Noninvasive Magnetic Resonance Imaging of Microvascular Changes in Type 1 Diabetes

PGC-GdDTPA-F was obtained from PharmaIN (Seattle, WA).

Five-week-old female BALB/c mice (n = 6, ∼20 g) were rendered diabetic by injection of 200 mg/kg i.p. of the β-cell toxin STZ. On the following day, diabetic animals were imaged before as well as 1, 17, and 40 h after intravenous injection of PGC-GdDTPA-F (0.2 mmol gadolinium/kg). Age-matched nondiabetic animals injected with PGC-GdDTPA-F and imaged at the same time points served as controls (n = 4).

Fasting blood glucose levels were determined 24 h after STZ administration using Glucometer Elite Testing System (Bayer Diagnostics, Tarrytown, NY). All animal experiments were performed in compliance with institutional guidelines approved by the Subcommittee on Research Animal Care at the Massachusetts General Hospital and in accordance with the National Institutes of Health Principles of Laboratory Animal Care (publ. no. 85-23, revised 1985).

MRI was performed on a 9.4T Bruker horizontal bore scanner (Billerica, MA) equipped with a home-built radio frequency transmit and receive 3 × 4 cm elliptical surface coil and using ParaVision 3.0 software. For in vivo imaging of diabetic and control mice, we obtained T1 maps, T1-weighted images, and low- and high-resolution 3D angiograms. Details on image sequences and image analysis can be found in the online appendix (available at http://dx.doi.org/10.2337/db07-0822).

PGC-GdDTPA-F was synthesized to carry a fluorescent label (FITC). Fluorescence microscopy on pancreata from experimental and control animals was performed in two channels (FITC for PGC-GdDTPA-F and DAPI for nucleus). Consecutive sections were stained with hematoxylin and eosin and analyzed by light microscopy. Additional tissue sections were stained for insulin and analyzed by fluorescence microscopy. Details on the staining procedures are presented in the online appendix.

To evaluate microvascular changes in the diabetic pancreas, we used the long-circulating paramagnetic T1 contrast agent, PGC covalently linked to GdDTPA labeled with fluorescein. PGC-GdDTPA-F was administered to STZ-induced diabetic mice (fasting blood glucose 287.8 ± 186.9 mg/dl [range 24–450]) and nondiabetic controls (114.0 ± 22.1 mg/dl [range 94–141]). Our results demonstrated sufficient accumulation of the agent in the pancreas for detection by MRI. On T1-weighted images obtained 17 h after injection of the contrast agent into STZ-induced diabetic animals, the pancreas was clearly enhanced compared with preinjection images due to the T1-shortening effect of GdDTPA (Fig. 1A).

Quantitative time-course analysis of T1 relaxation times was based on inversion-recovery T1 maps (Fig. 1B). These studies indicated accumulation of PGC-GdDTPA-F in the pancreata of both diabetic animals and nondiabetic controls, as early as 1 h postinjection, reflective of presence of the contrast agent in the blood pool. The peak of contrast agent bioavailability in the pancreas was reached 17 h after injection, followed by a gradual washout by 40 h (Fig. 1C). While we observed the expected vascular enhancement in both animal groups, the levels of PGC-GdDTPA-F in the pancreata of diabetic animals were significantly higher than in those of nondiabetic control animals (P < 0.05), consistent with increased blood volume and vascular permeability. This difference was reflected by a lower T1 in diabetic mice compared with controls and was seen at 1 h (diabetic T1 385.9 ± 30.1 ms and nondiabetic T1 501.3 ± 43.3 ms) and 17 h (diabetic T1 193.7 ± 3.5 ms and nondiabetic T1 342.5 ± 52.8 ms) after injection of PGC-GdDTPA-F (Fig. 1C).

To monitor the whole-body vascular distribution of PGC-GdDTPA-F, we performed low-resolution 3D angiography of the same animals at the specified time points. As seen in Fig. 2, using a 0.4 × 0.4 × 0.4 mm/pixel resolution, we could see significant enhancement in the upper left abdomen, consistent with the location of the pancreas, in diabetic animals but not in nondiabetic controls. The time course of this enhancement closely matched our T1 map analysis. It became apparent at 1 h postinjection of PGC-GdDTPA-F, increased at 17 h, and became negligent by 40 h after contrast agent delivery (Fig. 2).

For more precise identification of the origin of this signal, we imaged diabetic mice and controls using high-resolution 3D angiography. Brightly enhancing arteries and less-enhancing veins labeled with fluorescein could be distinguished at the 0.1 × 0.1 × 0.1 mm/pixel resolution. The hepatic vasculature was clearly identifiable in both experimental and control animals, as a result of contrast agent availability in the blood pool. Notably, we observed a clear and remarkable enhancement in the area of the pancreas in diabetic animals due to the vascular leakage of the agent as early as 1 h after injection, followed by a washout from the pancreas by 40 h (Fig. 3 and online appendix supplemental movies 1 and 2). No such enhancement in the pancreas was observed in control animals. As expected based on the long circulation half-life of PGC-GdDTPA-F (14 h in rodents [12]), residual levels of contrast agent in the larger hepatic arteries caused the enhancement of these vessels to persist even at the 40-h time point.

Finally, to independently validate our findings, we correlated our imaging conclusions with gold standard histological measurements of the tissue distribution of the contrast agent. On hematoxylin and eosin sections, we could identify pancreatic islets adjacent to islet-feeding pancreatic blood vessels (Fig. 4A). In STZ-induced diabetic mice, however, pancreatic islets were less abundant and islet morphology appeared abnormal (Fig. 4A). The detrimental effect of STZ on islet architecture was further confirmed by staining for insulin (online appendix supplemental Fig. 1). Fluorescence microscopy demonstrated a significant accumulation of PGC-GdDTPA-F in the pancreata of diabetic animals. There was bright green fluorescence, associated with the presence of PGC-GdDTPA-F, in blood vessels feeding pancreatic islets. Fluorescence intensity was markedly enhanced in experimental versus control animals, consistent with increased vascular volume (Fig. 4B and supplemental Fig. 2). Diffuse green fluorescence associated with the islet interstitium and present exclusively in tissue sections from diabetic but not control animals was consistent with extravascular leakage of the contrast agent. In control animals, fluorescence derived from PGC-GdDTPA-F was restricted to the islet periphery (Fig. 4C).

Accumulating evidence suggests that the vascular endothelium is of crucial importance for the development of inflammatory responses in the pancreas. In models of type 1 diabetes, modifications of the pancreatic microvasculature accompany the initiation and progression of disease. Vascular swelling and increased blood flow precede insulitis in nonobese diabetic mice and in STZ-induced diabetes (5,7,9,19,20). Consequently, pancreatic islet vascular dysfunction is a crucial element of the pathology of type 1 diabetes and therefore represents an early and potentially predictive biomarker for the loss of β-cell mass. In addition, vascular changes have also been reported in type 2 diabetes. Augmented islet blood flow has been demonstrated after short-term modest hyperglycemia, as well as in several animal models of type 2 diabetes, including the Goto-Kakizaki (GK) rat (21) and the obese hyperglycemic mouse (22).

Overall, pancreatic vascular dysfunction seems to play a critical role in both type 1 and type 2 diabetes and represents a complex multifactorial system that cannot be studied without a comprehensive systematic approach. In vivo MRI has been widely used to characterize vascular events in the brain and in various types of tumors (23–25). Here, we apply this methodology to study the islet vasculature. To the best of our knowledge, this represents the first application of magnetic resonance angiography to visualize islet vascular dysfunction using a paramagnetic blood pool imaging agent. Differential accumulation of PGC-GdDTPA-F in the pancreatic area detected by this technique was clearly identifiable in STZ-induced diabetic animals but not in healthy controls. In our experiments, we used the exaggerated single–high-dose STZ model of diabetes because it provides a solid framework for the establishment of proof-of-principle for our imaging method. The next logical step in our studies, however, is to test the feasibility of this approach in the multiple–low-dose STZ model, nonobese diabetic mice, as well as in models of type 2 diabetes. These models more closely resemble human pathology and therefore such studies would help assess the sensitivity of this method and its feasibility in a potential future clinical scenario. Furthermore, because the observed effects are the result of two separate events (increased vascular volume and permeability), further studies would be necessary to unravel the relative input of each. We believe that this imaging strategy represents a valuable research tool for studying 1) the time course of the disease, 2) the role of the vasculature in the insulitic process and islet cell death relevant to type 1 diabetes, and 3) the vascular response to metabolic dysregulation in both type 1 and type 2 diabetes. In a therapeutic scenario, the described methodology has implications for the early diagnosis of diabetes, the monitoring of events relevant to islet transplantation and other diabetes therapies, and the design of image-guided individualized treatment strategies.

The authors acknowledge John Moore and Pamela Pantazopoulos (Athinoula A. Martinos Center for Biomedical Imaging, MGH) for their excellent technical support with animal surgery.

This work was supported, in part, by the National Institute of Diabetes and Digestive and Kidney Diseases National Institutes of Health Grant DK069727 (to E.B.).