“Paradoxical” findings of tumor vascularity and oxygenation in recurrent glioblastomas refractory to bevacizumab

Anti-angiogenic therapy induces the apparent normalization of vascular structure, decreases microvessel density (MVD), and improves tumor oxygenation in glioblastomas (GBMs). Six initial and recurrent tumor pairs after bevacizumab (Bev) treatment were compared with GBMs from nine patients resected under neoadjuvant Bev treatment with regard to histological characteristics; MVD; MIB-1 index; and expression of vascular endothelial growth factor (VEGF) and its receptors, hypoxia markers (hypoxia-inducible factor 1 alpha, carbonic anhydrase 9), and nestin as a marker of glioma stem-like cells. In recurrent tumors post-Bev treatment, while the MVD remained low compared with the paired initial tumors (pre-Bev tumors), the expression of hypoxic markers were increased and were even higher in expression compared with the paired pre-Bev tumors in three of the six cases. MIB-1 indices were similar among the initial GBMs, neoadjuvant group, and recurrent tumors post-Bev treatment. The nestin-positive cell ratio of the post-Bev recurrent tumors was as high as that of the pre-Bev tumors. The expression of VEGF and VEGFR1 was increased in the post-Bev recurrent tumors in three and four cases, respectively, compared with the paired pre-Bev tumors. In the majority of Bev-refractory GBMs, tumor hypoxia was present with a paradoxical decrease in MVD. These findings suggest that re-activation of tumor angiogenesis is not initially involved in the acquisition of resistance to Bev.


INTRODUCTION
Glioblastoma (GBM) is a highly vascularized malignant tumor, and thus, inhibiting angiogenesis is an attractive treatment strategy. Vascular endothelial growth factor (VEGF) promotes endothelial cell proliferation and migration, and consequently, tumor angiogenesis and tumor growth. VEGF expression correlates with prognosis in glioma patients and histological grade of malignancy [1].
Bevacizumab (Bev) is a recombinant, humanized monoclonal antibody that inhibits VEGF-A. Data from randomized clinical trials have suggested that Bev administration is associated with favorable eventfree survival and improvement in patients with a poor performance status [2][3][4], and it is increasingly being used

Research Paper
for the treatment of newly diagnosed and recurrent highgrade gliomas in Japan. However, the response of GBM to Bev is invariably transient, and recurrent tumors after Bev treatment are associated with a more aggressive and invasive phenotype [5][6][7]. Because the targets of Bev are not tumor cells but are vascular endothelial cells, the mechanism of resistance to Bev is likely different from those of other chemotherapeutic agents that target tumor cells. However, the mechanisms of resistance/ refractoriness as well as the response to Bev treatment have not been fully elucidated [8]. In situ observations of surgical specimens are likely to be crucial for clarifying these issues. We previously demonstrated that Bev induces the apparent normalization of vascular structure, decreases microvessel density (MVD), and improves tumor oxygenation in human GBMs resected under Bev treatment [9].
In this study, we used histopathological specimens of initial and recurrent tumor pairs after Bev treatment and GBMs resected following neoadjuvant Bev treatment to investigate the molecular/histopathological differences in tumors under Bev treatment compared with refractory tumors, as well as those of initial tumors versus refractory tumors. To the best of our knowledge, this is the first report of in situ observations showing the difference between tumors that responded and those that acquired resistance to anti-angiogenic therapy.

Clinical characteristics of the patients
The clinical course of the nine patients in the neoadjuvant group is summarized in Table 1. Specifically, the Karnofsky performance status of these patients dramatically improved from 50-80 to 80-100 with one to three doses of pre-operative Bev. A median sum of perpendicular diameter (SPD) change in tumor volume induced by Bev treatment was -43% (-8% to -61%). Seven newly diagnosed patients were all alive as of April 2017, and overall survival ranged from 10 to 26 months following the first administration of Bev. The clinical characteristics of the six patients from whom paired pre-Bev and post-Bev tumors were obtained are summarized in Table 2. Three of the six patients underwent a second surgery at the time of recurrence following Bev treatment, and the remaining three patients were deceased at the time of recurrence. The tumors from the latter three patients were post-mortem. Regarding the radiological progression patterns of the post-Bev recurrent tumors, cases 10 and 11 had contrast-enhancement at progression (cT1 flare-up), whereas cases 12-15 had non-enhanced tumors (case 12: T2 circumscribed; cases 13-15: T2 diffuse). The interval between final Bev treatment and post-Bev resection or death ranged from 31 to 208 days.

Histological findings
Comparison of MVD and proliferative activity among neoadjuvant, pre-Bev, and post-Bev recurrent groups

Tumors resected under Bev treatment (neoadjuvant group)
Tumor cells showed pleomorphic cells with nuclear atypia and mitotic activity, but typical microvascular proliferation and palisading necrosis were inconspicuous, and "glomeruloid" microvasculature was occasionally collapsed in cases 1-9. Immunohistochemistry for R132H-mutant isocitrate dehydrogenase (IDH)-1 were negative in all nine tumors. Histopathological diagnosis was high-grade glioma, equivalent to grade III or grade IV. The tumor in case 9, the minimal responder to Bev (-8%; Table 1), was mainly composed of medium-sized astrocytic cells with increased cellularity and abundant mitotic figures. Palisading necrosis was occasionally observed and microvascular proliferation was absent. The histopathological diagnosis was GBM. The mean MVD was 24.0/5 high-power field (HPF) (12.8 to 62.6), and the mean MIB-1 index was 21.3%. These data are shown in Figure 1A and 1D.

Pre-Bev initial tumors
The pre-Bev initial tumors had the typical appearance of GBMs with increased cellularity and abundant mitotic figures. Palisading necrosis and microvascular proliferation were present, and immunohistochemistry for R132H-mutant IDH-1 were negative in all six tumors. The histopathological diagnosis was GBM. The mean MVD was 69.9/5 HPF (13.2 to 113), and the mean MIB-1 index was 23.3%. These data are shown in Figure 1B and 1E.

Post-Bev recurrent tumors
The recurrent tumors after Bev treatment were characterized by microbleeding, which was significant in two of the six tumors, and there was less microvasculature compared with the paired initial tumors. Palisading necrosis was observed in four tumors. The mean MVD was 26.8/5 HPF (17.6 to 47.2), and the mean MIB-1 index was 18%. The MVD of the post-Bev recurrent tumors was higher than that in the neoadjuvant group; however, it was still significantly lower than the paired initial tumors (mean MVD: neoadjuvant, 24.0; initial, 69.9; post-Bev recurrent, 26.8; p = 0.00871 by one-way ANOVA followed by post-hoc test; p < 0.05 for neoadjuvant vs. initial, not significant (NS) for neoadjuvant vs. post-Bev recurrent, p < 0.05 for initial vs. post-Bev recurrent). In contrast, there was no significant difference in the MIB-1 proliferative indices among the initial GBM group, neoadjuvant group, and post-Bev recurrent group (p = 0.894 by one-way ANOVA followed by post-hoc test; NS for neoadjuvant vs. initial, NS for neoadjuvant vs. post-Bev recurrent, NS for initial vs. post-Bev recurrent). These data are shown in Figure 1C, 1F and Table 4.
Comparative analyses of VEGF/VEGFR expression among neoadjuvant Bev, pre-Bev, and post-Bev groups VEGF expression was negative in the three best responders (-61% for case 1, -57% for case 5, -52% for case 8) of the nine patients in the neoadjuvant group ( Figure 2A, Tables 1, 3), and was negative or very faint in four of the six patients with pre-Bev tumors ( Figure  2B, Table 3). VEGF expression was clearly observed in the other 14 tumors including 6 post-Bev recurrent tumors. VEGFR1 expression was negative or faint in 5 of the 9 patients in the neoadjuvant group, but was clearly observed in the other 16 tumors (Table 3). Similarly, VEGFR2 expression was faint in 3 of the 9 cases in the neoadjuvant group, but was clearly observed in the other 18 tumors. Interestingly, in comparing the initial and post-Bev paired tumors, VEGF and VEGFR1 expression was increased in the post-Bev tumors in four and five of the six cases, respectively, including two cases that had a cT1 flare-up pattern ( Figure 2C-2F, Table 3).

Comparative analyses of hypoxia marker (HIF-1α and CA9) expression among neoadjuvant Bev, pre-Bev, and post-Bev groups
CA9 is a hypoxia-inducible enzyme as well as a useful biomarker for predicting the poor prognosis of GBM [10]. Transcriptionally active HIF-1α is bound to the hypoxia response element in the promoter area of the CA9 gene, thus inducing CA9 expression [11,12]. Expression of HIF-1α was not observed in four of the nine cases in the neoadjuvant group, and CA9 expression was faint or negative in three of the nine tumors ( Figure 2G). Thus, improved oxygenation was suggested in all but one minimal responder (-8% for case 9; Table 1).
In contrast, both HIF-1α and CA9 were clearly expressed in all of the pre-Bev initial and post-Bev recurrent tumors ( Figure 2H, 2I), suggesting that tumor oxygenation was improved with Bev treatment but returned to the original hypoxic condition in tumors refractory to Bev. Interestingly, in comparing pre-Bev initial and post-Bev paired tumors, CA9 expression was increased in post-Bev recurrent tumors in three (cases       10,11,13) of the six cases including two of the three enhancing patterns of recurrence ( Figure 2G-2I, Table 3).

Comparative analyses of stem cell marker (nestin) expression among neoadjuvant Bev, pre-Bev, and post-Bev groups
Nestin is a stem cell marker, and is also an angiogenesis marker of proliferating vascular endothelial cells [13]. Nestin-positive cells were significantly less frequent in tumors resected under neoadjuvant Bev than in pre-Bev initial or post-Bev recurrent tumors, and there was a significant difference in the nestin-positive cell ratio between pre-Bev initial and post-Bev recurrent tumors ( Figure 2J-2L, Table 3

DISCUSSION
In this study, we compared paired initial and recurrent GBMs after Bev treatment, in addition to GBMs resected following neoadjuvant Bev treatment.
In the neoadjuvant group, as previously reported [9], improvement of tumor oxygenation was noted in eight of the nine tumors, and the MVD and nestin-positive cell ratio were significantly decreased compared with the pre-Bev initial tumors. On the other hand, in the post-Bev recurrent tumors, while the MVD remained decreased compared with the paired initial tumors, the hypoxic markers were re-expressed and were even more prominent in expression compared with the paired initial tumors in three of the six cases. There was no significant difference in MIB-1 indices between the pre-Bev initial, neoadjuvant, and post-Bev recurrent groups. The nestin-positive cell ratio of the post-Bev recurrent tumors was as high as that of the initial tumors. The expression of VEGF and VEGFR1 was increased in the post-Bev recurrent tumors compared with their paired initial tumors in three and five of the six cases, respectively.
The observed decreased HIF-1α expression in the neoadjuvant group supports previous data that vessel regression induced by anti-VEGF therapy is associated with an increase in median pO 2 and a decrease in hypoxia below 5 mmHg [14]. The improvement of tumor oxygenation could be explained by a decrease in the number of tumor and endothelial cells consuming oxygen. The in situ observation in pre-and post-Bev GBMs is crucial to elucidate the mechanism of action as well as resistance to Bev; however, only a few such studies have been reported to date [9,[15][16][17].
DeLay et al. [17] performed comparative analyses of the paired pre-treatment and post-Bev recurrent tumors from 21 GBM patients. The authors classified the cases into two subtypes based on radiographic characteristics of the post-Bev recurrent tumors (i.e., enhanced and non-enhanced Bev-resistant GBMs). According to their findings, the non-enhancing pattern of recurrence revealed reduced MVD, the increased expression of hypoxic markers, and unchanged proliferative activity compared with the pre-Bev paired tumors. In contrast, the enhancing pattern of recurrence was characterized by an unchanged MVD, unchanged expression of hypoxic markers, and increased proliferation compared with pre-Bev paired tumors. The expression of VEGF and VEGFR2 was similar between the paired pre-and post-Bev tumors with a non-enhancing or enhancing recurrence pattern.
However, we did not observe clear differences in the immunohistochemical findings in recurrent GBMs between enhanced and non-enhanced as described above, although the reason and mechanism remain unclear. It is possible that the heterogeneity among patients in the different treatment groups caused the differences in MVD and hypoxic marker expression in the Bev-resistant GBMs between our study and the previous one, since more than half of the cases in the previous study had various chemotherapeutic agents combined with Bev and all six patients in this study had radiotherapy and temozolomide in addition to Bev.
Nonetheless, this study, together with previous studies [9,16,17], confirmed that MVD is significantly decreased with improvement of tumor oxygenation in GBMs during effectiveness of Bev, and that in the majority of Bevrefractory GBMs, tumor hypoxia is regained to a similar or even higher level compared with the paired pre-Bev GBMs while the MVD is still low. HIF-1α directly regulates the expression of VEGF; therefore, the degree of tumor oxygenation and tumor vascularity are generally parallel [18,19]. However, in post-Bev recurrent tumors, tissue oxygenation and MVD were paradoxically opposite. These findings indicate that the re-activation of tumor angiogenesis is not involved in the first stage of refractoriness to Bev.
In this study, there was a relative trend towards upregulation of VEGF and VEGFR1 in the post-Bev recurrent tumors compared with their paired initial tumors. Hypoxia in tumor tissue is caused by oxygen consumption without balance with its supply, which is the most dynamic effect on tissue pO 2 among several factors such as irregular vascular geometry, vascular density, and altered blood viscosity [20]. Vascular density in tumor tissue usually decreases after antiangiogenic therapy, which should cause oxygen concentrations to decrease; however instead, paradoxical oxygenation is induced [9,14,17,21,22]. During responsiveness to Bev, probably because oxygen demand by the tumor decreases, hypoxic markers (HIF-1α and CA9) and the marker for stem cell and vascular proliferation (nestin) are decreased [9], suggesting that tumor oxygenation is preserved; thus, VEGF and its receptors are relatively reduced. Whereas, when the tumor represents refractoriness to Bev, oxygen demand by the tumor increases and the microenvironment of the tumor tissue becomes hypoxic, causing stemness with chemoresistance and radioresistance. In addition, VEGF and its receptors are relatively increaesd under hypoxia, leading to inhibitory effects on tumor immunity.
It is also possible that activation of alternative angiogenic factors occurs at the time of recurrence during Bev therapy. Platelet-derived growth factor, fibroblast growth factor, ephrin, angiopoietin, and invasion-mediating genes such as integrin and matrix metalloproteinase (MMP) 2 and 9 are induced during resistance of VEGF blockade [8,[23][24][25]. Stromal cell-derived factor is also involved in rebuilding of vasculature at the time of tumor recurrence post-Bev treatment [8].
The salvage angiogenic signaling pathway involving KRAS and NF-κB was activated via upregulation of interleukin-8 [26]. VEGF/HIF-1α/ mammalian target of rapamycin (mTOR) signaling pathway could be related to the change of tumor metabolism; thus expression of mTOR might be altered by Bev therapy [27]. In addition, the bone marrowderived mononuclear cells, CD11b-positive myeloid cells [28,29], and Tie2-expressing monocytes [30] were induced at the time of refractoriness to anti-VEGF therapy, accompanied with production of MMP 2 and 9 under hypoxic condition resulting in invasive change. These molecules/pathways might indeed be involved in the initial mechanism of resistance, and among the therapeutic targets to prevent the acquisition of resistance to Bev. With respect to involvement of an alternative angiogenic pathway following Bev failure, various molecules and signal pathways as described above should be investigated in the future.
The major limitation of this study was the small sample size due to the rarity of cases that underwent surgery following tumor progression after Bev therapy. Moreover, the tumors resected under neoadjuvant Bev and the Bev-refractory tumors were not from the same patients. Future studies will ideally use pairs of tumors resected under neoadjuvant Bev and post-Bev refractory tumors to confirm our results.
In conclusion, the results of this study together with previous studies demonstrated that in GBMs under Bev treatment, MVD is significantly decreased with improvement of tumor oxygenation, and that in the majority of Bev-refractory GBMs, tumor hypoxia is recovered with a paradoxical decrease in MVD. These in situ comparative analyses of samples from the same patients pre-and post-Bev treatment will help to understand the mechanism of resistance to anti-angiogenic therapy and to develop optimal therapy that can be used in the clinic.

Patients and tissues
This study used 21 glioblastoma tissues from 15 patients obtained at three different settings: nine tumors resected under neoadjuvant Bev (defined as the neoadjuvant group, n = 9) and paired tumors of pre-and post-Bev treatment from another six patients (defined as pre-Bev initial group and post-Bev recurrent group, respectively). The nine patients in the neoadjuvant group (cases 1-9) had been treated with one to three courses of pre-operative Bev at a dose of 10 mg/kg, and resection was performed 21-36 days after the last administration of Bev ( Table 1). Eight of the nine patients (cases 1-8) were newly diagnosed cases. One patient (case 9) was treated with Bev at tumor recurrence, and resection was performed under Bev treatment.
From another six patients, the paired pre-Bev initial and post-Bev recurrent tumors were available. Four of the six patients (cases 10, 11, 12, and 14) were treated with the Stupp regimen (radiotherapy (60 Gy, 30 fractions) concomitant with temozolomide at a dose of 75 mg/m 2 ), temozolomide (150 mg/m 2 ), and Bev following initial resection, and two patients (cases 13 and 15) were treated with Bev at the tumor recurrence following Stupp regimen and temozolomide (150 mg/m 2 ). All six recurrent tumors were refractory to Bev despite initial response to Bev; three were resected tumors following progression after Bev and three were postmortem (Table 2). Bev was administered at a dose of 10 mg/kg every two weeks.
All 21 tumors were examined for histological features, MVD, VEGF/VEGFR, HIF-1α/CA9 as markers of hypoxia in the tumor tissue, and nestin as a marker for glioma stem-like cell population. The

Radiological assessment
Tumor response to Bev was assessed by the SPD [31]. Patterns of recurrence were classified as cT1 flare-up (two patients), T2 diffuse (one patient), or T2-circumscribed (two patients) as follows [32]. (1) cT1 flare-up: characterized by an initial decrease in contrast enhancement (CE) on T1-weighted images after treatment initiation, and an increase (flare-up) of CE again at tumor progression. T2 signal stays stable or increased. (2) T2 diffuse: characterized by a signal increase on T2-weighted images with a poorly defined border despite the fact that CE on T1-weighted images remains decreased. Hypointensity on T1-weighted images is faint and disproportionally smaller than T2 hyperintensity. (3) T2 circumscribed: characterized by a signal increase on T2-weighted images with a bulky structure and sharp borders that correspond to a T1 hypointense signal. CE on T1-weighted images remains decreased, or only a few faintly speckled CE lesions are visible.
VEGF-A expression in the tumor cytoplasm or stroma was assessed as follows: ++, diffuse intense staining; +, diffuse faint staining; −, negative staining [9]. The expression of VEGFR1 and VEGFR2 on endothelial or tumor cell membrane/cytoplasm was assessed as the following: ++, staining in both vascular endothelial cells and tumor cells; +, staining only in vascular endothelial cells; −, negative staining [9].
The expression of nestin was examined using immunohistochemistry with an anti-nestin antibody (1:100, 10C2, Chemicon, Tokyo, Japan), and was assessed as a positive cell ratio analyzed in more than 1000 tumor cells from more than three areas, showing the representative appearance of each tumor [9].

Statistical methods
One-way ANOVA with post-hoc analysis was used to compare MVD, and indices of MIB-1 and nestinpositive cell ratio. Analyses were performed with IBM SPSS statistics.