CD47 and thrombospondin-1 regulation of mitochondria, metabolism, and diabetes

Abstract

Thrombospondin-1 (TSP1) is the prototypical member of a family of secreted proteins that modulate cell behavior by engaging with molecules in the extracellular matrix and with receptors on the cell surface. CD47 is widely displayed on many, if not all, cell types and is a high-affinity TSP1 receptor. CD47 is a marker of self that limits innate immune cell activities, a feature recently exploited to enhance cancer immunotherapy. Another major role for CD47 in health and disease is to mediate TSP1 signaling. TSP1 acting through CD47 contributes to mitochondrial, metabolic, and endocrine dysfunction. Studies in animal models found that elevated TSP1 expression, acting in part through CD47, causes mitochondrial and metabolic dysfunction. Clinical studies established that abnormal TSP1 expression positively correlates with obesity, fatty liver disease, and diabetes. The unabated increase in these conditions worldwide and the availability of CD47 targeting drugs justify a closer look into how TSP1 and CD47 disrupt metabolic balance and the potential for therapeutic intervention.

INTRODUCTION

Metabolic dysregulation characterizes numerous acquired diseases including metabolic syndrome, nonalcoholic fatty liver disease (NAFLD), obesity, and diabetes and contributes to vascular disease, hypertension, myocardial infarction, and stroke. The incidence of these diseases is associated with increase in weight coincident with minimal activity. As of 2016, the World Health Organization reported that close to 2 billion adults were overweight, and 34% of these were obese (1). That an estimated 340 million children were likewise obese indicates that this process begins early. More topically, obesity is one of the leading risk factors for COVID-19-related mortality (2).

At the cellular level, metabolic health depends upon modulated delivery of oxygen and nutrients concurrent with removal of waste products such as excess nitrogen and carbon dioxide. For multicellular organisms, this requires complex layers of organization and systems with feed forward and backward regulation. Some relevant cues derive from the need of cells in multicellular organisms to adhere and respond to mechanical and chemical signals in their extracellular matrix environment. Cells build and constantly modify this extracellular matrix. Cells also secrete bioactive proteins that bind to the matrix and cell surface to provide yet another layer of regulation. Secreted proteins that serve as modulatory rather than structural links between the extracellular matrix and the cell surface are termed matricellular proteins (3). Thrombospondin-1 (TSP1) is a classic example of a function-modifying matricellular protein. The known regulatory functions of TSP1 have expanded from disrupting cell adhesion to modulating inflammation, responses to hypoxic and genotoxic stress, redox signaling, stem cell self-renewal, and autophagy. These functions of TSP1 are mediated by its interactions with several cell surface receptors, structural components of the extracellular matrix, and secreted factors such as transforming growth factor beta (TGF-β) (4). Among its cell surface receptors, a growing body of data indicate that TSP1 interactions with the cell surface receptor CD47 play a key role in regulating cell stress and fundamental metabolic processes (5) (Fig. 1).

Figure 1.

Direct and indirect effects of thrombospondin-1 (TSP1) on CD47 signaling. TSP1 is secreted under stress conditions and binds with high affinity to CD47, which is expressed on all cells. TSP1 binding induces CD47 signaling that alters cell metabolism, gene expression, and responses to stress. CD47 also interacts with signal regulatory protein-α (SIRPα) on different cells (trans interaction) and induces SIRPα signaling by altering the activity of phosphatases SHP1 and SHP2. Functional cis interactions between CD47 and SIRPα on the surface of a single cell have been reported but require further verification. TSP1 can also prevent cis and trans SIRPα signaling by inhibiting its binding to CD47. Conversely, therapeutic agents that target CD47 or SIRPα may also alter TSP1 interactions with CD47 to alter metabolism of glucose, fatty acids, and reactive oxygen species (ROS).

THROMBOSPONDIN-1 AND CD47 REGULATION OF MITOCHONDRIA

TSP1 is known to drive cell death through CD47, and this is associated with mitochondrial dysfunction. Treatment of human T cells with TSP1, a CD47 antibody (clone 1F7) or a TSP1-derived peptide that bound specifically to CD47, and a CD3 antibody led to cell death (6). Of note, T cells treated with the CD47 antibody showed a decrease in cyclic adenosine monophosphate (cAMP) levels, although it is not clear if the antibody employed in these studies blocked TSP1 binding to CD47. Cells lacking CD47 were resistant to cell death, whereas re-expression of CD47 in null cells promoted mitochondrial dysfunction and Fas antibody killing (7). TSP1 negatively regulates cAMP signaling in several cell types (8–11), and pretreatment of T cells with cell-permeant cyclic adenosine monophosphate (cAMP) provided protection from TSP1-CD47-mediated cell death (6). In smooth muscle cells, cAMP regulation is independent of phosphodiesterases (10). Treatment of leukemic cells with TSP1, or the C-terminus domain of TSP1 that was shown to bind CD47 (12), resulted in loss of mitochondrial membrane potential, increased mitochondrial reactive oxygen species (ROS) formation, and promoted cell death (13). This is likely of import, as mitochondria are a major source of intracellular ROS. In addition, the TSP1-mediated degradation of the mitochondrial membrane potential would be expected to disrupt normal mitochondrial oxidative phosphorylation and thus broadly impact cell metabolism. TSP1 and a CD47 antibody stimulated B cell death. The antibody treatment rapidly damaged mitochondria, in part, via translocation of the dynamin-related protein 1 (Drp1) from the cytoplasmic domain of CD47 to mitochondria (14). Here too, a human CD47 antibody (clone B6H12) induced loss of mitochondrial membrane potential and ROS production. B6H12 is known to block TSP1 binding to CD47 (12), but its ability to induce B cell death may be independent of that activity. Contrasting this, we found that B6H12 provides protection to cells and blood vessels to a range of stress (15–17), consistent with the agonist and antagonist activities in this and other CD47 antibodies (18). Another TSP1 receptor, CD36, likely mediated mitochondrial dysfunction through altering free fatty acid (FFA) uptake (19) and, as we found, TSP1, under some conditions, interfered with FFA uptake (20).

Mitochondrial biogenesis is stimulated via peroxisome proliferator-activated receptor γ coactivator-1 (PGC-1α), which is increased by low concentrations of nitric oxide (NO) (21) and cAMP (22). We reported that TSP1, via CD47, potently inhibited NO production and signaling (23) and cAMP signaling (10). Consistent with this, skeletal muscle from young Cd47-null mice, which have increased basal NO/cyclic guanosine monophosphate (cGMP) signaling (24), showed increased mRNA and protein levels of mitochondrial-associated genes such as PGC-1α, cytochrome b and c, and nuclear respiratory factor 1 (NRF1) compared with muscle from wild-type controls (25). These differences were age-sensitive and tissue-specific and not found in the brain, liver, kidneys, heart, aorta, or brown fat. On electron microscopy, fast-twitch and slow-twitch skeletal muscle from young Thbs1- and Cd47-null mice showed more and larger mitochondria. Indeed, null skeletal muscle cells contained approximately two times more mitochondria. This is far reaching, as skeletal muscle harbors the bulk of all mitochondria in individuals, and metabolic dysfunction in skeletal muscle is a fixture of metabolic syndrome, obesity, and diabetes (26). Despite the differences in mitochondrial mass between control and Cd47-null skeletal muscles, the function of electron transport chain complexes and the ATP/O₂ ratios were comparable between control and Cd47-null skeletal muscle mitochondria. Young Cd47-null mice had increased exercise capacity but consumed less oxygen compared with age-matched wild-type animals (25). Of relevance to age-related obesity, Cd47-null mice were leaner throughout the life cycle compared with controls. Volumetric analysis of fat from the null and control mice was not performed. This is relevant as visceral and subcutaneous fat expand and contract asynchronously. Similarly, white fat adipocytes from young Cd47-null mice showed increased levels of the downstream NO second messenger cGMP (27). However, baseline mitochondrial oxygen utilization was similar in null and wild-type white fat adipocytes (27). In contrast, brown fat adipocyte mitochondria from Cd47-null mice were larger and longer and, during fatty acid oxidation, consumed more oxygen compared with brown fat mitochondria from wild types. The latter enhancement is expected to increase the thermogenic activity of brown fat Cd47-null mitochondria in keeping with the role of brown fat to serve as a buffer against lowering of environmental temperature (28).

TSP1 AND CD47 IN METABOLISM

Mitochondria are essential for oxidative metabolism. In human T cells, loss of CD47 increased mitochondrial mass, oxygen consumption, and spare respiratory capacity (29). The implications of this on several metabolic pathways were demonstrated when a somatic T-cell mutant lacking CD47 (JinB8 cells) were subjected to metabolome analysis. More than 25% of assessed metabolite levels differed between CD47-deficient and control Jurkat T cells under standard normoxic growth conditions, and the CD47-deficient cells had a higher extracellular acidification rate (29). In addition, loss of Thbs1 altered basal metabolites in mouse livers and the response of null mice to dietary fat (30). In human colorectal cancer cell lines, CD47 overexpression was associated with increased mRNA levels of genes involved in glycolysis and gluconeogenesis (31). The mechanisms by which CD47 regulates basal metabolism remain to be defined. However, several findings suggest a role for the central transcription factor avian myelocytomatosis viral oncogene homolog (cMyc). TSP1-CD47 signaling, through regulation of phosphorylation, destabilized cMyc (16). As well, Thbs1- and Cd47-null cells had higher basal levels of cMyc mRNA and protein along with increased proliferation with asymmetric cell division that preserves stem cells (32). Furthermore, enhanced cMyc activity was maintained in Cd47-null cells and animals under stress conditions (16) and with normal aging (33). cMyc plays critical roles in regulating glycolysis, fatty acid homeostasis, and metabolism (34), suggesting that TSP1-CD47 widely and negatively controls cellular metabolism by decreasing the expression of cMyc.

In contrast to nonmalignant cells, Myc is not inhibited by TSP1-CD47 signaling in cancer cells where the MYC promoter is disrupted (32). In triple-negative breast cancer cells, a CD47 antibody that blocks TSP1 binding decreased epidermal growth factor (EGF) receptor signaling (35), and EGF itself regulates glycolysis (36). Mitochondria require oxygen for aerobic glucose metabolism, whereas oxygen, through effects upon the master transcription factor hypoxia inducible factor (HIF), controls TSP1 (37) and CD47 expression (38). The ability of TSP1-CD47 to control metabolism under hypoxic conditions requires verification. However, some effects of hypoxia-mediated induction of TSP1-CD47 signaling on gene expression in primary human cells persisted even after reoxygenation (39).

TSP1 AND CD47 IN METABOLIC SYNDROME, NONALCOHOLIC FATTY LIVER DISEASE, AND OBESITY

Metabolic Syndrome and Obesity

Excess calorie intake and inactivity are causative of several metabolic abnormalities that left unchecked, result in obesity, nonalcoholic fatty liver disease (NAFLD), and diabetes. These conditions rarely exist as isolated processes, and NAFLD is found in a majority of obese and diabetic individuals (40). As well, individuals with obesity have a markedly increased risk for developing diabetes (41). Metabolic syndrome encompasses earlier manifestations of all of these diseases and includes insulin resistance, along with elevated blood lipids and blood pressure and chronic inflammation (42). Roles of TSP1 and CD47 in each of these conditions have lately emerged (Fig. 2).

Figure 2.

Thrombospondin-1 (TSP1) impinges adversely on metabolic pathways and endocrine systems. TSP1 is upregulated by hypoxia, inflammation, hyperglycemia, high-fat intake, and age. Through activation of CD47 signaling, TSP1 adversely potentiates metabolic disease and diabetes via effects on mitochondrial homeostasis, metabolism, and survival in adipocytes and pancreatic islets. TSP1 signaling mediated by CD47 on vascular cells modulates angiogenesis and blood flow, and on macrophages and T cells modulates innate and adaptive immunity. CD47 interacting with signal regulatory protein-α (SIRPα) or SIRPγ further regulates innate and T cell immunity to impact metabolic disease.

Adipose tissue TSP1 mRNA levels were increased in high-fat diet (HFD)-fed mice with obesity (43). Interestingly, mice on a normal diet showed increased TSP1 transcript levels in adipose tissues simply with aging. TSP1 protein was increased largely in the perivascular areas of adipose tissues, and this was not associated with decreased vascularity (43). This latter observation is notable, given the primary role of TSP1, via CD47, to limit proangiogenic signals such as those induced by vascular endothelial growth factor (VEGF) (44) and NO (45–48). Conversely, lack of TSP1 protected from diet-induced weight gain and adiposity, especially in male mice, and was associated with fewer inflammatory cells in adipose tissue. The benefits noted in HFD-fed Thbs1-null mice were posited to be independent of activity and appetite since control and null animals on a standard chow diet showed similar energy expenditure and food intake. Interestingly, FFA and cholesterol levels were higher in HFD-fed Thbs1-null mice. Similarly, Thbs1-null mice had less HFD-mediated skeletal muscle fibrosis and more fat browning than controls (49). Noteworthy, muscle inflammation and fibrosis occur in individuals with obesity (50) which may adversely alter metabolism through decreased glucose uptake. Increased TSP1 expression was also noted in the visceral abdominal fat of genetically obese rats (51) and in the serum and adipose tissues of HFD-fed rats with obesity (52). Interestingly, isolated adipocytes from these animals showed high glucose-mediated suppression of TSP1. Although blood glucose levels were not characterized in the HFD-fed rats with obesity, the data in cultured adipocytes stand in contrast to results in multiple primary cell types that showed increased TSP1 expression under high-glucose culture conditions (53–55). Furthermore, differentiated murine adipocyte-like 3T3-L1 cells cultured under hypoxia showed decreased TSP1 mRNA and protein expression (56), again in contrast to the known effects of hypoxia to increase TSP1 in other cells types, animals (37, 57–60) and human arteries (15, 61). Together, these data suggest atypical control of TSP1 expression in rodent adipocytes. Localization of TSP1 to the perivascular area of murine subcutaneous adipose tissue was confirmed (62). Interestingly, HFD-fed mice lacking endothelial argonaute-1 (AGO1) had less TSP1 expression in adipose vascular endothelial cells along with less weight gain and more brown fat (62). The role adipocyte AGO1 plays in metabolic diseases such as fatty liver, and potential regulation of CD47 expression remains to be determined.

Nonobese individuals without overt chronic disease had comparable TSP1 mRNA expression in visceral and subcutaneous fat (63). Conversely, in individuals with severe obesity, especially women, TSP1 expression was higher in visceral compared with subcutaneous fat (64, 65). In Asian women, increased circulating TSP1 positively correlated with abdominal obesity, hyperglycemia, and hypertension (65). Adherence to a weight reduction diet for 6 mo tended to lower circulating TSP1. Consistent with this, extracellular vesicles (EVs) derived from the visceral fat of individuals with obesity expressed increased TSP1 protein compared with vesicles from subcutaneous fat from the same individuals (66). Further studies are needed to determine whether TSP1 on the surface of EVs engages receptors to actively signal. Separately, TSP1 mRNA expression in subcutaneous abdominal adipose tissue correlated positively with obesity, as defined by body mass index (BMI) (63). In these individuals, circulating levels of plasminogen activator inhibitor-1, a biomarker of metabolic syndrome (67), positively correlated with adipose TSP1 expression. Although the primary source(s) of circulating TSP1 under basal and disease conditions have not been confirmed, it should be pointed out that platelet alpha granules contain large amounts of preformed TSP1 (68) that is released with platelet activation, a common event during blood sampling. Thus, accurate assessment of soluble blood-borne TSP1 requires efforts to limit platelet activation before, and during, separation of the plasma from other blood components (69). Peripheral blood mononuclear cells (PBMCs) and abdominal wall subcutaneous adipose tissue from obese men showed increased TSP1 expression compared with cells and adnominal wall fat from lean individuals (70). Here too, TSP1 levels positively correlated with BMI. Exercise (40 min at 65%–80% of maximum heart rate daily for 3 mo) lowered TSP1 expression in PBMCs and fat from obese subjects while not altering BMI. Although the cohort of Nordic men in this study was small (38 obese and 11 lean subjects) and homogenous, the finding that moderate exercise, and possibly weight reduction (65), decreased circulating and tissue TSP1 justifies confirmation of these findings in a large clinical trial.

TSP1 and thrombospondin-2 (TSP2), which shares homology in its C-terminus to TSP1 and which binds CD47, albeit with less affinity (12), were increased in visceral and gonadal fat of diet- and genetic-driven mice with obesity (71). The implications of this remain unclear, as HFD-fed Thbs2-null mice, while weighing less than controls, showed no differences in adiposity, adipose-associated inflammatory cell numbers, and vascularization (72). We noted that Thbs2-null mice mimicked wild-type mice under ischemic stress (12). It is possible that TSP1 expression in Thbs2-null mice accounted for the lack of phenotypic differences in adipose tissue from HFD-fed control and Thbs2-null animals.

Young Cd47-null mice on a HFD showed less weight gain, less inflammation, and lighter livers compared with controls (73). Decreased adipose inflammatory cell numbers in the absence of CD47 is consistent with other reports that showed, under certain stress conditions, less inflammatory cell tissue invasion in Cd47-null mice (74). Circulating levels of cholesterol and FFAs were lower in HFD-fed Cd47-null mice. Lean muscle volume was comparable between groups, but controls showed more fat mass (73). We found that 18-mo-old Cd47-null mice resisted weight gain on a HFD compared with controls (33). It was not known if intake and activity varied between these groups. However, the latter is unlikely since we reported no variation in daily activity between young Cd47-null and control mice (75). This is germane as metabolic syndrome and nonimmune diabetes are diseases of aging, decreased activity, and excess caloric intake whereas TSP1 is maladaptively increased with aging (76). Both HFD-fed young (73) and aged (33) Cd47-null mice were more insulin sensitive compared with controls and showed less blood glucose excursion after glucose bolus. Thus, CD47 promotes metabolic imbalance throughout the life cycle.

Liver Disease

NAFLD is progressive and, through dysregulated lipid metabolism and chronic inflammation, results in hepatic sclerosis, organ failure, and death. To promote fatty liver disease, control and Thbs1-null mice were fed a choline deficient l-amino acid-defined high-fat diet (77). Inconsistent with other reports, body and liver weights of mice after 12-wk on the special liver fat-promoting diet were the same among groups. Unfortunately, animal age and sex were not reported. Still, Thbs1-null mice were protected from liver inflammation. Specifically, collagen protein and transforming growth factor-β (TGF-β) mRNA levels were less in null mice fed a fatty liver-promoting diet. The latter finding is consistent with the ability of TSP1 to activate latent TGF-β (78). Also, genes involved in metabolism, mitochondrial function, and cell proliferation varied between control and Thbs1-null mice. Expanding on this, mice with macrophage-specific loss of Thbs1 had less fatty liver disease on a HFD (79). Curiously, lack of adipocyte TSP1 did not alter the development of diet-driven NAFLD, or nonalcoholic steatosis (NASH) (79). Wild-type mice fed a HFD had increased TSP1 in macrophages and bile duct cholangiocytes. This latter finding warrants further inquiry given that cholangiocytes regulate bile cholesterol composition, likely, in part, through CD36 (80), another TSP1 receptor (81). Interestingly, in individuals with obesity, weight reduction after bariatric surgery is, in part, due to modifications of bile acid homeostasis (82) and the gut microbiome (83). However, contrasting these data, HFD-fed Thbs1-null/adenomatous polyposis coli-Min mutant (Apc^Min/+) mice showed abnormal liver lipid accumulation (30). In keeping with data that found a link between activity and TSP1 expression, wild-type mice with diet-mediated NAFLD showed reversion of liver TSP1 expression following daily exercise (1 h of swimming for 10 wk) (84). Exercise-mediated suppression of TSP1 was linked to increased Ak strain transforming (AKT) signaling. Of note, AKT-mediated activation of endothelial nitric oxide synthase increases NO production, which suppresses TSP1 expression (85).

Individuals with NASH-associated cirrhosis showed markedly increased TSP1 expression in the liver (86). This finding was not specific to NASH, as individuals with alcohol- and hepatitis-associated cirrhosis also showed increased liver TSP1. Human hepatocyte cells (HuH7 cells), cultured with FFAs showed increased lipid accumulation, ROS, and TSP1 (87). Adipocyte-secreted adiponectin limits NAFLD (88) and increases insulin sensitivity. Parenthetically, TSP1 bound to adiponectin in blood samples from healthy individuals and diabetics (89), although the physiologic implications of this are not known. The role of CD47 in clinical NASH and NAFLD remains to be assessed.

TSP1 AND CD47 IN DIABETES

In the United States, 1 in 10 people have overt diabetes, whereas one in three have incipient disease (90). Worldwide, diabetes is the seventh leading cause of death, a primary driver of cardiovascular, ocular, and renal disease, and is predicted to increase worldwide by 48% (629 million) over the next 25 years (91). In 2021, the National Institute of Diabetes and Digestive and Kidney Diseases requested some $2 billion in support (Congressional Budget Justifications, NIDDK). Although implicated in metabolic diseases, little is known about the role of TSP1 and CD47 in diabetes, particularly at the level of human islets. This latter deficit in knowledge, to some degree, likely arises from difficulties in obtaining fresh human islets.

Among individuals with type 1 diabetes (T1D, n = 30, mean age 31 ± 3 yr, duration of diabetes 8 ± 4 yr) plasma TSP1 was significantly elevated (136.6 ± 14.2 ng/mL) compared with controls (91.2 ± 14.3 ng/mL) (n = 15; mean age 29 ± 2 yr) (92). Plasma TSP1 levels were even higher in the subset of individuals with T1D-associated complications (150.4 ± 23.7 ng/mL). The ∼59% elevation in plasma TSP1 levels above healthy subjects was striking. Importantly, subjects with T1D with tighter control of blood glucose, as defined by lower glycated hemoglobin A1c (HbA1c), had lower plasma TSP1 than those with poor blood glucose control (i.e., higher HbA1c). As well, individuals with type 2 diabetes (T2D, n = 149) had elevated circulating TSP1 compared with those without diabetes (n = 225) (93). In type 2 diabetics, elevated plasma TSP1 independently predicted coronary artery disease (CAD). The study cohort, while relatively large, was ethnically uniform. Still, the clinical finding that TSP1 predicted CAD in individuals with T2D is in line with preclinical data implicating TSP1, via CD47, as a driver of systemic (33, 94) and pulmonary (15, 57, 69) vasculopathy through limiting endothelial VEGF and NO while stimulating vascular remodeling and atherosclerosis. In contrast, women with polycystic ovary disease, elevated blood glucose, and insulin resistance had lower circulating TSP1 compared with controls (95). However, those individuals treated with metformin displayed increased circulating TSP1 levels (95). This later observation is interesting since metformin is administered for several diseases including diabetes.

Although these results show an association in people between increased TSP1 and diabetes and diabetes-related complications, they do not indicate what role TSP1 plays in the disease process. β-klotho-mediated suppression of TSP1 in adipose-derived stem cells from individuals with T2D resulted in increased cell proliferation and matrix production (96), suggesting improved wound healing capacity. It would be interesting to see if changes in klotho signaling associate with age-related increases in TSP1. Fifteen-week-old HFD-fed male Thbs1-null mice were more glucose tolerant and sensitive to insulin compared with wild-type HFD-fed mice (97). Null mice had fewer inflammatory cells in adipose tissue and lower blood triglyceride levels than wild types, but both groups gained weight equally on the HFD. Male Thbs1-null mice were more resistant to HFD weight gain, but these differences only appeared after longer exposure to the special diet (5 mo) (43). Bone marrow-derived macrophages from the HFD-fed Thbs1-null mice were less sensitive to lipopolysaccharide and migrated less (97). It was suggested that CD36 signaling did not account for the migratory deficit in Thbs1-null macrophages. However, this finding was based on macrophages from Cd36-null mice on a regular diet. In contrast, CD36 was noted to retard macrophage migration to oxidized low-density lipoprotein (LDL) (98), although what role TSP1 had in this was not tested. CD36 also promotes foam cell formation (99). Furthermore, we found that TSP1-CD47 signaling stimulated nicotinamide adenine dinucleotide phosphate oxidase-1 (Nox1) to increase macrophage uptake of native LDL (100). This TSP1-CD47 effect was associated with changes in the macrophage actin cytoskeleton. Related to this, increased cell stiffness, which is influenced by the rigidity of the extracellular matrix, overcame CD47 inhibition of phagocytosis (101). Noteworthy, macrophage phagocytosis may contribute to T1D (102). CD47 negatively regulated dermal TGF-β1 mRNA expression and TGF-β1 activation, and fibrosis (103), whereas pancreata from individuals with T1D showed increased fibrosis (104, 105). It would be useful to apply biomechanical methodologies to determine the stiffness of T1D and T2D pancreata and, if possible, islets. Although data under the metabolic stress of a HDF suggests a deleterious role for TSP1, information under basal conditions indicates differently. Pancreas weight was similar in young 12-wk-old wild-type and Thbs1-null mice. Furthermore, islet and β cell mass were greater in organs from Thbs1-null mice (106). Despite this, Thbs1-null mice had less basal insulin and more glucose intolerance compared with wild type, findings that did not change in 52-wk-old animals. It is unknown if TSP1 stimulates β cell insulin release.

Streptozotocin (STZ) is an alkylator that is toxic to several cell types, but especially to insulin-producing islet β cells, and animals treated with STZ develop diabetes (107). Young wild-type mice treated with STZ (40 mg/kg daily for 5 days) had elevated plasma glucose and lower insulin levels with slower increases in body weight compared with untreated mice (108). In pancreata from treated mice, islet CD47 expression was decreased along with increased numbers of pancreatic macrophages. As cell-specific CD47 expression or cell death was not determined, the meaning of lower CD47 expression is unclear. In vitro, STZ treatment also decreased islet CD47 expression and was judged to increase phagocytosis of endocrine-like Min6 cells. Although preliminary, these data suggest that rapid alteration in CD47 expression may prime islet cells for macrophage recognition. However, it is not clear if this is associated with altered interactions with TSP1, which can enhance phagocytosis (109). Interestingly, PBMCs from individuals with T2D displayed decreased levels of CD47 compared with cells from nondiabetics (110). Diabetic nephropathy is a primary source of renal failure. In endothelial nitric oxide synthase (Nos3)-null mice, STZ-induced diabetes was associated with worse nephropathy and increased expression of renal TSP1 compared with controls. Conversely, mice treated with cinaciguat, to activate the NO cytoplasmic receptor soluble guanylate cyclase and increase cGMP levels, showed less diabetic nephropathy and less renal TSP1 (111). The latter is consistent with our prior finding that human cells treated with low amounts of NO showed decreased TSP1 expression (85). Diabetes is also characterized by a general decrease in tissue vascularity. STZ-treated diabetic mice showed increased tissue TSP1 and decreased capillaries, findings which were partially reverted by hyperbaric oxygen (112). Related to this, hyperbaric oxygen increased endothelial Nos3 expression and putatively NO production (113).

In contrast to the finding that loss of CD47 was associated with diabetes, enhanced CD47 signaling through altered binding affinity with the cell membrane signal regulatory protein-α (SIRPα) was linked to development of autoimmune T1D in nonobese diabetic (NOD) mice (114). NOD SIRPα bound CD47 with higher affinity compared with SIRPα from nonobese diabetes resistant mice, which were protected from T1D. Stronger binding resulted in T cell activation and increased diabetes when cells were transplanted. Interestingly, single-polymorphism nucleotides in the SIRPγ locus were linked to increased human T cell activation and killing (115). This has importance, as CD47 binds both SIRPγ and SIRPα and builds on our finding that variation in CD47-SIRPα affinity dictated dendritic cell immune response to allo-antigen (116). These findings predict possible untoward effects from agents that target CD47 binding with interacting partners such as SIRPα (18). Antibodies that engage other immune cell inhibitory receptors such as CTLA-4 and PD-1 were associated with rapid onset type 1-like diabetes (117, 118). It remains to be seen if cell-to-cell trans CD47-SIRPα/SIRPγ interaction or CD47 acting in a cis manner with SIRPα on the same cell membrane also regulates such events (119). What role TSP1 plays in this should be queried given that TSP1 inhibits SIRPα binding with CD47 (12, 120).

In BioBreeding rats, which develop autoimmune T1D, skeletal muscle TSP1 protein levels were increased and muscle capillary density decreased (121). Little is known about the skeletal muscle microcirculation of individuals with T1D. As insulin and glucose delivery is predicated upon a robust functional capillary network, it would be useful to characterize skeletal muscle and islet capillarization along with TSP1 and CD47 in individuals with T1- and T2-diabetes.

Inflammation resulted in enzymatic shedding of the extracellular domain of SIRPα in monocytes (122), which suggests an alternative mechanism for immune cell activation that could be important in the inflammatory microenvironment of T1D and T2D. Incidentally, these findings cast doubt upon CD47-SIRPα signaling as a dominant inhibitor of phagocytosis. Expression of a mutant nonfunctional SIRPα that lacked recruitment capacity in the cytoplasmic domain led to lower plasma insulin levels in standard diet-fed mutant mice and this was exacerbated in HFD-fed animals (123). Furthermore, HFD-fed mutant mice had less glucose sensitivity. Treatment with an α2-adrenergic antagonist partially improved glucose sensitivity. Insulin levels in pancreata from mutant and control mice were comparable as was insulin release, suggesting a possible role for SIRPα in regulating peripheral glucose homeostasis. Interestingly, in skeletal muscle, SIRPα interacted with the insulin receptor to decrease insulin signaling (124). It is not known if this extends to β cells.

The function of TSP1 and CD47 in mammalian and human islets and β cells remains to be fully explored. Isolated human islets from donors without overt diseases and normal BMI (23.6 ± 1.8 kg/m²) showed increased TSP1 mRNA after culture in high-glucose medium (125). As well, relatively high amounts of TSP1 protein were noted in the endocrine and exocrine tissues of human pancreata (126, 127). In contrast, CD47 protein was less abundant in endocrine tissues of human pancreata (128). However, cell-specific characterization of TSP1 and CD47 in human pancreata and islets remains undone. To this end, single-cell mRNA sequencing and FACS analysis will be helpful.

Expression levels in tissue samples indicate the presence of TSP1 and CD47 in human pancreata and islets (Fig. 3), but what this means under basal function or in stress is not clear. Treatment with the free fatty acid palmitate stimulated β-like INS-1 cell death, which was less after TSP1 knockdown (129). These results are provocative, as we showed TSP1 limited free fatty acid uptake via CD36 (20). The effect of TSP1 in INS-1 cells and human islets was postulated to be secondary to limiting hydrogen peroxide-mediated injury. This is contrary to data that TSP1-CD47 promoted cell injury by stimulating increased Nox1-mediated superoxide production (48, 57). Pertinent to this, β cells express Nox1 (130). In rodent and human islets, TSP1 was proposed to limit cytokine- and thapsigargin-mediated injury (131). Islets isolated from pancreata from young Thbs1-null mice secreted more insulin under low, but not high, glucose conditions (106). Islets from wild-type mice, on transplantation into Thbs1-null mice, showed lower glucose-mediated insulin secretion (132). A deficit in islet mitochondrial function was postulated to explain the altered response of Thbs1-null islets. Given these findings, it would be enlightening to assess Thbs1-null islet oxygen consumption. Interestingly, islet endothelial cells from wild-type mice and rats showed markedly greater TSP1 mRNA levels compared with other islet cells, including β cells. Increased islet EC TSP1 may explain the decreased vascularity of islets from diabetic Zucker rats (133), the increased vascularity in Thbs1-null islets compared with wild types (134), the better post-transplantation engraftment of Thbs1-null islets (135), and the improved islet angiogenesis following TSP1 blockade (136). Separate from this, islets expressing a nonfunctional extracellular domain of CD47 showed improved engraftment (137), probably through suppression of post-transplantation innate immunity, a process that negatively impacts islet transplantation (138). Modifying CD47 expression as a means of decreasing immune activation in transplantation was reported in other organs (139). Lastly, in a number of studies using isolated islets, a signaling role for TSP1 was inferred through knockdown strategies rather than employing exogenous TSP1 to interrogate ligand-receptor interactions. This may explain the discordant results in relation to the large body of data showing an injurious role for TSP1-CD47 signaling (140–142).

Figure 3.

Thrombospondin-1 (TSP1) and CD47 in the pancreas. In isolated human islets, TSP1 is increased by elevated glucose, and CD47 is expressed in human pancreata. However, cell-specific expression analysis of both genes and their protein products is wanting. This is important, since islets are organoids with a unique vasculature and neural input, multiple endocrine cell types, and resident inflammatory cells. The signaling effects of altered TSP1 expression will depend on the source (endogenous vs. blood stream) and cell type. Given its immune modulatory actions, manipulation of TSP1 signaling with clinical antibodies targeting CD47 may unbalance innate immunity with possible adverse effects in pancreatic islets.

Loss of β cell function and mass is seen in T1D and T2D (143). Approaches to restore functional β cell mass include stimulation of proliferation and transformation of precursor cells into β-like insulin producing cells. Interestingly, the former approach was successful in isolated human islets and was associated with increased cMyc levels (144). How realistic this goal may be is uncertain, since, along with increased cMyc expression, proliferating β cells lose the ability to secrete insulin (145), which presumably would be restored with maturation. Nonetheless, TSP1-CD47 signaling potently suppresses cMyc and other self-renewal factors (Oxt3/4, Sox 2, Klf4) and could represent an inherent barrier to precursor cell transformation.

FUTURE DIRECTIONS

Preclinical data indicate a role for TSP1 and CD47 in metabolic homeostasis and disease (Table 1). Clinical data finds increased TSP1 expression associated with metabolic dysfunction and disease. However, some findings require confirmation or resolution of discrepancies, which may identify new questions to address. It is not clear if TSP1-CD47 signaling impedes mitochondrial function in all cells or just in those with increased metabolic function. Furthermore, do mechanisms in addition to DRP1 transduce the cell surface TSP1-CD47 signal to mitochondria? Bcl-2 homology 3 (BH3)-only protein 19 kDa interacting protein-3 (BNIP3) regulation of calcium is another possible mediator of CD47 signaling (146). In endothelial cells and cardiac myocytes, TSP1 altered calcium-mediated signaling (147, 148). Some CD47-dependent changes in metabolites under stress are independent of changes in mitochondrial function. It would be useful to test if therapeutic disruption of TSP1-CD47 signaling improves mitochondrial function. In relation to diet-mediated diseases such as obesity and FLD, the role of TSP1-CD47 may be dependent upon cross talk with CD36, as was noted in relationship to amyloid effects on cGMP signaling (149).

Table 1.

Cell and animal responses to TSP1, CD47, SIRPα, and/or metabolic stress

Cell/Animal Type	Treatment/Stress	Responses
Human cells
T cells	TSP1, CD47 ab, TSP1 peptide	↓ cAMP, proliferation, activation↑ Cell death
Leukemic B cells	TSP1, CD47 ab, C-terminal domain	↓ Miitochondrial membrane potential↑ Cell death, reactive oxygen species
Adipocyte EVs	Obesity	↑ TSP1
Endothelial, vascular smooth muscle, retinal pigment epithelial, & renal tubular epithelial cells	High glucose	↑ TSP1
CD47-deficient T cells		↑ Mitochondria mass↑ Respiratory capacity
Rat cells
Adipocytes, mesangial cells	High glucose	↑ TSP1
INS-1 β-like cells	TSP1 KOpalmitate	↓ Cell death
Murine cells/tissues
Islets	TSP1 KO	↑ Insulin release to low glucose↑ Islet vascularity
Brown adipocytes	CD47 KO	↑ Mitochondrial oxygen consumption
3T3-L1 adipocytes	Hypoxia	↓ TSP1
Skeletal myocytes	TSP1 and CD47 KO	↑ Mitochondria number and size↓ Reactive oxygen species (CD47 KO)
Endothelial cells	AGO1 KO	↓ TSP1
Animals
Mice	AGO1 endothelial KO	↓ Weight gain
Mice	TSP1 KOHigh-fat dietLiver-injury diet	↓ Weight gain, muscle fibrosis, liver inflammation, blood triglycerides↑ Glucose tolerance
Mice	CD47 KOHigh-fat diet	↓ Weight gain, inflammation, cholesterol, free fatty acids↑ Insulin sensitivity
Mice	SIRPα nonfunctional mutationHigh-fat diet	↓ Plasma insulin worsened by high-fat diet

KO, knockout; SIRPα, signal regulatory protein-α; TSP1, thrombospondin-1.

Questions also remain in the realm of glucose balance and islet function. The cell-specific expression of TSP1 and CD47 should be defined in islets and in the exocrine pancreas. Especially relevant would be to determine islet and non-islet cell expression patterns in relation to the presence of disease and glucose control. Loss of pancreata neural innervation was noted in individuals with T1D (150). This suggests an area for further inquiry since astrocytes cultured in high glucose showed decreased TSP1 secretion despite elevated intracellular TSP1, which impaired neural synapse proteins in cocultures (151). However, the positive effect of TSP1 on synapse formation is mediated by α2δ1 rather than CD47 (152). Effects of TSP1 on glucose uptake by skeletal muscle cells may be through promoting muscle inflammation and fibrosis (49) or through effects on insulin-like growth factor (153), muscle growth, and metabolic demand. The effects of CD47 on islets and islet β cells require additional study. Changes in CD47 expression led to increased neural syntaxin-1 (154). Syntaxin-2 and syntaxin-4 are involved in β cell insulin secretion (155, 156), although they have yet to be related to CD47. Given the variations between human and rodent islets, future studies are best done in freshly isolated human islets. As well, immune cell contributions to T1D suggest that TSP1 and CD47, through suppressing immune cell functions in some situations, could participate in limiting cell-specific inflammation. Going forward, the implications of this may be revealed in individuals with cancer treated using CD47 targeting agents (18). These agents are intended to block CD47-SIRPα signaling, but effects upon TSP1-CD47 signaling may inadvertently alter metabolism (Fig. 3). Conversely, TSP1 and CD47 likely promote T2D via several mechanisms and almost certainly hasten disease complications if simply through deleterious effects on blood flow and vascularity (76, 157).

GRANTS

This work was supported by the Intramural Research Program of the NIH/National Cancer Institute (Project ZIA SC009172, D.D.R.).

DISCLOSURES

J. S. Isenberg serves as Chief Science Officer of Radiation Control Technologies, Inc. J. S. Isenberg and D. D. Roberts are co-inventors of CD47 technology patents assigned to the United States of America.

AUTHOR CONTRIBUTIONS

D.D.R. and J.S.I. prepared figures; drafted manuscript; edited and revised manuscript; and approved final version of manuscript.