Glioblastomas: Epidemiology, Pathology, and Treatment Outcomes

C

1 I

1.1 Glioblastomas: Epidemiology, Pathology, and Treatment Outcomes

Gliomas are the most prevalent primary intracranial tumor of adults, representing 81%

of brain tumors and the most common malignant glioma histology is glioblastoma, which accounts 45% of all gliomas, follow by anaplastic astrocytoma [1]. There are annually 150 to 200 glioblastoma cases which accounts for about 37% to 47% of adult primary brain tumor in Taiwan, and the male-to-female ratio of incidence is around 1.3 to 1.5 [2-4]. The histological features of glioblastoma include increased cellularity, mitotic activity, microvascular proliferation, nuclear atypia, and necrosis [5], which are correlated to the typical image findings of central areas necrosis with extensive peritumoral edema [6].

Clinically, common prognostic factors for glioblastoma patients include age, Karnofsky performance status (KPS), neurologic status, and tumor resection extent [7-9]. The molecular biomarker of isocitrate dehydrogenase 1 (IDH1) mutation significantly predict patient overall survival (OS) [10, 11]. Before the invention of temozolomide, an alkylating agent for treating malignant glioma, surgical resection followed by radiotherapy (RT) resulted in a better OS than resection without RT [12]. Nowadays, surgical resection followed by concurrent chemoradiotherapy (CCRT) with oral temozolomide became the standard treatment for glioblastoma. The methylation status of O⁶-methylguanin-deoxyribonucleic acid (DNA) methyltransferase (MGMT) promotor substantially identifies patients most likely to benefit from CCRT with temozolomide [10]. The median OS and 2-year survival rates for patients receiving definitive or adjuvant CCRT with temozolomide are 13.4–16.0 months and 26.5%–

31%, respectively [13-15].

2

1.2 The Clinical Unmet Needs for Glioblastomas: Individualized Treatment Strategies According to the Progression Patterns

The most common progression patterns after CCRT with temozolomide are local and in-field (72%–96.8%), and the rates of distant and out-field recurrence range from 2% to 28%

[16-18]. However, the clinical prognostic factors and available molecular biomarkers do not correlate with the glioblastoma progression patterns, in-field or out-field, after CCRT.

Clinically, there is no effective factor available to predict whether the tumor progresses confined to the tumor bed or spreads out of tumor bed. To develop individualized treatment strategy for glioblastomas of different progression patterns, effective factors are required for disease classification [19].

Currently, RT guidelines adopted by the European Organisation for Research and Treatment of Cancer (EORTC) and the Radiation Therapy Oncology Group (RTOG) for target delineation and dose prescription of the peritumoral edematous areas of glioblastoma are diverse [20]. The necessity of irradiating peritumoral edema of glioblastomas remains controversial. The EORTC uses a 2-cm volumetric expansion of the gross tumor volume (GTV) to generate the clinical target volume (CTV) in a single phase of 60 Gy in 30 fractions, which is based on published data stating that > 80% of recurrences occur within a 2-cm margin of contrast-enhanced lesions by computed tomography (CT) or magnetic resonance imaging (MRI) scans [20]. The RTOG defines CTV1 as the surgical resection cavity plus any residual enhancing tumor plus surrounding edema within a 2–2.5-cm margin, which should receive 46 Gy in 23 fractions, followed by a cone-down boost to the tumor bed with a 2-cm margin, while CTV2 should receive an additional 14 Gy in seven fractions [20], considering that pre- or postmortem findings have demonstrated high rates of glioblastoma cells at

3

peritumoral edema areas, as observed in CT and MRI scans [21].

Some studies reported that glioblastomas with subventricular zone (SVZ) or corpus callosum (CC) invasion are associated with adverse prognoses and diverse progression patterns [22-24]. Patients with glioblastoma involving the SVZ are highly associated with a significant decline in progression-free survival (PFS) and OS [24-26]. Some published data demonstrated that glioblastoma with SVZ invasion is associated with local recurrence as well as spreading to the ventricles, distant areas, and multifocal progression [22, 24, 27, 28].

Similarly, glioblastoma with preoperative contrast-enhanced tumors or edema involving the CC [9, 29] and butterfly glioblastoma, a tumor involving the bilateral hemispheres through the CC, [30, 31] are poor prognostic factors for OS. In addition to SVZ and CC invasion, extensive peritumoral edema observed on T2-weighted or fluid-attenuated inversion recovery (FLAIR) MRI images is associated with a poor prognosis for glioblastoma patients [32].

The tumor bed is the most common recurrence area of glioblastomas and one of the treatment strategies to increase the local tumor control is intratumoral drug delivery which has the advantage of reducing the systemic toxicity and circumventing the blood-brain barrier [33]. Three modalities, including wafer, convection-enhanced delivery (CED), and hydrogel, were proposed for intratumoral anti-cancer drug delivery [33-35]. Nowadays, only carmustine wafer was utilized clinically for glioblastoma as intratumoral drug delivery [34, 35].

The individualized treatment strategies for glioblastoma patients according to their progression patterns remained undetermined. Figure 1 illustrates our study concept to satisfy the clinical unmet needs for glioblastomas, we proposed an extensive study, including

4

clinical and basic investigations, to explore imaging biomarkers for glioblastoma classification and intratumoral drug injection for glioblastoma treatment, respectively.

1.3 Clinical Investigation of Imaging Biomarkers: Hypothesis and Purpose

The neurogenesis of adult mammalian brain occurs in the SVZ on the walls of the lateral ventricles and the subgranular layer of the dentate gyrus in the hippocampus [36-38]. In humans, the anterior, occipital, and temporal horns of lateral ventricles comprise different astrocytes and ependymal, proliferating cells, and migratory patterns [36]. Cerebral commissures, including the CC and the anterior commissure, contribute interhemispheric connections. The anterior (the genu) and posterior (the splenium) sections of the CC connect the bilateral frontal and occipital lobes through the radiating fibers, respectively. The ventral surface of the CC forms the roof of the lateral ventricles close to the SVZ [39]. The anterior commissure comprises a bundle of axons, which crosses the midline in the lamina terminalis and traverses the corpora striata, and supplies communication between the temporal lobes [39]. The peritumoral edema of glioblastoma correlates with cancer cell infiltration [21] or effusion resulting from blood-brain barrier damage [40].

The SVZ hosts potential neural progenitor cells, the CC provides the interhemispheric connections, and the preoperative edema (PE) correlates with cancer cell infiltration. Using the imaging biomarkers including the anatomical factors (SVZ and CC) [23] and pathophysiological factors (PE), we investigated the outcome of glioblastoma after CCRT and classified patients according to the survival and progression patterns. The classification

5

categorized glioblastoma with in-field (confined near the tumor bed) or out-field (spreading beyond the tumor bed), which requires individualized treatment strategy.

From the anatomical perspective, we hypothesized that glioblastomas with synchronous SVZ and CC (sSVZCC) invasion have distinct progression patterns associated with the interhemispheric and lateral ventricular involvement. From the pathophysiological perspective, we hypothesized that glioblastomas with extensive PE (EPE) have high tumor migration ability. For glioblastoma patients, the interactions between the anatomical factor sSVZCC and the pathophysiological factor EPE and the associated clinical impacts, including tumor migration ability and directions, remain undetermined (Figure 2). We analyzed the associations between these imaging factors and survival and tumor progression patterns after CCRT to classify glioblastomas for further proposing individualized treatment strategies.

1.4 Basic Investigation of Intratumoral Drug Injection: Hypothesis and Purpose

Clinically, carmustine (an alkylating agent) is delivered by local wafer implantation during surgery after tumor resection to improve the survival [34, 35]. Carmustine is released from wafers over a period of approximately 5 days and wafers degrade completely over a period of 6 to 8 weeks when in continuous contact with interstitial fluid [41]. However, the adverse effects of carmustine wafer, such as healing abnormalities, cerebral edema, cerebrospinal fluid leaks, and intracranial infection were reported [34]. Big resection cavity size required for adequate drug dosage or implant dislodgement result in the technique

6

difficulty [42]. Besides, the synergistic effect of combining carmustine with RT for malignant gliomas was limited [49].

Carboplatin is a platinum-based antineoplastic agent, which is widely combined with radiotherapy for clinical cancer treatment [43-45]. Carboplatin binds to DNA to form intrastrand and interstrand cross links with the purine bases [46] and thus enhances the formation of cluster damage to DNA by ionizing radiation [47]. Several animal studies demonstrated the synergistic effect of combined carboplatin and radiation for glioma treatment [48-50]. However, carboplatin delivered by intratumoral infusion via the CED is limited in clinical practice due to the complications and delivery difficulty [49, 51]. The complications caused by CED include increased brain edema, infection, bleeding, and seizures [52]. The difficulty in catheter placement surgery depends on the brain lesion locations [52]. Moreover, neurotoxicity can be induced by the infusate backflow in the catheter, which can not be completely prevented by any insertion method [53].

The oxidized hyaluronic acid/adipic acid dihydrazide (oxi-HA/ADH) hydrogel is a biocompatible and thermogelling material [54], which can transform from liquid form into a gel-like matrix within 1–8 min, depending on the operational temperature [54]. Oxi-HA/ADH hydrogel is stable at body temperature. It maintains its gel-like state for up to 5 weeks after in vivo injection and is degraded gradually by hyaluronidase [54]. To meet requirements, including slow and steady release and radiosensitizing effect and to avoid the difficulties of wafer implantation and CED catheter placement [41, 48-50, 52], some studies investigated the combination of hydrogel-loaded anticancer agents for glioma treatments, including temozolomide, gemcitabine, cisplatin, and methotrexate etc. [42].

7

The ability of oxi-HA/ADH hydrogel to carry hydrophilic drug by crosslink achieves slow and steady drug release. Carboplatin has more intensity to produce DNA double-strand breakage than cisplatin as regard to the synergistic effect with ionized radiation, [47].

Combining these features, thermogelling oxi-HA/ADH hydrogel is potential to load carboplatin for single intratumoral injection to simplify the delivery method and frequency without compromising the synergetic effect of RT [55-58].

In Figure 3 [59], we compared the drug release and safety features of intratumoral delivery modalities (wafer, thermogelling hydrogel, and CED) [33-35] and the radiosensitizing effects among anti-cancer drugs (carmustine, carboplatin, and cisplatin) [47, 60]. In the current study, we proposed a novel combination of carboplatin loaded by thermogelling oxi-HA/ADH hydrogel and RT to satisfy the unmet clinical needs for treating malignant gliomas. The effectiveness, convenience, and safety of intratumoral delivery of oxi-HA/ADH hydrogel carboplatin (briefly hydrogel carboplatin) combined with RT for treating gliomas remains undetermined. From the biomedical engineering perspective, we hypothesized the sustained release of hydrogel carboplatin simplifies the drug delivery method and frequency and remains the synergistic effect with RT and without severe toxicity.

Through the comprehensive biomaterial, cell, and animal experiment design, we intended to demonstrate that the synergistic effect with RT by using single injection of hydrogel carboplatin is comparable to multiple injections of aqueous carboplatin. In addition, we used different regimens, including drug forms and dosages, to find the optimal CCRT parameters to improve the local tumor control and survival for mice glioma treatment, which makes intraoperative single drug injection with subsequent RT a feasible and potential clinical treatment for glioblastomas.

8

C

2 M

M

2.1 Clinical Study of Imaging Biomarkers

2.1.1 Patient eligibility

Patients from a single institute (National Taiwan University Hospital), who were pathologically confirmed as having glioblastoma between August 2004 and December 2015, were retrospectively evaluated after receiving approval from the institute’s institutional review board (201405076RINC). Patients who received complete CCRT with temozolomide were enrolled and were excluded if they were younger than 18 years, had active concomitant malignancies, received an RT dosage of < 54 Gy, or did not receive follow-up MRI or CT scans after CCRT.

2.1.2 Treatment modalities

The extent of tumor excision was classified as gross total and subtotal resection. Biopsy was employed for unresectable lesions.

RT techniques [23], including three-dimensional conformal RT, intensity-modulated radiation therapy, volumetric-modulated arc therapy, and tomotherapy, were employed using 6-MV linear accelerators. GTV1 was defined as gadolinium-enhanced lesions on T1-weighted images and hyperintense lesions on FLAIR or T2-T1-weighted images. GTV2 was defined as gadolinium-enhanced lesions on T1-weighted images. CTV1 and CTV2 were defined, respectively, as the GTV1 and GTV2 plus a 1.5–2-cm margin for potential microscopic disease with a margin reduced to 0.5 cm around natural boundaries or the optic nerve/chiasm. The planning target volume (PTV)1 and PTV2 were CTV1 and CTV2 plus a 0.3–0.5-cm margin, respectively. An RT dosage of 46 Gy—administered at a daily dose of 2

9

Gy once per day for 5 days per week—was prescribed for PTV1, and an additional 14 Gy was prescribed for PTV2. Uninvaded SVZ was not irradiated in this study.

Concurrent chemotherapy with temozolomide was administered at a daily dose of 75 mg/m² for 7 days per week, from the first to the last day of RT. After a four-week break, the patients received up to six cycles of adjuvant temozolomide for 5 days over a 28-day period.

The dose was 150 mg/m² for the first adjuvant cycle, and this was increased to 200 mg/m² at the beginning of the second cycle [15].

2.1.3 Anatomical features of preoperative imaging

Neuroimaging (MRI or CT) findings were interpreted by neuroradiologists.

Preoperative imaging findings, including tumor size, edema extent, and anatomical invasion into the SVZ and CC, were evaluated. PE was defined as a hyperintense area observed through T2 or FLAIR MRI or a hypointense area observed through CT. We used the MRI hyperintense signal on T2W or FLAIR sequence with different sections, including axial, coronal, and sagittal sections, to measure the edema distance along the normal line from tumor margin to the distal edema edge. The method of evaluating the preoperative edema extent is illustrated in Figure 4. First, we selected the image section that presented the maximum diameter of the preoperative tumor (Figure 4A). Then, we created a 2-cm expansion of the tumor edge along the SVZ (the lateral wall of lateral ventricles) or CC if these structures are invaded (Figure 4B); otherwise, we drew tangential lines to the tumor edge and then used their normal lines to expand a 2-cm margin of the tumor. The 2-cm expansion margin of the tumor was used as the reference boundary to evaluate the PE extent.

The 2-cm expansion margin of the preoperative tumor was applied on the similar image

10

sections after concurrent CRT to evaluate the PD extent.

EPE was defined as PE extending ≥ 2 cm from the tumor margin (Figure 5, A1).

sSVZCC invasion was defined by the presence of contrast-enhanced lesions synchronously involving the lateral walls of the lateral ventricles and cerebral commissures (the CC or anterior commissure) [23] as illustrated in Figure 5, A2. The neural fiber tracts (NFTs) included the CC, forceps frontalis, and forceps occipitalis, defined by anatomical structures [39] (Figure 5, A2). For tumors with sSVZCC invasion, the EPE along the CC (CCEPE) and along the SVZ (SVZEPE) were defined as edema extending ≥ 2 cm from the tumor margin along the CC and the SVZ, respectively (Figure 5, A2). Tumor size was defined by the maximum diameter determined through preoperative T1-weighted gadolinium-enhanced MRI or contrast-enhanced CT scans [61].

2.1.4 Tumor progression patterns after CCRT

Neuroimaging after CCRT was performed primarily through MRI at intervals of 3–6 months. Tumor progression was assessed according to the imaging definition of Response Assessment in Neuro-Oncology criteria and involved either of the following: (1) ≥ 25%

enlargement of contrast-enhanced lesions compared with the smallest tumor measured either at the baseline (if no decrease) or at best response, or (2) appearance of any new contrast-enhanced lesions [62, 63].When early progressive disease is suspected on the images within 12 weeks after completion of concurrent CRT, we used the following criteria to differentiate it from the pseudoprogression: the presentation of new enhancement beyond the radiation field or the high-dose region [62, 63]. Reoperation for pathological confirmation was

11

accepted according to neurosurgeons’ clinical judgement for differentiating tumor progression from pseudoprogression and radiation necrosis [62, 63].

Extensive progressive disease (EPD) was defined as the progression of tumors continuously extending or discretely spreading ≥ 2 cm from the preoperative tumor margin along the PE areas (Figure 5, A3) [64]. For tumors with sSVZCC invasion, EPD in the SVZ (SVZEPD) and in the CC (CCEPD) were defined as tumors continuously extending or discretely spreading with PE areas along the SVZ and the CC, respectively, for ≥ 2 cm from the preoperative tumor margin (Figure 5, A3). We adopted the two imaging factors of EPE and sSVZCC invasion to categorize patients into four groups (Figure 5; B1,2 and C1,2) to analyze the survival, EPD rate, and tumor progression patterns (Figure 5, B3 and C3). The tumor progression areas were categorized as local (tumors involving the original tumor bed), regional (tumors involving the preoperative edematous areas and located beyond the original tumor bed), and distant (tumors located beyond the original tumor bed and preoperative edematous areas) [23].

2.1.5 Statistical analysis

Data analysis and statistical tests were performed using SPSS (version 19; SPSS, Chicago, IL, USA). OS was calculated on the basis of the date of first histological diagnosis to the date of death. PFS was calculated on the basis of the date of first histological diagnosis to the date of disease progression, including death or tumor progression proven through imaging Patients lost to follow-up or those who were alive but were without progression at the time of analysis were censored from the analysis. Survival was calculated using the Kaplan-Meier product-limit method. Differences in survival were compared between the

12

groups by using the log-rank test. The variables of the anatomical features and other adverse prognostic factors, including age, KPS, extent of tumor resection, and tumor size, were used for the univariate and Cox regression analyses [7-9, 61, 65]. Patient age was stratified as <

50 versus ≥ 50 years, KPS as ≤ 70 versus ≥ 80, and tumor size as < 5 versus ≥ 5 cm. The extent of tumor excision was classified according to the findings of postoperative neuroimaging or neurosurgeon records. The hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated for the survival data. Fisher’s exact test was used to examine the significance of the association between EPE and sSVZCC invasion and tumor progression patterns and areas. All tests were two-sided and results with p < 0.05 were considered statistically significant.

2.2 Basic Study of Intratumoral Hydrogel Carboplatin Injection

The workflow of our comprehensive study design, including biomaterial, in vitro, and in vivo investigations, is illustrated in Figure 6 [59].

2.2.1 Biomaterial investigation

The materials and reagents used for the preparation of hydrogel carboplatin were:

hyaluronic acid (HA, average molecular weight 1.5–1.8 × 10⁶ Da), adipic acid dihydrazide (ADH), sodium bicarbonate, and sodium chloride (Sigma-Aldrich, Missouri, USA), ethylene glycol and sodium periodate (NaIO4) (RDH Chemical Company, California, USA), dialysis bag MWCO 6000–8000 (Spectrum Laboratories Incorporated, California, USA), and

13

PARAPLATIN® (carboplatin aqueous solution, 10 mg/mL) (Bristol Myers Squibb Company, New York, USA).

2.2.1.1 Preparation of oxi-HA/ADH hydrogel and hydrogel carboplatin

The procedures of oxi-HA/ADH hydrogel and hydrogel carboplatin preparation were illustrated in Figure 7A and 7B [59]. One gram of HA powder was dissolved in 100 mL of double-distilled water to form a weight percentage 1% (w/v) aqueous solution of hyaluronic acid, and then 15 mL of sodium periodate solution (NaIO4, 2.67%) were added gradually with stirring. The molar ratio of NaIO4 to HA was 1:1 to achieve oxidation degrees [54].

After 24 h at room temperature, the oxidation reaction was ended by adding ethylene glycol for 30 minutes. The oxi-HA solution was poured into a dialysis bag MWCO 6000-8000 with water changed twice daily for 3 days. The final products were dried by a freeze dryer (FDU-1100, EYELA, Tokyo, Japan) for 3 days to yield a white fluffy product, oxi-HA [54]. The oxi-HA was dissolved in a phosphate-buffered saline (PBS, pH 7.4) to a concentration of 6%

(w/v) at a 4 °C overnight. At the same temperature, 0.24 mg of ADH dissolved in 3 mL of aqueous carboplatin (10 mg/mL) were homogeneously mixed to form ADH carboplatin solution. Afterwards, 12 mL of oxi-HA was added into the mold, and then 3 mL of ADH carboplatin solution was slowly poured and mixed to form hydrogel carboplatin [66], with the carboplatin concentration of 2 mg/mL.

2.2.1.2 Characterization of oxi-HA/ADH hydrogel and hydrogel carboplatin by Fourier

transformation infrared (FTIR) analysis

14

An FTIR spectrometer, FTIR-4200 (JASCO Incorporated, Maryland, USA), with ATR PRO450-S was used to distinguish the functional groups of oxi-HA, ADH, oxi-HA/ADH, and hydrogel carboplatin. Freeze-dried samples were ground into power, placed in well plates, and then pressed down gently by the pressure tip. The FTIR spectra were recorded by 16 scans between 2500 and 500 cm^-1 with a resolution of 8 cm^-1 [54].

2.2.1.3 Gelling time and temperature of oxi-HA/ADH hydrogel by rheometer

A rheometer DHR-3 (TA instruments, USA) with cone and plate geometry were used to estimate the rheological properties of oxi-HA/ADH hydrogel. The elastic modulus G’ and the viscous modulus G” were record and analyzed by TRIOS software (TA instruments, USA) [67]. The crossover point (called the gel point) of G’ and G” determined the gel formation and the gelling time of substance as the time elapsing from liquid state to gel state [54]. The G’ and G’’ crossover point of oxi-HA/ADH hydrogel was used to evaluate the gelling time at different temperatures ranging from 4 to 40 °C, which was measured by oscillation time sweep mode at 0.1 Hz, 10 Pa and terminated after 15 min [67].

2.2.1.4 Degradation property of oxi-HA/ADH hydrogel

In PBS under 37 °C, 5% CO2, 0.3 mL of liquid-state oxi-HA/ADH solution was poured into the cylinder mold and put for 10 min to form a gel-like matrix. Then the cylinder of

In PBS under 37 °C, 5% CO2, 0.3 mL of liquid-state oxi-HA/ADH solution was poured into the cylinder mold and put for 10 min to form a gel-like matrix. Then the cylinder of