TSPO imaging using the novel PET ligand [¹⁸F]GE-180: quantification approaches in patients with multiple sclerosis

The classic diagnosis of multiple sclerosis (MS) is based on clinical and paraclinical documentation of the dissemination of CNS lesions in time and space. Such lesions and their evolution over time are commonly detected by magnetic resonance imaging (MRI). This forms not only the basis of the diagnosis but is also used to monitor disease activity and inform the decision on appropriate therapeutic strategies. While in MRI, the disruption of the blood-brain barrier (BBB) is used as proxy of disease activity, positron emission tomography (PET) imaging of activated microglia or macrophages with the 18-kDa translocator protein (TSPO) visualises one of the hallmarks of neuroinflammation and thus might provide a more direct approach to assess disease activity in MS. TSPO is primarily expressed in activated microglia, astrocytes, endothelial cells, and infiltrating macrophages [1] and is therefore associated with nervous system inflammation [2]. The prototypic TSPO radioligand [¹¹C](R)-PK11195 has been frequently investigated in various PET imaging studies [3]. However, quantification with [¹¹C](R)-PK11195 has been shown to be challenging due to a low free fraction in plasma, a significant binding to plasma proteins, and a low extraction fraction in brain with a limited signal-to-background ratio [4, 5]. This led to the development of second-generation TSPO radioligands with lower non-specific binding and higher affinity and specificity.

Preclinical data of the third-generation TSPO radioligand [¹⁸F]GE-180 have demonstrated a higher specific signal in affected brain regions and a lower non-specific binding in healthy tissue than [¹¹C](R)-PK11195 in models of stroke [6] and neuroinflammation [7, 8]. Our own preclinical experience with this tracer indicated a very good applicability for monitoring neuroinflammatory disease as well [9]. First-in-human studies with healthy controls (HC) found a low first-pass extraction resulting in low uptake of [¹⁸F]GE-180 in healthy tissue [10, 11]. Various compartmental models with and without an extravascular component that takes into account tracer binding to endothelial cells were investigated and the authors suggested a two-tissue compartment model without an extravascular component as the preferred method for [¹⁸F]GE-180 quantification in healthy controls and 90 min as the optimal scan length for reliable estimation of volumes of distribution (V _T) [10, 11]. Distribution volumes from Logan plot and semi-quantitative SUVs (60–90 min p.i.) correlated well with V _T from 2TC model [10, 11]. Although the so far available pre-clinical data are promising, the performance of [¹⁸F]GE-180 as a tracer for neuroinflammatory diseases in human patients still needs to be verified.

This is the first study investigating relapsing-remitting MS (RRMS) patients with [¹⁸F]GE-180 PET with the aim of quantifying the uptake in various anatomical brain regions and in focal lesions. In particular, we focused (1) on the identification of a pseudo-reference region (PRR), which is challenging in diseases with widespread inflammation within the brain [12], and (2) on the comparison of parameters obtained from dynamic and static data, the latter one avoiding long scan times and demanding data processing steps, with the goal of providing a quantification procedure which is suitable for routine clinical use.

This study aimed to provide a robust, clinically suitable quantification approach for the third-generation TSPO ligand [¹⁸F]GE-180 in MS patients. The investigated static 60–90 min imaging containing a PRR-based SUVR quantification correlated well with DVR from modelling by application of the Logan reference tissue model on dynamic 90 min and thus proved suitability for clinical TSPO PET application, when patient compliance and economic aspects have to be considered. The presence of non-saturated lesion TACs suggests that a prolongation of the scan duration, at the cost of a lower count statistic, might allow for an improved assessment of equilibrium and tracer wash-out.

Binding potentials derived with reference modelling (BP = DVR − 1) reported previously for the prototypic TSPO ligand [¹¹C](R)-PK11195 in healthy controls, and MS patients were in a similar range as results presented here for [¹⁸F]GE-180: the lowest BP was found in normal-appearing white matter and the highest BP in the thalamus and the brainstem [23‐25]. In agreement with our SPM analysis results, PET signal was significantly elevated compared to HC in contrast-enhancing lesions, thalamus, parts of the brainstem, and in white matter frequently following white matter fibre tracts [23, 25‐27].

MS lesions exhibited a high [¹⁸F]GE-180 uptake and contrast, enabling a visual detection of focally elevated tracer accumulations (Fig. 4). The lesion-to-WM-background ratio (up to a threefold increase in mean lesion SUVR) appears to be high for [¹⁸F]GE-180 compared to other TSPO radioligands previously used in MS patients [23‐30]. [¹¹C]PK11195 signal in static images normalised to cortical grey matter was significantly higher in lesions with CE in MRI compared to normal white matter (up to a factor of 1.4) [26]. For [¹⁸F]FEDAA1106, lesions with CE in MRI were not detectable in SUV and Logan V _T images, probably due to a high non-specific uptake [30]. Both [¹⁸F]PBR111 and [¹¹C]PBR28 showed an increased V _T in some lesions with CE in MRI [24, 28, 29]. However, for [¹¹C]PBR28, static SUV_90–120 images were too noisy for visual detection of MS lesions, most probably due to the short half-life of ¹¹C in combination with a high-resolution PET tomograph [29].

The critical aspect for robust and reliable lesion quantification is the choice of the reference region. It is difficult to propose a standard reference region for all neurological diseases since patterns of affection vary widely. In MS, immune cell infiltration is predominantly localised in focal white matter lesions. However, as the disease progresses, inflammatory changes spread throughout the CNS and no region can be assumed to be unaffected. The corrected frontal cortex seems to be a suitable pseudo-reference region, at least for RRMS patients, since grey matter was reported to be less affected than white matter in early stages of MS [24‐26, 28]. In order to identify the least affected regions in RRMS patients, we compared our patients with a group of young HC, in which no CNS inflammation should be present. Although the SPM analysis revealed a non-negligible fraction (24%) affected by disease in the frontal cortex, time activity curves were not significantly different for HC and RRMS patients in this region, and it was feasible to exclude suspicious voxels in this relatively large and well-defined region. The VOI-based comparison of PRR SUVs with HC data showed good concordance. Normalisation of PET data by the corrected frontal cortex uptake reduced inter-patient variability in grey (from 11 to 7%) and white matter (from 13 to 10%) signals.

Alternative methods recommended for reference tissue TAC generation of TSPO tracers are data-driven clustering (DC) [31, 32] or supervised clustering (SC) on dynamic data preselected with a brain mask [33, 34]. SC has been validated for [¹¹C](R)-PK11195 and also tested for [¹⁸F]GE-180 [8]. However, DC and SC require dynamic PET studies and previous publications show that SC might not sufficiently exclude affected voxels in some cases and that other methods might be superior for the exclusion of affected voxels [8, 35]. A promising approach using fixed thresholds for the definition of affected voxels was applied to BP images derived with SC reference tissue [23] but also needs dynamic imaging.

The reported high inter- and intra-subject variability found for second-generation radioligands in other studies was attributed to differences in binding affinity status and in plasma protein binding [5, 36, 37]. For [¹¹C](R)-PK11195, in vitro and in vivo data show no significant differences between binding affinity groups. Unpublished in vitro work by D. Owen with cold GE-180 displacing [³H]PK11195 has shown a binding affinity ratio of 15:1 between HABs and MABs [11]. However, Feeney et al. [11] found no significant differences between MABs and HABs similar to [¹¹C](R)-PK11195 in healthy brain tissue of human subjects [16]. This is in line with the results of our current study in which we found no differences between MAB and HAB. Although in vitro prediction of differences in specific binding can differ from the relation observed for in vivo data [15, 38], it is questionable whether this can explain the results. Such a discrepancy may be explained by the high dependency of the in vitro studies on experimental conditions like temperature, fluid composition, and presence of intact mitochondria [15]. Furthermore, brain microvascular endothelial cells change BBB properties in vitro [39]. Fan et al. [10] suggest that the finding that no differences could be observed in vivo may be caused by a lower TSPO affinity of [¹⁸F]GE-180 compared to other second-generation TSPO tracers. Another explanation proposed by both previous studies is the low brain tissue uptake [10, 11]. The high fraction of ligand bound to plasma proteins, probably resulting from a relatively high lipophilicity (logD at pH 7.4 is 2.95 [40]), may be the reason for the slow propagation into tissue and the constantly high activity concentration in blood vessels dominating signal in healthy tissue (suggesting similarities to [¹¹C](R)-PK11195 in vivo) [41, 42]. Another reason for low uptake in brain may be a fast clearance by efflux pumps.

Yet, we observe a high contrast in MS lesions and also in gliomas as published recently [43]. The important question is what are the underlying processes leading to this contrast. Does it reflect specific binding to TSPO or rather other processes like a BBB breakdown? For gliomas, we could demonstrate that [¹⁸F]GE-180 uptake patterns do not correlate with contrast enhancement in T₁-weighted MRI images and that even for some gliomas, the highest [¹⁸F]GE-180 uptake can be found in non-contrast-enhancing tumour areas [43]. Still one might hypothesise that those areas exhibit micro BBB breakdown without apparent enhancement of contrast agent in MRI [44], which may allow the passage of [¹⁸F]GE-180 through the leaky BBB. However, we also observe areas with CE in T₁-weighted MRI, i.e. with BBB breakdown, but without elevated [¹⁸F]GE-180 uptake [43]. Even if micro BBB breakdown might ease the supply of [¹⁸F]GE-180, the [¹⁸F]GE-180 signal intensity does not correlate with severity of BBB breakdown, leading to the assumption that the dominant process resulting in the observed high range of [¹⁸F]GE-180 binding should be attributed to TSPO expression levels rather than mere BBB breakdown. Nevertheless, the lack of micro BBB breakdown in some regions may lead to an underestimation of TSPO expression.

One limitation of this study may be the usage of data from two different PET/CT devices in which the PET part is identical, but the CT data may yield a slightly different attenuation correction. Also, for future clinical studies, it would be beneficial to gather a larger database of HC with varying age to account for age-related changes. Furthermore, for an encompassing and comprehensive interpretation of the underlying processes, it is indispensable to perform in vivo blocking studies in combination with pharmacokinetic modelling with a metabolite corrected arterial input function and a longer scan duration.

The possibility of static PET imaging provided by the proposed method in contrast to dynamic PET imaging will greatly increase the acceptance by patients, as 30-min scans are usually well-tolerated and the imaging protocol does not include blood sampling, which is often perceived as an invasive, displeasing method by patients and is therefore often avoided in clinical settings. With these tools, TSPO PET with [¹⁸F]GE-180 may enable straightforward clinical assessment of neuroinflammatory activity in MS beyond the scope of structural MRI and seems to be a highly promising imaging method to assess disease activity and therapy response in RRMS patients.

In patients suffering from RRMS, the new TSPO ligand [¹⁸F]GE-180 presented a highly elevated signal up to a threefold increase in SUVR of focal lesions compared to surrounding background. Our data demonstrate a high correlation between parameters obtained from dynamic PET imaging with simple SUV ratios extracted from static 60–90 min [¹⁸F]GE-180 PET scans using the corrected frontal cortex as pseudo-reference region.