Kinetic modeling without accounting for the vascular component impairs the quantification of [(11)C]PBR28 brain PET data

Abstract

The positron emission tomography radioligand [(11)C]PBR28 targets translocator protein (18 kDa) (TSPO) and is a potential marker of neuroinflammation. [(11)C]PBR28 binding is commonly quantified using a two-tissue compartment model and an arterial input function. Previous studies with [(11)C]-(R)-PK11195 demonstrated a slow irreversible binding component to the TSPO proteins localized in the endothelium of brain vessels, such as venous sinuses and arteries. However, the impact of this component on the quantification of [(11)C]PBR28 data has never been investigated. In this work we propose a novel kinetic model for [(11)C]PBR28. This model hypothesizes the existence of an additional irreversible component from the blood to the endothelium. The model was tested on a data set of 19 healthy subjects. A simulation was also performed to quantify the error generated by the standard two-tissue compartmental model when the presence of the irreversible component is not taken into account. Our results show that when the vascular component is included in the model the estimates that include the vascular component (2TCM-1K) are more than three-fold smaller, have a higher time stability and are better correlated to brain mRNA TSPO expression than those that do not include the model (2TCM).

Figures

Figure 1

Interaction of TSPO ligand with brain parenchyma. The TSPO has a heterogeneous distribution in the normal brain parenchyma. The endothelium expresses high concentrations of TSPO, which has implications for PET kinetics. The free [11C]PBR28 concentration in tissue is not expected to be generally uniform but distributed along a gradient from the capillary space to the microglia, passing through the endothelium into the extracellular space. In the brain parenchyma the ligand exchanges with lipids (molecular bilayer) and possibly other proteins. It then enters the microglial inner cellular space to reach the mitochondria. All these interactions lower the concentration of the [11C]PBR28 amenable to bind the TSPO. This creates a change in the apparent affinity of the tracer of the target. Thus, the apparent affinity would be higher closer to the vascular space, with resulting slower kinetics for the bound fraction (e.g. irreversible trapping).

Figure 2

2TCM and 2TCM-1K model structure. (A) The 2-tissue compartmental model (2TCM) model is composed of two exchangeable tissue compartments, one for the non-displaceable component (Cnd) and one for the specific binding (Cs). K1 and k2 are the rate constants for transport from plasma to tissue and back, respectively. k3 and k4 are the rate constants from the non-displaceable compartment to the specific one and back, respectively. (B) 2TCM-1K includes also a vascular component (Cvasc), with Kb as the rate constant from plasma to the vascular compartment.

Figure 3

Model fit comparisons. Example of model fits to the parietal (A) and occipital cortex (B) for a representative MAB subject. Black circles represent the measured time–activity curves, while dotted and solid gray lines represent the model description provided by 2TCM and 2TCM-1K, respectively. The zoom of the initial phase of the curve (in the interval 0–20 minutes) is reported in the center of each panel. On the right side are the weighted residuals of 2TCM and 2TCM-1K (dotted and solid gray lines, respectively).

Figure 4

Distributions and correlation of regional VT and Kb estimates. Results are reported for 2TCM (VT, panel A) and 2TCM-1K (VT and Kb, panels A and B, respectively). Homozygous high-affinity binders (HAB, 8 subjects) and heterozygous mixed-affinity binders (MAB, 10 subjects) are shown separately. (C) Correlation analysis of 2TCM versus 2TCM VT estimates of HAB subjects (black diamonds) and MAB subjects (gray squares). The equations of the regression lines and the Pearson's correlation coefficients (R2) are reported in the same chart.

Figure 5

Application of 2TCM in simulated tissues with vascular trapping. We simulated different scenarios of VT with the 2TCM-1K model (baseline and from 10% to 50% VT increases) for both HAB and MAB cases (dark and light gray bars, respectively). The figure reports the VT variation (mean±s.d.) estimated by the 2TCM model when applied to these data.