Noninvasive MRI of β-cell function using a Zn²⁺-responsive contrast agent

Abstract

Elevation of postprandial glucose stimulates release of insulin from granules stored in pancreatic islet β-cells. We demonstrate here that divalent zinc ions coreleased with insulin from β-cells in response to high glucose are readily detected by MRI using the Zn²⁺-responsive T₁ agent, GdDOTA-diBPEN. Image contrast was significantly enhanced in the mouse pancreas after injection of a bolus of glucose followed by a low dose of the Zn²⁺ sensor. Images of the pancreas were not enhanced by the agent in mice without addition of glucose to stimulate insulin release, nor were images enhanced in streptozotocin-treated mice with or without added glucose. These observations are consistent with MRI detection of Zn²⁺ released from β-cells only during glucose-stimulated insulin secretion. Images of mice fed a high-fat (60%) diet over a 12-wk period and subjected to this same imaging protocol showed a larger volume of contrast-enhanced pancreatic tissue, consistent with the expansion of pancreatic β-cell mass during fat accumulation and progression to type 2 diabetes. This MRI sensor offers the exciting potential for deep-tissue monitoring of β-cell function in vivo during development of type 2 diabetes or after implantation of islets in type I diabetic patients.

The pancreatic islet is a highly vascularized, multicelled organelle capable of rapidly sensing changes in blood glucose. An increase in postprandial glucose stimulates release of insulin from granules stored in islet β-cells. The mass of pancreatic β-cells in humans generally increases throughout adult life (1), whereas the functional sensitivity of β-cells to high glucose tends to decline with age (2) and can be reduced significantly before type 2 diabetes is diagnosed clinically (1, 2). We show here that divalent zinc ions (Zn²⁺), coreleased with insulin from β-cells in response to high glucose, are readily detected by MRI using the Zn²⁺-responsive T₁ agent GdDOTA-diBPEN (3). Previous studies have shown that the Zn²⁺ concentration in the extracellular space of islets approaches 475 μM during glucose-stimulated insulin secretion (GSIS) (4), well above the detection limit of GdDOTA-biBPEN. Thus, this MRI sensor offers the potential to monitor β-cell function (release of insulin and Zn²⁺) in vivo during development of type 2 diabetes or after implantation of islets in type 1 diabetic patients.

Noninvasive imaging of islet mass and function in vivo has been an elusive goal for many years. Zn²⁺ ions, required for proper storage of insulin in granules in β-cells (≈11.6 Zn²⁺ ions are packaged per insulin hexamer), are coreleased from β-cells during exocytosis of insulin (5, 6). The role of Zn²⁺ in glucose homeostasis, including insulin packaging in granules and as a signaling species, has been widely studied (5, 7, 8). When released with insulin, the free Zn²⁺ ions bind weakly to extracellular matrix proteins in the immediate vicinity of β-cells to be available for recycling back into β-cells for further insulin storage and to prevent β-cell necrosis that can occur if the pancreas accumulates too much Zn²⁺ (4, 5). The Zn²⁺ concentration in insulin secretory granules has been estimated to be ≈20 mM (5, 9), so release of insulin from granules elevates Zn²⁺ in the extracellular space of the β-cell to a range of 400–500 μM (4), an order of magnitude higher than the basal level of Zn²⁺ in serum [40 μM or less (10, 11)].

Many elegant molecular designs have been proposed for Zn²⁺ sensors, most notably fluorescence-based sensors (12–14) and a few MRI sensors (3, 15–18). Unlike fluorescence spectroscopy, MRI is generally considered by many to be too insensitive for detection of Zn²⁺ at physiological levels. However, this low sensitivity could potentially be used to advantage if a MRI sensor does not respond to low levels of Zn²⁺ present during euglycemia but does respond by showing increased contrast in the MR image when the sensor becomes exposed to higher concentrations of Zn²⁺ typical of that present near the surface of β-cells during hyperglycemia. A MRI sensor with these properties may allow whole-body detection of Zn²⁺ only in those tissues where the extracellular concentration is high and physiologically important. Antkowiak et al. (19) recently reported the first example of imaging β-cell function by showing that Mn²⁺-enhanced MRI may be used to detect temporal differences in T₁ enhancement of the pancreas between normal and streptozotocin (STZ)-treated mice after a bolus injection of glucose. MRI enhancement of the pancreas in that study was thought to reflect influx of Mn²⁺ into islet β-cells through voltage-gated Ca²⁺ channels because those cells respond to high glucose. Here we report another approach for imaging β-cell function by demonstrating that Zn²⁺ ions can be detected in T₁-weighted MR images by use of a gadolinium-based Zn²⁺ sensor, GdDOTA-diBPEN, because these ions are released from β-cells during GSIS. It was previously shown that albumin has a low affinity for this agent in the absence of Zn²⁺ but a moderate affinity for the ternary GdDOTA-diBPEN–(Zn)₂ complex that readily forms in the presence of Zn²⁺ ions (3). The resulting GdDOTA-diBPEN–(Zn)₂ also has an enhanced water proton relaxivity (r₁) when bound to albumin (Fig. 1), the magnitude of which depends on field strength. In the absence of Zn²⁺, the r₁ of GdDOTA-diBPEN is comparable to other low molecular weight Gd³⁺ complexes used clinically (3), so one can administer the agent at a concentration below its MRI detection limit and anticipate detecting an enhanced water signal only in those tissues where the extracellular Zn²⁺ rises well above background (≈40–50 μM). This offers the potential to detect image enhancement only in those tissues having a high local concentration of Zn²⁺ ions. Several examples of Gd³⁺-based agents that are detectable only when bound to a target protein or after polymerization have appeared in the literature (20–23).

Fig. 1.

Schematic illustration of the steps involved in insulin (brown spheres) and Zn²⁺ (blue triangles) release from β-cells during GSIS. Zn²⁺ released from β-cells during GSIS forms a 2:1 complex with GdDOTA-diBPEN (yellow squares) which, upon forming a complex with albumin [human serum albumin (HSA), pink], shows a ≈3.5-fold increase in r₁ at 23 MHz³ and a ≈1.5-fold increase in r₁ at 400 MHz. The two plots show the increase in water proton relaxivity, r₁, of GdDOTA-diBPEN (1 mM) after incremental addition of ZnCl₂ either in the presence (squares) or absence (diamonds) of albumin (titrations were performed in 100 mM Tris buffer, pH 7.6, 37 °C, ±600 μM albumin at 23 and 400 MHz).

Results and Discussion

MRI of Islets ex Vivo.

The MR sensitivity of the Zn²⁺ sensor was first tested in freshly isolated mouse islets (ex vivo) bathed in HBSS buffer containing 50 μM GdDOTA-diBPEN and 600 μM serum albumin in a total volume of 80 μL (37 °C, pH 7.4). In the experiment shown in Fig. 2 A and B, ≈50 islets occupied the bottom of each well. In subsequent experiments, the number of islets in each well was varied to quantitatively evaluate the effect of increasing glucose on the total amount of released Zn²⁺ and insulin (Fig. 2 C and D). T₁-weighted images from a 2-mm slice were collected just above the level of the islets containing either a non–insulin-stimulating concentration of glucose (2.5 mM) or an insulin-stimulating concentration of glucose (17.5 mM). Within 10 min, the images of wells containing islets and high glucose showed enhanced image contrast due to a decrease in water T₁, whereas images of wells containing islets exposed to basal levels of glucose remained unchanged. Independent analytical measures of Zn²⁺ and insulin from samples collected above the layer of islets confirmed that both were significantly higher in wells containing 17.5 mM glucose than in wells containing 2.5 mM glucose (Fig. 2C). A separate analytical measure of the water proton T₁ relaxation rate changes (∆T₁⁻¹) between low and high glucose samples (Fig. 2D) confirmed that the enhanced MRI contrast in images of samples containing high glucose, GdDOTA-diBPEN, and albumin indeed reflected a decrease in T₁ initiated by binding of Zn²⁺ to the sensor. Collectively, these ex vivo results demonstrate that 50 μM GdDOTA-diBPEN detects release of Zn²⁺ from islets presented with a stimulatory concentration of glucose in T₁-weighted MR images.

Fig. 2.

Ex vivo imaging of freshly isolated islets. (A) Each well contains ≈50 islets in Krebs-Henseleit-Hepes buffer, pH 7.4, plus either high (17.5 mM) or low (2.5 mM) glucose at 37 °C. (B) T₁-weighted fast spin-echo images of six wells containing ≈50 islets each (repetition time 500 ms, echo time 10 ms, averaging 6; 1-mm-thick slice above the level of the islets as shown). The images were collected ≈20 min after exposure of islets to the components indicated next to each image. For those wells containing the Zn²⁺ sensor and albumin, the concentrations were 50 μM of GdDOTA-diBPEN and 600 μM albumin. (C) Analytical amounts of Zn²⁺ and insulin in buffer solutions collected from a different set of wells containing variable numbers of islets. Data are expressed in pmol Zn²⁺ or insulin divided by the number of islet equivalents (IEQ) to normalize the data. (D) The change in water proton relaxation rate (ΔT₁⁻¹) in buffer solutions collected from the same wells as in C containing variable numbers of islets. As in C, the relaxation rate changes before and after addition of glucose (Δ1/T₁ in s⁻¹) was divided by the number of IEQ to normalize for differences in islet number in each well. Data are presented as average ± SD per IEQ (i.e., the islet quantity normalized to a 150-μm-diameter islet). All wells contained 50 μM GdDOTA-diBPEN and 600 μM albumin.

Imaging β-Cell Function in Mice by MRI.

Twelve-week-old, 24-h fasted, male C57/blk6 mice were anesthetized and positioned in a 38-mm volume coil for imaging at 9.4 T. Fasting blood glucose and insulin measured in representative animals before imaging averaged 6.34 ± 0.74 mM and 0.25 ± 0.20 ng/mL, respectively. Because the pancreas does not have a well-defined shape or position in the mouse anatomy, any single MR slice will detect only regions of the pancreas (this is illustrated nicely in a photograph of a mouse with overexpressed ACTB-DsRed*MST protein in its pancreas; Fig. S1). After collection of a series of anatomical images (14 slices, 2-min acquisition time) to locate the pancreas, a single bolus of saline or glucose (50 μL of a 20% solution to reach a target blood glucose concentration of ≈17 mM) was injected into the peritoneum (i.p.), followed by collection of an additional 14-slice image data set. Approximately 10 min after addition of saline or glucose, a bolus of GdDOTA-diBPEN (25 μL of a 25-mM solution) was injected via a tail vein catheter (i.v.) and, after an additional 10-min period to allow distribution of the agent, a third 14-slice image data set was collected. The image intensity differences in each slice before and after addition of the agent reflect the enhancement that occurred as a result of Zn²⁺ ions binding to GdDOTA-diBPEN (Fig. 3). An increase in MR image intensity was observed only in tissues corresponding to pancreas and only in animals provided with the glucose bolus (Fig. S2). Substitution of the nonresponsive MR contrast agent Gadoteridol (Prohance, Bracco Diagnostics Inc.) for GdDOTA-diBPEN at an equivalent concentration did not produce MR enhancement of the pancreas (Fig. S3). This indicates that the contrast enhancement observed here is agent specific and does not reflect an increase in blood flow or volume to the pancreas in response to high levels of glucose. The grayscale anatomical image shown as background in Fig. 3 A and B reflects a single slice through the abdomen before injection of the contrast agent, whereas the color overlay reflects a composite of those pixels in each of the 14 slices where the water image intensity increased by threefold or more over the average noise. The contrast enhancement detected in the kidneys after the contrast agent (CA) injection was excluded from this analysis. This 3D data set provided an estimate of the total functional volume of β-cells in each animal.

Fig. 3.

Representative grayscale T₁-weighted MR images of a single slice through the abdomen that contains a portion of pancreatic tissue (1 mm slice without fat saturation) of 12-wk-old control animals after injection of either saline (A) or glucose (B) followed by GdDOTA-diBPEN. The colored overlays represent a 3D composite of those pixels in each of the 14 slices where the water image intensity increased by threefold or more over the average noise (N) after injection of saline plus agent or glucose plus agent.

The total functional volume of β-cells in 12-wk-old mice averaged 34.0 ± 8.5 mm³ (Fig. 4C). Given that the total mass of β-cells in a 12-wk-old mouse averages 1.9 mg (24) (and assuming 1 mg ≈ 1 mm³), one can conclude from these imaging data that GdDOTA-diBPEN senses Zn²⁺ over a substantially larger volume of tissue than that defined by the volume of β-cells alone. This is consistent with diffusion of Zn²⁺ ions away from the β-cell surface into a volume of extracellular space that is ≈18-fold (34.0 mm³/1.9 mm³) larger than the volume occupied by the total mass of β-cells. No attempt was made to measure actual in vivo T₁ changes in this experiment, owing to the time required for multislice T₁ experiments, but the gain in signal intensities observed here was consistent with those anticipated for a change in T₁ initiated by the binding of 50 μM GdDOTA-diBPEN–(Zn)₂ to albumin at 400 MHz (Fig. 1). For example, Fig. 1, Lower Right shows that the relaxivity enhancement one measures at 400 MHz upon binding of GdDOTA-diBPEN–(Zn)₂ to albumin is substantially smaller (r₁ increases from 4 mM⁻¹s⁻¹ to 6 mM⁻¹s⁻¹) than the increase observed at 23 MHz, as expected for a complex undergoing slow rotation. In any typical single MR slice that contains a portion of pancreatic tissue, the water intensity increased by ≈15% when the ternary GdDOTA-diBPEN–(Zn)₂–albumin complex was formed. If one assumes a T₁ of 1.5 s for the normal pancreas before agent injection, a 15% increase in water intensity in a spin-echo experiment would correspond to a decrease in T₁ to ≈1.1 s. This ≈27% decrease in T₁ is consistent with a local concentration of ≈50 μM agent with an r₁ = 6 mM⁻¹s⁻¹. It is interesting to note that the MR intensity increases observed here as a result of Zn²⁺ binding to GdDOTA-diBPEN do not seem to be uniform over the entire pancreas. This suggests that either Zn²⁺ release may not be uniform throughout the pancreas, or the distribution of agent or albumin is not uniform. Other glucose and agent injection protocols were also tested (e.g., simultaneous i.v. injection of glucose and agent and agent before glucose), but i.p. delivery of glucose followed by i.v. infusion of the agent gave the most consistent image enhancements from one animal to another. For this reason, all subsequent experiments were conducted using this protocol.

Fig. 4.

Representative grayscale T₁-weighted MR images of the abdomen (1-mm slice without fat saturation) of 24-wk-old mice fed either a standard 10% fat diet (A) or a 60% fat diet (B) over 12 wk. The colored overlays reflect a 3D composite of those pixels in each of the 14 slices where the water image intensity increased by threefold or more over the average noise after injection of glucose plus GdDOTA-diBPEN. (C) Total functional volume (mm³) of pancreatic tissue (regions with enhanced signal intensity by threefold or more summed over 14 separate MRI slices) in the three groups of animals as detected by GdDOTA-diBPEN. The average functional volumes, SDs, and P values are shown above each bar (n = 10 for control_12-wk; n = 6 for control_24-wk; n = 6 for DIO_24-wk).

Functional Volume of β-Cells Increases in Animals Exposed to a High-Fat Diet.

To test whether this MRI technique is sensitive enough to detect expansion of β-cells (neogenesis) known to accompany accumulation of excess fat, mice were fed either a control diet consisting of 10% fat or a high-fat diet consisting of 60% fat for 12 wk and imaged at 0, 6, and 12 wk using the Zn²⁺ sensor. The 60% fat diet has been shown to induce obesity, glucose intolerance, hyperinsulinemia, and expanded β-cell mass in C57BL/6 mice (25, 26). Representative microscopy of islets showed a marked increase in islet diameter in diet-induced obese (DIO) mice vs. control (Fig. S4). Color overlays of functional β-cell volume as reported by the Zn²⁺ sensor in select slices of control vs. DIO mice at 24 wk are compared in Fig. 4 A and B (images of mice after 6 wk on the high-fat diet did not differ significantly from the images at baseline). As before, these images were not uniform throughout the pancreas, with some regions showing larger changes in T₁ (more of the combined red, orange, and yellow voxels) than others, indicating nonuniform release of Zn²⁺. These images show quite clearly that the volume of functional β-cells was considerably larger in the DIO animals compared with controls at 24 wk. These differences are shown in graphical form in Fig. 4C. The functional β-cell volume in mice on the control diet (10% fat) increased from 34.0 ± 8.5 mm³ to 44.3 ± 5.0 mm³ over 12 wk, presumably reflecting normal growth of the animal over this period. Other studies have shown that β-cell mass increases in proportion to body weight in the same strain of mice kept on a similar low-fat diet (27). The mice maintained on the 10% fat diet in the present study gained on average 13 g over the 12 wk (from 20 ± 2 g to 33 ± 4 g). Thus, the 65% increase in body weight experienced by mice fed a control diet is not matched by the increase in functional volume of β-cells as measured by Zn²⁺ release (30%). In comparison, mice maintained on the 60% fat diet over 12 wk gained on average 24 g (from 20 ± 2 g to 44 ± 5 g), a 120% increase, whereas the volume of functional β-cells in these mice increased from 34.0 ± 8.5 mm³ to 58.0 ± 9.4 mm³, a 71% increase (Fig. 4C). As before, the number of pixels approaching an eightfold increase in signal intensity compared with noise was qualitatively much larger in the DIO mice than in the control animals. This indicates that not only is there a substantially larger volume of functioning β-cells in animals fed a high- vs. low-fat diet, but those cells also seem to release more Zn²⁺ and insulin in response to elevated glucose. The insulin levels in DIO mice at 24 wk were 18-fold higher than in animals maintained on a low-fat diet (Fig. S5), whereas blood glucose was not significantly different (5.9 ± 0.6 mM vs. 6.8 ± 1.7 mM). The 1.7-fold increase in functional β-cell volume observed here was coincidentally identical to the increase in β-cell mass and the proliferation index reported by Peyot et al. for C57BL/6 mice on this same high fat diet (26). Visual inspection of the pancreas before and after 12 wk of high-fat feeding showed clear changes in both the size and shape of the pancreas, and histology confirmed the expected increase in islet size (Fig. S4). These findings demonstrate that imaging of Zn²⁺ release accurately detects β-cell expansion known to occur during high-fat feeding in mice. Although no attempt was made to correlate the observed increase in MR signal with total β-cell mass in these animals, the image enhancement reported by GdDOTA-diBPEN clearly responds to changes in functional β-cell volume as anticipated for this mouse model.

Functional Imaging of Zn²⁺ Release in a Type I Diabetes (STZ-Treated) Model.

To determine whether Zn²⁺ imaging is also sensitive to loss of β-cell function that accompanies Type 1 diabetes, a separate group of mice was treated with the β-cell toxin STZ before imaging. Images of STZ-treated and age-matched control mice are compared in Fig. 5. Here, the contrast enhancement differences in images of the two groups of animals were dramatic, with essentially no contrast enhancement detected in images of the STZ-treated animals, consistent with a near total loss of β-cell function in these animals. This illustrates that Zn²⁺ release from functional β-cells during GSIS is an absolute requirement for producing the contrast enhancements detected by MRI.

Fig. 5.

Images of Zn²⁺ release during GSIS in a control (A) vs. a STZ-treated mouse (B). The color overlay represents the tissue areas where a contrast enhancement was observed after a bolus injection of GdDOTA-diBPEN and glucose. The colored image overlays reflect the same changes as noted in Figs. 3 and 4. F, fundus stomach; K, kidneys; S, spleen. Images were collected from the same mouse before and 5 d after a single high-dose treatment of STZ.

Conclusions

MR images of mice were enhanced largely in regions corresponding to the pancreas after injection of a bolus of glucose followed by a low dose of the Zn²⁺ sensor GdDOTA-diBPEN. Images of the pancreas were not enhanced by the agent in the absence of added glucose (euglycemia) but were enhanced after glucose was elevated to a level that stimulates insulin secretion. It has been well established that insulin is packaged as a Zn²⁺/insulin complex in β-cell granules, so these observations are consistent with MRI detection of Zn²⁺ released into the extracellular space of β-cells as a result of GSIS. The agent does not enhance MR images of the pancreas in euglycemic mice or in mice pretreated with STZ. Serial MR images of mice collected during a prolonged period of high-fat (60%) feeding showed a dramatic increase in contrast enhancement throughout the abdomen, consistent with expansion of the pancreas and a concomitant overall increase in β-cell function. We conclude that the Zn²⁺-responsive MRI agent GdDOTA-diBPEN detects both expansion and loss of β-cell function in vivo. Importantly, both events herald important pathological changes in the progression of type 1 and type 2 diabetes in humans. Thus, imaging agents that respond to biological function such as the Zn²⁺ sensor demonstrated here can potentially provide invaluable tools for clinical assessment of disease progression and to identify more effective treatment regiments for type 1 and type 2 diabetes.

Methods

All MR images (fast spin-echo multislice sequence) were obtained using a 9.4 T (400-MHz) horizontal-bore Varian INOVA imaging system with a dual-channel 38-mm-diameter birdcage volume coil. All data reported are averages with corresponding ±SD and n as indicated (Tables S1 and S2). Numerical data were obtained from images before any transformation or file conversion for figure printing purposes (e.g., image overlay representations). Detailed methods not discussed above, including statistical parameters and all associated references, are available in SI Methods.