Thrombospondin-1, CD47, and SIRPα display cell-specific molecular signatures in human islets and pancreata

Abstract

Thrombospondin-1 (TSP1) is a secreted protein minimally expressed in health but increased in disease and age. TSP1 binds to the cell membrane receptor CD47, which itself engages signal regulatory protein α (SIRPα), and the latter creates a checkpoint for immune activation. Individuals with cancer administered checkpoint-blocking molecules developed insulin-dependent diabetes. Relevant to this, CD47 blocking antibodies and SIRPα fusion proteins are in clinical trials. We characterized the molecular signature of TSP1, CD47, and SIRPα in human islets and pancreata. Fresh islets and pancreatic tissue from nondiabetic individuals were obtained. The expression of THBS1, CD47, and SIRPA was determined using single-cell mRNA sequencing, immunofluorescence microscopy, Western blot, and flow cytometry. Islets were exposed to diabetes-affiliated inflammatory cytokines and changes in protein expression were determined. CD47 mRNA was expressed in all islet cell types. THBS1 mRNA was restricted primarily to endothelial and mesenchymal cells, whereas SIRPA mRNA was found mostly in macrophages. Immunofluorescence staining showed CD47 protein expressed by β cells and present in the exocrine pancreas. TSP1 and SIRPα proteins were not seen in islets or the exocrine pancreas. Western blot and flow cytometry confirmed immunofluorescent expression patterns. Importantly, human islets produced substantial quantities of secreted TSP1. Human pancreatic exocrine and endocrine tissue expressed CD47, whereas fresh islets displayed cell surface CD47 and secreted TSP1 at baseline and in inflammation. These findings suggest unexpected effects on islets from agents that intersect TSP1-CD47-SIRPα.

NEW & NOTEWORTHY CD47 is a cell surface receptor with two primary ligands, soluble thrombospondin-1 (TSP1) and cell surface signal regulatory protein alpha (SIRPα). Both interactions provide checkpoints for immune cell activity. We determined that fresh human islets display CD47 and secrete TSP1. However, human islet endocrine cells lack SIRPα. These gene signatures are likely important given the increasing use of CD47 and SIRPα blocking molecules in individuals with cancer.

INTRODUCTION

Thrombospondin-1 (TSP1, THBS1) is a secreted protein that interacts with several cell membrane receptors and the extracellular matrix. TSP1 binds to cell receptor CD47 to alter self-renewal (1), metabolism (2), angiogenesis (3), and blood flow (4, 5). Skeletal muscle from Cd47-null mice showed increased levels of mitochondrial genes, including cytochrome b and c, nuclear respiratory factor 1, and peroxisome proliferator-activated receptor-gamma coactivator-1 alpha (6). Mitochondria from Cd47-null mice produced less reactive oxygen species, and null mice had more exercise capacity than wild-type mice (6). With aging, Cd47-null mice gained less weight and had less adiposity (6, 7). Furthermore, Thbs1-null mice on a high-fat diet were more glucose tolerant and insulin sensitive (8). Thus, TSP1-CD47 signaling appears to impact metabolism and glucose balance negatively.

The clinical role of these genes in metabolism is evolving. Individuals with type 1 diabetes (T1D) had elevated plasma TSP1 compared with nondiabetics (136.6 ± 14.2 ng/mL vs. 91.2 ± 14.3 ng/mL) (9) and better control of blood glucose was associated with lower plasma TSP1 (9). TSP1 was also elevated in individuals with type 2 diabetes and correlated with cardiovascular disease (10). Moreover, human islets cultured in high glucose showed increased TSP1 mRNA (11).

In addition to TSP1, CD47 binds to signal regulatory protein α (SIRPα) (12). In this capacity, it limited phagocytosis of transfused red blood cells (13) and provided a check on immune cell activity (14). Individuals treated with immune checkpoint-blocking antibodies against PD-1 or PD-L1 developed endocrinopathies (15) including insulin-dependent diabetes (16–18). Beyond the immune system, SIRPα was inferred to be expressed in β cells to support insulin secretion and cell survival (19, 20).

Herein, we characterized THBS1, CD47, and SIRPA in freshly gathered human islets and pancreata from nondiabetic individuals and in islets before and after exposure to inflammatory cytokines. We found that CD47 transcript and protein were expressed in all islet and nonendocrine islet cells, that THBS1 and SIRPA transcript were restricted primarily to endothelial cells and macrophages, and that TSP1 and SIRPα proteins were undetectable anywhere in the pancreas. Human islets showed cell surface CD47 and secreted soluble TSP1, both of which were maintained in the presence of inflammatory cytokines. These findings may be relevant as clinical trials employing CD47 blocking antibodies and SIRPα fusion molecules are underway (21).

MATERIAL AND METHODS

Cell Lines

THP-1 human leukemia monocytic cells were obtained from American Tissue Culture Collection (ATCC TIB-202, Manassas, VA) and cultivated in RPMI-1640 medium with 2-mercaptoethanol (0.05 mM) and 10% fetal bovine serum. Recombinant human TSP1 (Cat. No. 3074-TH) was from R&D Systems (Minneapolis, MN).

Human Organs

Pancreata from deceased donors were provided by OneLegacy (Los Angeles, CA), stored in preservation solution on wet ice, and transported to a Good Manufacturing Practices facility at City of Hope National Medical Center (Duarte, CA) for processing. Isolated islets were also distributed by the Integrated Islet Distribution Program. Informed consent was obtained from donor next of kin. Organs were approved for research by the Institutional Research Board of City of Hope National Medical Center (IRB # 01046). Only organs from nondiabetic individuals (HbA1c <6.5%) were included. Deidentified donor demographic information is summarized in Table 1.

Table 1.

Donor demographic information and pancreas and islet processing details

Study	Donor ID	Age, yr	Sex	BMI, kg/m2	HbA1c, %	Warm Ischemia Time, h	Cold Ischemia Time, h	Islet Purity, %	Islet Viability, %	Total Culture Time, h
scRNA-seq	1180	57	F	23.6	4.4	0	7.0	85	95	27.9
scRNA-seq	1206	51	F	25.1	5.6	0	6.5	90	95	20.5
scRNA-seq	1231	54	F	26.8	5.0	0	11.6	98	95	16.5
IF	1107	45	M	26.5	4.6	0	7.0	80	94	76.4
IF	1172	20	M	26.4	5.5	0	6.7	48	97	52.9
IF	1143	21	M	27.1	5.7	0	5.8	70	97	19.7
WB	1199	63	F	26.8	5.5	25	5.0	80	92	51.2
WB	1186	33	M	28	4.8	0	11.1	68	79	97
WB	1228	50	F	30.5	4.9	0	14.1	70	94	76.2
WB	1229	26	M	29.2	5.4	0	8.6	62	94.7	37.1
WB FC	1232	45	F	25.2	4.9	23	9.7	73	97.8	92.9
WB	1236	50	M	21.5	5.8	14	6.5	90	96	93.6
WB FC	1237	25	M	23.7	4.9	0	8.2	88	97	24.9
WB	1238	34	M	31.8	5.6	0	6.2	85	95	40
FC	1239	37	M	31.3	5.9	0	6.5	85	98	43

BMI, body mass index; FC, flow cytometry; HBA1c; hemoglobin A1c; IF, immunofluorescent staining; sc-RNAseq, single-cell RNA sequencing; WB, Western blot.

Islet Isolation

Pancreata were debrided of fat and connective tissue and perfused via the pancreatic duct with collagenase/protease solution using an automatic perfusion system (Biorep Technologies, Miami Lakes, FL). Organs were cut into pieces and digested using a Ricordi digestion chamber (22). Islets were purified by density gradient centrifugation and a COBE 2991 cell processor system (Terumo BCT, Lakewood, CO) (23). Islet morphological scoring and viability were determined as published (24–26). Organ processing and islet quality information are summarized in Table 1.

Single-Cell mRNA Sequencing

Human islets (2,000 IEQ/donor) were dissociated into single cells using TrypLE Express containing 1 mM EDTA (ThermoFisher Scientific, Cat. No. 12605028) at 37°C for 12–15 min. Cells were washed twice with PBS containing 0.1% human serum albumin (Baxalta, NDC 0944-0493-01). Five to seven thousand cells from each donor preparation were used for droplet-based single-cell mRNA sequencing analysis on the 10x Genomics Chromium platform of 3′ Library and Gel bead Kit V3.1. mRNA sequencing was conducted using the Illumina NovaSeq6000 platform. Fastq files were generated by demultiplexing raw data using cellranger (v6.0.2) with reference (GRCh38.p12). The output matrix for each sample was imported for integration in the R statistical environment using the Seurat software package (v4.0.3). For quality control, events with >9,000 unique genes, <500 total number of molecules, and >20% reads that mapped to the mitochondrial genome were filtered out. Counts were then normalized (method = LogNormalize, scale factor = 1,000) and the top 2,000 most variable features selected. Data were then scaled, and principal component analysis was performed up to the top 20 components. The k-nearest neighbors of each cell were determined, and shared nearest-neighbor (SNN) graphs were constructed. The clusters were identified using the SNN modularity optimization-based clustering algorithm (resolution = 0.5). Uniform manifold approximation and projection dimensionality reduction on the first 20 principal components, annotation of identified clusters, and generation of plots were performed.

Immunofluorescence Staining

Formalin-fixed, paraffin-embedded pancreas tissue blocks were cut into 5-µm thick sections and deparaffinized using Histoclear (National Diagnostics, HS-200). Samples were dehydrated with an alcohol gradient, hydrated with deionized water for 5 min, treated with citrate-based antigen retrieval solution (Vector Laboratory, H-3300) for 40 min, blocked with blocking buffer (Invitrogen, 00-8120) for 20 min, and incubated with primary antibodies overnight at 4°C. Samples were washed with Super Sensitive Wash Buffer (Biogenex, HK583-5K) three times and treated with secondary antibodies at room temperature (RT) for 30 min. Samples were treated with FluoroShield with DAPI (Sigma, F6057) for 10 min, coverslips were applied, and maintained at 4°C. As a negative control, tissue sections were incubated with secondary antibodies alone (Supplemental Fig. S1). Antibody details are provided in Table 2. Tissue sections were cut from samples taken from the head of the pancreas.

Table 2.

Antibody information

Primary antibody (Ref.)	Assay	Company	Cat. No.	Animal Source	Dilution
Anti-Insulin (27)	IF	Abcam	ab195956	Guinea pig	1:200
Anti-CD47*	IF	Invitrogen	PA5-116827	Rabbit	1:200
Anti-TSP1 (28)	IF	Invitrogen	MA5-133398	Mouse	1:100
Anti-SIRPα (29)	IF	Abcam	ab191419	Rabbit	1:1,000
Anti-CD47	WB	Invitrogen	PA5-116827	Rabbit	1:1,000
Anti-TSP1	WB	Invitrogen	MA5-133398	Mouse	1:250
Anti-SIRPα	WB	Cell Signaling	13379	Rabbit	1:1,000
Anti-PD-L1	WB	Cell Signaling	13684S	Rabbit	1:1,000
Anti-β-actin	WB	Cell Signaling	3700	Mouse	1:1,000
Anti-CD47 BV605	FC	Biolegend	323120	Mouse	1:20
Anti-SIRPα PE	FC	Biolegend	372103	Mouse	1:20
Secondary antibody
Anti-Guinea pig Alexa 488	IF	Jackson	706-545-148	Donkey	1:200
Anti-Mouse Alexa 647	IF	Jackson	715-606-150	Donkey	1:200
Anti-Rabbit Alexa 647	IF	Jackson	715-606-152	Donkey	1:200
Anti-Mouse IgG HRP	WB	Cytiva	N931	Sheep	1:3,000
Anti-Rabbit IgG HRP	WB	Cytiva	N934	Donkey	1:3,000

FC, flow cytometry; IF, immunofluorescent staining; sc-RNAseq, single-cell RNA sequencing; WB, Western blot. *This antibody was validated by using U2 OS cell (IF), mouse testis tissue (IHC), and Western blot from manufacturer: https://www.thermofisher.com/antibody/product/CD47-Antibody-Polyclonal/PA5-116827.

Image Acquisition and Quantification

Stained tissue slides were examined with a confocal microscope (Zeiss LSM 900 with Airyscan 2). Images were acquired under a ×10, ×20, and ×40 objective lens and processed with Zen 2 Blue software. Analysis of fluorescence intensity was performed using ImageJ software (National Institutes of Health, Bethesda, MD). For each image, islets were selected using the polygon selection tool. The RGB Measure option from the Plugins/Analyze menu was set, and values of fluorescence intensity and area of interest were obtained. The fluorescence signal was expressed as intensity relative to the negative control. One slide from each donor organ (total of 3 donors) was analyzed. Three to four separate areas of the tissue sample were characterized for each slide. Five or more islets were counted in each area. Protein expression in nonislet tissue was obtained by subtracting the fluorescence intensity of islets from the fluorescence intensity of islets + nonislet tissue. The final tissue fluorescence signal was expressed as intensity relative to the negative control.

Flow Cytometry

Isolated islets were cultured overnight in CMRL 1066 CIT Modification culture medium supplemented with 0.5% human serum albumin and IGF-1 (0.1 µg/mL). Islets were dissociated with TrypLE (Gibco, 12605036) containing heparin (10 U/mL) for 12–15 min at 37 °C. Dispersed cells were incubated with Human BD Fc Block (BD Biosciences, 564220) for 10 min and LIVE/DEAD Fixable Violet Dead Cell Stain (ThermoFisher, L34963) for 15 min at RT. Cells were then incubated with fluorochrome-conjugated monoclonal or isotype control antibodies for 30 min at 4 °C. Flow cytometry was performed using an Attune Nxt Acoustic Focusing Flow Cytometer System (Thermo Fisher). Live single cells were gated, and the percentage of CD47- and SIRPα-positive cells was determined using FlowJo software (BD Biosciences). In other studies, human THP-1 cells, which express TSP1 (30), CD47 (31), and SIRPα (32), were dissociated and stained with fluorochrome-conjugated anti-CD47 and anti-SIRPα antibodies (Supplemental Fig. S2). Antibody details are summarized in Table 2.

Western Blot

Isolated islets were incubated overnight in CMRL 1066 CIT Modification culture medium (Corning, 98-304-CV) supplemented with 0.5% human serum albumin (Baxalta, NDC 0944-0493-01) and IGF-1 (0.1 µg/mL) (Repligen, 10-1011-125). Islets were incubated for 24 h in medium without or with a mixture of proinflammatory cytokines: interleukin-1 beta (IL-1β; 50 U/mL), interferon-gamma (IFN-γ; 1,000 U/mL), and tumor necrosis factor-alpha (TNF-α; 1,000 U/mL) (Peprotech 200-01B, 300-02, and 300-01A, respectively). Following culture, islets were lysed with RIPA buffer (Millipore, 20-188) containing protease and phosphatase inhibitor (Roche, 11836170001 and 04906845001). Lysate protein concentrations were determined with a Pierce Rapid Gold BCA Protein Assay kit (ThermoFisher, A53226). Samples were resolved by electrophoresis on 4–12% NuPAGE Bis-Tris gradient gels (Invitrogen, NP0336BOX), transferred to nitrocellulose membranes, and incubated with blocking buffer (Sigma, W0138) for 30 min at RT, followed by primary antibodies overnight at 4°C. Membranes were washed and incubated with secondary antibody conjugated with horseradish peroxidase for 1 h at RT. Chemiluminescence reagent (ThermoFisher, 34075) was used to detect protein bands. Imaging was performed using a Bio-Rad ChemiDoc Imaging system. Densitometry characterization was done with ImageJ software, and values were normalized to β-actin. Primary antibodies to TSP1, CD47, and SIRPα were validated via Western blotting using lysates of human THP-1 cells, fresh islets, and recombinant TSP1 protein (Supplemental Fig. S3). Complete full-sized Western blots are provided in Supplemental Fig. S4. Antibody details are summarized in Table 2.

Soluble TSP1

Human islets were cultured in standard islet medium (CMRL 1066 CIT Modification) supplemented with 0.5% human serum albumin and IGF-1 (0.1 µg/mL) with or without inflammatory cytokines (IL-1β, IFN-γ, and TNF-α) for 24 h. Fresh and conditioned medium was characterized for soluble human TSP1 protein employing a Luminex assay (Cat. No. HSP3MAG-63K-01, Millipore, Burlington, MA).

Statistics

Data were expressed as means ± SD. Mean values were compared using a paired Student’s t test. All reported P values were two-tailed and considered significant if <0.05. Analysis was performed, and figures were generated using GraphPad Prism version 9.0 (GraphPad, San Diego, CA).

RESULTS

Single-Cell Expression of CD47, THBS1, and SIRPA mRNA in Human Islets

Human islets were isolated from pancreata from nondiabetic individuals, and single-cell mRNA sequencing was performed. As shown in the uniform manifold approximation and projections (UMAPs), cell-type clustering was consistent between the three donor preparations (Fig. 1A). Marker gene analysis identified distinct cell types in each cluster (Fig. 1B). CD47 transcript was found expressed in all islet endocrine and nonendocrine cells (Fig. 1C). THBS1 transcript was not found in endocrine cells but was expressed mostly in endothelial and mesenchymal cells and to a lesser extent in acinar and ductal cells. SIRPA transcript was found predominantly in macrophages but not in endocrine cells (Fig. 1C). mRNA of the immune checkpoint genes programmed cell death-ligand 1 (PD-L1, CD274), programmed cell death protein 1 (PD-1, PDCD1), and cytotoxic T lymphocyte-associated protein 4 (CTLA4) was not found in islet cells (Fig. 1C). Beta-2 microglobulin (B2M) was expressed in all islet cells and human leukocyte antigen DR alpha (HLA-DRA) was found in macrophages (Fig. 1C).

Figure 1.

Single-cell expression of CD47, THBS1, and SIRPA mRNA in human islets. UMAP plots of cell clusters of each donor islet preparation. Graphs display donor cells, color-coded for each islet cell population (1,745 cells for donor 1, 4,334 cells for donor 2, and 5,441 cells for donor 3) (A). DotPlots show relative gene expression for each cell population shown on the y-axis [islet marker genes (B); CD47, THBS1, SIRPA, CD274, PDCD1, CTLA4, B2M, and HLA-DRA (C)]. DotPlot graphs display the combined data from all three donors. Dot size indicates the percentage of cells with detectable transcript and color indicates the average expression of each gene by z score. The UMAP and DotPlot graphs were generated using Seurat and RStudio (2022.02.2 Build 485 version). UMAP, uniform manifold approximation and projection.

Human β Cells Do Not Display Key Self-Renewal Transcription Factor mRNAs but Display Hypoxia-Inducible Factor mRNAs

We reported that TSP1-CD47 limited the essential self-renewal transcription factors POU5F1 (Oct3/4), SOX2, KLF4, and MYC (abbreviated OSKM) (1, 7, 33). However, it is unknown if this ensued in proliferative cells such as T cells and endothelial cells within human islets. Single-cell mRNA sequencing revealed that OSKM transcripts were restricted to mesenchymal cells (Supplemental Fig. S5). Beta and other islet endocrine cells did not show substantial OSKM transcripts. As well, islets did not show significant amounts of thrombospondin-2 (THBS2) mRNA. This is pertinent as TSP2 binds CD47, albeit as we showed with lower affinity, to alter cell function (34). Clinical T1D is typified by hypoxia (35, 36). Notably, THBS1 (37) and CD47 (38) are hypoxia-inducible factor (HIF)-responsive genes. Single-cell analysis delineated wide expression of HIF1A (HIF-1-α) mRNA with less coverage of EPAS1 (HIF-2-α) (Supplemental Fig. S5). HIF1A and CD47 mRNA expression and distribution were similar, emphasizing the contrary functions of each gene, the latter, through its suppression of blood flow, encouraging hypoxia (39, 40), and the former countering hypoxia (41). As well, islets expressed signal transducer and activator of transcription 3 (STAT3). This is relevant as TSP1, via STAT3, upregulated PD-L1 (42), whereas suppression of CD47 lowered STAT3 activation (43). Granulocyte-macrophage colony-stimulating factor receptor (CSF2RA), which participates in autoimmune disease (44) and is an activator of the TSP1 promoter (45), was found in islet-associated macrophages (Supplemental Fig. S5). Of note, CSF2RA was not identified in another single-cell mRNA analysis of human islets (46).

CD47 Protein Is Expressed in the Endocrine and Exocrine Compartments of Human Pancreata

Immunofluorescence microscopy showed CD47 protein strongly expressed in islets and colocalized with insulin (Fig. 2A). In keeping with single-cell mRNA sequencing results, TSP1 and SIRPα proteins were not found in islets (Fig. 2, B and C). CD47 was reported to modify intestinal epithelial barrier function (47, 48). However, its expression in pancreatic exocrine epithelial cells was unknown. Here, CD47 protein was also demonstrated in acinar cells (Fig. 2, D and E). Under higher magnification, CD47 expression in insulin-positive β cells was confirmed, whereas TSP1 and SIRPα proteins were unappreciated in ductal and acinar cells (Fig. 2, F and G).

Figure 2.

CD47 protein is expressed in the endocrine and exocrine compartments of human pancreata. Human pancreas tissue sections were costained for insulin and CD47 (A), TSP1 (B), or SIRPα (C). Confocal images were taken at ×10 and ×20 magnification with tile scanning. Enlarged images in the boxed areas are shown on the right. Quantification of the fluorescence intensity relative to the negative control (secondary antibody only staining) was analyzed for islets (D) or nonislet areas (E) using ImageJ software. One slide from each donor organ (total of 3 donors) was examined. Three to four separate areas of the tissue sample were characterized for each slide. Five or more islets were counted in each area. The scale bar in tile scanning images equals 100 µm and in enlarged images equals 50 µm. Human pancreas tissue sections from a single donor (Hu1172) were costained for TSP1, CD47, and SIRPα and confocal images were acquired at ×40 magnification and are shown on the far right of each panel (F). Images at ×40 magnification from tissue samples from 3 donors were quantified for protein expression in acinar and ductal cells (G). Data are presented as the means ± SD. Each dot represents an islet or the nonislet area of each donor tissue section. SIRPα, signal regulatory protein α; TSP1, thrombospondin-1.

CD47, but Not TSP1 and SIRPα, Protein Is Found in Human Islet Lysate

To further substantiate findings, islet lysates were subjected to gel electrophoresis and proteins were characterized by Western blot. Here too, CD47, but not TSP1 and SIRPα, protein was detected (Fig. 3). Thus, in islets and pancreata from donors lacking a history of diabetes, TSP1 and SIRPα proteins are absent. Gel electrophoresis and Western blot of human THP-1 cell lysate confirmed antibody recognition of TSP1, CD47, and SIRPα (Supplemental Fig. S3).

Figure 3.

CD47, but not TSP1 and SIRPα, protein is found in human islet lysate. Lysates of islet preparations from donors without diabetes (n = 5, Hu 1186, Hu 1199, Hu 1228, Hu 1229, and Hu 1232) underwent separation by gel electrophoresis and protein expression was characterized by Western blot. Representative Western blot showing CD47, TSP1, and SIRPα expression (A). Densitometry analysis of CD47, TSP1, and SIRPα expression relative to β-actin (B). Each dot represents a donor. Data are presented as the means ± SD. Significance was determined using one-way ANOVA for multiple comparisons. SIRPα, signal regulatory protein α; TSP1, thrombospondin-1.

CD47 Resides on the Surface of Islet Cells

CD47 on the cell surface membrane of nonimmune cells interacts with immune cell surface SIRPα to activate a cytoplasmic immunoreceptor tyrosine-based inhibitory motif (49). Flow cytometry analysis demonstrated CD47 on the cell surface membrane of dissociated islets (Fig. 4). However, contrary to speculation (19), SIRPα was not found on the surface of islet cells. The lack of SIRPα on the surface of islets contrasts with other human cells, such as renal tubular epithelial (50) and THP-1 cells that simultaneously express cell membrane CD47 and SIRPα. Flow cytometry of human THP-1 cells confirmed antibody recognition of cell surface CD47 and SIRPα (Supplemental Fig. S2).

Figure 4.

CD47 resides on the surface of islet cells. Islets from donors without diabetes (n = 3, Hu 1232, Hu 1237, and Hu 1239) were dissociated and stained with fluorochrome-conjugated anti-CD47 or anti-SIRPα antibodies and analyzed by flow cytometry. Gating strategy for single live cells (A). Representative analysis of CD47- (B) and SIRPα-positive cells (C) after sequential gating from A. Positive thresholds were set according to isotype controls. Quantification of the % CD47- and SIRPα-positive cells (D). Each dot represents one donor. Data are presented as the means ± SD. Significance was determined using paired t tests. **P ≤ 0.01. SIRPα, signal regulatory protein α.

Human Islets Secrete Soluble TSP1 and Express Cell Surface CD47 That Is Maintained under Inflammatory Stress

Murine muscle cells exposed to inflammatory cytokines showed increased SIRPα mRNA and protein (51). However, the effect of inflammatory cytokines on TSP1, CD47, and SIRPα expression in islets was unknown. Human islets were exposed to inflammatory cytokines (IL-1β, IFN-γ, and TNF-α) (52) for 24 h, mimicking a diabetic milieu. Protein levels of CD47 did not change and TSP1 and SIRPα remained undetectable (Fig. 5, A and B). Cytokine-treated islets did show increased PD-L1 (53) (Fig. 5, A and B), corroborating islet responsiveness and cytokine potency. However, TSP1 is a secreted protein (54). Pursuing this, we found that cultured human islets secreted substantial quantities of soluble TSP1, which continued after exposure to cytokines (Fig. 5C). Standard islet culture medium was free of detectable TSP1 (data not shown).

Figure 5.

Human islets secrete soluble TSP1 and express cell surface CD47, which is maintained under inflammatory stress. Human islets (500 IEQ) from donors without diabetes (n = 4, Hu 1229, Hu 1236, Hu 1237, and Hu 1238) were incubated in islet culture medium with or without IL-1β (50 U/mL), IFN-γ (1,000 U/mL), and TNF-α (1,000 U/mL) for 24 h. Islet lysates were prepared, protein separated by gel electrophoresis and expression of CD47, TSP1, SIRPα, and PD-L1 characterized by Western blot. Representative Western blot showing protein expression (A). Densitometry analysis of expression relative to β-actin (B). Conditioned medium from freshly gathered human islets at baseline and following exposure to inflammatory cytokines for 24 h (n = 4 four donor preparations) was assayed for secreted soluble TSP1 (C). Each dot represents one donor. Significance was determined using paired t test. ns, P >0.05, ***P < 0.001. SIRPα, signal regulatory protein α; TSP1, thrombospondin-1.

DISCUSSION

Analysis of human islets found differential expression of THBS1 and SIRPA mRNA in endothelial, mesenchymal, acinar, and ductal cells extending the identified roles that these genes play in regulating human cells, including endothelial, smooth muscle, dendritic, and glial cells (THBS1) (55) and renal tubular epithelial (50) and immune cells (SIRPA) (49, 56), among others. The absence of SIRPA mRNA in human β cells implies that it is not involved in routine β cell homeostasis as proposed (19). Expression of THBS1 mRNA was limited in islets. Rather, TSP1 was found to be an islet-secreted protein, thus implicating autocrine and paracrine signaling. Heterogeneity in secreted TSP1 levels among donor islet preparations is consistent with variation in TSP1 levels that we noted in platelet-poor plasma from individuals with vasculopathy (57). As TSP1, via CD47, suppressed immune cell activation (58, 59), the possibility that, in certain situations, TSP1 is islet-protective should be entertained.

CD47 mRNA was expressed in all pancreatic cell types, including β and other endocrine and exocrine cells, underscoring its general distribution in human cells and tissues (39). In islets, protein expression via Western and immunofluorescence staining paralleled transcript abundance with substantial CD47, while TSP1 and SIRPα were imperceptible. These patterns persisted in islets following a challenge with inflammatory cytokines. The lack of CD47 responsiveness to cytokines stands in contrast to human fibroblasts that showed an IL6-driven increase in CD47 (60). However, islets exposed to inflammatory cytokines showed persistence of CD47 and, as mentioned, secreted soluble TSP1, linking the TSP1-CD47 axis to diabetes. Of physiological relevance, flow cytometry analysis of islets showed cell surface CD47, but not SIRPα. This negates a role for cis CD47-SIRPα signaling (61) in islet endocrine cells. However, it does not exclude the potential for trans CD47-SIRPα signaling.

The pancreata included in this study were harvested and processed to minimize tissue stress. Only organs from adults without diabetes were included. There were 6 women and 9 men with a mean age of 40.7 (range 20–63). Most donors (73%) were overweight (body mass index ≥ 25.0 kg/m²) and a fifth were prediabetic (HbA1c between 5.7 and 6.4%). Still, all islet preparations showed high purity and >90% viability. Typically, TSP1 is minimally expressed in tissues and body fluids (62). But TSP1, and to an extent, CD47, are upregulated after injury, in disease (4, 63) and in aging (7). Thus, the lack of TSP1 protein on IF and Western blots of islets from donors without diabetes may not be surprising. However, TSP1 was revealed in cultured islets as a secreted soluble protein. Notwithstanding, determining the expression pattern of all three genes in pancreata and islets from individuals with chronic diseases would be interesting.

Of note, CD47 expression tended to be greater in pancreatic tissue sections from individuals with T1D (64). We reported that the TSP1 promoter was HIF regulated (37) and a HIF-1-α single nucleotide polymorphism (SNP rs11549465) was enriched in individuals with T1D (65). Together, these data suggest a possible alternative mechanism for TSP1-CD47 activation in diabetes. Still, as a checkpoint molecule that restrains immune activity, it is curious that increased CD47 expression was associated with autoimmune T1D (64). It was also found that rapid CD47 knockdown decreased the viability of β-like EndoC-βH1 cells (64). This confirmed our prior finding that aggressive suppression of CD47 was cytotoxic (1). Further supporting a homeostatic role for CD47, genome exon sequencing of over 60,000 individuals found only one loss-of-function SNP (66), whereas 11 were predicted (62), suggesting that CD47 is essential. Akin to this, exposing islets to CD47 antibodies could yield confounding results since such antibodies are promiscuous with effects upon TSP1 and SIRPα binding to CD47 (34). These points support the approach taken in the present study to employ freshly gathered human islets cultivated under conditions present in T1D as a means of denoting physiologically relevant signaling.

SIRPA mRNA expression was absent from islets but present in exocrine ductal and endothelial cells, consistent with its effect on cell adhesion, migration (67), and extravasation (68). SIRPα binding to CD47 limited immune activation (69), and SIRPα signaling altered the myosin-actin cytoskeleton to suppress phagocytosis (70). Inflammatory cytokines are increased in diabetes (71) and we noted that inflammatory cytokine-mediated degradation of cell surface SIRPα promoted monocyte activation (72). Herein, we found SIRPα mRNA and protein were excluded from islet endocrine cells. Contrary, in cancer cells, SIRPα expression was regulated, in part, by HIF-1-α (73). Related to this, with colleagues, we noted that polymorphisms of SIRPα that varied in their binding affinity to CD47 dictated innate immune response (74). Alluding to a role in T1D, a SIRPα variant promoted diabetes in NOD mice (75). In this circumstance, the mutant SIRPα putatively bound with abnormal affinity to CD47. Together, these data raise the intriguing idea that variation in SIRPα binding contributes to immune-mediated diabetes. Correspondingly, SNPs in signal regulatory protein gamma (SIRPG), which altered interaction with CD47, were associated with increased human CD8 T-cell activity and cell killing (76). Furthermore, a SIRPG SNP (rs2281808) was associated with decreased T-cell SIRPγ in individuals with T1D (77). The implications of this finding remain unclear as individuals without T1D also displayed the variant. In addition, an antibody believed to disrupt SIRPγ-CD47 binding altered immune activation in mice (78). These findings should be considered in light of more than 10-fold greater affinity between human TSP1 and CD47 (34) versus human SIRPα and CD47 (79). Indeed, we showed in vitro that TSP1 bound CD47 with high affinity and, in so doing, disrupted SIPRα binding to CD47 (34). This begs the question whether TSP1 alters the interaction between SIRP (α or γ) and CD47, and whether TSP1 intercedes in checkpoint blocker molecules that target CD47 or SIRPα.

Beyond CD47, other checkpoint molecules, including PD-L1, PD-1, and CTLA-4, provide a natural brake on immune cells (80). As with CD47, these pathways have been leveraged to stimulate immune cell attack of cancers. However, insulin-dependent diabetes occurred in individuals following treatment with PD-1 checkpoint blockers (81). This suggests that similar side effects could arise from agents that disrupt islet-expressed CD47 interaction with immune cell SIRPα (21).

In considering the results of this study, several points bare mentioning. First, sample availability did not allow for perfect uniformity in analysis. Still, all samples were obtained from individuals without diabetes. Second, while showing the most abundant transcripts, single-cell mRNA sequencing does not prove the absence of others. In this regard, deeper sequencing and bulk mRNA analysis may be beneficial. However, bulk analysis would preclude assigning expression to specific cell subsets. Third, regarding immunofluorescence analysis, secondary antibody staining was used as a negative control. Fourth, fluorescence microscopy did not allow precise localization of proteins to cell organelles. The latter would likely prove informative in understanding further the signaling activities of cell surface receptor CD47 (12, 62). As well, tissue sections were obtained from the head of pancreas. Samples from other anatomical areas may yield variation in transcript and protein expression. Last, stress-induced changes in protein suggest, but are not sufficient, to prove ligand-receptor signaling.

To summarize, we characterized the molecular signature of THBS1, CD47, and SIRPA mRNA and protein in fresh human islets and pancreata from nondiabetic individuals. Cell, tissue, and stress-related expression patterns were identified. Considering that these genes regulate immunity and metabolism (2, 7, 82), more in-depth analysis is warranted.

DATA AVAILABILITY

Data will be made available upon reasonable request.

SUPPLEMENTAL DATA

https://doi.org/10.6084/m9.figshare.22006958.v1.

https://doi.org/10.6084/m9.figshare.22006943.v1.

https://doi.org/10.6084/m9.figshare.21861684.v1.

https://doi.org/10.6084/m9.figshare.21861678.v1.

GRANTS

This study was made possible by support from the Wanek Family Project for Type 1 Diabetes.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.Q., I.H.A., B.O.R., and J.S.I. conceived and designed research; N.E., K-T.C., M.Q., Y.Z., and I.G. performed experiments; N.E., K-T.C., M.Q., Y.Z., X.W., and I.G. analyzed data; N.E., K-T.C., M.Q., X.W., I.G., H.T.K., E.M., I.H.A., F.K., B.O.R., and J.S.I. interpreted results of experiments; N.E., K-T.C., M.Q., Y.Z., and I.G. prepared figures; N.E., K-T.C., and J.S.I. drafted manuscript; M.Q., H.T.K., E.M., I.H.A., F.R.K., B.O.R., and J.S.I. edited and revised manuscript; N.E., Y.Z., X.W., I.G., H.T.K., E.M., I.H.A., F.R.K., B.O.R., and J.S.I. approved final version of manuscript.