Technical feasibility of [¹⁸F]FET and [¹⁸F]FAZA PET guided radiotherapy in a F98 glioblastoma rat model

Glioblastoma (GB) is the most common primary malignant brain tumor. It has been given the highest grade of WHO classification of glioma (WHO grade IV) and is a highly invasive solid tumor [1‐3]. Histologically, GBs are characterized by increased vascularity, nuclear atypia, necrosis and increased mitotic activity. Tumor cell infiltration deep into the surrounding brain parenchyma renders a complete surgical resection elusive leading inevitably to tumor recurrence in most cases [4, 5].

Several factors such as lesion size, location and growth rate determine the clinical presentation of patients with GB. Headache, seizures, personality changes and focal neurological deficits are common symptoms of GB [1]. The clinical course of GB is rapidly progressive, with a mean survival time between six to fourteen months. Standard medical treatment consists preferably, and if possible, of a maximal safe surgical resection. Subsequently radiotherapy (RT; 60 Gy in 30 fractions) and concomitant chemotherapy with temozolomide (TMZ; 75 mg/m² during six weeks) followed by adjuvant chemotherapy with TMZ (150–200 mg/m² × 5 days, during 6 cycli of 28 days) are administered. Even with this treatment strategy, patients with GB typically develop fast post-operative progression resulting in fatal outcome for the majority of them [1, 6].

To improve this poor survival rate, new treatments are crucial. RT is an essential part of (postoperative) GB treatment, as has been demonstrated in the past [7]. RT planning is currently based on contrast-enhanced computed tomography (CT) and magnetic resonance imaging (MRI). The International Commission on Radiation Units and Measurements Report No. 50 (ICRU-50) proposed to delineate three different target volumes based on diagnostic imaging (Fig. 1) [8]. The gross tumor volume (GTV) reflects the macroscopical tumor volume as assessed by contrast-enhanced anatomical imaging. To include the microscopic infiltration into the adjacent healthy brain tissue, an extra margin is proposed being the clinical target volume (CTV). Finally, the planning target volume (PTV) also accounts for interfractional and intrafractional variations [9, 10].

In 2000, Ling et al. (10) introduced the biological target volume (BTV). The BTV is based on the concept that irradiation of the tumor region with an adapted dose based on proliferative activity, could result in a better outcome. To include BTV in the treatment protocol, a non-uniform dose can be delivered to the target volume based on functional imaging (i.e. dose painting) [11, 12]. Hence, additional biological information from positron emission tomography (PET) may have added value for defining this RT target tumor volume [12, 13]. The use of BTV in a (pre)clinical set up has been evaluated by Trani et al. (11) where they demonstrate the technical feasibility of preclinical dose painting strategies. However, more preclinical studies that confirm these results, or are aimed at testing similar biological hypotheses, are needed [12, 14].

O-(2-[¹⁸F]fluoroethyl)-L-tyrosine ([¹⁸F]FET) and 1-α-D-(5-deoxy-5-[¹⁸F]-fluoroarabinofuranosyl)-2-nitroimidazole ([¹⁸F]FAZA) were selected as tracers for PET-guided RT. Uptake of the amino acid PET tracer, [¹⁸F]FET, correlates with tumor cell density and proliferation and is therefore able to identify tumor regions susceptible to relapse [15, 16]. Uptake of the hypoxia tracer, [¹⁸F]FAZA, is correlated with tumor hypoxia, which can cause radioresistance [16, 17]. As such, a higher radiation dose targeting those specific regions might be beneficial [18].

Rat models have remained the mainstay of neuro-oncology research for over 30 years. The C6 glioma has been used extensively for a variety of studies, however, it has the potential to evoke an immune response and it is not syngeneic to any inbred strain [19, 20]. The 9 L gliosarcoma has been widely used but can be immunogenic in syngeneic hosts. The U251 xenograft model shows histological characteristics of human GB and displays genetic similarities to human GB but is criticized for an inadequate tumor-host immune response [21]. The U87 GB model displays key dissimilarities to human GB at the histopathological level. Unlike GB, U87 tumors show a non-diffusely infiltrative growth pattern, with more homogeneous and leaky vessels [21, 22]. It is also worth mentioning that patient-derived glioblastoma models would necessitate the use of nude rats, which are much more expensive. Taking these models in consideration, the F98 GB model was selected for this preclinical study based on the presence of all the necessary histologic characteristics, the costs and the extensive experience of our research group with this model [23‐26]. Recently, in 2017 Bolcaen et al. optimized a protocol for PET- and MRI-guided irradiation of a GB rat model using a micro-irradiator [23]. Using this approach, it is our goal to conduct a proof of concept for the application of PET-defined subvolume boosting using two different PET tracers and subsequently evaluate treatment outcome using MRI.

Given the poor survival of GB patients, there is an urgent need for improving current and developing new treatment methods. Multiple therapies have been proposed at the time of GB recurrence including cytotoxic chemotherapy, immunotherapy, angiogenesis inhibitors,… [30]. Unfortunately, none of these therapies turned out to be a big breakthrough [31]. Therefore, more research is being done in optimizing the currently used radiation therapy to provide the best possible outcome [18]. Although a dose-response relationship in GB is well known, escalation studies giving a higher dose to the entire tumor have not been successful [32]. PET-guided RT targeting the most metabolically active or radioresistant tumor region with RT might improve local tumor control without compromising normal surrounding brain and possibly lead to a better survival in GB patients. However, only limited studies have been performed using PET guided RT in GB rats. Menichetti et al. demonstrated the usefulness of [¹⁸F]FET-PET guided boron neutron capture therapy in F98 rats [33]. However, several groups have been incorporating optical molecular imaging techniques such as bioluminescence and fluorescence imaging, to guide RT successfully [34‐38]. Unfortunately, optical imaging has limited spatial resolution due to tissue absorption and scatter of visible light and background artifacts need to be taken into account caused by auto-fluorescence [39]. Most importantly, bioluminescent imaging cannot be easily translated to the clinic as it requires genetic manipulation of tumor cells. As such, the implementation of PET for RT planning looks promising [40].

In our F98 GB rat model, the TBR_mean value of [¹⁸F]FET was significantly higher compared to [¹⁸F]FAZA, suggesting that distinction between tumor and healthy brain tissue was more evident on the [¹⁸F]FET PET scan.

Unfortunately, there are only few studies available that describe the uptake of [¹⁸F]FAZA in GB patients or in a GB rat model. In order to characterize the biological features of the F98 GB model, Belloli et al. [41] studied the uptake of [¹⁸F]FDG and [¹⁸F]FAZA. Contrary to our study, the [¹⁸F]FAZA uptake in the study by Belloli et al. was centrally located and showed a TBR value of 2.2 ± 0.4, whereas in our study a lower TBR_mean for [¹⁸F]FAZA, i.e. 1.35 ± 0.14, was observed. In agreement with the study by Belloli et al., visual analysis of the PET scans with different tracers showed a more centrally located [¹⁸F]FAZA uptake in the tumor in comparison to [¹⁸F]FET uptake. The latter is in agreement with the knowledge that the most hypoxic tumor regions are centrally located whereas viable tumor cells, surrounding tumor necrosis are situated in the periphery of the tumor. As described by several research groups both [¹⁸F]FET and [¹⁸F]FAZA are suitable PET tracers for the purpose of sub-volume boosting [41, 42].

In agreement with the histological findings, GB is characterized by central tumor necrosis with a viable and rapidly infiltrating tumor rim whereas hypoxic regions are observed in the middle of the tumor and regions with elevated amino acid transport are detected at the border of the tumor [19].

RT was given using three non-coplanar arcs in order to limit the dose delivered to healthy brain tissue [29, 43]. However, some shortcomings of the SARRP system should be considered. The used collimators are rigid (square aperture of 1²–5² mm²) and therefore not conformed to the shape of the tumor. This might possibly be solved by developing a multi-leaf collimator as is being used in the clinic. In addition, this would also allow the use of intensity-modulated radiotherapy (IMRT), which is a common practice in the clinic. Dose painting by numbers, impossible without IMRT, was not possible due to unavailability of software needed for performing the calculations. Due to these restrictions, we opted for a set up with sub-volume boosting using a 1 × 1 mm² collimator to deliver an additional dose of 5 Gy to the tumor region corresponding with the highest PET signal.

By means of DVHs the influence of the extra 5 Gy on the total tumor dose was estimated for the groups receiving the PET based subvolume boost irradiation. Only the dose that was delivered to 2% of the tumor volume (which corresponds with the maximum dose) was significantly different. This can be explained by higher dose values in the groups with PET sub-volume boosting in comparison to dose values calculated based on the MRI based irradiation. The results illustrate that there is no significant difference regarding the delivered dose located at the tumor volume between the three therapy groups. This means that the PET sub-volume boosting does not result in a higher radiation burden for the adjacent brain. It should be noted that using a 1 × 1 mm² collimator to perform the boosting results in a very limited volume receiving the additional 5 Gy. If a larger target volume would be used, the DVHs would differ considerably.

Tumor volumes at different timepoints were compared with the Mann-Whitney U test. Soon after initiation of therapy, the first effect of treatment was visible showing a significant difference in tumor volumes from day 5 until day 12 between the treated and control groups. However, no significant differences were observed between the MRI based and [¹⁸F]FET- vs. [¹⁸F]FAZA-PET based RT groups. Tumor growth in the MRI based RT group corresponded to our previously published results [29], where GB was treated with RT and TMZ in the F98 rat model. The growth of the tumor remained stable until day 9 after therapy, while a continuous increase in tumor volume was observed in the control group.

Three important limitations of this study include the resolution of the PET scanner (1.2 mm) and the limited number of rats that completed the study. A more accurate delineation of the BTV might be possible by using a small animal PET scanner with better spatial resolution. Secondly, taking into account the humane endpoints, several animals had to be euthanized early due to clinical deterioration as result of large tumors as previously described by Schültke et al. [44]. This issue will influence the average values significantly. To prolong the follow-up period of the rats, inoculation of the tumor in the thigh could be an alternative [14]. However, subcutaneous glioma models lack the specific CNS microenvironment that has an essential role in glioma biology and progression [45]. Finally, the implementation of preclinical PET guided irradiation is very time consuming. A timeline is shown in Fig. 5. Using our GB rat model, we are obliged to perform the imaging and irradiation subsequently, while the rat is continuously fixed to a multimodality bed with multimodality markers fixed to the head. This is necessary to be able to perform exact manual image fusion using the radiation planning software or PMOD, as automatic image fusion is still not reliable in small animals. Patients don’t have their imaging and irradiation subsequently as image fusion software is perfectly able to fuse MRI/CT/PET images. Furthermore, manual segmentation of the CT image is necessary for dose calculation, while this is not the case in patients. Also, the preclinical RT planning has to be performed immediately after the CT to exclude a change of the rats position and to limit the total anesthesia time. In patients, the radiation therapist starts to design a RT plan after receiving the multi-modilty images and subsequently the first treatment of the patient can be initiated. As such, the timeline of the imaging and RT planning of a patient is covered over days while preclinically all these steps have to be performed in a few hours during anesthesia of the rat.

Taking the limitations into account, our data demonstrates the feasibility of dose targeting in a F98 rat GB model using MR, [¹⁸F]FET- and [¹⁸F]FAZA-PET imaging. However, technical advances are needed to apply a non-uniform dose to a target region more accurately.

In this study, we showed the feasibility of PET guided subvolume boosting in a rat GB model. Comparing the effect of therapy applying MRI-based RT versus PET-based sub-volume boosting, we concluded that only a significant difference in tumor growth was found between active therapy and sham therapy respectively, while no significant differences were found when comparing the three treatment groups. Improvements for dose targeting in rodents are mandatory to apply a non-uniform dose to the more resistant regions of the tumor which might then be beneficial in terms of local tumor control. The latter combined with ne targeted drugs for GB treatment is, in our opinion, the most promising strategy.