FMISO-PET and immunohistochemistry veri�ed tumor oxygenation, stemness, and immunosupportive microenvironment after preoperative neoadjuvant bevacizumab for newly diagnosed glioblastoma

Background Gadolinium-enhanced magnetic resonance imaging and T2-weighted imaging/uid-attenuated inversion recovery imaging are used to determine the e�cacy of bevacizumab (Bev) against glioblastoma (GBM). Positron emission tomography (PET) using 18 F-uoromisonidazole (FMISO) re�ects hypoxia in the tumor microenvironment (TME). This study compared FMISO-PET �ndings for alterations in tumor oxygenation in the TME of GBM during Bev treatment.


Background
Hypoxic regions within the tumor microenvironment (TME) show reduced radiosensitivity and also increase the risk of refractory tumors via the activation of transcription factors such as hypoxia-inducible factor-1α (HIF1-α) and, further downstream, vascular endothelial growth factor (VEGF) [1,2].Bevacizumab (Bev), a monoclonal antibody against VEGF, is known to inhibit angiogenesis in tumor tissue and normalize vascular structures in tumor tissue to improve oxygenation and reduce brain edema around the lesion by decreasing vascular permeability.Bev is currently used in glioblastoma multiforme (GBM) treatment [3].Recently, indirect effects have also been reported to reduce immunosuppressive immune cells such as regulatory T cells and tumor-associated macrophages within the TME of GBM by improving oxygenation [4,5].However, the e cacy of Bev therapy varies from person to person and the duration of e cacy is limited [3].Furthermore, contrast enhancement on magnetic resonance imaging (MRI) disappears during the course of treatment with Bev, representing one of the issues to be addressed in the future for proper imaging evaluation of therapeutic e cacy [6].
Evaluating alterations in glioma metabolism has become important when anti-angiogenic therapies such as Bev have been included along with standard chemoradiotherapy.Hypoxia plays a pivotal role in tumor malignancy, aggressiveness, and refractoriness against chemoradiotherapy.Positron emission tomography (PET) using the tracer 18 F-uoromisonidazole (FMISO) detects hypoxic TMEs, which speci cally accumulate the tracer.FMISO-PET has therefore been used to visualize hypoxic regions in GBM [7,8].Previous research has shown that alterations depicted on FMISO-PET have prognostic value due to their responsiveness to the induction of tumor oxygenation during Bev therapy for speci c recurrent gliomas [9][10][11].
Although the optimal timing and duration of Bev administration remain controversial, we have proposed that preoperative neoadjuvant Bev (neo-Bev) for newly diagnosed GBM contributes to reducing intraoperative blood loss and improving preoperative performance status, and might increase sensitivity to postoperative radiation therapy (RT) [2,12,13].To date, the only FMISO-PET studies that have evaluated the e cacy of Bev in the treatment of gliomas have been reports of recurrent cases.Furthermore, all observation periods to date have been short-term, and no reports have provided long-term observations of changes on FMISO-PET leading to the acquisition of resistance to Bev treatment in GBM over time.
The aim of the present study was to investigate whether FMISO-PET and immunohistochemistry (IHC) could verify our hypothesis that preoperative neo-Bev add-on neoadjuvant chemoradiotherapy might induce tumor oxygenation in patients with newly diagnosed GBM, while the TME in recurrent GBM would become hypoxic again when recurrent tumor becomes resistant to Bev therapy.

Patients and Treatment
This retrospective study included patients with primary IDH-wildtype GBM who underwent surgery and preoperative neoadjuvant and postoperative adjuvant therapies along with FMISO-PET at Kagawa University.Of these patients, those who underwent craniotomy with neo-Bev followed by RT and temozolomide (TMZ) as preoperative chemoradiotherapy were de ned as the preoperative neo-Bev group, and those who underwent craniotomy without preoperative treatment were de ned as the Control group.
In the preoperative neo-Bev group, all patients received two courses of Bev at 10 mg/kg, followed by RT (40 Gy in 16 fractions or 60 Gy in 30 fractions) and TMZ at 75 mg/m 2 for 42 days (Fig. 1).Craniotomy was performed at least 5 weeks after neo-Bev.After craniotomy, TMZ at 200 mg/m 2 and postoperative Bev at 10 mg/kg were continued as maintenance therapy (Table 1).

FMISO-PET
FMISO-PET was performed: 1) before preoperative neo-Bev; 2) after preoperative neo-Bev (5 weeks after end of Bev, 1 week before surgery); 3) after surgery (only in some cases); and 4) at recurrence (only in cases that recurred) (Fig. 1).The examination protocols were performed using a previously described method [14,15] PET studies were performed using a Biograph mCT PET/CT scanner (Siemens Medical Solutions, Knoxville, TN, USA).PET scans were acquired in the 3-dimensional model, and PET images were reconstructed as described in our previous study (simultaneous acquisition of 51 transverse images per eld of view [FOV], with an intersection spacing of 3 mm, for a total axial FOV of 15 cm) [16].PET radiotracers were produced using an HM-18 cyclotron (Sumitomo Heavy Industries, Tokyo, Japan).The radiochemical purity of FMISO was > 95% [17].Transmission and regional emission images of the brain were obtained as described in our previous study [10].Fasting was initiated 6 h before all PET studies, and the examination schedule was as follows: MRI, including contrast examination, on day 1; and FMISO on day 2.

PET data analysis
Uptake of FMISO in the brain tumor was semiquantitatively assessed by evaluating the maximum standardized uptake value (SUVmax).FMISO-PET images were converted into average venous blood concentrations of FMISO to obtain tumor-to-blood ratios (TBRs), allowing 3-dimensional pixel-by-pixel calculation of maximum TBR (TBRmax) for SUVmax.A cutoff value of 1.2 for FMISO TBR was used to determine the hypoxic volume (HV) [14,18].PET and MRI datasets were transferred to a Linux workstation and co-registration of FMISO/MRI was performed using Dr. View/Linux version R2.5 (AJS, Tokyo, Japan).Before reaching histopathological and molecular diagnoses, two radiologists analyzed the data to lower the risk of observer bias as much as possible.

Statistical analyses
Paired-samples t-testing was used to compare FMISO TBRmax and HV before and after Bev treatment and before and after recurrence in recurrent cases.To compare FMISO TBRmax and HV between two groups, t-tests were used.For pathological markers, the Mann-Whitney U test was used for comparisons between two groups.Correlations between FMISO TBRmax and various markers were also analyzed using Spearman's rank correlation for the sample at the time of rst surgery.All p-values were two-sided with the signi cance level set to p < 0.05.

Patients
FMISO-PET was performed on 7 patients with newly diagnosed GBM during the observation period.Of these, 3 were included in the preoperative neo-Bev group and 4 in the Control group (Table 1).
The preoperative neo-Bev group included 2 men and 1 woman with a mean age of 69.7 years (range, 65-78 years).The Control group included 3 men and 1 woman with a mean age of 59.8 years (range, 32-76 years).In the preoperative neo-Bev group, two cases (Cases 1 and 2) showed recurrence during the observation period.

FMISO-PET ndings
Comparing FMISO TBRmax and HV in PET scans taken before (pre-Bev) and after Bev (post-Bev) in the preoperative neo-Bev group, no signi cant difference was seen in TBRmax (p = 0.13), but HV was signi cantly lower after Bev (p = 0.03) (Table 1; Fig. 2a, b).No differences in FMISO TBRmax (p = 0.56) or HV (p = 0.09) were evident between the preoperative neo-Bev and Control groups on PET scans taken before craniotomy (Fig. 2c, d).The two patients in the preoperative neo-Bev group who developed recurrence showed higher values of FMISO TBRmax at the time of recurrence compared to before tumor removal (Table 2).

Immunohistochemistry analyses
To on FMISO-PET and in the TME of GBM during Bev therapy, expressions of a marker of hypoxia (CA9), markers of stemness (nestin, FOXM1), and immunoregulatory molecules (CD163, PD-1, and PD-L1) were evaluated (Table 2).Interestingly, CA9 and PD-L1 scores correlated with FMISO TBRmax (CA9: r = 0.90, p = 0.006; PD-L1: r = 0.88, p = 0.009) (Fig. 3a, b).Next, we analyzed differences between specimens from initial surgery in the preoperative neo-Bev and Control groups.All but one case in the Control group showed strong expression of CA9, corresponding with high FMISO TBRmax (Fig. 4a, b).A case showing negative CA9 expression in the Control group revealed lower FMISO TBRmax than cases appearing strongly positive for CA9.On the other hand, the neo-Bev group tended to show faint positive expression of CA9 with lower FMISO TBRmax compared with the Control group (Fig. 4d, e).Besides, after comparing other candidate markers, FOXM1 showed a trend toward higher expression in the Control group (p = 0.06; Fig. 4c, f).Next, we analyzed the difference between initial and reoperation specimens in the preoperative neo-Bev group.We found that CA9 was only faintly positive at initial surgery, although to the extent that expression of CA9 was not re ected in the score in Case 1, staining was more prominent at reoperation with increasing FMISO TBRmax (Fig. 4e, h).Moreover, FOXM1 in both cases increased at reoperation (Fig. 4f, i).However, alterations of other markers varied (Table 2).
Regardless of statistical signi cance, expressions of the endothelial marker (CD34) and the immunoregulatory molecules (CD163, FOXP3, PD-1, and PD-L1) tended to be lower in the preoperative neo-Bev group than in the Control group (Fig. 5a-f).

Discussion
We report here the rst study using FMISO-PET in the follow-up of GBM patients treated with preoperative chemoradiotherapy including Bev before and after preoperative neo-Bev and for a long period of time up to the time of recurrence.In this study, FMISO-PET was used to con rm that even though initial treatment with Bev improved the oxygenation of hypoxic tumor tissues, hypoxic changes in TME occur at recurrence despite the administration of Bev.Similar changes were also con rmed pathologically.
As hypoxia in tumor tissues decreases radiosensitivity, FMISO-PET was initially used to identify hypoxic areas for systemic cancers before and after radiotherapy [1].After Valk et al. rst applied FMISO-PET to scanning brain tumors, this modality has often been used to delineate hypoxic regions within brain tumors, particularly gliomas [22].FMISO-PET has been compared to and used in conjunction with other modalities such as other PET scans, dynamic susceptibility contrast MRI and contrastenhanced MRI in relation to blood ow and metabolism.Previous studies have reported that the degree of FMISO accumulation is low in low-grade gliomas and high in high-grade gliomas, indicating consistency between grade and FMISO-PET ndings [20,23,24].Cher et al. suggested that FMISO-PET may be useful for detecting recurrent disease in clinical practice for patients with glioma [24].Kawai et al. found that HV for FMISO correlated strongly with volume as assessed by contrast-enhanced MRI, and PET ndings correlated with VEGF, re ecting a hypoxic environment [14,15].
After Bev became available as an inhibitor of angiogenesis, vascular normalization by Bev was reported, then FMISO-PET was applied as a method to evaluate therapeutic e cacy [8].Together with ndings from contrast-enhanced MRI, Yamaguchi et al. reported that patients with improved oxygenation and loss of contrast enhancement effect after Bev administration show good prognosis [9].According to Gerstner et al., the effects of Bev treatment varied among individuals, with no signi cant difference in FMISO accumulation before and after Bev treatment for recurrent GBM [10].Similarly, in another study, we evaluated the relationship between prognosis of patients with recurrent GBM under Bev treatment and FMISO-PET ndings.
The results showed that median progression-free survival (PFS) and overall survival (OS) were longer in GBM patients with decreased TBR for FMISO after Bev therapy than in other GBM patients with increased TBR for FMISO.The patient group with reduced HV for FMISO after Bev therapy showed better results than the other group with increased HV in terms of both PFS and OS [in press].In other words, decreased TBR and HV for FMISO may offer useful biomarkers and could better predict outcomes in GBM patients receiving Bev therapy.A recent study used FMISO-PET to evaluate the therapeutic effects of evofosfamide, a prodrug of nitromisonidazole that is used to improve hypoxic areas of recurrent GBM refractory to Bev treatment [11].Thus, all reports related to the use of Bev for GBM have been for recurrent cases and have been limited to short-term observations of the course of treatment.
In our previous reports, preoperative neo-Bev was useful to control intraoperative hemorrhage due to reductions in tumor vascularity [12].We have veri ed the safety and clinical bene ts of preoperative neo-Bev in a multicenter exploratory prospective clinical phase I/II study [manuscript under preparation].We have also compared the status of TME, including tumor oxygenation and stemness, by immunohistochemical analyses of paired samples obtained from the same patients who underwent repeated surgeries following preoperative neo-Bev administration as well as at the time of recurrence [4,12,13,19].In this context, we have described the effects of Bev therapy in improving hypoxic and immunosuppressive TMEs.By adding FMISO-PET to preoperative neo-Bev-based GBM therapeutics, we were able to obtain PET ndings before and after Bev administration and clarify hypoxic changes during long-term follow-up, even in patients who initially responded well to treatment.In the current study, we identi ed a positive correlation between FMISO TBRmax and in situ CA9 expression.Furthermore, we were able to visualize alterations in tumor oxygenation throughout the clinical course.This nding may support therapeutic interventions.
Regardless of the statistical signi cance of immunohistochemical markers in the present study between the preoperative neo-Bev and Control groups, a hypoxic marker (CA9), stem cell markers (nestin and FOXM1), immunosuppressive cell markers (CD163 and Foxp3) and an immune checkpoint molecule (PD-L1) tended to decrease during effective Bev therapy.CA9 is known to correlate with FMISO accumulation, and the present results are also consistent with previous reports that PD-L1 correlates with hypoxia [25][26][27].This study con rmed that CA9 and PD-L1 expressions correlated signi cantly with FMISO TBRmax.Thus, FMISO-PET could allow the identi cation of the hypoxic niche preoperatively and may effectively detect recurrent lesions during post-neo-Bev follow-up, which would presumably be useful for comprehensive evaluation in combination with contrast-enhanced MRI.
Some limitations to the current study need to be kept in mind.Since this was a retrospective study with a small number of cases, veri cation in a larger number of cases is essential to draw conclusions.Furthermore, the addition of RT and TMZ as preoperative therapeutic interventions makes the simple interpretation of the effects of preoperative neo-Bev more di cult.
Tumor oxygenation and associated alterations to the TME (including stemness, recruitment of immunosuppressive cells and induction of immune checkpoint molecules) should play a pivotal role in assessing the clinical bene ts of Bev therapy.Given that RT and TMZ would have little effect on hypoxia and immunoregulatory molecules as determined by previously reported immunohistochemical ndings [28], the current immunohistochemical and PET ndings during Bev therapy might facilitate elucidation of the mechanisms underlying the effectiveness of and resistance to Bev therapy, and identi cation of suitable biomarkers for therapeutic response and clinical outcomes for the combination of immunotherapy with Bev.

Conclusion
This study revealed that preoperative neo-Bev induced tumor oxygenation that could be visualized on FMISO-PET.In addition, FMISO-PET was useful in monitoring the deterioration of oxygenation within the TME during the course of GBM treatment.This study strengthens the concept that preoperative neo-Bev may promote an immunosupportive TME with reductions in stemness markers and immunoregulatory molecules.Further research is required to con rm whether FMISO-PET is valid for assessing the duration and intensity of effects due to Bev treatment for GBM.Schema representing the schedule of treatment and FMISO-PET in the current study.In the preoperative neo-Bev group (neo-Bev) group, all patients received two courses of Bev at 10 mg/kg, followed by radiation therapy (40 Gy in 16 fractions or 60 Gy in 30 fractions) and temozolomide (TMZ) at 75 mg/m 2 for 42 days.FMISO-PET was performed: 1) before preoperative neo-Bev; 2) after neo-Bev (5 weeks after end of Bev); 3) after surgery (only in some cases); and 4) at recurrence (only in cases that recurred).In the Control group, FMISO-PET was performed before surgery alone.Figure 4 FMISO-PET (a, d, g) and ndings from immunohistochemical analysis (b, e, h: CA9; c, f, i: FOXM1) for representative cases from the Control group (a-c: Control 1) and preoperative neo-Bev group (d-f: Case 2 at initial surgery; g-i: Case 2 at second surgery).Comparing specimens from the preoperative neo-Bev group at initial surgery and reoperation, CA9, as a marker re ecting a hypoxic environment, was only faintly positive at initial surgery and was more prominent at reoperation.e, h) For FOXM1, a trend toward higher FOXM1 was seen in the control group, but the difference was not signi cant (c, f, i: p=0.06).

Table 1
Patient pro les of both groups and details of preoperative treatment