Soluble Insulin Receptor Ectodomain Is Elevated in the Plasma of Patients With Diabetes

We developed two kinds of ELISA systems to specifically measure hIRα and full-length IR, respectively (details available in an online appendix at http://dx.doi.org/10.2337/db07-0394).

Healthy Japanese volunteers with no diabetic history or familial diabetic history (in relatives within the third degree) residing in the Tokushima district (n = 123) and confirmed to have normal glucose tolerance by 75-g oral glucose tolerance according to the World Health Organization guidelines for diabetes (10) were enrolled as control-1 subjects. Healthy Japanese volunteers with no diabetic history or familial diabetic history residing in the Nagano district (n = 120) and confirmed to have normal fasting plasma glucose (FPG) and A1C levels were enrolled as control-2 subjects. Outpatients seen at the Shiga University of Medical Science in the Shiga district (type 2 diabetic, n = 474), the University of Tokushima Affiliated Hospital in the Tokushima district (type 2 diabetic, n = 162), and the Tokushima University Hospital in the Tokushima district (type 1 diabetic, n = 53) were enrolled as T2DM-1, T2DM-2, and T1DM patients, respectively.

All clinical laboratory data for the T2DM-1 and T1DM patients were obtained at the Central Clinical Laboratory of the University Hospital of Shiga University of Medical Science or the Central Clinical Laboratory of the University of Tokushima. Plasma glucose was measured by the glucose oxidase method. Insulin and urine C-peptide immunoreactivity were estimated using ELISA methods, while the A1C level was measured by high-performance liquid chromatography. Total cholesterol, triglycerides, HDL cholesterol, free fatty acid, lactate, and glycoalbumin levels were determined using standard enzymatic methods.

In this study, we used two strains of transgenic mice expressing hIR (details in online appendix).

Immunoblotting was carried out using specific antibodies as indicated in the figure legends and as previously described except using Can-get-signal solution (Toyobo, Osaka, Japan) for primary antibody dilution (8).

All other reagents were of analytical grade and obtained from Sigma or Nacalai Tesuque (Kyoto, Japan).

All study protocols and designs were approved by the ethics committees of the University of Tokushima (approval #171) and/or the Shiga University of Medical Science (approval #16–36). We also obtained written informed consent from all participants. Animal experiments were approved by the animal ethics committees of the University of Tokushima (approval #16–57 and #05052) and carried out in accordance with international care regulations.

Based on the hypothesis that soluble IRα may exist in human plasma, we attempted to detect IRα in human plasma. After partial purification by wheat-germ agglutinin (WGA)-conjugated affinity chromatography (GE Healthcare Bio-Sciences, Piscataway, NJ) and immunoprecipitation with an anti–hIRα-specific antibody (5D9), IRα (135 kDa) was detected by immunoblotting with an IRα-specific antibody (N-20). A band corresponding to hIRα, but not the IR precursor, was observed, and its peak signal intensity corresponded to the peak absorbance in hIRα ELISA analyses (A₄₅₀) (Fig. 1A). Furthermore, intact hIRβ was not detected by immunoblotting with an anti-IRβ antibody that recognized the COOH-terminal 19 amino acids of IRβ (data not shown), suggesting that the hIRα observed in human plasma was probably derived by the cleavage of the ectodomain (α-subunit and a part of β-subunit) of the receptor from cell surfaces, rather than by release from damaged tissues (cells) or microvesicles (online appendix Fig. 3 and in-text Table 1). Furthermore, we also compared the molecular weights of the soluble hIR ectodomain in human plasma with standard hIR-ectodomain protein (α-subunit and a part of β-subunit) (8). The standard protein was derived from CHO-IR-SspI cells (8) and stably introduced the cDNA encoding the whole hIR ectodomain except for just three amino acids of the peritransmembrane region (amino acids 1–953), and its logical molecular weight was ∼370 kDa (8). As shown in Fig. 1B, the retention time of immunoreactive hIR ectodomain in human plasma was identical to that of standard hIR-ectodomain protein, suggesting that the molecular weight of both proteins under nonreducing conditions were almost same using the Superdex-200 gel-filtration column. In addition, the apparent molecular weight was ∼370 kDa. If the soluble hIR ectodomain exists as α-subunit homodimers, the molecular weight would be ∼270 kDa. If the soluble hIRα exists with intact β-subunits as an intact heterotetramer, the molecular weight would be ∼460 kDa. Given all these data, the soluble hIRα appeared to exist with parts of the extracellular region of β-subunits, and proteolytic cleavage (“shedding”) appeared to occur at a site in the extracellular peritransmembrane region (online appendix Fig. 3).

Based on the above results, we established ELISAs for both hIRα and full-length hIR that were highly accurate, specific, and unaffected by the presence of either insulin or hemolysis (details in research design and methods). According to the receiver operator characteristic analysis using the data of control-1 and T2DM-1 groups, the area under the curve was 0.794 ± 0.217 (P < 0.0001, online appendix Fig. 1G), suggesting that this hIRα ELISA system has significant specificity and sensitivity in separating these two groups. We also used a commercially available IRβ ELISA system, as well as a full-length IR ELISA, to confirm that the plasma hIRα was not derived from damaged cells or microvesicles. To investigate the clinical significance of the presence of IRα in human plasma, we measured the hIRα levels in samples from patients with diabetes. First, we examined the variations in the hIRα level in normoglycemic individuals throughout the day and found that the level did not change in normoglycemic subjects, even in the postprandial state (online appendix Fig. 1F). Next, we measured plasma samples obtained from both the T2DM-1 and T2DM-2 groups, as well as from the control-1 group, who had been confirmed to be normoglycemic by oral glucose tolerance tests according to World Health Organization criteria (10). As shown in Fig. 2A, the T2DM-1 group exhibited a significantly elevated plasma hIRα level compared with the control-1 group (2.26 ± 0.80 [n = 474] vs. 1.59 ± 0.40 ng/ml [n = 123], respectively; P < 0.001). On the other hand, the levels of plasma intact IRβ and full-length IR were negligible in the type 2 diabetes plasma samples with a high hIRα level (Table 1), indicating that the plasma hIRα existed without intact IRβ and was derived from the cell surface by cleavage rather than by release from damaged cells. These results are also supported by Fig. 1B. The plasma hIRα level was also increased in the T1DM group (2.00 ± 0.60 ng/ml [n = 53], P < 0.001) versus the control-1 group (Fig. 2A), and the levels of plasma intact IRβ and full-length IR were also negligible (Table 1). More than 30% of the diabetic patients exhibited an hIRα level higher than the cutoff value (2.39 ng/ml, mean + 2 SD of control-1 group values). There were no differences in the hIRα levels among people from different districts. On a molar basis, the hIRα concentration in diabetic patients was ∼20% of the fasting plasma insulin concentration. We also evaluated the percentages of insulin-bound hIRα in plasma from 11 cases of type 2 diabetic patients by measuring the changes in the immunoreactive insulin (IRI) level after depletion of hIRα by 5D9 (anti-hIRα)-affinity beads. Under conditions of almost complete hIRα depletion, a comparable amount (∼10–20%) of the IRI was depleted by the anti-hIRα antibody beads. Therefore, at most, 10–20% of the IRI binds to the plasma hIRα (Table 2).

Next, we analyzed the correlations of the hIRα level with the clinical parameters of the diabetic patients (online appendix Table 1). There were significant positive correlations between the hIRα and the blood glucose levels, including fasting blood glucose (P < 0.001) and A1C (P < 0.001) levels (online appendix Table 1 and online appendix Fig. 2A and B). On the other hand, the hIRα level was not correlated with markers of insulin secretion, including fasting IRI level, 24-h urinary C-peptide immunoreactivity excretion, or homeostasis model assessment of β-cell function. The hIRα level also exhibited a very weak correlation with age (P < 0.001) in the T2DM-1 group, but not in the control-1, control-2, or T2DM-2 groups. As shown in online appendix Table 1, there were no correlations with other standard clinical parameters. Moreover, there were no significant differences in the hIRα levels among the types of treatments the patients were receiving. To further confirm the relationship between hIRα and blood glucose levels, we examined 18 patients whose A1C level had changed greatly (≥1.8%) within the last 3 years. As shown in Fig. 2B, there was a strong positive correlation between the changes in the plasma hIRα level and those in the A1C level (n = 18, P < 0.001, R = 0.84). To examine whether hyperglycemia induced the increase in plasma hIRα, we followed the clinical courses of eight new-onset type 1 diabetic patients (Fig. 2C), all of whom were sent to the Tokushima University Hospital due to the onset of type 1 diabetes (T1DM group). At admission, FPG, A1C, and hIRα levels were 350 ± 82 mg/dl, 12.6 ± 2.1%, and 5.5 ± 1.8 ng/ml, respectively. At discharge after 6–10 days of intensive insulin therapy, their blood glucose levels had almost normalized (124 ± 31 mg/dl) and hIRα levels had decreased to 3.3 ± 1.0 ng/ml. After 1 month, the patients' glycemic control levels were maintained with insulin therapy and their A1C, glycoalbumin, and hIRα levels were 9.2 ± 1.3%, 24.9 ± 2.4%, and 2.3 ± 0.8 ng/ml, respectively. At that moment, only hIRα almost reached plateau, but glycoalbumin and A1C levels were further decreased at 2 months. The results presented in Figs. 2B and C strongly suggest that plasma hIRα levels promptly parallel any changes in blood glucose levels and changed more rapidly than glycoalbumin or A1C. In addition, there was a significant positive correlation of hIRα levels with blood glucose levels including blood glucose, glycoalbumin, and A1 C (online appendix Table 1, in-text Fig. 2D, online appendix Fig. 2A, and online appendix Fig. 2B). The regression values were higher in the type 1 diabetic patients (Fig. 2D) than in the type 2 diabetic patients (online appendix Figs. 2A and B). Furthermore, during long-term follow-up of a type 1 diabetic patient, hIRα level completely paralleled with glycoalbumin and A1C levels and changed more rapidly and dynamically than glycoalbumin and A1C levels (Fig. 2E).

To further confirm these clinical observations in vivo, we initially used transgenic (TG) mice that systemically express kinase-deficient hIR (hIR^K1030MTG) (11,12). Although these TG mice express a kinase-deficient mutant hIR (K1030M; Lys→Met at residue 1,030 in the kinase domain of the β-subunit), they do not show diabetic phenotypes (11,12). Furthermore, the mutation did not appear to affect the receptor release based on analyses of cultured cells (data not shown). In addition, we recently generated and analyzed wild-type (WT) hIR–expressing TG mice that systemically expressed at least fourfold higher amounts of hIR than hIR^K1030MTG mice (data not shown).

Diabetes was induced in either TG or control mice (nontransgenic [NTG] littermates) by intraperitoneal streptozotocin (STZ) injection. After 4–7 days, diabetic hIR-TG mice showed a markedly higher hIRα level, and this was strongly correlated with their blood glucose levels (n = 14, P < 0.001, R = 0.80; Fig. 3C). On the other hand, NTG diabetic mice or nondiabetic hIR-TG mice showed negligible hIRα levels, similar to those of NTG nondiabetic mice. In the NTG diabetic mice, the endogenous mouse IRα level was expected to increase but was not detected with the hIRα-specific ELISA systems. The level of full-length IR was also confirmed to be negligible using these ELISA systems, indicating that the plasma hIRα was not derived from damaged cells (Fig. 3B). Furthermore, to investigate whether the IRα responses were attributed to the changes of glucose level, we treated the STZ-induced diabetic mice with insulin (subcutaneous injection twice a day of NPH human insulin, Novolin-N; NovoNordisk, Bagsv\jrd, Denmark). As shown in Fig. 3D, the IRα levels were promptly decreased by insulin therapy paralleled with blood glucose level. Moreover, after a transient pause (3 days) of insulin treatments, both blood glucose and IRα levels were reelevated and then promptly redeclined by the resumption of insulin therapy. Notably, some mice failed to develop diabetes even with same dose of STZ and did not exhibit elevated IRα levels. We also treated nondiabetic mice with sustained-release insulin implants (Linshin Canada, Toronto, ON, Canada), and these mice showed normoglycemia and marked weight gain, but the IRα levels were changed little (data not shown), indicating that neither STZ nor insulin alone changed plasma IRα levels. We further estimated the half-life of the IRα in blood using the WT-hIR-TG mice. After diabetes induction by STZ, insulin therapy was initiated and plasma IRα levels decreased, parallel with glucose levels. As shown in Fig. 3E, the circulating plasma IRα levels decreased to 49.3 ± 13.8% (n = 9) at 6 h after the treatment, 28.7 ± 17.5% (n = 19) at 24 h, and 19.0 ± 12.3% (n = 19) at 48 h.

In this study, we provide evidence that soluble hIR ectodomain (α-subunit and a part of β-subunit), but not intact hIRβ or whole hIR, exists in human plasma. Furthermore, patients with type 2 diabetes, as well as those with type 1 diabetes, showed a significantly elevated plasma hIRα level compared with control subjects, as measured using newly established hIR-specific ELISA systems (i.e., hIRα-specific and whole hIR–detectable systems). Plasma hIRα level positively correlated with blood glucose, glycoalbumin, and A1C levels. In addition, comparable amounts of plasma insulin appeared to bind to hIRα. Moreover, hyperglycemia was confirmed to induce hIRα release in STZ-induced diabetic mice transgenically expressing hIR.

The ectodomains of receptors for several cytokines and growth factors have been found to circulate in plasma (1–3). The existence of soluble IR in human plasma has previously been suspected, since several studies have reported shedding of IR from cultured cells (e.g., IM-9 human lymphoblasts, MCF-7 human breast cancer cells, HepG2 human hepatoma cells, and human lymphocytes, as well as 3T3-L1 mouse fibroblasts transfected with hIR) (4,6,7). We have also observed hIRα release from CHO-hIR cells, HepG2 cells, hIR-expressing L6 myocytes, and hIR-expressing 3T3-L1 adipocytes (T. Obata, K. Yokoyama, E. Okamoto, T. Yuasa, Y. Ebina, unpublished data).

Although Pezzino et al. (5) reported the detection of both IRβ and IRα in healthy human plasma that presented insulin-stimulated autophosphorylation activity without tyrosine kinase activity against exogenous substrates, we did not observe a protein corresponding to the β-subunit, at least in its intact form, in either of the ELISA (full-length and β-subunit) systems or in immunoblotting (data not shown). We showed that IRα is released concomitantly with part of the extracellular domain of IRβ (online appendix Fig. 3) into the plasma of diabetic patients using Superdex gel-filtration column (Fig. 1B). In fact, many membrane proteins (e.g., tumor necrosis factor [TNF] receptor, epidermal growth factor receptor, and interleukin [IL]-6 receptor) have soluble ectodomains that are usually cleaved at a site in the stalk region between the transmembrane segment and the globular extracellular domain. The distances of the cleavage sites from the plasma membrane are ∼1–43 amino acids (13). In accordance with this report, as shown in Fig. 1B, the retention times of immunoreactive soluble hIR in human plasma and standard hIR ectodomain protein derived from CHO-IR-SspI cells were similar, suggesting that the molecular weight of both proteins under the nonreducing condition were almost same. Moreover, the apparent molecular weight was ∼370 kDa. Considering all of this information, the ectodomain shedding appeared to occur at a site in the stalk region of the extracellular peritransmembrane region, as in the case of other soluble receptors; as a result, the soluble IRα in human plasma appeared to exist with parts of the extracellular region of β-subunits (online appendix Fig. 3). Since the actual cleavage site is unclear at this time, further experiments are necessary. Recently, the crystal structure of IR-ectodomain has been clarified (14), therefore these data must be helpful.

In the present study, the clinical results indicated that the release of hIRα into plasma may be augmented by hyperglycemia, and the subsequent in vivo study using hIR-TG mice supports these findings. On the other hand, plasma hIRα level was not correlated with markers of insulin secretion, suggesting that the release of hIRα into plasma may be regulated by blood glucose level, rather than secreted insulin. The concentration of hIRα in patients with diabetes was on average ∼20% (20.3 ± 51.2%, n = 88) of the fasting IRI levels on a molar basis. Comparable amounts (∼10–20%) of insulin were immunodepleted by anti-IRα antibody (Table 2), suggesting that appreciable amounts of plasma insulin appeared to bind to plasma soluble hIRα. In turn, the absolute amount of plasma insulin sequestered by hIRα in diabetic patients seemed to be much larger than that in normoglycemic subjects. In cases of the binding of insulin by insulin autoantibody, the antibody first sequesters insulin and then releases insulin after a time (15). Therefore, the apparent insulin-bound hIRα fraction may be underestimated. In addition, although we expected released soluble IR to participate in insulin resistance as one of the factors that contribute to glucose toxicity by sequestering plasma insulin, we found no correlations between hIRα level and other parameters reflecting insulin resistance (i.e., homeostasis model assessment of insulin resistance) or fasting IRI level in the present study.

According to clinical data (Fig. 2C and E), hIRα level changed more rapidly and dynamically than A1C or glycoalbumin levels. Therefore, we estimated the half-life of hIRα using STZ-induced diabetic TG mice (Fig. 3E). The half-life was estimated to ∼6 h and was much shorter than that of A1C (30 days) or glycoalbumin (∼17 days), suggesting that IRα could be a more rapid glycemic marker. We examined the daily profile of hIRα levels and observed little change throughout the day (online appendix Fig. 1F) in normoglycemic subjects. However, considering such a short half-life, it may be possible to observe daily change in patients with diabetes, especially those with brittle diabetes. Furthermore, the shedding of IR appeared to depend on biological response to hyperglycemia, while A1C or glycoalbumin are elevated in response to hyperglycemia simply by nonenzymatic mechanisms. Thus, plasma hIRα levels may reflect biological response to hyperglycemia.

Many membrane proteins have soluble ectodomains that are subject to proteolytic release, i.e., the process known as shedding (16). In most cases, shedding is caused via proteolytic cleavage by members of the ADAM (a disintegrin and metalloprotease) family of membrane-tethered zinc metalloproteinases (MMPs) (16). We have also observed suppressive effects of a general MMP inhibitor (GM-6001) on IRα shedding (T. Obata, K. Yokoyama, E. Okamoto, Y. Ebina, unpublished observations), suggesting the involvement of MMPs in this process. Most MMPs have COOH-terminal cytoplasmic tails that have been shown to be phosphorylated by various protein kinases including protein kinase C (PKC) (17,18). This phosphorylation subsequently activates protease activity (17,18). Under high-glucose conditions, the activation of PKC by de novo synthesis of diacylglycerol is well known (19,20), including in endothelial cells (21), which also express IR (22). All these data suggest a vicious cycle for glucose toxicity; high glucose stimulates PKC activation, which in turn activates MMPs that cause the release of soluble IR into the plasma, which sequesters insulin, thereby further raising glucose levels. The hypothesis regarding the vicious cycle is supported by our prior studies showing that the injection of purified hIRα into mice with elevated blood glucose levels in vivo (8). At this time, the source organ of soluble IR is unclear. Of note, although we detected plasma soluble IR in hIR-TG mice, the expression level of IR in the liver was extremely lower than that in other organs.

Regarding the shedding of other membrane proteins in the diabetic state, Lim et al. (23) reported that soluble CD40 ligand and soluble P-selectin levels were increased in plasma from patients with diabetes, and they proposed that the increments promoted atherothrombotic complications in cardiovascular disease. In the diabetic condition, other receptor proteins as well as IR are possibly shed from the cell membrane. However, the shedding of these membrane proteins, except IR, was not associated with glycemic control. In our study, we measured plasma soluble TNF-αR1 and soluble IL-6R levels and compared them with plasma hIRα levels (online appendix Fig. 4A and B). Plasma soluble TNFα-R1, but not soluble IL-6R, levels significantly correlated with hIRα levels (P < 0.001), suggesting that there may exist both common and distinct mechanisms between the shedding of IR and these receptors. Thus, the possible existence of various mechanisms for shedding needs to be considered.

In general, the physiological roles of ectodomain shedding of membrane receptors were diverse and complex. In the case of growth hormone receptor (GHR), the shedding of GHR generates growth hormone binding proteins that interact with growth hormone with high affinity and therefore downregulate the availability of the ligand (24). As in the case of soluble GHR, the majority of soluble receptor ectodomains appeared to inhibit the actions of their ligands (25), but soluble IL-6 receptor seems to act agonistically upon IL-6 binding (26). In contrast, in the case of TrkA (the receptor for nerve growth factor), residual membrane-associated fragments were found to be phosphorylated after ectodomain shedding, associated with intracellular signaling molecules, potentiating its action (27). In this case, the process appeared to play a compensative positive role. In fact, expression of ectodomain-truncated IR has also been shown to exhibit ligand-independent activation (28). At present, the physiological role of the process of IR ectodomain shedding remains unclear. Even in normoglycemic subjects, some basal hIRα was detected. Further study is necessary to clarify this issue.

Toshiyuki Obata,¹ Ichiro Yokota,² Kazuhiro Yokoyama,¹ Eiji Okamoto,³ Yoshiko Kanezaki,¹ Yoshinori Tanaka,¹ Hiroshi Maegawa,⁴ Kiyoshi Teshigawara,¹ Fumiko Hirota,⁵ Tomoyuki Yuasa,¹ Kazuhiro Kishi,¹ Atsushi Hattori,¹ Seiichi Hashida,⁸ Kazuhiko Masuda,⁶ Mitsuru Matsumoto,⁵ Toshio Matsumoto,⁷ Atsunori Kashiwagi,⁴ and Yousuke Ebina.¹

From the ¹Division of Molecular Genetics, ⁵Division of Molecular Immunology, Institute for Enzyme Research, ²Department of Pediatrics, ⁷Department of Medicine and Bioregulatory Science, Graduate School of Medicine, The University of Tokushima, Tokushima, Japan; ³Medical & Biological Laboratories Co. Ltd., Nagoya, Japan; the ⁴Department of Medicine, Shiga University of Medical Science, Otsu, Shiga, Japan; ⁶Naruto Hospital, Naruto, Tokushima, Japan; and the ⁸Division of Life Style Diseases, Institutes for Health Science, Tokushima Bunri University, Tokushima, Japan.

T.O. and I.Y. contributed equally to this study. T.O. is currently affiliated with the Department of Medicine, Shiga University of Medical Science, Otsu, Shiga, Japan.

T.O., K.K., and Y.E. are supported by research grants from the Ministry of Education, Science, Technology, Sports and Culture of Japan. T.O. is supported by research grants from the 21st Centers of Excellence program of the University of Tokushima, the Insulin Research Foundation of Novo Nordisk Pharma, the Uehara Memorial Foundation, the Mitsui Social Welfare Foundation, the Japan Diabetes Foundation, and the Takeda Science Foundation.

The authors thank Dr. R. Roth (Stanford University, Stanford, CA) for anti-IRα antibody (5D9), K. Wakamatsu, M. Yamanaka, and C. Takehara for technical and secretarial assistance, and Dr. Y. Shintani (University of Tokushima, Tokushima, Japan) for assistance in blood sampling. We also thank Dr. J. Miyazaki (Osaka University, Osaka, Japan) for pCAGGS expression vector for TG mice.