Introduction
Glutamate, the major excitatory neurotransmitter in the brain, plays an essential role in a variety of physiological processes. Among several ionotropic and metabotropic receptors, the metabotropic glutamate receptors (mGluRs) are G protein-coupled receptors involved in the modulation of synaptic transmission and neuronal excitability [
1]. The mGluRs of the group I are located post-synaptically, and they include the mGluR type 1 (mGluR1) and type 5 (mGluR5).
Both mGluR1 and mGluR5 have been linked to a number of neurological disorders, including epilepsy, stroke, fragile X syndrome, Huntington’s disease, obsessive-compulsive disorder, Alzheimer’s disease, Parkinson’s disease, and drug addiction [
2]. Thus, given the relevance of group I mGluRs for the evaluation of potential therapeutic interventions, there is a growing interest for
in vivo monitoring of group I mGluRs in the living brain which can be achieved by means of positron emission tomography (PET) imaging. Although mGluR1 and mGluR5 share a high degree of homology, they are characterized by a distinct cerebral expression pattern, with mGluR5 mainly distributed in the striatum, hippocampus, and cortex, whereas mGluR1 primarily located in the thalamus and in the cerebellum [
3,
4].
While a large body of literature on preclinical and clinical mGluR5 PET imaging is available [
5], only a limited number of studies in disease models have been reported for mGluR1 PET imaging [
6]. In particular, the application of mGluR1 PET imaging in the mouse brain has been extremely limited to date, an important shortcoming given the relevance of mouse models in understanding the pathophysiology of neurological disorders.
Among the radiotracers for mGluR1 described in the literature, N-[4-[6-(isopropylamino)-pyrimidin-4-yl]-1,3-thiazol-2-yl]-N-methyl-4-[
11C]methylbenzamide ([
11C]ITDM) [
7] is one of the most promising and better characterized radiotracers in the preclinical (rats and rhesus monkeys) settings [
7‐
9]. Notably, Yamasaki and colleagues [
8] claimed that the pons is a suitable reference region for noninvasive kinetic modelling of [
11C]ITDM in rats. [
11C]ITDM has also been employed to investigate changes in mGluR1 levels in a mouse model of Huntington’s disease using the pons as reference region [8]. However, as
in vivo validation has not been performed in mice, it is not clear yet whether the pons is a receptor-free region suitable as reference region. A proper validated reference region is of the utmost importance since erroneous selection of reference region for quantifying PET signal may lead to significant misinterpretation of PET data.
For these reasons, the first aim of the present study was to validate the specific binding of [
11C]ITDM and to investigate whether a suitable a reference region exists in the mouse brain. To this end, we performed
in vitro blocking of [
3H]ITDM as well as
in vivo blocking and displacement of [
11C]ITDM with the specific mGluR1 antagonist YM-202074 [
10]. Since invasive arterial blood sampling in mice presents several challenges and limitations in the perspective of longitudinal studies, the second aim of the study was to investigate whether an image-derived input function (IDIF) could be used as a valid alternative approach for noninvasive quantification of [
11C]ITDM PET imaging in the eventual absence of a reference region. The third aim was to characterize suitable kinetic models for [
11C]ITDM quantification and to assess possible semi-quantitative approaches.
Discussion
The present study described in vivo and in vitro validation of [11C]ITDM specific binding to mGluR1 in the mouse brain, investigated the use of an IDIF for noninvasive kinetic modelling for future longitudinal studies, and characterized pharmacokinetic models for [11C]ITDM imaging in mice.
Since, to the best of our knowledge, no
in vivo validation of [
11C]ITDM binding to mGluR1 has been reported in the mouse brain, the primary aim of this work was to investigate whether a region devoid of specific [
11C]ITDM binding existed and thus reference region-based kinetic models could be applied. We demonstrated lack of a suitable reference region in the mouse brain for [
11C]ITDM. As a first indication,
in vivo pretreatment with the highly selective mGluR1 antagonist YM-202074 resulted in a significant blockade in the whole brain, primarily in the mGluR1-rich regions (thalamus and cerebellum), but also in low mGluR1 density regions (pons). This evidence was further supported by the
in vivo displacement study, which clearly showed a manifest decline in [
11C]ITDM binding following the administration of YM-202074 in the mouse brain, including pons. Based on these findings, we conclude that reference region-based kinetic modelling is not possible for [
11C]ITDM in mice since no region devoid of specific tracer binding exist [
18]. To the best of our knowledge, only one study investigated [
11C]ITDM binding changes in the mouse brain, quantifying mGluR1 in a model of Huntington’s disease using a reference region-based (pons) kinetic modelling [
8]. However, as no
in vivo validation was performed, the findings based on such analysis should be interpreted with care. Indeed, if neurological disorders, ageing, or their combination have an effect on [
11C]ITDM specific binding in the pons, this could severely affect reference region-based [
11C]ITDM quantification and lead to significant misinterpretations of the outcome.
Small animal PET imaging features a relative limited resolution for
in vivo imaging in relation to the size of the rodent brain; thus, partial volume and spill over effects (PVE) might occur between high contrast regions [
19]. Given the anatomical proximity of cerebellum and pons, high and low uptake regions for [
11C]ITDM respectively, it cannot be excluded that at least a portion of the changes observed in pons might be related to spill in from the cerebellum. In order to investigate this eventuality, we performed
in vitro blockade of [
3H]ITDM with different doses of YM-202074, proving blockade of [
3H]ITDM specific binding occurred in a dose-dependent manner in all regions, pons included. Therefore, since
in vitro validation supported the
in vivo findings, we could exclude that the changes observed in the pons were PVE-related.
Noteworthy, when injected in rats, the radiolabelled YM-202074 ([
11C]YM-202074)
in vivo metabolism was fairly rapid [
20]. Thus, it is conceivable that the extent of
in vivo blockade measured (79 %) could even be underestimated in relation to the total amount of YM-202074 injected. Nonetheless, as metabolites of YM-202074 do not apply to an
in vitro setting [
20], the agreement observed between the
in vitro and
in vivo blockade studies confirmed the lack of suitable reference region. In order to further investigate the extent of
in vivo specific binding in the pons, we are currently testing an equally potent but more stable compound to block [
11C]ITDM binding.
The lack of suitable reference region implies the need of measuring an input function in order to perform kinetic modelling of [
11C]ITDM. However, invasive arterial blood sampling in mice presents several challenges and limitations in the perspective of longitudinal studies due to the small amount of blood collectable and since it is an end of life procedure. Hence, an attractive approach to perform noninvasive quantification is the application of an IDIF [
21] to bypass the need for an invasive input function as we previously validated in mice over an atereriovenous shunt for the mGluR5 radiotracer [
11C]ABP688 [
11]. Similar to the findings for [
11C]ABP688 [
11], the IDIF overestimated the tail of the [
11C]ITDM input function, thus resulted in lower
VT values. However,
VT (IDIF, Uncorr) values showed excellent correlation with the
VT using the metabolite-corrected plasma AV shunt as input function (
VT (AV shunt, Corr)) (
r = 0.977,
r2 = 0.954,
p < 0.0001), supporting the applicability of this noninvasive approach for [
11C]ITDM quantification.
Importantly, mGluRs have been described in the rodent heart [
22]. In order to exclude that the higher uptake measured with the IDIF compared to the AV shunt was related to specific binding in myocardiocytes, we compared the IDIF of the same animals during baseline and pretreatment as well as displacement with YM-202074. As for both paradigms no difference in the IDIF was observed (Suppl. Fig.
5, see ESM), the higher values in the IDIF are likely to be due to non-displaceable activity spilling-in from the myocardium, which becomes visible with the decline of blood activity. Indeed, although other organs surround the heart, lungs are characterized by extremely low uptake and liver is not in close proximity to the VOI.
While only intact radiotracer has been described in the rat brain,
in vivo metabolism of [
11C]ITDM has been reported in the plasma of rats and non-human primates [
7], with nearly 40 % of intact radiotracer 30 min p.i. in parallel with the generation of polar metabolites that do not penetrate the blood-brain barrier. Accordingly, we measured around 31 % of intact radioligand at 30 min p.i. in mice. However, the small blood volume in mice does not allow collection of multiple blood samples for individual metabolite correction with standard techniques. For this reason, we generated a population-based metabolite curve in order to adjust for the peripheral metabolism and to validate the applicability of a noninvasive IDIF over the metabolite-corrected plasma input function. A limitation in the use of a population-based correction is the need of high inter-individual reproducibility; otherwise, under- or overestimations in the quantification may be introduced since the same curve is applied to all the subjects as we recently reported for the radioligand [
11C]UCB-J [
13]. Since correcting the noninvasive IDIF for metabolism did not improve the agreement to
VT (AV shunt, Corr), we will use the uncorrected IDIF as noninvasive input function. Nonetheless, before studying the radioligand in disease-specific animal models, it is recommended to verify whether animals with different genotype or diseased condition are characterized by altered metabolism.
Kinetic analysis was performed by comparing 1TCM, 2TCM, and Logan plot. According to visual assessment and model selection approaches (AIC and MSC), 1TCM was excluded as it did not fit the data, while both 2TCM and Logan plot proved to be valid alternatives. In addition, time stability of the
VT (AV shunt, Corr) and
VT (IDIF, Uncorr) estimations for both 2TCM and Logan plot were investigated. Shortening the scan duration destabilized
VT (AV shunt, Corr) and
VT (IDIF, Uncorr) outcomes with both models. This was expected given the fairly gradual brain uptake of [
11C]ITDM and its slow wash-out profile similar to the findings reported in rats [
8].
Finally, we explored whether a different approach simpler than VT (IDIF, Uncorr) could be employed to measure [11C]ITDM uptake. While SUV was not a reliable measurement, the use of SUVR resulted in more accurate values. Interestingly, SUVR (IDIF, Uncorr) (r2 = 0.907) performed almost as good as VT (IDIF, Uncorr) (r2 = 0.954) when comparing both measurements to the VT (AV shunt, Corr). While this might indicate that a static PET scan 60 to 90 min p.i. could be sufficient, it is important to underline that this approach might still need dynamic PET imaging in order to define the VOI for the IDIF following radioligand injection, unless a different strategy to extract a reliable IDIF is identified. Alternatively, it may be possible that a single blood sample collected during the PET scan could be sufficient to measure blood activity to derive the SUVR (AV shunt, Corr). Yet, such approach appeared less reliable (r2 = 0.692), invasive, and likely to be sensitive to noise due to the low activity in blood following decay of the radioligand at these later time points p.i.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.