In Vivo Imaging of Insulin Receptors in Monkeys Using ¹⁸F-Labeled Insulin and Positron Emission Tomography

Abstract

We previously described a prosthetic group methodology for incorporating ¹⁸F into peptides and showed that ¹⁸F-labeled insulin (¹⁸F-insulin) binds to insulin receptors on human cells (IM-9 lymphoblastoid cells) with affinity equal to that of native insulin (1). We now report studies using ¹⁸F-insulin with positron emission tomography to study binding to insulin receptors in vivo. Positron emission tomography scans were performed in six rhesus monkeys injected with 0.3–1.4 mCi of ¹⁸F-insulin (~0.1 nmol, SA 4–11 Ci/μmol). Integrity of the tracer in blood, determined by immunoprecipitation, was 94% of control for the first 5 min and decreased to 31% by 30 min. Specific, saturable uptake of ¹⁸F was observed in the liver and kidney. Coinjection of unlabeled insulin (200 U, ~1 nmol) with the ¹⁸F-insulin reduced liver and increased kidney uptake of the labeled insulin. Liver radioactivity was decreased by administration of unlabeled insulin at 3 min, but not 5 min, after administration of the tracer, while some kidney radioactivity could be displaced 5 min after injection. Clearance of ¹⁸F was predominantly in bile and urine. ¹⁸F-insulin is a suitable analogue for studying insulin receptor-ligand interactions in vivo, especially in the liver and kidney.

Previously we reported a prosthetic group methodology for synthesizing an ¹⁸F-insulin derivative, designed for studying the insulin receptor in vivo using positron emission tomography (PET;1). The tracer binds to human (IM-9) lymphoblastoid cells in vitro with an affinity for the receptor that is at least equal to that for ¹²⁵I-insulin. Here we report on studies involving ¹⁸F-insulin in nonhuman primates and show that the tracer binds to receptors in the liver and kidney.

RESEARCH DESIGN AND METHODS

Macaca mulatta monkeys (3–5 kg) were maintained on a standard diet by the Veterinary Resources Program of the National Institutes of Health. On the morning of the study the animals were anesthetized using ketamine, pentothal, and isoflurane. An intravenous catheter was inserted in the saphenous vein and a cutdown was performed for placement of an arterial line in the femoral artery. Patency of the arterial catheter was maintained by pressurized infusion of heparinized saline (60 U heparin/mL).

The Posicam 6.5 body scanner (Positron, Houston, TX) was used for the studies. Depending on the experiment design, the animals were positioned to obtain transverse or sagittal plane tomographs. A transmission scan with ¹⁸F was performed for later attenuation correction. ¹⁸F-Insulin was prepared as previously described (1). Briefly, a derivitized insulin with a bromomethylbenzoyl prosthetic group covalently linked to the amino terminus of the β-chain (B1) was prepared in advance. On the day of the scan, ¹⁸F was introduced by the F- reaction, and the material was purified by high-performance liquid chromatography. The tracer dose and specific activity are given in the Figure captions.

¹⁸F-Insulin was injected as a bolus over 1 min, and arterial blood samples (0.3 mL) were taken at 30-sec intervals at the start of the procedure, decreasing to every 5–10 min by the end of the procedure, for determination of blood radioactivity and tracer integrity. Coincidence data were acquired in list mode and were later reformatted and reconstructed to produce images with in-plane resolution of 9 mm and axial resolution of 12 mm. Each procedure lasted 1–2 h, depending on experiment design. Human recombinant insulin (Humulin, Lilly, Indianapolis, IN) was administered by bolus injection as indicated in the Figure captions. Blood glucose concentrations were maintained in the normal or hyperglycemic range by infusion of dextrose after administration of insulin.

Scans were analyzed using the MIRAGE image analysis system (Department of Nuclear Medicine, NIH, Bethesda, MD). Time-activity curves for elliptical regions of interest representing liver, kidney, gall bladder, and urinary bladder were analyzed. Data were expressed in nCi/mL of tissue radioactivity. When sequential injections were performed, the same animal was used for both studies, the tracer used was from the same batch, and injections were ~1 h apart. The specific activities corrected for decay to the time of the injection are given in the figure captions. The tissue activity from the first study and the second study (corrected for differences in injected activity, residual counts in the tissues at the time of the second injection, and radioactive decay) are plotted with a common origin.

Tracer integrity was determined by immunoprecipitation with antiinsulin antibodies. At each time point, an aliquot of plasma was immunoprecipitated under conditions of excess antibody. Guinea pig antiporcine insulin antibody (first antibody; preparation 619, Indiana University School of Medicine, Indianapolis) was diluted in assay buffer (1/2500 final in veronal buffer, pH 8.5 with 2% human serum albumin and rabbit fraction II (ICN, Lisle, IL). For each time point, an aliquot (100 μL) of primate plasma was added to tubes (12-by-75-mm tubes, final volume 0.5 mL) containing 100 μL first antibody and assay buffer. The tubes were centrifuged and incubated for 105 min at 4°C. Guinea pig serum (1 μL) was added prior to the addition of 50 μL antiguinea pig serum antibody (second antibody; Linco, St. Louis, MI). In the presence of the second antibody, the tubes were incubated for an additional 45 min at 4°C. The supernatants were pipetted into a separate tube, and the radioactivity in both the precipitate and supernatant were counted sequentially in a gamma counter (Packard Cobra Gammacounter Model 5003, Packard, Downers Grove, IL). Data are expressed as a percent of counts bound compared to the binding of the tracer in control tubes (%B/B_O); control tubes contained tracer and an aliquot of primate plasma obtained prior to injection of the tracer.

RESULTS

When ¹⁸F-insulin was injected intravenously, radioactivity accumulated rapidly in the liver and kidney (Fig. 1). Uptake in the liver was maximal 2–6 min after administration of the tracer, and in later images (15–20 min) diminished in liver parenchyma and increased in the gallbladder. In images taken 50–55 min after injection of the tracer, the activity was predominantly within the gall bladder and was seen in the small intestine as well (image not shown).

FIG. 1.

Summed positron emission tomography scans of liver (top row) and kidney (bottom) at 2–6, 15–20, and 50–55 min from transaxial scans after intravenous injection of ¹⁸F-insulin. Gray scale is shown at right, with white representing the greatest radioactivity.

Quantitative data, expressed as nCi/mL of tissue, are shown for the liver in Fig. 2. Two sequential studies were performed in the same animal on the same day. In the first study, tracer alone was administered (Fig. 2A, solid circles); in the second study, unlabeled insulin (200 U, 1 nm) was coadministered with the tracer (Fig. 2A, open triangles). Coinjection of unlabeled insulin with the tracer prevented receptor binding of the ligand and reduced the uptake of radioactivity by 50%.

FIG. 2.

Time vs. ¹⁸F radioactivity curves for liver (A and B) and plasma (C) after sequential injections of ¹⁸F-insulin. Two sequential tracer doses from the same batch were administered to animals on the same day. In each case the first injection (●) was tracer alone. In A, the first dose was 0.674 mCi (SA 5.4 Ci/μmol), and the second dose (0.656 mCi, SA 2.98 Ci/μmol) was coadministered with 200 U of unlabeled insulin (△). In B, the first dose was 0.462 mCi (SA 7.06 Ci/μmol) and the second injection of tracer (0.592 mCi, SA 3.86 Ci/μmol) was followed after 3 min by injection of 50 U of unlabeled insulin (△). C shows the plasma radioactivity curve from the same study as B. In both studies, the residual radioactivity from the first injection has been subtracted from the second scan by exponential extrapolation.

We did several studies to examine reversibility of the binding of tracer by the liver. In these studies, as in the previous studies, the animals were injected sequentially with two doses of tracer on the same day. Both the first injection (Fig. 2B, solid circles) and the second injection (Fig. 2B, open triangles) were with tracer alone, but the second injection was followed after 3 min by injection of 50 U of unlabeled insulin. We observed a sharp decrease in liver activity coincident with injection of unlabeled insulin and a concomitant increase in plasma radioactivity (Fig. 2C). This is consistent with discharge of insulin from the receptor into blood by saturation of binding sites with unlabeled insulin.

In a similar experiment, injection of 10 U of unlabeled insulin 3 min after the second tracer dose blocked subsequent uptake by the liver and slightly increased blood activity but did not discharge activity from the liver (data not shown).

Images of renal radioactivity are shown in Fig. 1. We observed uptake of radioactivity in the early images (bottom) in a region corresponding to renal parenchyma, while in later images the activity moved into the collecting system (15–20 min) and was subsequently visualized in the ureters and bladder. Quantitative data for kidney parenchyma, shown in Fig. 3, are from the same studies upon which Fig. 2 is based. As observed in the liver, when tracer alone was administered, the kidney parenchymal activity was maximal after 4–8 min (Fig. 3A, closed circles). The pattern of uptake in the kidney after coinjection of the tracer with unlabeled insulin (200 U, 1 nm) was different, with a gradual increase in activity that reached a maximum after 20 min (Fig. 3A, open triangles). The difference between the two curves represents a saturable component of renal parenchymal uptake. The increased amount of radioactivity in the kidney after the second injection of tracer can be explained by the increase in blood radioactivity.

FIG. 3.

Time vs. ¹⁸F radioactivity curves for kidney parenchyma after sequential injections of ¹⁸F-insulin. Data are from the experiments also used to plot Fig. 2A arid 2B, respectively. In 3A, the second injection of tracer (△) was with 200 U of unlabeled insulin. In 3B, the second injection of tracer was followed after 3 min by injection of 50 U of insulin.

Reversibility of binding in the kidney was also observed. When 50 U of unlabeled insulin was injected 3 min after the tracer dose, a sharp decrease in renal parenchymal activity was observed (Fig. 3B, open triangles). This was coincident with the decrease in liver activity (Fig. 2B, open triangles), and the increase in plasma, activity (Fig. 2C, open triangles) observed in the same study. Injection of 10 U of insulin at 3 min prevented further uptake by kidney tissue, but did not discharge counts from the tissue.

An additional experiment was performed to compare reversibility of binding in the liver and kidney. In this experiment, an animal was injected with tracer and, after 5 min, 100 U of unlabeled insulin was administered (Fig. 4). We observed a decrease in kidney activity (Fig. 4, open triangles) but not in liver activity (Fig. 4, closed circles) immediately after administration of the unlabeled insulin. This suggests that in the liver the counts are sequestered and cannot be discharged into the blood within 5 min of tracer administration, whereas in the kidney a component of the tracer is not irreversibly bound.

FIG. 4.

Time vs. ¹⁸F radioactivity curves for kidney (△) and liver (●) in an experiment in a single animal. Tracer (0.331 mCi, SA 6.7 Ci/μmol) was injected at time zero, and 100 U of unlabeled insulin was injected after 5 min.

In the studies plotted in Figs. 2A and 3A, the amount of activity in the liver and kidney cortex 5 min after injection was calculated to be 16% and 10%, respectively, of the total activity administered. Organ weights for the liver and kidney were estimated at 21 and 4 mL/kg, respectively. Residual blood activity (using blood volume = 75 mL/kg) accounted for 11% of total activity. Thus, 62% of the total activity administered was presumed to be taken up by muscle and other tissues and in the renal collecting system. We did not design these studies to determine total recovery, nor to study skeletal muscle uptake. We did, however, observe saturable uptake in the lateral head and neck that could represent uptake in skeletal muscle or salivary gland or both (data not shown). We observed that the integrity of the tracer, as measured by immunoprecipitation with antiinsulin antibodies, was 95% of control for the first 5 min after administration, then decreased rapidly (Fig. 5). This is consistent with the uptake of tracer by the liver and kidney with subsequent metabolism of the insulin analogue.

FIG. 5.

Immunoprecipitation of ¹⁸F-labeled insulin from serum by antiinsulin antibodies.

Quantitative estimates of receptor number can be derived from the in vivo data presented here by determining the bound/free ratio. Consider the liver data in Fig. 2, in which tissue radioactivity accumulation after tracer alone and after tracer plus a saturating amount of unlabeled insulin are shown. If the coinjection of 200 U of insulin is sufficient to block all specific binding, the amount of specifically bound tracer can be estimated from the concentration difference between the blocked first and second scan. This approach underestimates the specific binding because of the increase in blood activity during the scan with unlabeled insulin. The concentration of free tracer can be estimated from the concentration in plasma, after correction for metabolites of the tracer.

Assuming a single receptor class that obeys conventional bimolecular kinetics, bound/free = B_max/K_D where B_max is the concentration of free receptor during the tracer study and K_D is the dissociation equilibrium constant. For this study B_max/K_D for the liver was estimated to be 4.2. The derivation assumes that the free pool in tissue is in rapid equilibrium with that in plasma and that equilibrium has been reached. These assumptions were supported by constant bound/free ratios >4 between 10 to 25 min for the liver. Constant bound/free values were not attained for the kidney. Note that for the kidney the concentration of nonspecifically bound tracer in the study with coinjection of unlabeled insulin exceeded that of the study with tracer alone at 20 min, due to increased blood availability of the tracer resulting from decreased receptor-mediated uptake of the ligand. A more complete understanding of the fate of the tracer and a more appropriate mathematical model is required to produce more precise estimates of receptor number and affinity.

DISCUSSION

Studies of insulin clearance with iodinated insulin tracers have shown that 50% of insulin in portal vein blood is cleared by the liver (2). The clearance mechanism works by binding insulin to specific receptors on the surface of hepatocytes, followed by an internalization of receptor-ligand complexes (3,4); it is prevented by unlabeled insulin. Autoradiography of hepatocytes has clarified the sequence of events by which labeled insulin is taken up by the liver. After intraportal injection of labeled insulin activity is localized to hepatic plasmalemma for the first few minutes (5), the receptor-ligand complexes are subsequently internalized by endocytosis, with radioactivity reaching a maximum in 5–10 min at body temperature. Endosomes containing insulin are found associated with the Golgi region and a substantial proportion fuse with bile canaliculi, releasing their contents into the bile ducts (5). Intact insulin is present in bile in greater amounts than in systemic venous blood, and the concentration of insulin in bile mirrors changes in endogenous insulin secretion (6–8).

We observed rapid, specific uptake of ¹⁸F-insulin by the liver and found that hepatic activity could be discharged 3 min after tracer administration, but not after 5 min, Our findings are consistent with the rapid sequestration of iodoinsulin tracers by the liver, resulting from internalization of insulin-receptor complexes by hepatocytes (5). The subsequent appearance of ¹⁸F in bile that we observed is also consistent with the rapid, biliary excretion of insulin and its metabolites seen in previous studies (5). Excretion of activity in bile was not observed in the studies of Sodoyez et al. (9,10), although activity in the duodenum is seen in some published images. Duodenal activity could be explained on the basis of the gastroenteric cycle of iodine, in which case the activity should be decreased by iodine loading prior to tracer administration.

Radioactivity observed in bile in our studies is probably a metabolite of the analogue. Whether there is a component of intact insulin analogue in bile cannot be determined from our studies. Metabolism of ¹⁸F-insulin occurs during passage through the hepatocyte (5). Endosomes containing internalized insulin-receptor complexes during passage through the hepatocyte have been observed to fuse with lysosomes and directly with bile canaliculi (5). Metabolism of insulin within bile does not appear to occur.

Renal clearance of insulin is complex and is mediated in part by insulin receptors (11), which, along with a large amount of nonsaturable binding, are present in preparations of luminal (brush border) and contraluminal membranes (12,13). Both luminal and contraluminal sites are involved in insulin degradation by the kidney. Ninety-nine percent of insulin filtered at the glomerulus is taken up by proximal tubule cells, and degradation of insulin is accomplished in part by insulin-receptor-mediated endocytosis and in part by charge-specific mechanisms that do not involve the insulin receptor (11). Insulin that reaches the contraluminal membrane of tubular cells via the postglomerular, peritubular capillaries is also degraded, but the process does not appear to involve endocytosis (14,15). Contraluminal membrane receptors for insulin are located throughout the nephron, particularly along the ascending loop of Henle and the distal convoluted tubule (11).

Our observation of specific binding to kidney cortex is not inconsistent with studies using iodinated tracers in rats and man with low affinity insulin and in subjects with absent insulin receptors (9,10,16). As in studies with iodoinsulins, we observed increased total renal activity with co-injection of labeled and unlabeled insulin. We attribute this to an increase in the amount of tracer filtered at the glomerulus and degraded by non-saturable mechanisms, resulting from the marked reduction in hepatic uptake. We have not done experiments with low-affinity insulin or in patients with defective insulin binding, but we would expect to see the same results as observed with a saturating amount of unlabeled insulin. In our experiment, we used the images from the PET scan to choose a region of interest representing predominantly renal cortex. The saturable and reversible component of kidney binding was seen in this compartment.

We observed a two- to fourfold increase in specific kidney binding at early time points compared with liver binding. Because we did not measure renal blood flow, it is difficult to make quantitative comparisons between liver and renal specific binding. A decrease in renal blood flow would decrease the nonspecific filtration and degradation and increase the apparent specific binding in kidney. The increase in kidney-specific binding could also be caused by increased receptor number, affinity, or both. We believe the difference is more likely caused by higher affinity of the renal receptor for the insulin analogue. Although we have not studied the affinity of our insulin analogue in kidney or liver, we do know that the affinity for the human lymphoblastoid cell (IM-9) insulin receptor is equal to or perhaps slightly greater than for porcine insulin (1). Significant differences in binding affinity between hepatic and extrahepatic insulin receptors have been observed, with significantly lower hepatic binding for some tracers (17). Further studies are needed to determine the reasons underlying the relative increase in renal uptake.

A component of renal uptake was reversible 5 min after injection of ¹⁸F-insulin, at a time when hepatic activity was sequestered and could not be displaced by unlabeled insulin. These findings are consistent with known differences in insulin receptor endocytosis in the two tissues. Hepatocytes rapidly internalize cell-surface insulin-receptor complexes, so that by 5–10 min after administration of insulin tracers the activity is sequestered in endosomes (4,5). Endocytosis of insulin-receptor complexes by luminal membranes of the proximal tubule cells also occurs, but receptor-mediated insulin degradation by contraluminal membranes is not endocytosis mediated and is unaffected by agents that inhibit internalization (14,15). The displaceable component of binding in kidney we observed at a time when hepatic activity was completely sequestered is consistent with these known differences in receptor internalization in the two tissues.

In each experiment the blood radioactivity curve showed an exponential decrease with time without an increase in blood radioactivity at later time points, as has been observed with studies using iodoinsulin. In the studies where unlabeled insulin was injected after the tracer, we always observed a sharp increase in blood radioactivity proportionate to the amount of insulin injected. Furthermore, subsequent clearance of labeled insulin was prolonged compared with the control injections. This is consistent with displacement of labeled insulin from the receptor compartment into blood, with subsequent delayed clearance due to saturation of receptor-mediated uptake. Tracer integrity, as assessed by immunoprecipitation, was 90–100% of control for the first 5 min after injection, then fell rapidly. Thus, during the time that tissue uptake by the liver and kidney occurred, the tracer was intact, and only after uptake did significant metabolism occur.

The ¹⁸F-labeled insulin appears to be a suitable analogue for studying insulin receptors in vivo. The timing and reversibility of uptake and excretion of the analogue closely mimics the known physiological processes by which insulin is cleared. Tracer concentrations estimated to be in the range of 50–100 pM yielded excellent images, consistent with a high-affinity ligand-receptor interaction. The methodology for labeling insulin with ¹⁸F (fluoromethylbenzoyl derivation of amines) is now shown to be suitable for in vivo studies and is sufficiently stable to metabolism to apply to other peptides of biological interest. PET does, however, carry some disadvantages that limit its application. The isotopes are short-lived and must be produced on-site with a cyclotron. The radiochemical synthesis involves large activities and must be completed within 2–3 half-lives of the end of the cyclotron run. These factors increase the cost and difficulty of using positron-labeled tracers, which may be offset by the increased resolution PET tomographs offer.