequential MACS protocol. First, negative selection was performed with the Human NK Cell Isolation Kit (Miltenyi Biotec, #130-092-657) to deplete CD3⁺ T cells and other non-NK lineages. The resulting cell fraction was then subjected to positive selection using human CD56 MicroBeats (Miltenyi Biotec, #130-050-401), yielding a highly pure CD3⁻CD56⁺ NK cell population (identified by flow cytometry) [15]. Freshly isolated NK cells were cultured at 0.5–1.0 × 10⁶ cells/mL in serum-free NKMACO^™ Medium (Lonza, #BE001NC), supplemented with 500 IU/mL IL-2 (PeproTech, #200-02) and 10 ng/mL IL-15 (PeproTech, #AF-200-15), and maintained at 37 °C with 5% CO₂. Half-medium changes were performed every 2–3 days using fresh cytokine-supplemented medium, maintaining cell densities between 0.5 and 2.0 × 10⁶ cells/mL. For EV isolation, supernatants were collected and sequentially processed by 0.22 m filtration, followed by centrifugation at 12,000 × g for 30 min to remove debris and large vesicles. NK-EVs were then pelleted via ultracentrifugation at 110,000 × g for 70 min, resuspended in exosome-free PBS [16], and stored at −80 °C for subsequent use.

Characterization of NK-EVs

The purified NK-EVs were characterized using multiple complementary approaches. Protein content was quantified by BCA assay, and nanoparticle tracking analysis (NTA) was used to determine particle concentration and size distribution. Morphology was visualized by transmission electron microscopy (TEM), which showed typical cup-shaped vesicles. Western blotting further confirmed the EV-specific markers, including CD63 (Abclonal, #A5271), CD9(Abclonal, #A24130), CD81(Abclonal, #A4863), and ALIX (Abcam, #ab275377), as well as the NK cell-specific effector granzyme B (Cell Signaling Technology, #4275S), verifying the successful isolation of NK-derived extracellular vesicles.

NK-EVs uptake assay

For fluorescence labeling, isolated NK-EVs were incubated with DiD membrane dye (Thermo Fisher Scientific, #V22887), with unincorporated dye removed by size-exclusion chromatography. DiD-labeled NK-EVs were then added to the cultures and incubated with T cells for 12 h. Following fixation, cells were stained for cytoskeletal markers and counterstained with DAPI before visualization using confocal microscopy (Zeiss LSM900).

Isolation and stimulation of T cells

The procedure for in vitro T cell culture and differentiation was performed as previously described [17]. Briefly, CD4⁺ T cells were isolated from PBMCs using CD4⁺ microbeads (Miltenyi Bio, #130-045-101). The purified cells were cultured in DMEM medium at 37 °C under 5% CO₂. Cells were divided into three experimental groups: a control group cultured in medium without Th17-polarizing cytokines; a model group stimulated under Th17-polarizing conditions: soluble anti-CD3 at 1 µg/mL (PeproTech, #05121-25-500), soluble anti-CD28 at 1 µg/mL (PeproTech, #10311-25-500), IL-6 at 20 ng/mL (R&D, #206-IL-010), TGF-β at 2 ng/mL (PeproTech, #100 − 21), anti-IFN-γ at 5 µg/mL(PeproTech, #500 − 32), and anti-IL-12 at 5 µg/mL (BioLegend, #501802); and a treatment group that additionally received 20 µl of NK-EVs (6 × 10¹¹ particles/mL).

Proteinase K and RNase treatment to NK-EVs

To identify the functional components responsible for the biological activity of NK-EVs, aliquots of purified EVs were subjected to selective enzymatic digestio [18]. Briefly, NK-EVs were freezing and thawing over times and digested with Proteinase K (1 mg/mL) at 37 °C for 30 min for protein-free EVs. For RNA-free EVs, samples were incubated with 0.1% Triton to permeabilize the vesicle membrane, followed by digestion with RNase A (100 µg/mL) at 37 °C for 30 min to degrade accessible RNA cargo.

Flow cytometry

For Th17 cell analysis, cells were first stimulated with 50 ng/mL PMA and 1 µg/mL ionomycin (Thermo Fisher Scientific, #00–4970-03) for 4 h. Subsequently, cells were stained with a corresponding anti-CD4 antibody, followed by fixation and permeabilization using specific buffers. Finally, intracellular staining was performed with APC-conjugated anti-IL-17 A antibody (Elabscience, #F1173E). Th17 cells were defined as the CD4⁺ IL-17 A⁺ population. To direct differentiation towards the Treg lineage, interleukin-related cytokines need to be changed. Cells were activated and polarized in culture medium supplemented with IL-4 (10 µg/mL; BioLegend, #500904) and TGF-β (10 ng/mL). Cells were cultured under these conditions for 4–5 days at 37 °C in a humidified atmosphere with 5% CO₂, with a possible replenishment of fresh medium containing IL-2 (300 IU/mL) on day 2. To assess immune cell populations in animal models, splenic single-cell suspensions were prepared by mechanical dissociation through a 70 μm strainer. Following red blood cell lysis and PBS washes, surface staining was performed using an anti-CD4 antibody. Cells were then fixed and permeabilized with a commercial kit (Thermo Fisher Scientific, #00–5523-00). Intracellular cytokine staining for IL-17 A was conducted to identify Th17 cells (defined as CD4⁺IL-17 A⁺, Abclonal, #A25652), while Treg cells (defined as CD4⁺Foxp3⁺CTLA4⁺, Abclonal, #A27685 and #A27774) were characterized by subsequent staining for Foxp3 and CTLA4. All populations were quantified by flow cytometry.

RNA extraction and quantitative real-time PCR (qPCR)

Total RNA from cells and animal tissues was extracted using the Cell Total RNA Isolation Kit (Forgene, #11000120) and Animal Tissue Total RNA Isolation Kit (Forgene, #11000130), respectively. Subsequently, the extracted RNA was reverse transcribed into cDNA using the PrimeScript^™ RT Master Mix (Takara, #RR036A). Primers were synthesized by Tsingke Biotechnology (Supplementary Table 1) and qPCR reactions were performed on a LightCycler^® 480 II System. Gene expression levels were analyzed using the 2^−ΔΔCt method.

ELISAs

The concentrations of inflammatory cytokines were quantified in serum and cell culture supernatant samples using commercially available enzyme-linked immunosorbent assay (ELISA) kits. Levels of human IL-1β (BioLegend, #430904), IL-6 (BioLegend, #430504), and TNF-α (BioLegend, #430204) were measured. Correspondingly, levels of murine IL-1β (MeiMian, #MM-0040M1), IL-6 (MeiMian, #MM-0163M1), and TNF-α (MeiMian, #MM-0132M2) in serum were determined. For all assays, sample concentrations were calculated by interpolation from standard curves prepared with the recombinant protein standards provided with each kit, following the manufacturers’ protocols.

Small RNA sequencing and bioinformatic analysis

Small RNA sequencing was performed by RiboBio Co., Ltd. (Guangzhou, China), which provides specialized exosomal small RNA sequencing services. Total RNA was extracted from NK-EVs. Sequencing libraries were constructed using a platform-specific small RNA library preparation kit. This process typically involves steps such as 3’ and 5’ adapter ligation, reverse transcription, and PCR amplification to generate cDNA libraries compatible with high-throughput sequencing. The quality and size distribution of the final libraries were assessed using an Agilent 2100 Bioanalyzer. Qualified libraries were then sequenced on an Illumina NextSeq 500 platform to generate single-end reads.

Differential expression analysis of miRNAs between experimental groups was performed using DESeq2, a widely used R package based on a negative binomial distribution model. To ensure accurate inference, the analysis accounted for potential confounding factors. Specifically, the EV isolation method was included as a covariate in the statistical model to estimate and control for its effect. MiRNAs with an adjusted p-value (FDR) < 0.05 and an absolute log2 fold change > 1 considered statistically significant and differentially expressed. Further bioinformatic analyses, such as target gene prediction and pathway enrichment for key miRNAs, were conducted based on the differential expression results.

Combined analysis of GEO database

To handle data sparsity, NK-EV miRNAs were retained only if detected in at least 50% of all samples, yielding 617 miRNAs for analysis. To further identify disease-relevant miRNAs, we analyzed an independent cohort from the GEO database (https://www.ncbi.nlm.nih.gov/geo/, accession GSE157293), which profiles gene expression in LN renal tissues. The kidney-specific gene profile was extracted from the dataset using Perl. A core set of candidate miRNAs was defined by intersecting the differentially expressed miRNAs from our NK-EV sequencing with the LN-associated miRNAs from GSE157293, using R package. The expression of these key miRNAs was then quantified in a validation cohort by qPCR, and their correlations with key clinical disease indicators were assessed in Prism 8.0.

Clinical samples

This study enrolled a cohort of individuals including healthy controls and patients diagnosed with SLE. All SLE patients fulfilled the American College of Rheumatology (ACR) classification criteria. For the clinical validation of core miRNAs, a sub-cohort comprising 26 healthy individuals and 26 SLE patients was analyzed. Furthermore, the disease correlation analysis specifically targeting NK-EV-derived miR-1290 was performed using an expanded sample set of 53 subjects, which included 25 patients with LN. Briefly, PBMCs were isolated from fresh whole blood samples. NK cells were subsequently purified from PBMCs using positive immunomagnetic selection targeting CD56⁺/CD3⁻ populations. The freshly isolated NK cells were seeded at an initial density of 0.5–1.0.5.0 × 10⁶ cells/mL and expanded as described above. Conditioned media were collected during the maintenance culture period EV isolation. The collected media were first subjected to differential centrifugation to remove cells and debris. NK-EVs were then isolated as previously described.

For miRNA expression analysis in NK‑EVs, total miRNA was isolated using the miRcute Serum/Plasma miRNA Isolation Kit (Tiangen, #DP160526). To control for extraction efficiency and normalize technical variation, a known amount of synthetic C. elegans miRNA, cel‑miR‑39 (Qiagen, #219610), was spiked into each sample as an exogenous control prior to lysis. Reverse transcription was performed with the miRcute Plus miRNA First‑Strand cDNA Kit (Tiangen, #KR211), followed by qRT‑PCR using the miRcute Plus miRNA qPCR SYBR Green Kit (Tiangen, #FP411). For preliminary miRNA expression profiling, level of each miRNA was normalized to the spiked‑in cel‑miR‑39 within the same sample using the 2^‑ΔΔCt method. Results are presented as relative expression units. All miRNA detection primers used in this study were purchased from GeneCopoeia (Supplementary Table 2).

For absolute quantification of miR‑1290, a standard curve was generated using serial dilutions of a synthetic hsa‑miR‑1290 mimic. The absolute concentration of miR‑1290 in each NK‑EV sample was interpolated from the curve and expressed as fmol/L. To correct for extraction and reverse transcription efficiency, the absolute copy number of miR‑1290 was further normalized to the recovered copy number of cel‑miR‑39 spiked into the same sample.

Engineered NK-EVs construction

The high concentration NK-EVs for engineering used in this study were derived from human cord blood NK cells (CBNK) as maintained in our laboratory [19]. Briefly, CBNKs were isolated from umbilical cord blood and expanded in serum-free medium supplemented with IL-2 and IL-15 under feeder-free conditions. The supernatant was subjected to differential ultracentrifugation to harvest CBNK-EVs, which were subsequently used for downstream engineering. MiR-1290 mimic, inhibitor, and NC mimic for in vitro studies, as well as agomir and antagomir for in vivo applications, were synthesized by RiboBio (Guangzhou, China). NK-EVs were transfected using the Exo-Fect^™ Exosome Transfection Kit (SBI, #EXFT20A-1) according to the manufacturer’s instructions. Briefly, 50–300 µg of EV protein was resuspended in 50 µl PBS, mixed with Exo-Fect solution and the corresponding miRNA, and incubated at 37 °C for 10 min. The reaction was stopped with ExoQuick-TC reagent, incubated at 4 °C for 30 min, and centrifuged at 13,000 ×g for 3 min. The pellet was resuspended in 300 µl PBS and stored at −80 °C.

Distribution of engineered NK-EVs in LN mice

Animal experiments were approved by the Animal Ethics Committee of Southern Medical University (IACUC-AWEC-202601011). MRL/lpr mice were randomly divided into two groups (n = 6 per group). The treatment group received intravenous injection of 200 µl DiD-labeled NK-EVs (2 × 10¹² particles/mL), while control animals were administered equal-volume diluted DiD dye. Whole-body fluorescence imaging was performed at 2, 4, 8, and 24 h post-injection using the IVIS Lumina K Series III imaging system (PerkinElmer), with quantitative analysis of radiant efficiency ([p/s/cm²/sr]/[µW/cm²]) conducted using Living Image 4.4 software. For organ-specific distribution assessment, mice were euthanized via a pentobarbital sodium overdose at the indicated time points, followed by ex vivo imaging of organs to determine the regional fluorescence intensity.

Zeta potential detections

The zeta potential of NK-EVs was measured using a NanoBrook Omni analyzer (Brookhaven, USA) based on phase analysis light scattering (PALS). Samples were appropriately diluted in a low-conductivity buffer and loaded into a folded capillary cell. Measurements were performed at 25 °C, and each sample was analyzed in three independent runs. The average value was calculated and reported in millivolts (mV).

Animals study

Female MRL/lpr (8-week-old) and age-matched MRL/MpJ mice (served as MRL/wt controls) were purchased from Shanghai SLAC Laboratory Animal Co., Ltd. and acclimatized for two weeks before the experiment under specific pathogen-free conditions. Ten-week-old MRL/lpr mice were randomly assigned to three treatment groups (n = 8 per group): the NK-EVs NC group receiving control NK cell-derived extracellular vesicles, the NK-EVs agomir group receiving NK-EVs enriched with a specific miRNA agomir, and the NK-EVs antagomir group receiving NK-EVs carrying a specific miRNA antagomir. All groups received intravenous tail vein injections three times per week for four consecutive weeks, with each injection consisting of 200 µl PBS containing 2 × 10¹² particles/mL of the respective EVs. After 4 weeks, serum, kidneys, and spleens were collected for subsequent analysis.

For immune cell analysis, spleens were aseptically removed and gently dissociated through a 70 μm cell strainer to prepare single-cell suspensions. Splenocytes were isolated by density gradient centrifugation using mouse lymphocyte separation medium (Dakewe, #7211011). Subsequently, CD3⁺ T cells and CD3-depleted lymphocytes were obtained by MACS with mouse CD3 microbeads (Miltenyi Biotec, #130-094-973).

Urine biochemical analysis

Mice were housed in metabolic cages for 24-h urine collection. Urinary protein and creatinine levels were measured using specific quantitative assay kits (Jiancheng Bioengineering, #C035-2-1 and #C011-2-1), employing the Coomassie Brilliant Blue method and a standard enzymatic method, respectively. The urinary albumin-to-creatinine ratio (UACR) was subsequently calculated to evaluate the degree of albuminuria.

Histopathology

Kidneys were harvested from mice at 14 weeks of age and fixed overnight in 4% formaldehyde. Following fixation, tissues were washed in PBS, embedded in paraffin, and sectioned. Tissue sections were subsequently stained with hematoxylin and eosin (H&E), Masson’s trichrome, or Sirius red according to standard protocols. Images were acquired using a light microscope. Histopathological evaluation of kidney sections, including glomerular and interstitial inflammation, was performed based on established scoring criteria as described previously [20].

Immunofluorescence staining and analysis

Cells were washed with phosphate-buffered saline (PBS) and fixed with 4% paraformaldehyde on ice for 15 min. After fixation, cells were permeabilized with 0.1% Triton X-100 for 10 min at room temperature and then blocked with 5% goat serum for 1 h to reduce non-specific binding. The cells were subsequently incubated overnight at 4 °C with primary antibodies against mouse antigens, including IL-6 (Abclonal, #A0286), TNF-α (Abclonal, #A11534), IL-1β (CST, #12242), CD3 (Abcam, #ab237721), CD4 (Abcam, #ab183685), CD8 (Abcam, #ab17147), α-tubulin (Proteintech, #66031), NR4A2 (Proteintech, #10975), IL-17 A (Servicibio, #GB11110), IFN-γ (Abclonal, #A12450), TGF-β (Bosterbio, #BA0290) and F4/80 (Abclonal, #A1256). Following incubation, samples were washed with PBS and incubated with appropriate fluorophore-conjugated secondary antibodies (1:500 dilution) for 1 h at room temperature in the dark. Cell nuclei were counterstained with DAPI. For quantification, positively stained cells in each microscopic field were manually counted by two independent, blinded observers. The counts were averaged to yield a single value per field, with a minimum of three random fields analyzed per biological sample.

Western blot

Proteins were extracted from cells or renal cortical tissues using cold RIPA lysis buffer (Fdbio Science, #FD008) supplemented with protease and phosphatase inhibitors (Fdbio Science, #FD1001). Protein concentrations were determined using a BCA assay. Equal amounts of protein lysates were denatured in 1× loading buffer at 95 °C for 10 min, separated by 10% SDS-PAGE, and subsequently transferred onto nitrocellulose membranes using a wet transfer system. The membranes were then probed with specific primary antibodies against NR4A2 (Proteintech, #10975), AKT (CST, #4691), phospho-AKT (Ser473, CST, #4060), p65 (CST, #8242), and phospho-p65 (Ser536, CST, #3033). After incubation with appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies, protein signals were visualized using the FluorChem E imaging system (ProteinSimple, USA).

Dual-luciferase reporter assay

To investigate the targeting relationship between miR-1290 and the 3’UTR of NR4A2, a dual-luciferase reporter assay was conducted using the pmirGLO Dual-Luciferase miRNA Target Expression Vector (Miaoling Bio, China). DNA fragments corresponding to the wild-type (WT) and mutant (MUT) 3’UTR regions of NR4A2 were synthesized by Tsingke (China) and cloned into the pmirGLO vector downstream of the firefly luciferase gene. HEK293T cells were plated in 12-well plates and cultured overnight. The recombinant reporter plasmids were co-transfected with either miR-1290 mimic or negative control (NC) mimic into the cells using Lipofectamine 3000 reagent (Invitrogen, USA). After 48 h of transfection, luciferase activities were measured with the Dual-Glo Luciferase Assay System (Promega, USA). Firefly lucuciferase (fLuc) signals were normalized to Renilla luciferase (rLuc) values to account for variations in transfection efficiency. The fLuc/rLuc ratio was then calculated to represent the relative luciferase activity.

Statistical analysis

The normality of data distribution was assessed using the Shapiro-Wilk test, and the homogeneity of variances was confirmed using the Brown-Forsythe test. All data met the assumptions for parametric tests. Data are presented as the mean ± SD. Statistical comparisons between two groups were performed using an unpaired two-tailed Student’s t test. For comparisons across more than two groups, one-way ANOVA was applied, followed by Tukey’s post hoc test for detailed group-wise comparisons. Diagnostic performance was evaluated by receiver operating characteristic (ROC) curve analysis. Linear regression analysis was conducted to assess correlations between variables. All statistical analyses were performed using GraphPad Prism software (version 8.0). Significance levels are denoted as follows: ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Results

LN patients derived NK-EVs promoted Th17 differentiation via RNA cargo in vitro

Primary human CD4⁺ T cells from healthy donors via magnetic beads and identified by flow cytometry for further analysis (Fig. 1a and b). NK cells were purified from the peripheral blood of both healthy control (HC) and LN patients, purified NK cells were expanded (Fig. 1c) and the conditioned media were collected for the isolation of NK-EVs. The isolated NK-EVs were characterized using NTA (Fig. 1d), western blotting for exosomal markers (Fig. 1e), and TEM for morphological analysis (Fig. 1f). A fluorescence uptake assay confirmed that these NK-EVs were efficiently internalized by primary T cells (Fig. 1g and h). To assess their immunomodulatory function, CD4⁺ T cells were pre-treated with NK-EVs for 24 h, followed by polarization under Th17-skewing conditions. Flow cytometric analysis revealed that LN patients derived NK-EVs significantly enhanced the differentiation of IL-17 A⁺ Th17 cells compared to HC (Fig. 1j). This suggested that LN derived NK-EVs could act as pro-inflammatory mediators by modulating signals involved in Th17 cell fate. To identify the active component, we performed loss-of-function experiments (Fig. 1k). RNase A treatment of LN-NK-EVs, which degrades accessible RNA, completely abolished their capacity to promote Th17 differentiation. In contrast, proteinaseK-treated NK-EVs, which retain the RNA cargo, remained fully active in driving Th17 polarization. This was evidenced by an increased frequency of IL-17 A⁺ Th17 cells (Fig. 1l), elevated IL17A mRNA expression in T cells (Fig. 1m), and enhanced secretion of IL-17 and IFN-γ (Fig. 1n). Under these Th17-skewing conditions, however, the levels of IL-10 and TNF-α showed no significant differences (Fig. 1n). Together, these data pinpoint RNA as the primary functional cargo within lupus NK-EVs that drives pro-inflammatory Th17 polarization.

Fig. 1.

Characterization of NK-EVs and their effect on Th17 cell differentiation in vitro. (a) Schematic of the experimental procedure for treating primary T cells with NK-EVs. (b) Purity analysis of sorted CD4⁺ T cells by flow cytometry. (c) Purity analysis of sorted NK cells, showing the proportion of CD3⁻CD56⁺ NK cells exceeded 94%. (d) Size distribution of NK-EVs as determined by NTA. (e) Western blot analysis of exosomal markers and specific cargos in NK-EVs, with 293T-derived EVs as a control. (f) TEM analysis of NK-EVs showing cup-shaped bilayer membrane structures (indicated by black arrows, scale bar: 200 nm). (g) Immunofluorescence staining showing the uptake of NK-EVs by T cells (indicated by black arrows, scale bar: 10 μm) and (h) fluorescence intensity analysis of the local region indicated by white arrows. (i) Gating strategy and representative flow cytometry plots for CD4⁺IL-17 A⁺ T cells under Th17-polarizing conditions after EVs treatment. (j) LN-EVs significantly reduced the proportion of CD4⁺IL-17 A⁺ cells (N = 3). (k) Schematic of the procedures for generating Protein-free NK-EVs via protease treatment and RNA-free NK-EVs via RNase treatment. (l) Effects of LN-derived NK-EVs, Protein-free NK-EVs, and RNA-free NK-EVs on the proportion of Th17 cell differentiation. (m) Effects of LN-derived NK-EVs, Protein-free NK-EVs, and RNA-free NK-EVs on the IL17A mRNA level in T cells. (n) Effects of LN-derived NK-EVs, Protein-free NK-EVs, and RNA-free NK-EVs on the levels of inflammatory cytokines IFN-γ, TNF-α, IL-17 A, and IL-10 in T cell culture supernatants. Data are presented as mean ± SD (N = 3). *P < 0.05, **P < 0.01 (One-way ANOVA)

Identification of miR-1290 as a disease-associated miRNA in Lupus NK-EVs

MiRNAs are key epigenetic regulators selectively packaged into extracellular vesicles to mediate intercellular communication and precisely regulate recipient cell gene expression [21]. We next performed small RNA sequencing on NK-EVs to identify specific effector miRNAs. Compared to healthy controls, NK-EVs from lupus patients exhibited a distinct miRNA expression profile (Fig. 2a and S1a). MiRNAs regulate cellular functions by targeting specific mRNAs (Figure S1b). Gene Ontology analysis suggested these differentially expressed miRNAs could target genes involved in immune-regulatory pathways, including T cell signaling (Fig. 2b). To prioritize candidates relevant to renal pathology, we integrated our sequencing data with a public LN kidney transcriptomic dataset (GEO: GSE157293), which profiles differentially expressed miRNAs and mRNAs in renal biopsy tissues from LN patients. This cross-analysis identified 10 miRNAs associated with LN progression (Fig. 2c).

Fig. 2.

Identification and clinical significance of elevated miR-1290 in NK-EVs from LN patients. (a) Volcano plot of small RNA sequencing showing significantly altered miRNA expression profiles in LN NK-EVs (N = 5). (b) Disease Ontology (DO) enrichment analysis of the differentially expressed miRNAs. (c) Venn diagram of the integrated analysis between differentially expressed miRNAs and the GSE157293 dataset, identifying 10 core miRNAs. (d) Relative expression abundance of the 10 core miRNAs in NK-EVs derived from healthy controls and SLE patients. (e) Relative expression analysis of the 5 NK-EVs miRNAs showing statistically significant differences between SLE and LN patients. (f) Standard curve for qPCR analysis constructed using an miR-1290 mimic. (g-l) Correlation analysis between miR-1290 levels in NK-EVs and clinical parameters: (g) serum complement C3, (h) serum complement C4, (i) SLEDAI score, (j) plasma anti-nuclear antibody (ANA) levels, (k) plasma anti-dsDNA antibody levels, and (l) 24-h proteinuria levels. (m) miR-1290 levels were significantly elevated in NK-EVs derived from LN patients. (n) ROC curve analysis of the diagnostic efficacy of miR-1290, anti-C1q antibody levels, and their combination in distinguishing SLE patients with and without lupus nephritis. Correlations between miR-1290 levels and clinical parameters were assessed by Spearman’s rank correlation analysis. A P value of less than 0.05 was considered statistically significant

Validation in an independent clinical cohort confirmed significant upregulation of five miRNAs (miR-212-5p, miR-1246, miR-1290, miR-5100, and miR-6501-5p) in SLE patient NK-EVs (Fig. 2d). However, only miR-1290 demonstrated a significant further elevation in LN subgroup, whereas the other four miRNAs failed to differentiate SLE patients with LN from those without nephritis (Fig. 2e and S1c). The clinical relevance of miR-1290 was underscored by its strong correlation with key disease parameters: NK-EV miR-1290 levels showed a significant inverse correlation with serum complement C3/C4 levels (Fig. 2g and h), and a strong positive correlation with disease activity as measured by the SLE Disease Activity Index (SLEDAI) score (Fig. 2i), titers of anti-nuclear antibodies and anti-dsDNA antibodies (Fig. 2j and k), as well as 24-h proteinuria (Fig. 2l). Notably, miR-1290 was specifically elevated in patients with LN (Fig. 2m). Furthermore, combining NK-EV miR-1290 levels with anti-C1q antibody titers enhanced the diagnostic performance for distinguishing LN from non-renal SLE (Fig. 2n). Combining NK-EV miR-1290 with anti-C1q antibody levels significantly enhanced the diagnostic efficacy for distinguishing LN from SLE without LN (AUC = 0.791, sensitivity = 0.65 and specificity = 0.73) compared to anti-C1q alone (AUC = 0.726, sensitivity = 0.57 and specificity = 0.73). Collectively, these results identify miR-1290 as the key functional miRNA that drives the pro-inflammatory activity of LN patient-derived NK-EVs.

MiR-1290 Directly Targets NR4A2 to Disrupt Th17/Treg Balance

To elucidate the mechanism by which miR-1290 exerts its effect, we first predicted its potential mRNA targets by integrating our sequencing data with miRNA-mRNA association analysis from the GSE157293. Among the predicted targets were the transcription factors PGR and NR4A2, as well as the transcriptional cofactor DACH1 (Fig. 3a). Given the observed impact of NK-EVs on Th17 cell differentiation, we directly tested the function of miR-1290 by transfecting primary CD4⁺T cells with miR-1290 mimics (Fig. 3b). This manipulation promoted their polarization toward a Th17 phenotype while attenuating the formation of Tregs, thereby increasing the Th17/Treg ratio and recapitulating the pro-inflammatory shift induced by LN patient-derived NK-EVs (Fig. 3c and e). This functional outcome directed our focus to NR4A2, a transcription factor recognized for its role in promoting Foxp3 expression and maintaining Treg-mediated immune tolerance [22]. To investigate whether miR-1290 directly regulates NR4A2, we performed a dual-luciferase reporter assay. We constructed reporter vectors containing either the wild-type or a mutated version of the predicted miR-1290 binding site within the 3’UTR of NR4A2 (Fig. 3f). Co-transfection in HEK293T cells demonstrated that miR-1290 significantly suppressed the luciferase activity of the wild-type reporter but had no effect on the mutant construct, confirming direct targeting (Fig. 3f). Western blot analysis further validated this regulatory relationship, showing that miR-1290 mimics reduced, while a miR-1290 inhibitor increased, endogenous NR4A2 protein levels in T cells (Fig. 3g). We next examined the functional consequence of NR4A2 using lentiviral-mediated overexpression and knockdown in Jurkat cells (Fig. 3h). Overexpression of NR4A2 robustly enhanced the induction and stability of Foxp3⁺ Jurkat cells (Fig. 3i). RNA sequencing of T cells under inflammatory conditions revealed that NR4A2 overexpression led to a broad transcriptional suppression of genes (Figure S2a, S2b). KEGG pathway analysis revealed that the differentially expressed genes were associated with immune-related diseases at the category level (Figure S2c). At the signaling pathway level, it further highlighted significant downregulation of numerous innate immune and inflammatory pathways, including virus-associated signaling (e.g., COVID-19 and HSV-1 infection) as well as key immune-regulatory pathways such as JAK-STAT, mTOR, and T cell receptor signaling (Fig. 3j, S2d). Further Gene Set Enrichment Analysis (GSEA) corroborated that the inhibition of inflammatory and immune-related signatures was closely associated with NR4A2 activity, with notable suppression of innate immune pathways including RIG-I-like receptor and Toll-like receptor signaling (Fig. 3k, S2d). Given the reported dual roles of NR4A2 in T cells [22, 23], we further investigated its proximal signaling effects. Results showed that NR4A2 overexpression significantly reduced the phosphorylation and nuclear translocation of the NF-κB subunit p65 upon inflammatory stimulation (Fig. 4a). Consistent with this, western blot analysis confirmed that NR4A2 stabilization potently suppressed the phosphorylation of both p65 and AKT (Fig. 4b and d), aligning with recently reports [24]. This indicates that NR4A2 can inhibit the activation of the NF-κB pathway, a critical upstream driver of RORγt and Th17 differentiation [25]. This mechanism was conserved in mice, as transfection of miR-1290 mimics into murine primary CD4⁺ T cells similarly decreased NR4A2 protein (Fig. 4e and f). Collectively, these data establish NR4A2 as a key negative regulator that restrains Th17-promoting signals and promotes Treg stability. The upregulation of miR-1290 in LN facilitates a pro-inflammatory T cell shift, at least in part, through the direct suppression of NR4A2, revealing a novel miRNA-mediated mechanism of immune dysregulation in lupus nephritis.

Fig. 3.

MiR-1290 drives Th17/Treg imbalance by directly targeting NR4A2. (a) Integrated analysis of sequencing data and the GEO database to predict potential target genes of miR-1290. (b) Effect of miR-1290 mimic transfection on the secretion of IL-17 A and IFN-γ under Th17-polarizing conditions (N = 6). (c) Impact of miR-1290 mimic transfection on the Th17/Treg ratio under respective polarizing conditions (N = 3). (d) Gating strategy for identifying CD4⁺IL-17 A⁺ Th17 cells after transfection under Th17-polarizing conditions. (e) Gating strategy for identifying CD4⁺Foxp3⁺ Treg cells after transfection under Treg-polarizing conditions. (f) Dual-luciferase reporter assay verifying the direct targeting of miR-1290 to the 3’UTR of NR4A2 (N = 3). (g) Western blot analysis showing the effect of miR-1290 mimic or inhibitor on NR4A2 protein expression in T cells. (h) Modulation of NR4A2 protein levels in Jurkat cells via lentiviral transduction (overexpression, OE; knockdown, KD). (i) Effect of NR4A2 levels on the proportion of CD4⁺Foxp3⁺ Jurkat cells under Treg-polarizing conditions (N = 3). (j) KEGG pathway analysis of RNA sequencing data from NR4A2-overexpressing Jurkat cells versus controls. (k) GSEA of the same RNA-seq data, showing inflammatory signaling pathways suppressed by NR4A2 overexpression. Data are presented as mean ± SD. *P < 0.05, **P < 0.01 (One-way ANOVA)

Fig. 4.

NR4A2 inhibited the AKT/NF-κB pathway. (a) Confocal fluorescence microscopy images showing the effect of NR4A2 on the nuclear localization of p65 in Jurkat cells (sacle bar: 10 μm). (b) Western blot analysis of the effects of NR4A2 on the phosphorylation of p65 and AKT under inflammatory stimulation. (c) Grayscale analysis of the p-p65/p65 ratio (N = 3). (d) Grayscale analysis of the p-AKT/AKT ratio (N = 3). (e) Effects of miR-1290 mimics and inhibitor on NR4A2 protein expression in mice CD4⁺ T cells and (f) corresponding grayscale analysis (N = 3). Data are presented as mean ± SD. *P < 0.05, **P < 0.01 (One-way ANOVA)

Construction and distribution of engineered NK-EVs

Furthermore, we engineered NK-EVs to carry miR-1290 agomir. Using the exosome transfection kit, we loaded the agomir into NK-EVs (engineered EVs), with unmodified NK-EVs serving as a control. Normal NK cells exhibit average abundance of miR-1290 (Figure S3a and Table 3) and produce extracellular vesicles with a particle size below 200 nm (Figure S3b). NTA analysis revealed that agomir-EVs had a moderately increased average size of approximately 220 nm compared to control NK-EVs (Fig. 5a), further confirmed by TEM (Fig. 5b). And the absolute zeta potential of engineered EVs was decreased (Fig. 5c). Crucially, owing to shared membrane components, NK-EVs showed a natural tropism for T cells [26]. In vitro uptake assays demonstrated that T cells internalized significantly more engineered EVs than other immune cell types, highlighting their inherent targeting capability and biocompatibility (Fig. 5d). Successful loading was proved by the increased abundance of miR-1290 in agomir engineered EVs (Figure S3c). Following intravenous injection into MRL/lpr lupus-prone mice, DiD-labeled engineered NK-EVs exhibited a distinct temporal and spatial distribution profile. Time-course live imaging revealed robust accumulation of the EV signal in the liver and spleen at early time points (Fig. 5e and f). The signal intensity progressively declined over time, with a marked diminution observed in the spleen by 24 h (Figure S3d). Notably, EV signals were also detected in the brain throughout the observation period, albeit at lower intensity compared to the liver(Figure S3e), indicating that the engineered NK-EVs retained the ability to traverse the blood-brain barrier.

Fig. 5.

Characterization and distribution of engineered NK-EVs. (a) NTA size distribution analysis of NK-EVs before and after engineering. (b) TEM analysis of NK-EVs before and after engineering. Scale bar: 100 nm. (c) Zeta potential measurement of NK-EVs before and after engineering (N = 3). (d) Uptake of engineered EVs by mouse splenic lymphocytes, CD3⁺ T cells, and CD3-depleted lymphocytes (scale bar: 10 μm). (e) Representative in vivo images showing the temporal biodistribution of EVs. Mice received an intravenous injection of DiD-labeled EVs. Fluorescence images were acquired at the indicated time points (2, 4, 8, and 24 h) post-injection. (f) Representative ex vivo images of dissected major organs at corresponding time points. The color scale represents radiant efficiency ([p/s/cm²/sr]/[µW/cm²]). Data are presented as mean ± SD. *P < 0.05

Engineered NK-EVs Targeting miR-1290 Ameliorate Lupus Nephritis by Restoring Immune Balance and Attenuating Renal Inflammation

Next, we evaluated the therapeutic efficacy of engineered NK-EVs in the MRL/lpr mouse model, which recapitulates key features of human LN, including T cell dysfunction and a skewed Th17/Treg balance (Fig. 6a). After four weeks of treatment, mice receiving antagomir-EVs (NK-EVs loaded with miR-1290 antagomir) showed a significant reduction in the UACR compared to those treated with control NK-EVs, indicating improved renal function (Fig. 6b). Analysis of renal cortical RNA revealed that antagomir-EVs treatment suppressed the expression of pro-inflammatory cytokines, while agomir-EVs (loaded with miR-1290 agomir) exacerbated renal inflammation (Fig. 6c). Splenocyte analysis by flow cytometry showed that antagomir-EVs treatment shifted the Th17/Treg ratio toward immunoregulation, characterized by a reduction in CD4⁺ IL-17 A⁺ Th17 cells and an expansion of CD4⁺ Foxp3⁺ Tregs, whereas agomir-EVs aggravated this pathogenic imbalance (Fig. 6d and f).

Fig. 6.

Engineered NK-EVs reshape T cell balance and cytokine profile in lupus nephritis. (a) Experimental timeline of EVs administration in MRL/lpr mice. (b) Measurement and analysis of urine protein-to-creatinine ratios, reflecting renal function impairment (N = 8). (c) Detection of inflammation-related mRNAs in the renal cortex of mice (N = 8). (d) Gating strategy for flow cytometric analysis of the effects of consecutive administration of therapeutic antagomir or pathogenic agomir loaded NK-EVs on splenic T cell function. (e) Changes in the proportion of CD4⁺IL-17 A⁺ cells (N = 6). (f) Changes in the proportion of CD4⁺CTLA-4⁺FOXP3⁺ cells. (g) Representative images of CD4⁺ and CD8⁺ T cell infiltration in the kidneys of LN mice (scale bar, 50 μm) and (h) quantitative analysis of cell counts (N = 6). (i) Expression levels of inflammatory cytokines in mouse serum (N = 6). (j) Confocal microscopy images showing T cell infiltration and IL-17 A expression around kidney-resident macrophages (scale bar, 20 μm), and (k) quantitative analysis of cell counts (N = 3). Data are presented as mean ± SD. *P < 0.05, **P < 0.01 (One-way ANOVA)

This immunomodulatory effect translated to the target organ. Antagomir-EVs treatment markedly reduced the infiltration of both CD4⁺ and CD8⁺ T cells into the kidneys, whereas agomir-EVs increased it (Fig. 6g and h). This may be related to the systemic effect of engineered EVs following their distribution to organs like the spleen, which helped rebalance the systemic immune dysregulation. Notably, antagomir-EVs significantly downregulated the levels of serum inflammatory cytokines in lupus-prone mice (Fig. 6i), which could lead to therapeutic benefits against systemic lupus progression. Further analysis revealed that, compared to the model group, the antagomir group exhibited alleviated renal T cell infiltration and IL-17 A production. Conversely, in the agomir group, the infiltrating renal T cells showed a significantly enhanced Th17-driven inflammatory response (Fig. 6j and k).

To precisely quantify kidney-infiltrating, cytokine-producing T cells, we performed ELISPOT on sorted renal CD3⁺ T cells. Antagomir-EVs significantly reduced the numbers of IL-17 A⁺ and IFN-γ⁺ T cells, while agomir-EVs expanded these populations (Fig. 7a and b). This reduction is crucial as renal IL-17 A and IFN-γ directly promote inflammation and tissue injury.

Fig. 7.

Engineered NK-EVs ameliorate renal inflammation and fibrosis in lupus nephritis. (a) ELISPOT analysis of T cells infiltrating mouse kidneys to identify their potential for pro-inflammatory differentiation into Th17 (IL-17 A⁺) and Th1 (IFN-γ⁺) subsets. (b) Quantitative analysis of ELISPOT counts (N = 3). (c) Representative images of multiplex immunofluorescence staining showing the expression of IL-6, TGF-β, and IL-1β in the renal cortex (scale bar: 50 μm). (d) Percentages of IL-6-positive and TGF-β-positive areas (N = 6). (e) Local fluorescence intensity analysis showing a high degree of co-localization between IL-6 and TGF-β in renal cells (white arrows), indicative of a driver for inflammatory fibrotic transition. (f) Histopathological analysis of inflammation and fibrosis in the kidneys of LN mice (scale bar: 50 μm). Data are presented as mean ± SD. *P < 0.05, **P < 0.01 (One-way ANOVA)

Consistent with reduced inflammatory T cell influx, multiplex immunofluorescence of kidney sections showed that antagomir-EVs treatment substantially lowered the expression of the inflammatory mediators IL-6 and IL-1β, as well as the key fibro-genic cytokine TGF-β (Fig. 7c and d). And agomir-EVs treatment amplified these signals. The concurrent suppression of IL-6 and TGF-β is particularly significant (Fig. 7e), as their co-upregulation in tubular epithelial cells drives progressive fibrosis through the coordinated activation of STAT3 and Smad signaling pathways [27]. Finally, histopathological assessment confirmed that antagomir-EVs alleviated glomerular and interstitial inflammation and fibrosis (Fig. 7f and S4).

Discussion

The pathogenesis of LN is characterized by a complex interplay between innate and adaptive immunity [28]. Despite the recognized roles of Th17/Treg imbalance and NK cell dysfunction in LN [29], it remains unclear how these components are mechanistically linked. This study identifies a crucial link through EVs. We show that EVs derived from the dysfunctional NK cells of LN patients exhibit a distinct miRNA signature.

EVs have emerged as pivotal vectors of intercellular communication, enabling the precise transfer of proteins, lipids, and nucleic acids to modulate recipient cell function [30]. Within the immune landscape, EVs from various cellular sources are known to fine-tune T cell differentiation. For instance, Treg-derived exosomes can suppress Th1 responses via transfer of let-7d miRNA [31], while tumor-associated macrophage exosomes can disrupt the Treg/Th17 balance through miR-29a-3p [32]. Extending this paradigm to innate-adaptive immune crosstalk in autoimmunity, we identified a specific pathogenic signal emanating from NK cells in LN. Clinical studies have consistently documented quantitative and functional NK cell defects in SLE patients, including reductions in circulating subsets and diminished expression of inhibitory receptors, which correlate with disease activity [10, 33]. This altered NK cell compartment may thus be a critical source of pathogenic EV signals. We found that NK-EVs from LN patients, but not from healthy controls, possess a distinct miRNA cargo that potently drives naïve CD4⁺ T cells toward a Th17 phenotype while impairing Treg differentiation. This functional alteration underscores a novel mechanism whereby impaired NK cell function may actively contribute to disease progression not only through deficient cytotoxicity [34] but also via the release of immunologically active EVs. By combining loss-of-function experiments with a systematic comparison of small RNA profiles, we identified miR-1290 as the most pathologically relevant miRNA specifically enriched within LN-derived NK-EVs. The level of NK-EV miR-1290 strongly correlated with key disease activity indices and distinguished patients with active LN from non-renal SLE cases, highlighting its potential as a specific biomarker for renal involvement.

Mechanistically, miR-1290 directly targets and suppresses the transcription factor NR4A2 in CD4⁺ T cells. The role of NR4A2 in immunity is complex and exhibits significant context-dependent plasticity [35]. In certain autoimmune settings, NR4A2 has been associated with pro-inflammatory roles in T helper cells [36]. Conversely, other evidence underscores its anti-inflammatory capacity, such as contributing to Treg stability and Foxp3 expression [37]. This apparent contradiction may be reconciled by considering the integrated in vivo milieu, where the net effect of NR4A signaling is determined by the specific pathological context [38]. Regarding safety, the engineered EVs predominantly accumulated in the liver and spleen (Fig. 5f). In the tissues with low-level uptake, endogenous RNA degradation mechanisms provide a natural safety threshold [39]. We emphasize that further enhancing EV targeting through engineering strategies, such as membrane modification, is essential to minimize off-target distribution [40]. Furthermore, rigorous safety evaluations will be strictly implemented to validate the safety profile for clinical translation.

As immunomodulatory agents, NK-EVs have been explored in cancer immunotherapy [41] and other disease strategies [42]. Therefore, we sought to investigate the specific immunoregulatory functions of NK-EVs on T cells in the context of LN progression. While NK-EVs have been implicated great potential in T cell regulation [26, 43], our study directly demonstrates their preferential uptake by T cells. This intrinsic targeting specificity likely facilitates their natural biodistribution and underlies their regulatory function. Moreover, our ex vivo imaging results confirmed that NK-EVs could accumulate in the brain tissue of MRL/lpr mice. This capability to traverse the blood-brain barrier is a distinct advantage of endogenous vesicle carriers. Recent advancements in exosome-based therapies have demonstrated that engineered vesicles can effectively cross physiological barrier to deliver cargoes for immunotherapy [44, 45]. Furthermore, small EVs are known to mediate long-distance inter-organ communication, regulating gene expression in distal tissues [46]. Exploiting immune cell–derived exosomes as natural carriers to traverse the blood-brain barrier for targeted drug delivery is an attractive approach in translational neuroscience.

While intrinsic homing capabilities are beneficial, the integration of active targeting modules is essential for minimizing systemic exposure and enhancing delivery efficiency. Specifically, surface functionalization with specific antibodies has recently demonstrated significant potential in directing nanovesicles to distinct immune subsets for combinational immunotherapy [47]. Building on this, more advanced genetic engineering strategies, such as the surface display of glycosylated nanobodies, offer a robust platform for enhanced immune recognition and uptake, as demonstrated in recent work on engineered cellular vesicles [48]. Furthermore, the application of transformable supramolecular bispecific cell engagers represents a cutting-edge avenue to augment the precise interaction between therapeutic vesicles and target T cells or NK cells [49]. By adopting these progressive engineering strategies, future iterations of NK-EVs could achieve superior accumulation in target lymphoid tissues, thereby maximizing the therapeutic index.

This study has several limitations. Firstly, although the clinical cohort size was sufficient for initial correlation analyses, it remained moderate. Multi-center studies with larger and more diverse patient populations are required to validate NK-EV miR-1290 as a valuable biomarker across various SLE disease phases and prognostic subgroups in the future. Moreover, while the MRL/lpr model reproduces key LN features, its ability to fully recapitulate human disease heterogeneity remains limited. Further mechanistic validation in human-relevant systems, such as renal organoids, will be important to strengthen the translational relevance of the miR-1290/NR4A2 axis identified here. Additionally, NK cells were expanded in a feeder-free system with cytokine stimulation, a common practice to ensure a pure EV source. However, this ex vivo conditioning may have influenced the miRNA profile of dysfunctional patient-derived NK cells. This potential alteration could partially explain the limited number of differentially expressed miRNAs observed. Future investigations employing single-vesicle analysis technologies to concurrently profile vesicle origin and cargo in large-scale cohorts will be crucial to unequivocally define the role of NK-EV miR-1290 in LN pathogenesis.

In conclusion, this work uncovers a novel dimension of innate-adaptive immune crosstalk in LN. We describe a pathway in which functionally impaired NK cells in lupus release miR-1290-enriched EVs. These EVs remotely promote a Th17/Treg imbalance by directly targeting the NR4A2 axis in CD4⁺ T cells. This axis provides new insights into LN immunopathogenesis and facilitates new interventional strategies via engineered EVs.

Author contributions

CC and MML designed the study. CC, XL, RL, and TWG performed the animal experiments and in vitro experiments. XL, RL and QY collected samples and clinical information from healthy donors and lupus patients. CC and HHS performed the data analyses. CC, HHS, RL and CWO wrote the manuscript. CWO and MML supervised the project.

Funding

This work was supported by the China Postdoctoral Science Foundation (2025M771982 to C.C.), Guangdong Basic and Applied Basic Research Foundation (2024A1515110217 to C.C. and 2023B1515130005 to CW.O.), Dongguan Social Development Science and Technology Project (20231800932472 to CW.O.) and National Natural Science Foundation of China (32371428 to CW.O. and 82172346 to MM.L.).

Data availability

Data will be made available on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.