Proinflammatory Stress Activates Neutral Sphingomyelinase 2–Based Generation of a Ceramide-Enriched β-Cell EV Subpopulation

Abstract

β-Cell extracellular vesicles (EVs) play a role as paracrine effectors in islet health, yet mechanisms connecting β-cell stress to changes in EV cargo and potential impacts on diabetes remain poorly defined. We hypothesized that β-cell inflammatory stress engages neutral sphingomyelinase 2 (nSMase2)-dependent EV formation pathways, generating ceramide-enriched small EVs that could impact surrounding β-cells. Consistent with this, proinflammatory cytokine treatment of INS-1 β-cells and human islets concurrently increased β-cell nSMase2 and ceramide abundance, as well as small EV ceramide species. Direct chemical activation or genetic knockdown of nSMase2, chemical treatment to inhibit cell death pathways, or treatment with a glucagon-like peptide-1 (GLP-1) receptor agonist also modulated β-cell EV ceramide. RNA sequencing of ceramide-enriched EVs identified a distinct set of miRNAs linked to β-cell function and identity. EV treatment from cytokine-exposed parent cells inhibited peak glucose-stimulated insulin secretion in wild-type recipient cells; this effect was abrogated when using EVs from nSMase2 knockdown parent cells. Finally, plasma EVs in children with recent-onset type 1 diabetes showed increases in multiple ceramide species. These findings highlight nSMase2 as a regulator of β-cell EV cargo and identify ceramide-enriched EV populations as a contributor to EV-related paracrine signaling under conditions of β-cell inflammatory stress and death.

Article Highlights

• Mechanisms connecting β-cell stress to extracellular vesicle (EV) cargo and diabetes are poorly defined.

• Does β-cell inflammatory stress engage neutral sphingomyelinase 2 (nSMase2)-dependent EV formation to generate ceramide-enriched small EVs?

• Proinflammatory cytokines increased β-cell small EV ceramide via increases in nSMase2. Ceramide-enriched EVs housed distinct cargo linked to insulin signaling, and ceramide species were enriched in plasma EVs from individuals with type 1 diabetes.

• Ceramide-enriched EV populations are a potential contributor to β-cell EV-related paracrine signaling.

Introduction

Pancreatic β-cell failure is the cornerstone of diabetes development, and β-cells may not be innocent bystanders in this process (1). Rather, β-cells may exhibit maladaptive responses to inflammatory stress in different forms of diabetes, such as prolonged exposure to systemic elevations in proinflammatory cytokines and saturated free fatty acids in type 2 diabetes (T2D) or autoimmune attack of β-cells and cytokine exposure in type 1 diabetes (T1D) (2,3). Specifically, metabolic stress pathways intrinsic to the β-cell can hasten the progression to apoptosis or enhance β-cell responsiveness to inflammatory or immune-mediated destruction (2,4). Clear delineation of the contributions of intrinsic β-cell dysfunction to diabetes development is critical, as it may ultimately pave the way for novel β-cell–targeted therapies and novel biomarkers of disease.

Extracellular vesicles (EVs) are membrane-bound nanoparticles housing molecular cargo (5). EVs can be transferred to or interact with other cells as a means of cell-cell communication (6–8). The most abundant β-cell EVs are small EVs, predominantly composed of exosomes, or EVs formed from exocytosis of multivesicular endosomes (MVEs) (5,8). β-Cell EV cargo changes under stress and disease conditions, and some data suggest that β-cell EVs may serve as paracrine effectors in the islet microenvironment in health and disease (6,7,9–16). However, mechanisms of changes in β-cell EV cargo remain poorly defined.

In non-islet cells, diverse molecular mechanisms have been identified as regulating EV biogenesis and cargo loading, including ceramide-dependent MVE vesicle formation and release of ceramide-enriched small EV subpopulations (17–20). Ceramides are bioactive sphingolipids, molecules containing a sphingoid base backbone attached to a fatty acid side chain that can function as membrane signaling components in stress responses (21). Ceramide generation is linked to several β-cell stress pathways, including oxidative stress, endoplasmic reticulum (ER) stress, and proapoptotic signaling (22–28). Based on this, we hypothesized that islet inflammatory stress and increased ceramide generation would increase ceramide-enriched EV populations. We specifically sought to test whether this occurs due to increases in neutral sphingomyelinase 2 (nSMase2), an enzyme that hydrolyzes sphingomyelin to generate ceramide and has been linked to changes in EV miRNA cargo in other systems (17,18,29).

Research Design and Methods

Single-Cell RNA-Sequencing Data Set Analysis

To examine Smpd3 expression in disease models of primary β-cells, we used the mouse islet single cell atlas (30) downloaded from CELL×GENE (https://cellxgene.cziscience.com/collections/296237e2-393d-4e31-b590-b03f74ac5070). The R data serialization file was inputted into Seurat version 4.3.0 (31) in R version 4.2.2. We filtered the data set to include only “beta” cells from “normal,” “type 1 diabetes mellitus,” and “type 2 diabetes mellitus” samples. Visualization and differential expression analyses were performed using the VlnPlot and FindAllMarkers functions.

Culture of Cells and Islets

Specific reagents used are detailed in Supplementary Table 1. β-Cell lines (INS-1 823/13 and MIN6) were cultured as previously described (8,32) with 10% exosome-depleted FBS (purchased or prepared by ultracentrifugation at 120,000g for 20 h). Mouse islets (C57BL6/J; The Jackson Laboratory) were isolated using collagenase (32). Human islets were provided from the Integrated Islet Distribution Program and The Alberta Diabetes Institute IsletCore (Supplementary Tables 3 and 4) and cultured in DMEM with 10% FBS.

To model stress, cells were exposed to interleukin-1β (IL-1β) (5–10 ng/mL), cytokine mix (5 ng/mL IL-1β, 10 ng/mL tumor necrosis factor α [TNF-α], and 100 ng/mL interferon-γ [IFN-γ]) (8,32), or 5 nmol/L thapsigargin, 1 µmol/L tunicamycin, 50 nmol/L doxorubicin, or 10 µmol/L raptinal. Caspase and RIP3 kinase inhibition were performed using 10 µmol/L ZVAD or 5 µmol/L GSK-872. Caffeic acid phenethyl ester (CAPE) 5 µmol/L was used to activate nSMase2. To examine the effect of exendin-4, we pretreated cells with 10  nmol/L exendin-4 or vehicle for 15 min, then IL-1β with or without 10  nmol/L exendin-4 for another 24  h.

nSMase2 knockdown (KD) cells and scramble shRNA cells were generated via plasmid transfection (shRNA with GFP reporter), selected with puromycin, FACS sorted, expanded, and validated with immunoblot. Immunoblots were performed as previously described (32) using anti-nSMase2 (1:200) and anti-β-actin (1:1,000) antibodies, with secondary detection via Odyssey Fc (Li-Cor). RNA was extracted using QIAzol (QIAGEN), cDNA synthesized using the High-Capacity kit (Thermo Fisher Scientific), and quantitative RT-PCR conducted using SYBR Green (Bio-Rad) (32). Primers are listed in Supplementary Table 2.

For glucose-stimulated insulin secretion (GSIS), cells were preincubated with Krebs-Ringer bicarbonate buffer with 1 mmol/L glucose, then treated with 5.6 or 16.7 mmol/L glucose for 1 h at 37°C. Supernatants were collected and centrifuged, and insulin was measured using ELISA (Mercodia) normalized to protein content.

Cell death was assessed by treating cells with the described agents using media containing either Sytox green (S7020; Thermo Fisher Scientific) or Cytotox red (4632; Sartorius), followed by real-time imaging using Incucyte S3 (Sartorius). Fluorescence intensity was normalized to cell area and expressed as a percentage of vehicle control.

Flow Cytometry

Cells were harvested and filtered through 40-µm cell strainers to obtain single-cell suspensions. For surface ceramide detection, cells were blocked and stained in blocking buffer 5% FBS/0.1% BSA/PBS/2 mmol/L EDTA. For total ceramides, cells were fixed in 4% methanol-free formaldehyde solution, permeabilized in 0.1% saponin/PBS/2 mmol/L EDTA, and blocked in buffer with 0.1% saponin. Cells were stained with anti-ceramide antibody (1 µg/mL) and/or nSMase2 antibodies (1 µg/mL), followed by fluorescent secondary antibodies (1 µg/mL). The monoclonal anti-ceramide antibody (33) (clone MID 15B4; MilliporeSigma) used recognizes free and bound ceramides without cross-reactivity to other lipids. Cells were resuspended in 2% FBS/PBS/2 mmol/L EDTA for flow cytometry. Human islets were dispersed using 0.25% trypsin and neutralized with 10% PBS or by using enzyme-free cell dissociation buffer and filtered, fixed, and stained as above. Dead cells were excluded by forward scatter and side scatter gating, and doublets were removed using forward scatter height versus forward scatter area gating. A total of 10,000 events were analyzed per sample using FlowJo (Becton Dickinson). EVs were captured using immunoaffinity, stained with anti-ceramide antibody or isotype control (1 µg/mL for 1 h at 4°C), washed, stained with fluorescent secondary antibody (1 µg/mL, 30 min, 4°C), washed three times, and resuspended in 500 µL of bead wash buffer for analysis.

EV Isolation

EVs were isolated using ultracentrifugation, affinity purification using biotinylated tetraspanin antibodies (CD9, CD63, CD81), or size exclusion chromatography (SEC). For ultracentrifugation, conditioned media were spun at 3,000g for 10 min, filtered (0.22-µm filter), then centrifuged at 120,000g for 1.5 h at 4°C. Pelleted EVs were washed, respun, and resuspended in 100 µL of filtered PBS.

For immunoaffinity isolation, filtered media were concentrated and EVs captured by magnetic streptavidin beads coupled with biotinylated tetraspanin or ceramide antibodies (Exo-Flow Capture Kits). Captured EVs were stained with Exo-FITC (System Bioscience) and quantified using flow cytometry. Human plasma EVs were purified by SEC using qEVsingle/35-nm columns.

EV validation was performed by nanoparticle tracking analysis (NTA) (ZetaView). Transmission electron microscopy (TEM) was conducted by the Electronic Microscopy Core at the University of Nebraska Medical Center.

EV Treatment of Cells

INS-1 scramble cells and INS-1 nSMase2 shRNA cells were exposed to 24 h of IL-1β (5 ng/mL) or left untreated. A total of 5 × 10⁸/mL small EVs (isolated by ultracentrifugation, quantified using NanoSight LM10) was applied to wild-type INS-1 cells for 48 h.

Mass Spectrometry Analysis

EV pellets were resuspended in water and submitted to liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based lipidomics analysis as previously described (34), with eluting lipids submitted to data-dependent acquisition on an orbitrap mass spectrometer (Orbitrap Lumos; Thermo Fisher Scientific). Lipid species were identified, aligned, and quantified with MS-DIAL version 4.92 (35). Data preprocessing and ANOVA statistical analyses were performed using the pmartR package in R.

RNA Sequencing

Small RNA was extracted using QIAzol with on-column DNase digestion. Small RNA sequencing was performed on a NextSeq 500 platform, and total RNA sequencing was performed on the Illumina NovaSeq 6000 platform using the Illumina Total RNA library preparation protocol by the Indiana University (IU) Center for Medical Genomics (36). Differential expression was analyzed with edgeR (37), and network analysis was performed using QIAGEN Ingenuity Pathway Analysis (IPA).

Human Samples

Deidentified plasma from the IU Center for Diabetes and Metabolic Diseases biobank was used to test samples from children with type 1 diabetes or age-, sex-, and BMI-matched control children with euglycemia. All protocols were approved by the IU Institutional Review Board, and informed consent or assent was obtained in accordance with the Declaration of Helsinki.

Statistical Analysis

Data are plotted as mean ± SE. Statistical significance was determined using parametric or nonparametric testing in GraphPad Prism as indicated by Kolmogorov-Smirnov normality testing. P ≤ 0.05 was considered statistically significant. Participant sex was considered as part of the study matching design for human samples with matched control samples.

Data and Resource Availability

Data and resource sharing are available upon reasonable request to the authors.

Results

β-Cell nSMase2 and Ceramide, as Well as β-Cell EV Ceramide Content, Are Increased Under Conditions of Proinflammatory Cytokine Exposure

To examine the expression of sphingomyelin phosphodiesterase 3 (Smpd3), which encodes nSMase2, we first examined published data on β-cell RNA expression from the mouse islet single-cell atlas (30). Grouping by disease state, shown in Fig. 1A, identified a significant (P < 0.01) upregulation of Smpd3 in β-cells from the nonobese diabetic (NOD) mouse model of T1D. A smaller increase appeared to be present in β-cells from a model of T2D, but this was not statistically significant. Based on these findings, we tested whether proinflammatory cytokine exposure directly induced increases in nSMase2 by quantifying changes in protein levels. Using immunoblot and flow cytometry (Fig. 1B and C), we observed that 24-h IL-1β treatment of INS-1 cells increased nSMase2 expression. In concert with changes in nSMase2, IL-1β treatment significantly increased both total cellular ceramide expression (Fig. 1D) and cell surface ceramide expression (Fig. 1E). Similar patterns were also present in MIN6 cells after 24-h treatment with a proinflammatory cytokine mix of IL-1β, TNF-α, and IFN-γ (Supplementary Fig. 1). Findings were also verified in human islets treated with cytokine mix (Fig. 1F and G).

Figure 1.

β-Cell nSMase2 expression and ceramide production are increased in concert under conditions of T1D and cytokine exposure. A: Violin plots of β-cell Smpd3 expression in control, T1D, and T2D models show a significant upregulation in β-cells from the NOD mouse model of T1D (P < 0.01). B–E: In INS-1 β-cells, 24-h IL-1β increases nSMase2 as measured via immunoblot (B) (with quantification as shown) or flow cytometry staining (C), as well as total cellular ceramides (D) and surface ceramide flow cytometry staining (E). INS-1 experiments were performed using 3–11 biological replicates, as plotted. Data in C and E were compared using Student t tests, and data in D were compared using Kruskal-Wallis ANOVA with Tukey adjustment for multiple comparisons. F and G: Human islets were treated with a 48-h mix of IL-1β (5 ng/mL), TNF-α (10 ng/mL), and IFN-γ (100 ng/mL), then dispersed for flow cytometry–based quantification of nSMase2 (F) and ceramides (G). Data are shown for five unique donors for whom samples were tested in triplicate, and changes were compared using a Student t test. Significant differences were also present when using a paired t test to compare mean donor values between control and untreated islets (n = 5 per group). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001. Cyto, cytokine mix; MFI, mean fluorescence intensity; Veh, vehicle.

To understand whether cytokine-induced increases in β-cell nSMase2 and ceramide were also associated with increases in EV ceramide, we performed tetraspanin immunoaffinity-based small EV pulldowns and quantified INS-1 and human islet small EV ceramides using flow cytometry. Small EV isolations were validated using TEM and NTA (Fig. 2A and B) (38). For both INS-1 cells and human islets, treatment with IL-1β or cytokine mix also increased small EV total ceramide content (Fig. 2C and D). We also performed mass spectrometry on cytokine-treated human islet EVs and found significant upregulation of nine ceramide species with cytokine treatment (Table 1).

Figure 2.

In concert with cellular nSMase2 and ceramide, cytokine treatment increases β-cell EV ceramides. A and B: β-Cell small EVs were isolated using bead-based tetraspanin antibody pulldown, eluted, washed, concentrated, and verified using TEM (direct magnification ×110,000) (A) and NTA (B). C and D: Flow cytometry–based staining for ceramide was performed to assess EV ceramide content in control and cytokine-treated INS-1 cells (C) and human islets (D). INS-1 experiments were performed using 12 biological replicates, as plotted, and compared using Kruskal-Wallis ANOVA with Tukey multiple comparisons test. Human islet samples were tested in triplicate, and mean values for 10 unique donors were compared using a paired t test. Data are mean ± SEM. **P < 0.01, ***P < 0.001. Cyto, cytokine mix; MFI, mean fluorescence intensity; Veh, vehicle.

Table 1.

Mass spectrometry analysis of small EVs from human islets showed upregulation of multiple ceramide species after 48-h treatment with proinflammatory cytokine mix

Species	m/z	Adduct	Retention time (min)	Fold change	P
Cer 34:2;2O|Cer 18:2;2O/16:0	518.49451	[M + H − H2O]+	13.115	1.15	0.0163
Cer 40:1;2O|Cer 18:1;2O/22:0	622.61469	[M + H]+	17.821	1.15	0.0232
Cer 40:2;2O|Cer 18:2;2O/22:0	602.58997	[M + H − H2O]+	17.259	1.16	0.0403
Cer 41:0;2O|Cer 18:0;2O/23:0	638.64691	[M + H]+	18.303	1.21	0.0223
Cer 42:0;2O|Cer 18:0;2O/24:0	652.66278	[M + H]+	18.576	1.34	0.0421
Cer 42:1;2O|Cer 18:0;2O/24:1	650.64709	[M + H]+	17.983	1.23	0.0245
Cer 42:2;2O|Cer 18:1;2O/24:1	630.61969	[M + H − H2O]+	17.773	1.13	0.0110
Cer 42:3;2O|Cer 18:2;2O/24:1	628.60382	[M + H − H2O]+	17.239	1.27	0.0310
Cer 42:3;3O|Cer 18:1;2O/24:2;O	644.59631	[M + H − H2O]+	17.824	1.14	0.0199

Human islets (n = 3) were exposed to a cytokine mixture (5 ng/mL IL-1β, 100 ng/mL IFN-γ) for 48 h. Ceramides were analyzed via mass spectrometry, and 24 unique species were identified. Species with significant differences between groups are shown.

Direct Manipulation of Cellular nSMase2 Impacts EV Ceramide Content

Next, we aimed to directly test the impact of increases in nSMase2 expression on ceramides in cellular and small EV ceramide content. First, we exposed INS-1 cells to 24-h treatment with CAPE, a chemical inducer of nSMase2 activity and intracellular ceramide accumulation (39). CAPE treatment increased cellular ceramides both at baseline and in response to IL-1β treatment (Fig. 3A). In parallel, treatment increased the total small EV number (Fig. 3B), as well as small EV ceramide content (Fig. 3C), suggesting that nSMase2-based increases in small EV ceramides may not only yield increases in individual EV ceramide cargo, but also increases in ceramide-enriched small EV populations.

Figure 3.

Modulation of nSMase2 activity impacts β-cell ceramide-enriched EV generation. A–C: INS-1 β-cells were treated with 24 h of CAPE (5 µmol/L), which increased cell ceramides (A), total EV secretion (B), and EV ceramide staining (C) with or without concurrent cytokine treatment. D: Stable nSMase2 KD INS-1 β-cells were developed using rat nSMase2 shRNA to reduce cellular nSMase2 levels at baseline and in response to cytokines (10 ng/mL IL-1β). E: nSMase2 KD also reduced cytokine-induced increases in cellular ceramide staining. F and G: nSMase2 KD reduced total EV secretion and drastically reduced EV ceramide. H: Raptinal treatment to directly induce apoptosis increased EV ceramide content in mouse islets. I: ZVAD treatment to inhibit caspase-induced cell death abrogated cytokine-induced increases in small EV ceramide. J: Treatment with 5 µmol/L GSK-872 for 24 h to inhibit RIPK3-induced cell death also abrogated cytokine-induced increases in small EV ceramide. K–M: INS-1 β-cells were treated with the diabetes therapeutic and GLP-1 receptor agonist exendin-4 at 20 nmol/L for 24 h. Exendin-4 treatment significantly abrogated IL-1β–induced cellular nSMase2 (K) and cellular ceramide staining (L). Exendin-4 also significantly decreased EV ceramide staining (M). Data in A, I, and J were compared using one-way ANOVA with Tukey multiple comparisons test, while a nonparametric ANOVA with Dunn multiple comparisons test was applied to K–M. Two-way ANOVA with Tukey multiple comparisons test was used to compare data in D and E. Data in B were tested with a Student t test, while C, F, G, and M were compared using a nonparametric Mann-Whitney test. Data are mean ± SEM (n = 3 biological replicates for NTA experiments and nSMase2 KD validation and 6–12 biological replicates for other panels, as plotted). *P < 0.05, **P < 0.01, ***P < 0.001. EX, exendin-4; MFI, mean fluorescence intensity; SCR, scramble.

To assess the impact of reduced nSMase2 activity, we performed genetic KD of nSMase2 in INS-1 cells using shRNA (Fig. 3D). Compared with scrambled shRNA control cells, phenotypic analysis of KD cells showed decreased cell death and increased basal insulin secretion, but similar insulin secretion in response to high glucose (Supplementary Fig. 2). RNA sequencing identified increases in multiple genes associated with β-cell identity and function (Supplementary Table 4). nSMase2 KD cells exhibited reduced expression of nSMase2 both at baseline and in response to cytokine treatment (Fig. 3D). Consistent with a direct effect of nSMase2 activity on cytokine-induced increases in cellular ceramides, cellular ceramides were not reduced at baseline, but cytokine-induced increases in cellular ceramides were abrogated in nSMase2 KD cells (Fig. 3E). In contrast to CAPE-induced nSMase2 activation, nSMase2 KD yielded a reduced total small EV number (Fig. 3F). This was associated with dramatic reductions in ceramide staining in small EV ceramides on flow cytometry (Fig. 3G), again supporting the idea that nSMase2 is linked to generation of ceramide-enriched small EV populations.

Modulation of Other Intrinsic Islet Signaling Pathways Also Impacts β-Cell EV Ceramide Content

Given that inflammatory stress causes β-cell death, which has also been linked to β-cell ceramides, we asked whether modulation of cell death would impact cytokine-induced changes in β-cell EV ceramide. To test this, we treated cells with raptinal to directly induce cell apoptosis. Here, we also observed an increase in EV ceramide (Fig. 3H). In contrast, cotreatment of cells with IL-1β and ZVAD to inhibit caspase activity (Fig. 3I) abrogated cytokine-induced increases in EV ceramide. Cotreatment with IL-1β and GSK-872 to inhibit RIP3 kinase also abrogated cytokine-induced increases in EV ceramide (Fig. 3J). In aggregate, these data suggest that increases in nSMase2 and cell death pathways contribute to cytokine-induced increases in EV ceramide.

Next, we applied treatment with exendin-4, a glucagon-like peptide 1 (GLP-1) agonist, which in addition to positive impacts on β-cell health and potentiation of insulin secretion, has been shown to inhibit activation of de novo ceramide synthesis (40,41). As shown in Fig. 3K and L, exendin-4 treatment did not impact INS-1 cellular nSMase2 and ceramide content at baseline and showed some borderline effects to abrogate IL-1β–induced increases in cellular nSMase2 and ceramide content. However, exendin-4 treatment significantly reduced EV ceramide content (Fig. 3M).

To test the impacts of activation of other β-cell intrinsic stress pathways that would be expected to be activated by cytokine exposure, we also exposed INS-1 cells to 24 h of tunicamycin or thapsigargin to model ER stress, or doxorubicin to induce DNA damage (validation of stress pathways in Supplementary Fig. 4). Tunicamycin (1 µmol/L) (Fig. 4A–C), thapsigargin (5 nmol/L) (Fig. 4D–F), and doxorubicin (50 nmol/L) (Fig. 4G–I) all showed similar trends toward increased cellular nSMase2, cellular and EV ceramides, as well as increased cell death (Supplementary Fig. 5).

Figure 4.

Tunicamycin, thapsigargin, and doxorubicin treatment increase β-cell nSMase2 and ceramide production in INS-1 cells. A–I: In INS-1 cells, 24-h exposure to tunicamycin (1 µmol/L) (A–C), thapsigargin (5 nmol/L) (D–F), or doxorubicin (50 nmol/L) (G–I) increased nSMase2 (A, D, and G) and ceramides (B, E, and H), as well as ceramides in EVs (C, F, and I). EVs were isolated by immunoaffinity using tetraspanin antibodies. Ceramide content on EVs was determined by flow cytometry using ceramide antibody. INS-1 experiments were performed using 3–12 biological replicates, as plotted, and compared using Student t test (A–D, and I) or nonparametric Mann-Whitney test (E–H). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001. Doxo, doxorubicin; MFI, mean fluorescence intensity; TG, thapsigargin; TUN, tunicamycin.

nSMase2 KD Abrogates Impact of EVs on GSIS After Cytokine Exposure

To understand whether nSMase2-related EV content could spread impacts of cytokine treatment to other cells, we obtained EVs from cytokine-exposed or -unexposed parent cells from either the nSMase2 KD line or scramble shRNA controls. Wild-type recipient INS-1 cells were then exposed to 48-h EV treatment using EVs from each of the four groups, followed by analysis of static GSIS in recipient cells. Compared with vehicle-exposed scramble shRNA control parent cells, EV treatment from cytokine-exposed scramble shRNA control parent cells yielded a negative impact on peak GSIS (Fig. 5A). To test whether nSMase2-derived ceramides were involved in this EV effect to reduce GSIS, this experiment was repeated using nSMase2 KD parent cell EVs and showed an abrogation of the negative impact of cytokine exposure (Fig. 5B).

Figure 5.

β-Cell cytokine-induced ceramide-enriched EV subpopulations impact recipient cell GSIS and carry differential miRNA cargo compared with global EV populations. A and B: INS-1 scramble control (A) or nSMase2 KD cells (B) were exposed to IL-1β or vehicle control for 24 h. Small EVs were isolated from each well of parent cells, then used for treatment of wild-type recipient INS cells at a 5 × 10⁸/mL concentration for 48 h. GSIS was measured in wild-type cells. Three biological replicates per group were compared using two-way ANOVA with Tukey posttest. C and D: Wild-type INS-1 cells were exposed to IL-1β or vehicle control for 24 h. Ceramide-enriched EVs and tetraspanin pulldown EVs were isolated using streptavidin magnetic beads and biotinylated antibodies against ceramide or tetraspanins. EV RNA was isolated and subjected to miRNA sequencing. C: One-dimensional plot of multidimensional scaling on miRNA showed baseline differences in miRNA expression profiles between ceramide-enriched EVs and global EVs. D: The highest scored network was generated using QIAGEN IPA with a set of cutoffs (down and up 1.5-fold of expression fold change (FC), P < 0.05 and FDR < 0.05). The network suggested that differential miRNAs were involved in a series of network nodes, including insulin. dim, dimension; HG, high glucose; LG, low glucose; Scr, scramble; Veh, vehicle.

Ceramide-Enriched EVs House Distinct miRNA Cargo Compared With Global EV Populations

We next sought to explore the idea that distinct small EV populations enriched for membrane ceramide exist. To test this possibility, we performed immunoaffinity-based small EV pulldowns using antibodies to tetraspanins (CD9, CD63, CD81, global EV membrane proteins) (42) or EV pulldowns using antibodies to ceramide, and performed bulk small RNA sequencing to assay differences in EV cargo with and without IL-1β treatment. To validate the EV pulldown using ceramide antibodies, we captured EVs on magnetic streptavidin microbeads by using ceramide antibody and IgM isotype control, then EVs were stained with Exo-FITC and analyzed by flow cytometry (Supplementary Fig. 6).

Here, multidimensional scaling (shown in Fig. 5C) comparing the small EVs isolated via tetraspanin pulldown versus EVs isolated via ceramide antibody pulldown suggested that at baseline, significant differences existed in cargo from small EVs isolated via tetraspanin versus ceramide antibody pulldown, again supporting the idea of ceramide-enriched small EV subpopulations. Changes in miRNA content induced by IL-1β treatment were overall similar within both subpopulations. As shown in Table 2, ceramide-enriched EVs exhibited significant increases or decreases in expression of at least a 1.5-fold difference of 69 miRNAs (P < 0.05 and false discovery rate [FDR] < 0.05) compared with global EV populations, with 34 downregulated and 35 upregulated miRNAs. IPA identified three highly scored networks highly impacted by ceramide-enriched EV miRNAs. The top scoring network included miRNAs and genes involved in insulin signaling (Fig. 5D). As shown in Supplementary Figs. 7 and 8, the other highly scored networks included Ago2 (argonaute RISC catalytic component 2), Pax3 (paired box 3), Foxo1 (forkhead boxO1), Abca1 (ATP-binding cassette subfamily A member 1), Braf (B-Raf proto-oncogene, serine/threonine kinase), Camta1 (calmodulin-binding transcription activator 1), Mtss1 (MTSS I-BAR domain containing 1), Nr0b2 (nuclear receptor subfamily 0 group B member 2), and Nr3c2 (nuclear receptor subfamily 3 group C member 2).

Table 2.

Differential miRNA profiles of ceramide-enriched EVs versus global EVs from INS-1 cells

MiRBase_ID	Fold change	P	FDR
rno-miR-500-3p	6.74	1.77937E-07	2.45552E-06
rno-miR-466b-5p	4.01	7.35357E-13	2.25509E-11
rno-miR-708-5p	3.07	7.65659E-19	1.05661E-16
rno-miR-5132-5p	2.97	0.000317306	0.001509938
rno-miR-143-3p	2.90	0.002745723	0.009021662
rno-miR-708-3p	2.80	0.000116241	0.000641648
rno-miR-344b-1-3p	2.49	0.000286652	0.001387997
rno-let-7f-5p	2.45	1.32009E-24	3.64344E-22
rno-miR-182	2.32	4.11349E-18	2.83831E-16
rno-miR-802-3p	2.27	1.35358E-05	9.83129E-05
rno-miR-298-5p	2.27	0.003120973	0.010016147
rno-miR-30b-5p	2.23	8.44323E-06	6.47314E-05
rno-miR-802-5p	2.18	1.87356E-05	0.000132591
rno-miR-328b-3p	2.13	0.00090559	0.003717219
rno-miR-138-5p	2.09	0.000234187	0.001219537
rno-miR-99a-5p	2.03	0.000746181	0.003126204
rno-miR-96-5p	2.02	9.44848E-13	2.60778E-11
rno-miR-296-3p	2.00	0.00315825	0.010019275
rno-miR-183-3p	1.99	0.001413566	0.004938534
rno-miR-676	1.98	1.55019E-16	7.13088E-15
rno-miR-351-5p	1.92	3.31263E-07	3.97515E-06
rno-miR-504	1.89	1.27883E-07	1.85767E-06
rno-miR-125a-5p	1.84	9.25012E-14	3.19129E-12
rno-miR-31a-5p	1.84	0.00409481	0.012842815
rno-let-7g-5p	1.81	5.79682E-12	1.45448E-10
rno-miR-322-5p	1.80	3.49859E-05	0.000224561
rno-miR-187-3p	1.71	4.90326E-10	1.12775E-08
rno-let-7d-3p	1.70	2.16393E-09	3.98162E-08
rno-miR-99b-5p	1.62	1.02577E-06	9.43709E-06
rno-let-7a-5p	1.60	7.73178E-10	1.52427E-08
rno-miR-103-3p	1.56	6.26766E-07	6.65336E-06
rno-miR-455-5p	1.55	0.00049492	0.002276633
rno-miR-532-5p	1.52	0.000274855	0.001354642
rno-miR-339-5p	1.51	0.000699707	0.003017486
rno-miR-30c-5p	1.51	0.001454632	0.00501848
rno-miR-101a-3p	−1.51	6.35543E-05	0.000389799
rno-miR-1839-5p	−1.51	1.31608E-08	2.01799E-07
rno-miR-218a-5p	−1.52	0.002460586	0.008281974
rno-miR-340-5p	−1.53	5.17357E-05	0.000324524
rno-let-7c-5p	−1.53	2.40965E-07	3.02301E-06
rno-miR-9a-5p	−1.54	0.000915837	0.003717219
rno-miR-361-3p	−1.56	2.26749E-05	0.000156457
rno-miR-30a-3p	−1.57	0.017033401	0.04435112
rno-miR-423-3p	−1.57	3.47116E-05	0.000224561
rno-miR-34c-5p	−1.59	0.001049111	0.004136496
rno-miR-1224	−1.60	0.000149112	0.000806961
rno-miR-18a-5p	−1.66	5.92065E-07	6.5364E-06
rno-miR-582-3p	−1.68	0.001497428	0.005102347
rno-miR-181c-3p	−1.70	0.011121444	0.030391273
rno-miR-328a-3p	−1.71	7.08683E-10	1.50459E-08
rno-miR-203a-3p	−1.76	4.50997E-09	7.7797E-08
rno-miR-9a-3p	−1.83	0.001224098	0.004454987
rno-miR-212-3p	−2.00	0.004738426	0.014062427
rno-miR-10a-5p	−2.01	0.001183534	0.00441426
rno-miR-196a-5p	−2.12	1.00787E-14	3.97387E-13
rno-miR-140-5p	−2.24	0.002499395	0.008311242
rno-miR-335	−2.25	3.30566E-18	2.83831E-16
rno-miR-122-5p	−2.39	6.89935E-07	7.05267E-06
rno-miR-33-5p	−2.48	0.013608189	0.036464663
rno-miR-382-5p	−2.58	1.25009E-08	2.01799E-07
rno-miR-300-3p	−2.58	0.015536459	0.041231373
rno-miR-338-3p	−2.81	0.000100115	0.000575658
rno-miR-134-5p	−3.01	0.000390699	0.001827678
rno-miR-7a-1-3p	−3.29	0.007183249	0.020438937
rno-miR-709	−3.87	1.33398E-05	9.83129E-05
rno-miR-146a-5p	−8.37	1.91999E-17	1.05983E-15
rno-miR-1b	−9.93	5.10324E-07	5.86872E-06
rno-miR-135a-5p	−10.39	2.68233E-06	2.37494E-05
rno-miR-100-5p	−16.23	3.87346E-06	3.14434E-05

A total of 69 miRNAs exhibited a >1.5-fold difference (up- or downregulated) between EV populations (P < 0.05 and FDR < 0.05).

Multiple Plasma EV Ceramide Species Are Increased in Children With T1D Compared With Euglycemic Pediatric Samples

Finally, we asked whether ceramide EV content was altered in the circulation of children with T1D. We obtained previously biobanked plasma samples from 26 children with new-onset T1D and matched these by age, sex, and BMI to nondiabetic control plasma samples (demographic characteristics in Supplementary Table 6). Small EVs were isolated using SEC (Supplementary Fig. 9). Although we did not detect differences in total plasma EVs using our total ceramide antibody, mass spectrometry analysis identified increases in six unique ceramide species in plasma EVs from children with diabetes compared with those without diabetes (Table 3).

Table 3.

Plasma small EVs from humans with T1D show increases in multiple ceramide species compared with plasma from control humans without diabetes

Species	m/z	Adduct	Retention time (min)	Fold change	P
Cer 40:0;2O|Cer 18:0;2O/22:0	624.62952	[M + H]+	18.022	1.56	0.0006
Cer 40:0;3O|Cer 17:0;2O/23:0;O	640.62665	[M + H]+	17.434	1.30	0.0214
Cer 41:0;2O|Cer 18:0;2O/23:0	638.6452	[M + H]+	18.303	1.58	0.0010
Cer 41:0;3O|Cer 17:0;2O/24:0;O	654.64062	[M + H]+	17.742	1.35	0.0094
Cer 41:2;2O|Cer 18:2;2O/23:0	616.6051	[M + H − H2O]+	17.587	0.58	0.0394
Cer 42:0;2O|Cer 18:0;2O/24:0	652.66272	[M + H]+	18.575	1.58	0.0008
Cer 42:0;3O|Cer 18:0;2O/24:0;O	650.64697	[M + H − H2O]+	17.966	1.47	0.0189
Cer 42:2;2O|Cer 18:2;2O/24:0	648.63184	[M + H]+	18.056	0.66	0.0365

A total of 48 human samples (26 per group) were evaluated. Ceramides were analyzed via mass spectrometry, and 31 unique species were identified. Species with significant differences between groups are shown.

Discussion

Findings from our laboratory and others have identified changes in β-cell EV cargo under disease conditions and the potential of islet EVs to serve as paracrine effectors, but mechanistic etiologies underlying β-cell EV generation and content are poorly understood (6,7,9–12,43,44). Here, we show that β-cell inflammatory stress yields increases in β-cell small EV populations enriched for the bioactive lipid ceramide housing distinct miRNA cargo. Cytokine-induced increases in β-cell EV ceramide content were mediated by changes in the expression and activity of nSMase2. Furthermore, inhibition of nSMase2 abrogated functional impacts of cytokine-exposed β-cell EV transfer to surrounding β-cells.

Several studies have suggested that islet EVs have potential as paracrine effectors in the islet microenvironment. Cytokine treatment of β-cells to induce inflammatory stress induces physiologic changes in EV RNA and protein cargo (9–11,16,44). β-Cells or EVs from islets can display immunostimulatory properties for antigen-presenting cells and CD8⁺ T cells that are enhanced by cytokine treatment of parent cells (9,16,45). Coculture of cytokine-treated MIN6 cells led to an miRNA-dependent increase in recipient cell apoptosis (10). CXCL10 on the surface of cytokine-treated islet EVs reduced GSIS and expression of identity genes in recipient β-cells, while inducing upregulation of chemokine signaling and proteins associated with antigen presentation (16). T-lymphocyte EVs can transfer miRNAs to β-cells that induce apoptosis and chemokine signaling (9,46). Islet EVs also stimulate endothelial cell angiogenesis (43) and have the potential to interact with other β-cells.

Small RNA sequencing identified a distinct set of miRNAs in ceramide-enriched EVs that were linked to key pathways involved in β-cell function and identity, including insulin signaling, FOXO1 (47), and miRNA action (48). Our observation that nSMase2 activation is associated with changes in EV recipient cell GSIS suggests that certain EV populations or physiologic conditions may potentiate the paracrine effects of β-cell EVs on surrounding β-cells. Future studies will assay the impacts of ceramide-enriched β-cell EV populations on other cell types in the islet microenvironment.

In other systems, nSMase2-mediated ceramide synthesis promotes MVE membrane ceramide accumulation, leading to inward vesicle budding and ultimately, release of miRNA-dense, ceramide-enriched EVs (17,19). Specifically, nSMase2 can be required for the release and transfer of certain exosomal miRNAs (17,18). Additionally, bioactive EV lipids themselves may potentially be transferred to recipient cells (49). Other mechanisms, such as the ESCRT (endosomal sorting complex required for transport)–dependent pathway, also regulate EV formation and likely contribute to β-cell EV formation (50). However, given prior descriptions of ceramide in β-cell stress pathways (22–28), a role for changes in stress response–induced regulation of β-cell EV content is logical, and our findings suggest that this is in part related to increased nSMase2 expression. Future work will interrogate the role of other regulators of cellular ceramide content and localization (51) and other EV biogenesis pathways.

Our finding that nSMase2 KD cells abrogated cytokine-induced increases in EV ceramide could point to a mechanistic link between nSMase2-based ceramide production, β-cell death and function, and EV ceramide content. This not only is consistent with prior reported relationships between ceramide accumulation and β-cell death, but also suggests that ceramide in EVs could serve as a propagation signal in the context of death induced by β-cell stressors (52). Experiments also showed that direct manipulation of cell death pathways impacted EV ceramide. Importantly, we only studied small EVs, which should exclude contributions of larger apoptotic bodies formed during cell death. The improvement in EV ceramide in response to exendin-4 is intriguing, consistent with studies showing that the GLP-1 receptor agonist liraglutide decreased total plasma ceramides in individuals with T2D (53), and may represent an additional therapeutic effect of this drug class.

This work has some limitations. EV isolations are well recognized to exhibit heterogeneity (38). Technical challenges associated with adequate EV isolations limited our ability to perform all mechanistic studies in primary cells. We tried to address this by isolating by size and specific membrane components to more selectively isolate subpopulations, and validating isolations using multiple techniques, with key experiments validated using human cells (5). Plasma sample analysis was limited by available blood volume and included systemic EVs from multiple cell types; therefore, the relative contribution of β-cell EVs was likely limited, especially considering reduced β-cell mass in children with T1D.

In aggregate, this work provides substantial novel mechanistic data informing the understanding of nSMase2-based generation of increased numbers of a subpopulation of ceramide-enriched β-cell EVs under conditions of inflammatory stress. Compared with global EVs, ceramide-enriched EVs contained distinct miRNA cargo, and blocking nSMase2 activity abrogated effects of β-cell EVs from cytokine-treated parent cells on GSIS, suggesting a role in paracrine effects of β-cell EVs. Targeting and modulation of this process may ultimately allow for novel therapeutics aimed at improving β-cell health in diabetes.

This article contains supplementary material online at https://doi.org/10.2337/figshare.29661020.