Antibody-Dye Conjugate Detects Epidermal Growth Factor Receptor Expression in High-Graded Gliomas: A Feasibility Study for Fluorescence-Guided Resection


 Background: The prognosis for high-grade glioma (HGG) remains dismal and extent of resection correlates with overall survival and progression free disease. Epidermal growth factor receptor (EGFR) is a biomarker heterogeneously expressed in HGG. We assessed the feasibility of detecting HGG using near-infrared fluorescent antibody targeting EGFR. Methods: Mice bearing orthotopic HGG xenografts with modest EGFR expression were imaged in vivo after systemic panitumumab-IRDye800 injection to assess its tumor-specific uptake macroscopically over 14 days, and microscopically ex vivo. EGFR immunohistochemical staining of 59 tumor specimens from 35 HGG patients during was scored by pathologists and expression levels were compared to that of mouse xenografts. Results: Intratumoral distribution of pan800 correlated with near-infrared fluorescence and EGFR expression. Fluorescence distinguished tumor cells with 90% specificity and 82.5% sensitivity. Target-to-background ratios peaked at 14 hours post panitumumab-IRDye800 infusion, reaching 19.5 in vivo and 7.6 ex vivo, respectively. Equivalent or higher EGFR protein expression compared to the mouse xenografts was present in 77.1% HGG patients. Age, combined with IDH-wildtype cerebral tumor, was predictive of greater EGFR protein expression in human tumors. Conclusion: Tumor specific uptake of pan800 provided remarkable contrast and a flexible imaging window for fluorescence-guided identification of HGGs despite modest EGFR expression.


Introduction
High-grade gliomas (HGGs) are the most common primary malignant brain tumors in adults and the leading cause of cancer-related deaths in children, carrying a poor prognosis despite intensive treatments with surgery, radiotherapy, and chemotherapy. 1,2 As extent of resection correlates with overall survival and progression free disease, [3][4][5] fluorescence guided surgery that targets specific HGG biomarkers can enhance intraoperative tumor visualization and delineation of tumor margins and improve overall survival of patients.
Epidermal growth factor receptor (EGFR) is an attractive biomarker for HGG imaging.
Significantly higher EGFR protein overexpression and gene amplification are more characteristic of high-grade gliomas compared to low grade gliomas, and are implicated in tumor cell migration and aggressiveness. 6,7 In a fluorescence imaging study, the IRDye800CW (Ex/Em: 774/789 nm) labeled therapeutic antibody, panitumumab-IRDye800, bound to EGFR positive rat glioma cells with higher affinity than the fluorescent EGFR ligand, EGF800. 8 Moreover, this fully humanized EGFR antibody has an improved safety profile compared to its chimeric counterpart, cetuximab. 9,10 A major barrier to the clinical translation of EGFR targeting imaging probes is the heterogeneity of EGFR protein expression which can vary by orders of magnitude in human HGGs. 11,12 Despite a widely recognized problem in cancer research, animal imaging studies often adopt subcutaneous xenograft models with exceptionally high target expressions that poorly represent actual malignancies in the brain. 12,13 This study examined the feasibility of detecting human HGGs via fluorescence imaging using panitumumab-IRDye800 in a preclinical orthotopic tumor model with only modest level of EGFR expression which was benchmarked against similar expression levels in patient HGG tissues. Since biopsies are not usually taken before resection surgeries, HGG specimens were characterized to stratify a patient population with positive EGFR protein expression that would benefit most from intraoperative targeted imaging to assist in resection.

Panitumumab-IRDye800 detects EGFR protein expressed in human glioma cells in vitro
Four human glioma cell lines, U251, H37, D2159 and D270, Fig 1A, Fig. S1. U251 was selected to establish the orthotopic brain tumor model in mice due to its modest EGFR expression in tumors and to test the feasibility of detecting HGG with panitumumab-IRDye800 in vivo.

U251 produces contrast enhancing orthotopic HGG xenograft tumor in mice
U251 cells implanted (arrow) in the left hemisphere of mouse brains (n = 5) were optically accessible through a glass cranial window (white box), Fig. 2A. The implant site was examined with intravital microscope on POD7 to confirm the xenograft was growing close to the brain surface (0-300µm), Fig. 2B. Intravenously injected dextran (3kDa) was confined to the lumen of the blood vessels as a vascular tracer and suggested an intact blood-brain barrier (BBB). The tumor area revealed marked hyperintense signal in T1-weighted MRI scans with gadolinium contrast on POD15, indicating a disrupted BBB to gadolinium (molecular weight: 1058 Da), and the nonenhancing region (arrowhead) within the tumor corresponding to the focus of necrosis, Fig. 2C.
The average tumor volume reached 12.1 ± 1.9 mm 3 with 3.5 ± 1.2 mm in the greatest dimension.
Significant post contrast enhancement between tumor and normal brain was seen on T1 imaging, from 1.21 ± 0.19 to 5.64 ± 0.98 (P = .00057, n = 5), Fig. 2D. Minimal contrast (1.13 ± 0.15) was present in the tumor area on T2-weighted MRI, indicating the absence of hydrocephalus. Tumor growth was monitored for one month via bioluminescence, Fig. 2E. Exponential growth was followed for a latency period of 12 to 18 days, with 16.7 ± 1.38-fold change from POD0 to POD30,

Panitumumab-IRDye800 detects HGG in mice in vivo and ex vivo on macroscopic near-infrared fluorescence imaging
In mice bearing contrast enhancing HGGs (n = 5), intratumoral fluorescence steadily increased after panitumumab-IRDye800 injection, peaking at 14 hours, and was maintained at a high level from 12 to 18 hours before tapering off, Fig. 3A. Mean fluorescence intensity (MFI) reduced to half of its 14-hour peak after 9 days, while the target-to-background ratio (TBR) reached a high point of 19.5 ± 1.3 at 14 hours and kept increasing after reaching a minimum at 48 hours, indicating a faster clearance of panitumumab-IRDye800 from normal tissue than tumor, Fig. 3B. In comparison, much lower panitumumab-IRDye800 accumulated in the brain of normal mice, reaching a peak at 3 hours (MFI = 54.6 ± 6.1 % compared to that of HGG bearing mice, TBR = 6.6 ± 0.6) and returned to the baseline after two weeks, Fig. 3A  was found to be specifically accumulated in HGG xenografts compared to surrounding normal brains (29.4 ± 2.6 vs. 3.9 ± 1.8 normalized to fat, contrast against normal brain = 7.6 ± 1.5, P = 8.03 ×10 -8 ), Fig. 3D. Resected HGG tumors were serially bisected and the contrast between tumor pieces against normal brain ranged between 1.22 and 9.30, Fig. 3E. Tumor pieces as small as 0.9 ± 0.3 mg could be readily detected on fluorescence imaging.

IRDye800
In H&E stained mouse brain sections, tumor cells were identified by their high nucleus to cytoplasm ratio, nuclear and cytoplasmic abnormal morphology, and their pleomorphism, Fig. 4A.
Minimal panitumumab-IRDye800 delivery was detected in normal brain characterized by a normal BBB and resulted in high contrast (TBR = 9.1 ± 1.7). NIR fluorescence from panitumumab-IRDye800 was able to detect tumor at 90% specificity and 83% sensitivity (area under the curve = 0.93), Fig. 4D.

EGFR protein is commonly expressed in a variety of human high-grade gliomas
To assess the potential of detecting EGFR with panitumumab-IRDye800 on imaging in HGG patients, EGFR protein expression was immunohistochemically characterized in human HGG tissues. 59 HGG tissue samples from a total of 35 (2.9%) patients were included in the study, Table 1 tissues from nine patients also contained normal brain structures (e.g. choroid plexus and cerebellum) with little, if any, non-specific EGFR immunoreactivity. As a biomarker, EGFR protein expression distinguished human HGG tumors against normal brain with high specificity (89%) and sensitivity (77%) at 18.3% cutoff IHC staining index (likelihood ratio = 6.9), Fig. 5C.

Supplementary
A majority of patients in various HGG disease subgroups expressed positive EGFR (anaplastic ependymoma, 100%; glioblastoma, 85%; atypical teratoid rhabdoid tumor, 83%; and diffuse midline glioma, 67%), Fig. 5D. A substantial number of these patients (anaplastic ependymoma and ATRT, 50%; diffuse midline glioma, 33%; and glioblastoma, 15%) expressed EGFR at a modest level (IHC score: 1+ and 2+) comparable to that found in the mouse xenografts. Two cases of medulloblastoma in this study expressed no EGFR. 60% of less common HGG diagnoses (grouped as "Other") expressed positive EGFR. These results suggest that a large number of HGG patients are expressing EGFR equal to or above the level detectable by panitumumab-IRDye800 imaging and could potentially benefit from fluorescence image-guided surgery.

Clinical features of EGFR positive HGG patients
Since patients with positive EGFR protein expression are more likely to benefit from EGFR-  Table 1. Patients with positive EGFR protein expression were older at diagnosis (median 24 versus 5 y, P = .0059), had tumors located more frequently in the cerebral hemisphere (74% vs. 37.5%, P = .019) and were more likely to be IDH wildtype (94% vs. 0%, P = .0048). Over two thirds (69.2%) of pediatric patients (age < 18 y) had HGGs that expressed detectable EGFR protein on IHC. Differences in gender, tumor size, classification, WHO grade, aggressiveness (Ki-67 proliferation index) and recurrence history were not observed between the two groups; the incidence of TP53 mutation and EGFR gene amplification were also similar.

Discussion
Intratumoral and intertumoral EGFR expressions in HGGs are known to be heterogeneous. 14 Many preclinical imaging studies were conducted on subcutaneous implants with monolithically highly positive expression of the molecular target of interest or a small number of specific patientderived xenografts without relating their target expression level with the larger patient population.
In this preclinical study, U251 cells were specifically chosen, from in vitro immunofluorescence assays, to establish the orthotopic brain tumor model in order to assess the imaging performance of panitumumab-IRDye800 in a challenging yet realistic environment where EGFR expression in tumors is more modest. As 77.1% of all human HGG tissues tested expressed EGFR at equal or higher levels, it is reasonable to believe that panitumumab-IRDye800 would be able to detect them in vivo as well given the right conditions. The fluorescence signal in mouse brain tumors reached its peak at 14 hours after systemic panitumumab-IRDye800 injection. The time frame of 12 to 18 hours after panitumumab-IRDye800 infusion at this particular dose is promising for clinical translation of fluorescence guided surgery, since patients can be infused with the imaging agent on the day before surgery. 10,15 This is a favorable window for imaging and gives enough time for observation of potential adverse effects. In comparison, FDA (U.S.) approved intraoperative optical imaging agent for glioma surgery, 5-ALA, 16 is administered orally 2 to 4 hours before anesthesia, which can be logistically challenging to align clinical workflow with the more narrow optimal imaging window. The prolonged high TBR measured over two days can better accommodate changes in clinical workflow. In vivo imaging would be less than ideal after 48 hours, since by then, overall fluorescence intensity will have dropped significantly despite TBR continuing to increase due to faster clearing of non-specific panitumumab-IRDye800 binding.
The EGFR specific distribution of panitumumab-IRDye800 was associated with increased vasculature density and reduced BBB tight junction protein expression in contrast enhancing HGG xenografts in mice. This result was consistent with a previous clinical study where intratumoral uptake of cetuximab-IRDye800CW was observed in contrast-enhancing HGGs in two patients, but not in a third patient with an astrocytoma that on MRI also did not demonstrate any contrast enhancement. 17 Taken together, these results suggest that compromised BBB integrity, indicated by contrast enhancing tumor on MRI with small molecule gadolinium, may predict intratumoral delivery of antibody-sized molecules, although further work is needed to validate this finding and elucidate the specific mechanism that may be involved in this phenomenon.
The advantage of panitumumab-IRDye800 is that this agent, due to its targeting specificity, could selectively bind to a tumor-specific biomarker of HGGs. When tested in the same animal model, panitumumab-IRDye800 demonstrated over 10% higher specificity for tumor and 30% higher comprehensive TBR than those of 5-ALA. 18 As an imaging probe for fluorescence-guided surgery, panitumumab-IRDye800 could potentially outperform 5-ALA which suffers from limited imaging contrast and low negative predicting value (16.7% for HGG). 19 Distribution of panitumumab-IRDye800 in resected tumor tissues could also have potential therapeutic implications such as monoclonal antibody delivery for targeted therapy although efficacy studies are beyond the scope of this study.
Population subgroups depending on intrinsic factors such as age, gender and ethnicity are commonly and intentionally included in efficacy and safety evaluation of clinical trial outcomes. 20 However, patients in HGG disease subgroups with lower incidence rates but at equally high-risk, for whom effective therapeutic concepts are desperately needed, tend to be underrepresented in clinical brain imaging studies. 17,21,22 As the most frequent malignant brain tumor in adults, glioblastoma accounts for 45-50% of all primary malignant brain tumors and occurs 30 times as frequently as anaplastic oligodendroglioma. 23 On the other hand, ATRT and diffuse midline glioma are rarely seen in adults but account for 10% and 20% of all pediatric primary tumors, respectively. 24,25 The variable clinical outcome of anaplastic ependymomas in both adults and children primarily depends on the extent of surgical resection and molecular group, pointing to the clinical relevance of molecular classification. 26 Ongoing clinical trials for fluorescence guided surgery of brain tumors using EGFR antibody or affibody (NCT03510208, NCT02901925 and In conclusion, significant imaging contrast was observed 14 hours after systemic panitumumab-IRDye800 administration in HGG xenografts with modest EGFR protein expression. Both target and tumor specific probe uptake was confirmed microscopically. In the majority of patient HGG tissues, EGFR protein expression levels were equivalent to or higher compared to that in the preclinical model, suggesting their detectability with intraoperative EGFR targeted imaging by panitumumab-IRDye800. Older patients with IDH-wildtype cerebral hemisphere tumors tend to express greater levels of EGFR protein. These findings demonstrate the feasibility to employ an EGFR-targeting antibody for fluorescence-guided surgery of HGGs, as well as the utility of clinical profiling in stratification of HGG patients most likely to benefit from antibody-based imaging strategies.

Patients
This single-institution study was approved by the Stanford University Institutional Review Board were randomly included from HGG disease subtypes. Cases with either scant tissue or incomplete medical records were excluded, resulting in the 35 patients in the study. Two board-certified pathologists reviewed hematoxylin and eosin (H&E) slides to outline regions of viable tumor and normal brain. Peripheral cases and those with scant tissues (size < 1mm) were excluded.

Cells
Human glioma-derived tumorigenic cell lines, U251 (luc+, from Dr. Lawrence Recht, Stanford University), H37, D2159 and D270 (from Dr. Vidya Chandramohan, Duke University) were maintained in improved MEM zinc option medium (Gibco) as previously described. 30 Head-andneck squamous cell carcinoma (SCC-1, University of Michigan) and human embryonic kidney cells (293T, ATCC) were positive and negative controls for EGFR expression, respectively. All cultures were routinely subjected to mycoplasma testing and only used for experiments when confirmed negative. Routine short tandem repeat analysis was performed using the Geneprint 10 system (Promega) to ensure cell line identity. Growth and morphology were observed with bright field (Evos XL Core, Life Technologies) and phase contrast (BZ-X710, Keyence).

Cranial window installation and orthotopic brain tumor mouse model
Animal experiments were carried out in accordance with Stanford University Institutional Animal Care and Use Committee guidelines (protocol #26548), and in compliance with the ARRIVE guidelines (PLoS Bio 8(6), e1000412, 2010). Cranial window installation and orthotopic brain tumor mouse model were established in twenty female NSG mice (005557, Jackson Laboratory) as previously described. 34 U251 human xenografts were implanted (5 μl, 1 x 10 6 cells/ml) and optically accessible through a 6.0 mm x 4.0 mm x 0.4 mm glass cranial window. Normal mice were prepared using the same protocol as controls, with sham injection of 5 μl saline instead of tumor cells. Tumor growth was monitored via bioluminescence (IVIS Spectrum, PerkinElmer) every three days for one month. Total photon flux (photons/s) over the mouse brains was quantified in Living Image (v4.5.5, PerkinElmer). To ensure growth of tumor close to the surface of the brain for subsequent imaging experiments, on postoperative day 7 (POD7), vascular elements in brain tumors were imaged through the cranial window with a two-photon intravital microscope (Nikon) following 3 kDa fluorescein-dextran injection (10mg/ml, i.v., ThermoFisher). Excitation laser (920 nm) powers was modulated between 10 and 30 mW with depth (0-300 µm, step size = 5 μm).
Image were reconstructed and analyzed with Image J (v1.52p).

Magnetic resonance imaging
Anesthesia was induced with 5% isoflurane and maintained with 2% isoflurane in room air and oxygen mixed at 3:1 ratio. Five HGG bearing mice were imaged in a 7 Tesla small-bore MR

Near-infrared fluorescence imaging
Tumor bearing mice (n = 5) and normal mice (n = 5) with cranial windows were imaged in intervals over a period of 14 days (starting on POD15) after receiving 20mg/kg panitumumab-IRDye800 (i.v.). Pharmacokinetics were measured as mean fluorescence intensity (MFI) and target-to-background ratio (TBR) calculated as MFI in each tumor region of interest, ROI-T, divided by MFI of normal tissue, ROI-B. Another five tumor bearing mice were sacrificed at the peak antibody-dye conjugate delivery time and MFIs of harvested organs were normalized to fat. Resected brain tumors were serial bisected into different sizes, weighed and imaged with normal brains. All images were collected in a near-infrared florescence imager (Ex/Em: 785/820 nm, Pearl Trilogy Imaging System, LI-COR) and processed with Image Studio (v5.2, LI-COR). Mouse brains (n = 5) removed 14 hours after panitumumab-IRDye800 injection were formalin-fixed overnight and paraffin-embedded. Brain sections were rehydrated, incubated in DyLight 488 labeled tomato lectin (10 µg/ml, Vector Laboratories) for 30 minutes and counterstained with DAPI (300nM, Invitrogen) before they were imaged with a custom-built fluorescence microscope (Leica). 31 MFIs were measured from 10 randomly selected cells in three views.

Immunohistochemical analysis
Formalin-fixed paraffin-embedded brain sections (4μm thick) were incubated with primary antibodies after heat mediated antigen retrieval. All immunohistochemical procedures were performed following preprogramed standard protocol on an automatized histostainer (Dako Autostainer, Agilent) with positive and negative controls in each operation. Immunoreactivity was visualization with diaminobenzidine and magenta chromogens (Dako EnVision). A staining index was calculated as the product of intensity and fraction of positive tumor cells using Aperio ImageScope (Leica). 35 The immunoreactivity of human brain tissues were scored by two boardcertified pathologists as previously described. 36

Statistical analysis
Data are expressed as mean ± standard error of the mean (SEM). Paired and unpaired t-tests (twotailed) were performed for T1-weighted MR imaging contrast and other group comparisons, respectively. Patient characteristics were compared between EGFR positive and EGFR negative groups using a Wilcoxon rank-sum test and Pearson's chi-square tests (GraphPad Prism 8.0) as appropriate. Significance was defined at P < 0.05.    Cldn5: tight junction protein, claudin-5; ERG: erythroblast-transformation-specific related gene (a vascular endothelial marker). Red boxes: primary HGG core and normal brain margin (black dotted line); Blue boxes: HGG cells infiltrating normal brain; Arrows: blood vessels; arrowheads:

Figure Legends
infiltrating HGG tumor cells. B, IHC staining index (left Y axis) and Cldn5/ERG ratio (right Y axis, blue) of tumor areas (n = 16) and normal brains (n = 10), respectively (*P < 0.001). C, NIRF microscopic imaging of intratumoral uptake of panitumumab-IRDye800 (pan800, green) in HGG mouse brain sections in the same field of view as A. D, Fluorescence intensity of panitumumab-IRDye800 fluorescence distinguishes tumor from normal brain in C with 90% specificity and 83% sensitivity. AUC: area under the curve; ROC: receiver operating characteristic.