Introduction
Radiological evaluation of relapsing remitting multiple sclerosis (RRMS) is mainly based on new T2 lesions and active gadolinium-enhancing lesions on magnetic resonance imaging (MRI) [
1]. However, detection of these disease-specific lesions by MRI only partly demonstrates the underlying pathophysiological processes in MS. As positron emission tomography (PET) can be used to visualise distinct molecular processes
in vivo, it may provide unique insights in the pathophysiology of neuroinflammation (and neurodegeneration) in MS [
2]. Over the past decades, the translocator protein 18 kDa (TSPO), present on the mitochondrial membrane of microglial cells, has been used as an
in vivo marker of neuroinflammation in several neurological disorders including MS [
3‐
5]. Although in general results have been positive, the use of TSPO as a marker for neuroinflammation has some limitations, such as an intracellular binding site, genetic polymorphisms, and additional binding sites on monocytes and vascular wall endothelium [
6,
7]. In addition, TSPO does not differentiate between resting state and pro-inflammatory and neuro-protective microglia subtypes [
8]. Therefore, new PET tracers are needed, which specifically target proteins that reflect the status of microglial cells. Recently, using well-characterised post-mortem tissues of patients with MS and activated human microglia, it has been demonstrated that the purinergic P2X
7 receptor is highly expressed on pro-inflammatory microglia and macrophages, is selectively expressed within MS lesions, and may be involved in the neuroinflammatory cascade [
8‐
10]. To a lesser extent, expression of P2X
7 receptor has also been found in grey matter on astrocytes, oligodendrocytes, and neurons [
11]. Moreover, it was shown that the radioligand [
11C]SMW139 selectively binds to the P2X
7 receptor with high affinity in preclinical
in vivo studies and in post-mortem human tissues [
8,
12‐
14].
However, so far, no human in vivo data are available for this tracer and its validity to visualise activated microglia and macrophages. The aims of this study were therefore to evaluate in vivo pharmacokinetic characteristics of [11C]SMW139 and to perform a proof of concept study assessing whether this novel tracer can be used for identifying neuroinflammation in RRMS.
Discussion
This study shows that binding of [11C]SMW139 to the P2X7 receptor tracer can be quantified using a two-tissue reversible plasma input model with fixed k4. Using this model, it was possible to identify neuroinflammation in both MS lesions and normal appearing brain tissue in RRMS.
In this study, simplified reference tissue models were not evaluated, as the pathophysiology of MS violates the assumptions underlying such models. Firstly, because of global low-grade inflammation in the MS brain, none of the brain regions are devoid of specific binding (P2X
7 receptors) and hence cannot be considered a reference region [
24]. Secondly, due to disruption of the blood-brain barrier in MS, a constant K
1/k
2 across the brain cannot be guaranteed [
25].
The rate of influx of [11C]SMW139 from plasma into the brain was sufficient for accurate data analysis. Regional TACs demonstrated fast efflux of [11C]SMW139, which is supported by high k2 estimates. Rapidly declining activity in non-displaceable and, consequently, specific compartments was observed, resulting in unreliable estimates especially k4 in smaller ROIs. As k4 = koff, it should be constant for the P2X7 receptor independent of its location. This limited inter-subject variation for k4 for the different ROIs enabled fixing this parameter for smaller regions to the more reliable whole-brain values. By reducing the number of fit parameters, the precision of the other fit parameters was improved, resulting in acceptable SEs for all estimated parameters in most ROIs when using the 2T4k_VB_k4 model.
As demonstrated in Table
2, K
1 was similar in patient and control groups, but k
2 was lower in patients, leading to a higher non-displaceable distribution volume (V
ND) in patients. Nevertheless, apart from V
T, BP
ND (= k
3/k
4) was also higher in patients than in controls, suggesting higher specific binding in the patient group than in the controls for both grey and white matter. This implies an increase in activated microglia in normal appearing brain tissue of these clinically active RRMS patients.
The patient who had the lowest BP
ND within the patient group for all ROIs had slightly different demographics: longest disease duration (5 years), highest Expanded Disability Status Scale (5.5), slowest timed 25-foot walk (8.0 s), second slowest 9-Hole Peg Test (25.4 s), lowest Symbol Digits Modalities Test (33 correct), and highest T2 lesion load (31.3 cm
2) with only one small gadolinium-enhancing lesions (0.7 cm
2). This implies that in this patient MS was more progressed compared with the rest of the group, which in turn could result in less active neuroinflammation and therefore lower P2X
7 receptor binding [
26]. When looking at the demographics of the healthy control with the highest regional BP
ND values, this subject’s relatively high age stands out (59.2 years). This could resemble an effect of age on P2X
7 receptor binding similar to the effect of age on TSPO binding [
27], although this clearly needs to be confirmed in a larger cohort.
In the subjects with gadolinium-enhancing lesions, VT was similar for non-lesional white matter and T2 lesions, but increased in the enhancing lesions. In contrast, BPND values decreased in T2 lesions, with a further decrease in enhancing lesions. As VT includes both non-displaceable and specific compartments (the latter represented by BPND), this indicates that decreased specific binding of [11C]SMW139 was compensated by an increased non-displaceable compartment (non-specifically bound or free tracer). This was reflected by an increase in the K1/k2 ratio in these gadolinium-enhancing lesions. The most likely explanation for this increased K1/k2 ratio is disruption of the blood-brain barrier in active MS lesions, resulting in increased vascular permeability. Consequently, to quantify [11C]SMW139 binding in these active lesions, BPND is required.
Interestingly, cerebral [
11C]SMW139 BP
ND was increased in RRMS patients compared with healthy controls, but counterintuitively specific binding was decreased in active MS lesions. An explanation for this could be found in the heterogeneity of neuroinflammation in MS. Microglia activation is a complex and dynamic process, and different microglia subtypes and macrophages could be involved in both focal and diffuse neuroinflammation [
28].
As carbon-11 has a half-life of 20 min, minimal radioactivity is present in both tissue and blood in the final stage of a 90-min scan. Together with the observed fast kinetics of [
11C]SMW139, this makes accurate measurements towards the end of the 90-min scans difficult. In addition, a 90-min scan can be quite challenging for both healthy controls and especially patients. Therefore, reducing the scan duration would be beneficial. As shown in Fig.
5, both V
T and BP
ND obtained from 60-min datasets correlated very strongly with those from 90-min datasets, despite a slight (systematic) underestimation, suggesting that 60-min scans may be sufficient to assess in vivo kinetics of [
11C]SMW139 without compromising quantitative accuracy (see Online Resource
3). In addition, there appeared to be two clusters in the Bland-Altman analysis of V
T. These clusters, however, were not present in the Bland-Altman plot of BP
ND. This suggests that the two clusters may be due to differences in non-specific binding.
As this is a first in-man study, the sample size is rather limited, which is the main limitation of this study. Larger study cohorts are needed to confirm these initial findings. In addition, the small sample size resulted in some demographic differences between patient and control groups, even though they did not reach statistical significance. A possible group difference that could be relevant is the number of smokers, as an effect of smoking on neuroinflammation (in MS) has been reported [
29‐
31]. If and how this could impact
in vivo [
11C]SMW139 binding is currently uncertain.
Moreover, a larger sample size would allow for an assessment of possible correlations between [11C]SMW139 binding and disease duration, clinical outcome measures such as the EDSS- and MRI-derived outcome measures like atrophy. In addition, [11C]SMW139 could potentially be used for longitudinal analyses, which would require additional studies such as test-retest evaluation.
To conclude, this first in-man study demonstrates that a reversible two-tissue plasma input model with fixed k4 is the preferred model to quantify [11C]SMW139 in healthy controls and RRMS patients.
Additional studies are warranted to further evaluate its clinical relevance as a novel neuroinflammation tracer.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.