Multi-tracer and multiparametric PET imaging to detect the IDH mutation in glioma: a preclinical translational in vitro, in vivo, and ex vivo study

Background This translational study explores multi-tracer PET imaging for the non-invasive detection of the IDH1 mutation which is a positive prognostic factor in glioma. Methods U87 human high-grade glioma (HGG) isogenic cell lines with or without the IDH1 mutation (CRISP/Cas9 method) were stereotactically grafted into rat brains, and examined, in vitro, in vivo and ex vivo. PET imaging sessions, with radiotracers specific for glycolytic metabolism ([18F]FDG), amino acid metabolism ([18F]FDopa), and inflammation ([18F]DPA-714), were performed sequentially during 3–4 days. The in vitro radiotracer uptake was expressed as percent per million cells. For each radiotracer examined in vivo, static analyses included the maximal and mean tumor-to-background ratio (TBRmax and TBRmean) and metabolic tumor volume (MTV). Dynamic analyses included the distribution volume ratio (DVR) and the relative residence time (RRT) extracted from a reference Logan model. Ex vivo analyses consisted of immunological analyses. Results In vitro, IDH1+ cells (i.e. cells expressing the IDH1 mutation) showed lower levels of [18F]DPA-714 uptake compared to IDH1- cells (p < 0.01). These results were confirmed in vivo with lower [18F]DPA-714 uptake in IDH+ tumors (3.90 versus 5.52 for TBRmax, p = 0.03). Different values of [18F]DPA-714 and [18F] FDopa RRT (respectively 11.07 versus 22.33 and 2.69 versus − 1.81 for IDH+ and IDH- tumors, p < 0.02) were also observed between the two types of tumors. RRT [18F]DPA-714 provided the best diagnostic performance to discriminate between the two cell lines (AUC of 100%, p < 0.01). Immuno-histological analyses revealed lower expression of Iba-1 and TSPO antibodies in IDH1+ tumors. Conclusions [18F]DPA-714 and [18F] FDopa both correlate with the presence of the IDH1 mutation in HGG. These radiotracers are therefore good candidates for translational studies investigating their clinical applications in patients. Supplementary Information The online version contains supplementary material available at 10.1186/s40644-022-00454-6.


Background
The isocitrate dehydrogenase 1 (IDH1) mutation is a key molecular feature of the World Health Organization (WHO) Central Nervous System (CNS) tumor classification. This molecular alteration is a significant biomarker in the diagnostic and prognostic evaluation of gliomas [1,2]. Gliomas expressing the IDH1 mutation are associated with better chemo-and radiotherapy responses and longer patient survival periods [2][3][4]. This IDH1 mutation leads to an intracytoplasmic accumulation of 2-hydroxyglutarate (2-HG) [5] which alters overall tumor metabolism [6].
Whilst several studies have used magnetic resonance imaging (MRI) to attempt to non-invasively characterize the IDH1 mutation in gliomas ( [4,[7][8][9][10][11][12], the potential for positron emission tomography (PET) as a complementary metabolic imaging tool, is a growing field [13,14]. PET imaging allows to depict different patho-physiological mechanisms to study changes in cellular metabolism. The diagnostic information acquired with 18 F-fluorodeoxyglucose ([ 18 F]FDG) allows to investigate glycolytic pathways albeit that this radiotracer is poorly suited to neuro-oncology applications owing to its high physiologic uptake by healthy brain tissue [14]. Amino acid PET tracers, such as 18 F-fluoroethyltyrosine ([ 18 F]-FET) or 18 F-fluorodihydroxyphenylalanine ([ 18 F]FDopa) however bind to the L-type amino acid transporter-1 (LAT-1) and may be used to specifically investigate amino-acid metabolism. These amino-acid PET radiotracers exhibit reduced uptake in healthy brain tissue, thereby enabling more accurate characterization of gliomas [15]. The [ 18 F] FDopa radiotracer is widely used in the US and some European countries. It is cost-effective because, unlike 18 F-FET, it can be used for other clinical neurology or oncology indications e.g., to detect parkinsonian disorders or neuroendocrine neoplasms [16]. Although IDH1 mutation status and its correlation with amino acid PET data remains ambiguous and heterogeneous [17], some recent analyses recommend using dynamic analyses for predicting the IDH mutation [18,19]. New generation radiotracers have been developed to better characterize other glioma properties by PET imaging. Notable among these is the tumor inflammation specific [ 18 F]DPA-714, a ligand of the translocator protein (TSPO), expressed in glioma cells [20][21][22]. It is this advance that has recently lead to reports of TSPO PET radiotracers as suitable options for the characterization of gliomas at initial diagnosis [23]. To the best of our knowledge, no study has to date combined multi-tracer static and dynamic PET acquisitions to investigate the metabolic changes induced by expression of the IDH1 mutation in gliomas.
Currently, preclinical studies represent the sole means of investigating the IDH1 mutation phenotype, independently of other mutations and tumor characteristics, in vivo. Previous preclinical studies have attempted to investigate the overexpression of the IDH1 mutation in vivo [24][25][26][27] although none with the original objective of specifically investigating its pathophysiological expression by introducing the mutation in the well characterized U87 human high-grade gliomas (HGG) cell lines using CRISPR/Cas9. This innovative genetic approach allows to evaluate the effects of the isolated, basal expression of the IDH1 mutation. We previously characterized in vivo physiologic and metabolic changes of the IDH1 mutation using multiparametric MRI and spectroscopy [28]. Such multiple MRI sequences are nevertheless difficult to apply in clinical routine practice which underpins the recent interest in currently available PET radiotracers, particularly radiotracers specific for inflammation and metabolism, that may be able to inform about additional tumor properties and more targeted pathophysiological mechanisms.
The aim of our current study is to characterize in vitro and in vivo metabolic changes induced by the expression of the IDH mutation in U87 human derived HGG cell lines in the rat orthograft model, by using multiparametric and multitracer PET imaging specific for glycolytic metabolism ([ 18

In vitro study
The IDH1+ and IDH1-U87 human derived HGG cells were resuspended in HBSS at a concentration of 2 × 10 5 / mL, and incubated (37 °C, 5% CO 2 /95% air) with 1 MBq/ mL of each individual radiotracer for 30-, 60-and 120min. Tracer uptake was stopped with the addition of icecold phosphate-buffered-saline (PBS). The cell substrate was then washed twice in cold PBS and resuspended in 1 ml of cold HBSS. The radioactivity, expressed as counts per minute and corrected for individual radiotracer decay, was measured using a calibrated gamma counter (Wizard, PerkinElmer ® , France). Cell uptake, expressed as uptake per million cells, was determined by trypan blue. Individual radiotracer experiments were performed in duplicate for each of the three time points (N = 6) examined per radiotracer and per cell line.

In vivo study
PET recordings were obtained with a camera dedicated to small animal studies (Inveon, Siemens Preclinical Solutions ® , Knoxville, USA).
Two sets of rat brain acquisitions were performed, static acquisitions only for the first set of rats and dynamic and static acquisitions for the second set of rats.
A flowchart of the in vivo procedure is detailed in Fig. 1.
For static acquisitions, a 30-min PET scan was performed with [ 18 F] FDG (injected activity of 74 MBq/mL) at 60 min p-i [30], [ 18 F] FDopa (30 MBq/mL) at 30 min p-i [31] as recommended in the literature [32]. For [ 18 F]DPA-714 radiotracer, a 20 min PET scan was acquired at 60 min p-i with an injected activity of 30 MBq/mL [33]. All PET images were reconstructed using the ordered-subset expectation maximization 3D algorithm (OSEM3D, 4 iterations, 16 subsets, zoom 1) together with scatter and attenuation corrections based on the transmission scan using a 57 Co source measurement. The final voxel size was 0.77 × 0.77 × 0.79 mm 3 .
List-mode acquisitions were started a few seconds prior to the tracer injection with a 120 min acquisition for all radiotracers, and all acquired PET data were subsequently reconstructed in 27 consecutive frames (i.e., 5 frames × 120-s, 22 frames × 300-s). Reconstructed parameters were similar to those reported for static acquisitions.
Volumes of interest (VOIs) were defined on static PET and PET/T 2 -w MRI fused images with dedicated software (Inveon Research Workplace 4.1, Siemens ® , Knoxville, USA) and centered on the maximal tumor uptake. From these VOIs, tumors VOIs were deduced from metabolic tumor volumes defined in the literature: with a 0.8 ratio of healthy brain tissue for [ 18 F] FDG, a 1.3 ratio for [ 18 F] FDopa [14], and a 1.8 ratio for [ 18 F]DPA-714 [23]. The normal brain reference was defined as a symmetrical VOI in the contralateral healthy brain.
Maximum and mean Standardized Uptake Values (SUV max and SUV mean , respectively) were determined Fig. 1 Flowchart of experimental design. Rats were orthotopically implanted with U87 human derived HGG IDH1+ or IDH1-cell lines. After tumor outgrowth, imaging sessions were performed sequentially during 3-4 days. The order in which radiotracers were imaged was predicated by their availability and was as follows [ 18 F] FDG PET on the 14th day post tumor cell graft (p-g), MRI and [ 18 F] FDopa on the 15th day p-g, [ 18 F]DPA-714 between the 16th day and 17th day p-g. Rats were subsequently sacrificed for tumor histology analyses from tumor VOIs. A partial volume effect correction (PVC) was applied to tumor SUV max and SUV mean values using the maximum and mean recovery coefficients (RCs) corresponding to the tumor volume, respectively. For each radiotracer, RCs were computed from phantom experiments using spherical inserts (volume range: [0.031 mL; 0.25 mL]) and a sphere-to-background contrast corresponding to the mean contrast observed between tumor and healthy brain. After PVC, tumor-tonormal-brain (TBR) ratios were computed as SUV mean or SUV max of the tumor uptake divided by the SUV mean of the normal brain (TBR mean and TBR max ). VOIs previously defined on static images, were identically defined on dynamic images. For dynamic analyses, reference Logan models were generated for [ 18 F] FDopa and [ 18 F]DPA-714 as they have been shown to be adapted to the pharmacokinetics of these radiotracers [34,35]. Two parameters, with regression starting at 30 min, were extracted from this model, namely the distribution volume ratio (DVR) and the relative residence time (RRT) computed respectively as the slope and the negative of the intercept. Additionally, the Dice Similarity Coefficient (DSC) [36] was further calculated at the individual rat level, to measure the index of similarity between different metabolic volumes obtained with different PET tracers.
The volume of interest (VOI) was manually delineated on the T 2 -w images of the MRIs of each rat brain to determine the tumor volume.

Statistical analyses
All analyses were performed using SPSS (IBM, SPSS Statistics version 25.0) and R version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria). The two-tailed significance level was set at p < 0.05. As the number of animals examined was not sufficient to assume a normal distribution, continuous variables were expressed as medians and interquartile ranges, except for in vitro variables which were expressed as means and standard deviations. PET and MRI parameters were compared using Mann Whitney with adjusted p values (Benjamini Hochberg correction) to reduce the false discovery rate. The ability of each individually extracted parameter to predict an IDH mutation was assessed using receiver operating characteristic (ROC) curves from which area under the curve (AUC), sensitivity, specificity and accuracy were computed. The optimal threshold was determined by selecting the point on the curve closest to (0,1).

Rat population
For static acquisitions, we induced 24 IDH1+ and 28 IDH1-rat brain tumors. For static and dynamic acquisitions of the second set of rats, 15 rats with IDH1+ and 14 rats with IDH1-tumors, were imaged. A detailed Table  (Table S1) of available imaging session samples for each radiotracer with both static and dynamic datasets is presented in the additional file.  Fig. 3. Examples of tumor activity curves and dynamic analyses are provided in Fig. 4.

Dice
The only exploitable results from the DICE coefficients (N ≥ 5 for direct comparisons between the two radiotracers) showed a low DICE coefficient between

Discussion
In this preclinical multiparametric and multi-tracer PET study, the IDH1 mutation is associated with lower glioma uptake of [ 18 F]DPA-714 in vitro, in vivo and ex vivo. In addition, the dynamic RRT [ 18 F] FDopa parameter is also associated with the presence of the IDH1 mutation. These preclinical results are consistent with recent results obtained in human patients [37,38]. In contrast, no significant differences were observed between IDH1+ and IDH1-tumors for static [ 18 F] FDG and [ 18 F] FDopa acquisitions.
Our study used an animal model based on human derived HGG cell lines. Even though the IDH mutation is predominantly observed in glioma patients with lowgrade tumors, one of the advantages of using a high-grade tumor model in preclinical studies is to reach a high enough rate of tumor growth potential in these models to accentuate any potential differences in uptake of the PET radiotracers investigated. Several experimental studies have endeveoured to develop candidate radiotracers for noninvasive PET imaging of the IDH1 mutation but most of these yielded inconclusive results due to instability of the radiotracer in blood, lack of radiotracer selectivity or because experiments were not performed under representative physiologic conditions [39][40][41]. Our current study is original since it employs a combination of three  different, easily clinically transposable, PET radiotracers to investigate specific pathophysiological pathways of glioma development that may be uniquely impacted by the IDH1 mutation, using an innovative application of CRISPR/Cas9. The radiotracer of interest from our current study is [ 18 F]DPA-714, because it targets TSPO receptors which are upregulated during neuroinflammation. A lower uptake of this radiotracer was observed in IDH1+ tumors in vitro, in vivo, and ex vivo thus confirming the specificity of this radiotracer. The lower tracer uptake values are consistent with the reduced level of neuroinflammation and by extension the decreased aggressiveness of IDH+ tumors, which is in complete accordance with reports from the literature [38,42]. Our recent study, which investigated multiparametric MRI, also found a spectroscopic profile of lower aggressiveness for IDH+ tumors, which further confirms results obtained in the current study [28]. Other mechanisms related to the IDH mutation may be suggested albeit based on assumptions that cannot be substantiated from our current data. In terms of metabolism, the IDH1 mutation may for instance lead to an intracytoplasmic accumulation of 2-hydroxyglutarate (2-HG) [5]. This may in turn alter the glioma epigenome by increasing DNA methylation [43,44], activating cellular malignant transformation [45,46], thereby altering overall tumor metabolism by inducing a downregulation of phospholipid biosynthesis [6,[47][48][49]. Such a mechanism may thus contribute to the less aggressive profile of tumors with an IDH mutation. Although amino-acid radiotracers have been increasingly studied in neuro-oncology [50], TSPO imaging seems to be particularly promising because it identifies glioma physio-pathological mechanisms that are distinct from those invoked by amino-acid tracers. This could explain the weak DICE index between TSPO and amino-acid uptake observed in our current study even though it was obtained from a small number of rats. TSPO imaging and IDH mutational status confer a trend towards higher 18 F-GE-180-uptake in IDH-wild-type gliomas in the overall group, albeit that this trend was not confirmed in identical WHO grade gliomas [23]. Interestingly, both static and dynamic parameters for [ 18 F] DPA-714 provided high predictive performance values to discriminate between the two tumor types, the best performances being obtained with the dynamic RRT [ 18 F]DPA-714 (AUC of 1 [1;1]). Blood brain barrier (BBB) penetrance remains an issue for TSPO imaging [22] but is not a limitation for brain tumors in general or for gliomas per se since the brain blood barrier is disrupted in the vast majority of HGGs. Our previous study showed that contrast-enhancements were observed in our animal models, confirming that the BBB was disrupted [28]. In any case, our current study underlines the fact that using multiparametric PET, i.e.  were associated with the IDH mutation, unlike any of the in vitro or ex vivo variables, it appears that dynamic parameters predominantly detect perfusion phase differences between the IDH1+ and IDH1-cell types. Indeed, wash-in-wash-out phases are typically observed in aggressive IDH-wildtype gliomas, which may correspond to tumors that not only express high concentrations of LAT transporters but also tumors that are characterized by more extensive tracer perfusion as discussed in our previous publication [34]. In contrast, our [ 18 F] FDG PET results are somewhat less surprising, and consistent with reports in the literature, since this radiotracer does not allow to differentiate tumor tissue from healthy brain cortex because of its physiologic uptake in healthy brain tissue, which limits the potential usefulness of this radiotracer in neuro-oncology [13] .
Our study does have several inherent limitations. Even though our tissue staining results confirmed the PET results, neo-angiogenesis and microglia/macrophages (GAMs), may have contributed to the TSPO signal [22]. This was not specifically tested in our current study. Further preclinical studies with post-autopsy histo-radiography analyses are required to better define the specific regions of radiotracer uptake. Moreover, it is known that TSPO polymorphisms can alter radiotracer uptake, which would limit the radiotracer's potential usefulness in translational human studies. Our current study is also limited by the fact that not all rats were able to be PET imaged for all three radiotracers, due to the differing availabilities of individual radiotracers delivered to the lab at the time. In addition, although the volumes of IDH1+ tumors were statistically similar to IDH1tumors, tumor growth varied between the groups and between radiotracer acquisitions. The concordance of in vitro, in vivo and ex vivo results at least for the [ 18 F] DPA-714 PET nevertheless suggests that the contribution of any such potentially confounding effects is limited.

Conclusion
To summarize, [ 18 F]DPA-714 and [ 18 F] FDopa are two PET radiotracers that are associated with the presence of the IDH1 mutation in HGG in this preclinical study. Their complementarity may lead to future translational studies exploring their clinical applications in patients.