REPLY: We would like to thank Dr. Kwee et al. for their interest and comments about our review on PET imaging of β-cell mass (1). They raise a major concern that PET with its limited spatial resolution may not be able to resolve small islets containing β-cells that are scattered throughout the pancreas. To address this concern, we would like to point out first that the ultimate goal of PET using radioligands such as 11C-dihydrotetrabenazine (11C-DTBZ) that target transporters and receptor proteins is to estimate average target protein density per unit volume of the tissue (pancreas) in which the target proteins are found (Bavail, in units of molar [often nmol·L−1, that is, nanomoles of receptor or transporter per 1,000 cm−3 tissue]) (2). Here, we are not estimating vesicular monoamine transporter 2 (VMAT2) per individual islet volume. The goal is to estimate VMAT2 density per unit pancreatic tissue volume or total pancreatic VMAT2 content (β-cell mass). To further clarify this point, an analogy can be made to PET imaging of neuroreceptors and transporters in the brain.
VMAT2 and dopamine transporters are found in the dopaminergic terminals in the striatum. It has been well established that PET using high-affinity radioligands such as 11C-DTBZ can quantify the binding potential that correlates with the number of target sites per unit volume of striatal tissue (Bavail) (3). The volume of the human striatum (caudate nucleus and putamen) is about 5.5 cm3 (4). Clusters (150–300 μm in diameter) of intrastriatal neurons (96% are medium spiny neurons) make up about 15% of the volume of the striatum and are embedded in surrounding “matrix,” which makes up about 85% of the volume of the striatum. This larger matrix compartment consists of a mesh of extensive dendritic trees of medium spiny neurons and afferent axonal terminals, as well as efferent axons. Each branch of these dendritic trees is packed with numerous small spines that receive several different afferent inputs including glutamatergic, dopaminergic, GABA-ergic, and cholinergic terminals. The major massive input in terms of quantity of axons is not dopaminergic but glutamatergic corticostriatal projections.
Scattered patches of nigrostriatal dopaminergic terminals (0.1–0.5 μm in diameter) are a minor component of this matrix in terms of the volume contribution and are almost homogeneously dispersed throughout the matrix of the striatum (5). Therefore, VMAT2 and dopamine transporters containing dopaminergic terminals are by no means densely packed in the striatum. Yet PET with high-affinity radioligands for these sites allows detection of excellent binding signals, and kinetic model–based quantification allows for the estimation of binding potential, which correlates with the VMAT2 or dopamine transporter density (Bavail = number of transporter sites per unit volume of striatal tissue).
On the other hand, the volume of the normal pancreas is about 80 cm3 (15 times the volume of the striatum) (6). The major component of the pancreas in terms of volume is exocrine tissue, and exocrine cells account for 98%–99% of the pancreatic parenchyma. The endocrine component of the pancreas is organized into clusters of islets of Langerhans containing 4 major cell types (producing different hormones), α-cells (glucagon), β-cells (insulin), δ-cells (somatostatin), and pancreatic polypeptide-producing cells. The islets (average 150 μm in diameter) are scattered almost homogeneously within the exocrine tissue and represent 1%–2% of the parenchyma (7).
From this analogy, it is apparent that the pancreas resembles the stratum in terms of the miniscule volumes of VMAT2 binding sites that are scattered in the tissue of interest. The volume of the pancreas is, however, about 15 times larger than that of the striatum. The target tissue volume is important, as pointed out by Kwee et al., for the recovery of PET signal (partial-volume effects) from the target tissue (pancreas or striatum) because of the limited spatial resolution of the PET system. Another important factor that influences the accuracy of the PET measurements relates to PET signal intensity. With 11C-DTBZ PET, the average peak pancreatic standardized uptake value was 9, which is comparable to that of 18F-FDG in the gray matter of the brain.
Our group recently started using 18F-fluoropropyl-(+)-DTBZ, which has high binding affinity for VMAT2 (inhibition constant, 0.11 nM) and is superior to 11C-DTBZ with lower affinity (higher inhibition constant) (0.97 nM) (8,9). In addition, this tracer can be shipped from a remote central radiopharmacy thanks to the much longer half-life of 18F (110 min). Our preliminary human data (unpublished) indicate that the average peak standardized uptake value of 18F-fluoropropyl-DTBZ in the human pancreas is 21, more than double that of 11C-DTBZ. Our VMAT2 PET brain and pancreas imaging studies using 18F-fluoropropyl-DTBZ in baboons suggest that the pancreas-specific VMAT2 binding signal is approximately 85% that of the striatum of the brain, suggesting that VMAT2 density in the pancreas is similar to that in the striatum (10). One important point here is that the pancreatic PET signal represents a combination of specific VMAT2 binding signal and nondisplaceable background signal consisting of free tissue and nonspecific binding signal from both endocrine and islet tissues of the pancreas (the latter accounting for up to 40% of the total pancreatic signal). 18F-fluoropropyl-DTBZ thus provides exquisitely intense target tissue (pancreas) signal of high PET image quality (high signal-to-noise ratios).
The discussion so far has not directly addressed the partial-volume effect in PET quantification of VMAT2 in the pancreas, or in the striatum for that matter. Given that our current PET/CT system (Biograph HD; Siemens) has a full width at half maximum of 2.5–4.0 mm, voxel signal at the periphery of the pancreas suffers from the partial-volume effect. However, signal from the islets located more centrally is completely recovered, because radioactivity sampling is performed over all voxels of the pancreas. Here, we are not quantifying signal from individual islets. Unlike the situation with detection by 18F-FDG PET of small lung nodules surrounded by air-filled lung tissue containing practically no activity, the pancreas is in the abdomen, and surrounding soft-tissue background activity mitigates the partial-volume effect for the peripheral voxels. However, implementation of partial-volume correction or resolution recovery would improve the accuracy of PET quantification of the pancreas. For PET of the brain, particularly for the thin cortical gray matter, partial-volume correction should improve the quality of PET brain images. For example, absolute quantification of glucose utilization of the thin cortical gray matter in tiny mouse brains (each brain measuring about 1 g) using small-animal 18F-FDG PET (1.5 mm in full width at half maximum) underestimated slightly by 8% that quantified by the autographic 2-14C-deoxyglucose method in the same animals (11).
Finally, we agree that 18F-FDG PET is an excellent method for tracking the viability of transplanted islet cells in the subcutaneous space. Although the subcutaneous location of the transplant is favorable in minimizing the partial-volume effects, the signal intensity of 18F-FDG uptake in the transplant is relatively low (standardized uptake value of 1–3). We believe that one key consideration for successful and accurate PET that targets islets and, more specifically, β-cells within the islets is the availability of tracers that can show high target-to-noise ratios. Our preclinical experiments suggest that PET targeting VMAT2 is also feasible for imaging transplanted islets and does not suffer from the need to label islets before transplantation (12).
- © 2011 by Society of Nuclear Medicine