Could ¹⁸ F-DPA-714 PET imaging be interesting to use in the early post-stroke period?

Cerebral stroke is a severe and frequently occurring condition that represents a leading cause of mortality and morbidity worldwide [1, 2], being also the main aetiology of adult-acquired disabilities [3, 4]. Stroke is haemorrhagic in 10% to 15% of cases but more often it is ischemic (85% to 90%) [3]. Due to arterial occlusion, ischemic stroke is a diagnostic and therapeutic priority; emergency treatment (e.g., intravenous thrombolysis) should be administered in the first hours after symptom onset. Unfortunately, no other emergency drug treatment has been validated, but an improved understanding of the pathophysiology of early cerebral ischemia should identify potential molecular targets, particularly for neuroprotective treatments. Therefore, promising neuroprotective drugs, which demonstrated effectiveness in animals, have been evaluated; however, these results remain unconfirmed in clinical trials [5, 6]. Moreover, there is a lack of reliable imaging strategies to assess brain neuroinflammation after stroke [7‐9].

Cerebral ischemia rapidly evolves to necrosis and a peri-necrosis area of ischemic penumbra, in which the brain tissue is still viable for a few hours. This cerebral tissue can be preserved if treatment is initiated quickly to restore the cerebral blood flow. Thus, this area of ‘darkness’ is the prime target for potential neuroprotective drugs. In animals, the infarcted area has been shown to expand in 24 h after the occlusion of a cerebral artery [10, 11].

Although different mechanisms are involved in the pathogenesis of stroke, increasing evidence suggests that inflammation, mainly involving the microglial and the immune system cells, account for its pathogenic progression [11].

Experimental studies have demonstrated that microglial cells are the first inflammatory cells activated after the onset of cerebral ischemia [12]. After breakdown of the blood-brain barrier (BBB), which accompanies cerebral ischemia, the perivascular microglia and macrophages are activated; this activation appears to be at least partially responsible for the inflammatory lesions in cerebral infarctions [12]. Microglia, which constitute up to 10% of the total brain cell population, change from a resting to an activated state in response to central nervous system insults, and this change stimulates these cells for phagocytosis. Several studies focusing on the inflammatory reaction during the first days after stroke have demonstrated that this inflammatory response changes dramatically over time [13, 14]. Therefore, the inflammatory markers correlated to the time course must be considered for any anti-inflammatory treatment approach in patients with acute ischemic strokes [14].

In humans, microglial activation can be assessed in vivo through neuroimaging of the 18-kDa translocator protein (TSPO) with selective TSPO radioligands. TSPO, formerly known as peripheral benzodiazepine receptor (PBR), is part of a multimeric ‘protein complex’ associated with the outer mitochondrial membrane of many cells [15]. TSPO is present in peripheral tissues and also in glial cells, but in the healthy brain, its expression is minimal [16]. TSPO may therefore be a valuable biomarker of inflammation, as it is highly expressed in phagocytic inflammatory cells.

A large number of positron emission tomography (PET) and single photon emission computerised tomography (SPECT) radioligands, selective for TSPO, have been developed, of which ¹¹C-PK11195 was the first to be evaluated [17]. Studies using ¹¹C-PK11195 demonstrated increased binding of this radiotracer around the outer border of ischemic lesions several days after stroke, as well as in distant areas from the lesion [18‐23]. This increased uptake was observed from 3 days after ictus, reaching its maximum at 7 days, and continuing for a period of 4 weeks [9].

However, the 20-min radioactive half-life of ¹¹C-PK11195 is a serious drawback to the increased accessibility of biomarkers for routine clinical purposes since the use of these markers is limited to centres with an on-site cyclotron. Fluorine-18-labelled ligands, therefore, appear to be the best alternative, as the 110-min half-life of fluorine-18 enables centralised production and loco-regional delivery. Several groups have also evaluated neuroinflammation models in rats using the TSPO radioligand ¹⁸ F-DPA-714, concluding that it provides accurate quantitative information of TSPO density after cerebral ischemia, herpes encephalitis, amyotrophic lateral sclerosis, and gliomas [24‐27]. Several studies have already demonstrated that the regional distribution of ¹⁸ F-DPA-714 aligned well with other PET studies using TSPO radioligands in the human brain [25, 26].

The aim of the present study was to assess microglial activation after a recent stroke using ¹⁸ F-DPA-714, and to evaluate the relationship between ¹⁸ F-DPA-714 uptake and the infarct volume, also analysed by magnetic resonance imaging (MRI).

The patients' clinical characteristics are summarised in Table 1. No adverse or subjective effects were observed for any of the nine subjects studied after injecting an average of 244.0 ± 27.4 MBq of ¹⁸ F-DPA-714. The imaging studies were performed 12.6 ± 2.7 days (min 8 days, max 18 days) after the acute injury.

Figure 1 shows an example of MRI and ¹⁸ F-DPA-714 images obtained from one patient (patient 4). For this patient, we observed that the radiopharmaceutical uptake is relatively comparable to injury tissue highlighted by MRI.

Figure 2 depicts a representative TAC of ¹⁸ F-DPA-714 uptake in the injured tissue and in the intact counterpart contralateral region and ipsilateral cerebellum (both used as reference regions) from a single subject. Figure 3 shows the evolution of the ratios calculated between the mean activity of the injured region and the two reference regions for the same subject. For all patients, the TAC revealed an initial rapid accumulation of radioactivity in the injured tissue, with a peak observed at 5 min post-injection (pi), followed by a gradual cerebral clearance from 10 to 20 min pi to the end of the PET acquisition. For the two reference regions, lower binding values of ¹⁸ F-DPA-714 were observed, with a maximum uptake at 5 min pi followed by two decreased phases from 5 to 20 min pi and from 10 to 20 min pi to the end of the acquisition. Moreover, the binding of ¹⁸ F-DPA-714 was lower for the contralateral region than for the cerebellum. For all subjects, the ¹⁸ F-DPA-714 uptake in the infarcted tissue remained higher than that observed for the other two regions during the entire acquisition.

Using the two reference regions at 40 and 80 min pi, the ratios obtained for the nine patients are presented in Table 2. The ratios using the mean activities were higher, although they were significantly different only at 40 min pi, when the contralateral VOI was used as the reference region, compared with the ratios using the cerebellum (IT/CL = 1.78 ± 0.31 and IT/Cerebellum = 1.55 ± 0.21 at 40 min, p = 0.020; IT/CL = 1.99 ± 0.56 and IT/Cerebellum = 1.79 ± 0.40 at 80 min pi, p = 0.155 respectively). Using the maximal activity calculated for each VOI, the ratios were also higher, however not significantly different, when using contralateral VOI (IT/CL = 2.05 ± 0.54 and IT/Cerebellum = 2.02 ± 0.64 at 40 min, p = 0.911; IT/CL = 2.44 ± 0.72 and IT/Cerebellum = 2.19 ± 0.54 at 80 min pi, p = 0.164, respectively).

The median lesion volume estimated by MRI was 13.1 cm³ [range 0.7 (subject 7) to 190 cm³ (subject 3)]. There was no statistically significant relationship between this volume and the ¹⁸ F-DPA-714 uptake, according to Spearman's test (p = 0.24). However, a strong correlation was observed between the lesion volume estimated by MRI and the VOI volume used to calculate the activity concentration (r = 0.95 end p = 0.0003, Spearman's test).

In this study, all imaging scans were performed within 12.6 ± 2.7 days after the ictus. Studies performed with the reference TSPO radioligand ¹¹C-PK11195 in stroke patients indicated that the binding became significant a few days after the stroke onset, subsequently increased in approximately 1 week, and declined after 3 to 4 weeks [18, 19, 29]. A more recent study with ¹¹C-PK11195, conducted 2 to 3 weeks post-stroke in a group of 18 subjects confirmed the local activation of microglia with increased radioligand uptake in the infarcted area [23]. Moreover, activation in the peri-infarct area appears to be more intense than in the core [30]. The value of our study is that it demonstrates the ability of a TSPO fluorine radioligand to map inflammatory process, in both the infarct and the surrounding surviving tissues. Indeed, one of the main difficulties in validating new radiopharmaceuticals as biomarkers of neuroinflammation lies in ruling out the possibility of non-specific binding, particularly passive leakage through a damaged BBB, as well as potential infiltration of peripheral circulating monocytes/macrophages within the brain parenchyma.

We observed that the increased uptake of ¹⁸ F-DPA-714 on PET images differed, even if very moderately, from the MRI gadolinium-enhanced regions, and was not limited to the infarct tissue with BBB breakdown; it also included non-infarcted cerebral tissue. This finding most likely reflects the activation of microglia with an increase in TSPO expression at the periphery of the necrotic lesion. On the other hand, we cannot forget that the PET spatial resolution is lower than that of the MRI's.

Moreover, the radiotracer uptake was not correlated with the infarct volume measured by MRI. This point is in accordance with a previous USPIO MRI study performed in subacute stroke patients demonstrating that the extent of the inflammatory response after stroke was unrelated to the lesion volume [31].

To our knowledge, only one study has compared two different TSPO radioligands in post-stroke patients (i.e., vinpocetine and PK11195, both labelled with carbon-11) [32]. The authors compared the diagnostic potential of these radioligands to visualise activated microglia in four post-stroke patients. They observed that the binding potential for ¹¹C-vinpocetine was greater than that obtained for ¹¹C-PK11195 in all evaluated regions, but the differences did not reach significance. ¹¹C-vinpocetine binding potential was greater in the peri-infarcted zone than in the ischaemic core, but the difference did not prove to be significant [32, 33].

However, a direct comparison between ¹⁸ F-DPA-714 and ¹¹C-PK11195 labelling, on a model of cerebral ischemia in rats, was done by Boutin et al. [34]. ¹⁸ F-DPA-714 displayed a higher signal-to-noise ratio than ¹¹C-PK11195, suggesting that with the longer half-life of fluorine-18, ¹⁸ F-DPA-714 could be a good alternative for TSPO imaging.

The TACs generated for each subject showed that ¹⁸ F-DPA-714 uptake in the lesion, as well as in the contralateral cerebral tissue and the cerebellum, reached peak values at 5 min pi. The TACs for the injured region showed a slow decrease in binding from the maximum uptake to the end of the acquisition. However, this kinetic analysis showed two decreased phases for the other two regions: (1) a faster phase, between 5 and approximately 20 min, and (2) a slower phase, from 20 min pi to the end of the acquisition. This evolution of the kinetics of ¹⁸ F-DPA-714 in the cerebellum and in the normal cerebral tissue aligned with that observed in a previous study performed in a group of seven healthy volunteers [25]. Indeed, in this previous study, we observed that the kinetics of the radioligand was similar for all cortical regions and the cerebellum.

On the other hand, stroke patients may show remote metabolic changes in the cerebellum resulting from diaschisis [25, 35]. Although microglial changes, as a consequence of cerebellar diaschisis, are not known, we have chosen to take into account only the cerebellar hemisphere ipsilateral to the stroke to define the reference region.

The evolution of the ¹⁸ F-DPA-714 uptake in humans differed from that reported by Martin et al. in an ischemia model in rats [36]. The authors showed that the ¹⁸ F-DPA-714 uptake in the lesion reached a peak value at 30 min and remained stable until the end of the acquisition. In our study, we observed an earlier peak at 5 min pi followed by a slower decrease.

We also demonstrated that the binding ratio of ¹⁸ F-DPA-714 was higher when the contralateral VOI was used as a reference region, compared to when the cerebellum was used as a reference region. Several authors have previously documented the fact that the cerebellum in humans was not devoid of TSPO [37] and it would be of interest in future TSPO PET imaging studies to focus on focal inflammation, including stroke, Rasmussen encephalitis, epilepsy, and traumatic brain injury, and to consider the contralateral area as a suitable and TSPO-expression free reference tissue. Nevertheless, Martin et al. also detected increased ¹⁸ F-DPA-714 binding in the contralateral side at 11 days after focal cerebral ischemia in rats. They concluded that this finding was probably due to different mechanisms related to the expansion of infarction, such as spreading depression [36]. Thus, the exact fraction of specific binding to TSPO, in both the contralateral region and the cerebellum, would have to be determined by a blocking study requiring substantial doses of a blocking agent, which was ethically inconsistent with the present study. Interestingly, the ratio values calculated in the present study between the injured tissue and the contralateral region at 40 min post-¹⁸ F-DPA-714 injection, using the mean activity values, were significantly higher (p = 0.020, Mann-Whitney U test) than those previously published by Gerhard et al. using ¹¹C-PK11195, obtained with three acquisitions from patients with the same specifications (interval in the range of 6 to 20 days between stroke ictus and PET acquisition) [18]. Although this comparison of two separate studies should be considered cautiously, it suggests that ¹⁸ F-DPA-714 might represent a good fluorine-18 alternative to ¹¹C-PK11195 for TSPO PET imaging, with the higher ratio of ipsilateral-to-contralateral uptake. Moreover, our results are in accordance with preclinical direct comparison of both radiopharmaceuticals in a rodent model of focal ischemia [34], demonstrating that ¹⁸ F-DPA-714 achieved in vivo a higher ratio of ipsilateral uptake to contralateral uptake than ¹¹C-PK11195.

Several studies have demonstrated the existence of different binder populations for ¹¹C-PBR28, for which approximately 10% of the population appeared to be non-binders [38]. Owen and colleagues evidenced 3 types of binding pattern: high-affinity binders (approximately 50%), low-affinity binders (approximately 20%) and mixed-affinity binders (approximately 30%), related to a single-nucleotide polymorphism (rs6971) within the human TSPO gene [39], and they extended this finding to other PET TSPO radioligands, including the carbon 11-radiolabelled derivative of ¹⁸ F-DPA-714, namely, DPA-713 [40]. One limitation of our study is that we did not directly identify the subjects with low or high affinity for TSPO. It is likely that ¹⁸ F-DPA-714 is also sensitive to this variable inter-individual affinity state. Nevertheless, all patients in our study demonstrated increased ¹⁸ F-DPA-714 uptake by the injured tissue.

Our results show that the uptake of ¹⁸ F-DPA-714 occurred not only due to the BBB breakdown, but also because of the activation of microglia in the area surrounding the infarct. We also showed that the kinetics of ¹⁸ F-DPA-714 differ between injured and normal tissues. The ability to assess microglial activation in vivo may improve our understanding of the mechanisms of neuroinflammation in acute disorders such as stroke, and should enable effective treatment monitoring. This study also demonstrates that a 10 to 20 min of PET emission acquisition, performed between 40 and 80 min pi, was able to differentiate acute tissue injury from normal brain tissue.