Preclinical evaluation of a humanized, near infrared fluorescent antibody for fluorescence-guided surgery of MUC16-expressing pancreatic cancer

Abstract

Surgery remains the only potentially curative treatment option for pancreatic cancer, but resections are made more difficult by infiltrative disease, proximity of critical vasculature, peritumoral inflammation, and dense stroma. Surgeons are limited to tactile and visual cues to differentiate cancerous tissue from normal tissue. Furthermore, translating preoperative images to the intraoperative setting poses additional challenges for tumor detection, and can result in undetected and unresected lesions. Thus, PDAC has high rates of incomplete resections, and subsequently, disease recurrence. Fluorescence-guided surgery (FGS) has emerged as a method to improve intraoperative detection of cancer and ultimately improve surgical outcomes. Initial clinical trials have demonstrated feasibility of FGS for PDAC, but there are limited targeted probes under investigation for this disease, highlighting the need for development of additional novel biomarkers to reflect the PDAC heterogeneity. MUCIN16 (MUC16) is a glycoprotein that is overexpressed in 60-80% of PDAC. In our previous work, we developed a MUC16-targeted murine antibody near-infrared conjugate, termed AR9.6-IRDye800, that showed efficacy in detecting pancreatic cancer. To build on the translational potential of this imaging probe, a humanized variant of the AR9.6 fluorescent conjugate was developed and investigated herein. This conjugate, termed huAR9.6-IRDye800, showed equivalent binding properties to its murine counterpart. Using an optimized dye:protein ratio of 1:1, in vivo studies demonstrated high tumor to background ratios in MUC16-expressing tumor models, and delineation of tumors in a patient-derived xenograft model. Safety, biodistribution, and toxicity studies were conducted. These studies demonstrated that huAR9.6-IRDye800 was safe, did not yield evidence of histological toxicity, and was well tolerated in vivo. The results from this work suggest that AR9.6-IRDye800 is an efficacious and safe imaging agent for identifying pancreatic cancer intraoperatively through fluorescence-guided surgery.

Graphical Abstract

INTRODUCTION

Pancreatic cancer is currently the fourth leading cause of cancer-related deaths in the United States, and has an extremely low 5-year survival rate of 10%.¹ Globally, the incidence of pancreatic cancer-related deaths is projected to increase, with predictions that pancreatic cancer will be the second leading cause of cancer-related deaths by 2030.^{2, 3} Surgery remains the only potentially curative option for patients with this disease, but only 20% of pancreatic cancer patients have resectable disease at the time of diagnosis.⁴ However, an additional 20-30% of patients present with borderline resectable or locally advanced disease. In these cases the tumor either abuts or invades adjacent vasculature and may involve locoregional lymph nodes but has not yet spread to distant organs. Neoadjuvant therapy has become increasingly implemented for this subset of patients in order to downstage tumors, decrease disease complexity, and increase eligibility for surgery.^{5, 6} Under current standard of care chemotherapy regimens, neoadjuvant therapy has shown efficacy in substantially increasing patient eligibility for resection.^7–11 Therefore, surgical resections continue to impact a growing patient population in a cancer that has no other potentially curative treatment options, highlighting the importance of successful resections.

Pancreatic cancer resections are made more difficult by infiltrative disease, peritumoral inflammation, and dense desmoplastic stroma. Surgeons are limited by visual and tactile clues to differentiate normal tissue from cancerous tissue.^12–15 While many preoperative imaging modalities like MRI, CT, PET, and ultrasound can provide initial staging and diagnostic information, translating these images to the intraoperative setting can be difficult, resulting in missed lesions.¹⁶ Furthermore, in the context of neoadjuvant therapy, traditional imaging modalities like CT may be unreliable in determining resectability and staging of PDAC, due to limitations in differentiating treatment-induced fibrosis from infiltrative disease.^{17, 18} Since therapeutic response and the presence of additional lesions may alter the course of treatment for the patient and preclude them from initial surgery, or may lead to disease recurrence, identification of the extent of the disease during surgery is of critical importance. R1, or incomplete resections, occur at high rates, reported as high as 70%. R0, or complete resections, in which there is a distance greater than 1 mm between the tumor and the surgical margin, have demonstrated an improved survival benefit.¹⁹ Currently, up to 85% of patients that undergo surgical resection succumb to disease recurrence, due to both undetected lesions, and incomplete resections. Thus, there is an unmet need for intraoperative methods to detect lesions for surgical resections in pancreatic cancer.

Fluorescence-guided surgery (FGS), or the use of fluorescent contrast agents and cameras in the surgical suite to detect tumors, has demonstrated efficacy for intraoperative identification of cancer in many clinical trials for a variety of cancer types. Several clinical trials have been conducted investigating FGS for PDAC, and have demonstrated initial safety, efficacy, and feasibility for improved surgical resections.^{12, 14, 20} These studies have largely employed antibody-based probes to target specific biomarkers for imaging of pancreatic cancer. However, due to the characteristic heterogeneity of pancreatic cancer, several of these studies have suggested a need for additional biomarkers to be investigated for FGS to increase available targeted agents. In our preliminary studies, we showed that MUCIN16, or MUC16, a glycoprotein that is expressed in 60-80% of pancreatic cancers, has potential as a novel target for FGS of pancreatic cancer with a murine MUC16-targeted antibody conjugated to a NIR dye, termed AR9.6-IRDye800.²¹ Our initial studies showed significantly improved contrast enhancement of tumors with AR9.6-IRDye800 as compared to a non-specific IgG control in subcutaneous and orthotopic xenograft models, and, based on current recommendations for developing new FGS agents, warranted further investigation to refine and evaluate the agent for clinical translation.²²

Herein, our objective was to improve translational potential, assess the preclinical efficacy of AR9.6-IRDye800 to support potential clinical translation, and to investigate the role of antigen expression and tumor microenvironment on accumulation and contrast. Furthermore, we sought to evaluate the preliminary safety, biodistribution, and toxicity profile of huAR9.6-IRDye800 to lay the groundwork evidence for clinical translation. To that end, we developed a humanized variant of this antibody conjugate to minimize potential immunogenicity and undesirable adverse reactions.²³ We assessed this agent for feasibility of clinical translation by optimizing the dye to protein ratio in vitro and in vivo, and dynamic contrast enhancement over time. To address the impact of variable biomarker expression on tumor contrast, we evaluated three subcutaneous tumor models with differential expression of MUC16. We assessed acute, 14-day, and long term biodistribution, and evaluated organ pathology for evidence of toxicity. Finally, we incorporated a patient-derived xenograft model to recapitulate clinical disease presentation and evaluate efficacy of utilizing huAR9.6-IRDye800 for FGS.

MATERIALS AND METHODS

Humanized Antibody Generation

The humanization and production of recombinant AR9.6 (huAR9.6) has been described previously.²⁴ Individual plasmids (pcDNA3.4) containing huAR9.6 or chAR9.6 heavy chain (human IgG1, and kappa for huAR9.6, murine IgG1 and kappa for chAR9.6) were used for transient transfection of expiCHO cells according to the manufactures’ instructions (Thermo Fisher). Transfections were carried out using a 2:1 ratio of light chain to heavy chain plasmids. Antibody production was monitored by SDS-PAGE and 7-9 days post-transfection, the culture supernatant was harvested, filtered and antibody purified to homogeneity in a single step using Protein G affinity chromatography.

Cell Culture

Pancreatic cancer cell lines including HPNE, T3M4, and Colo357 were obtained 2/2018 from Dr. Michael A. Hollingsworth. OVCAR3 cells were obtained 3/2018 from Dr. Adam Karpf. Panc1 cells were obtained 8/2019 from Dr. Joyce Solheim. All cells were grown in RPMI 1640 (Corning, 10–040-CV; Tewksbury, MA), supplemented with 10% fetal bovine serum, 100 I.U./ml penicillin, and 100 μg/ml of streptomycin (P/S) (Corning, 30–002-CI). Cells were maintained at 37 °C in a humidified incubator with 5% CO₂. Cells were regularly tested for mycoplasma (Myco-Sniff Mycoplasma PCR Detection Kit, 093050201, MP Biomedicals; Irvine, CA).

Antibody Conjugation

HuAR9.6 was conjugated to IRDye800 N-hydroxysuccinimide (NHS) ester (0.5 mg, LI-COR Biosciences, 929-70020; Lincoln, NE) according to the instructions of the manufacturer. Briefly, 100 μl of 1 M potassium phosphate was added to each mg of antibody to raise the pH to 8.5. Dye was dissolved in 50 μl of nanopure water and added to 1 mg of antibody and incubated for 2 hours at room temperature. Addition of 0.12 mg of dye consistently resulted in ~1:1 dye:protein. Excess dye was removed by Zeba spin desalting columns (ThermoFisher Scientific, 89891). Antibody diluted 1:5 in 1:1 PBS:methanol was loaded into a 1 cm cuvette (Eppendorf, E0030106300; Hauppauge, NY). Conjugation ratios were determined spectrophotometrically with a Thermo Scientific Evolution 220 UV-visible spectrophotometer, and fluorescence was confirmed with a FluoroMax 4 spectrofluorometer (Horiba Scientific; Irvine, CA). An SDS page gel (4-20% gradient gel, Bio-Rad, 4568094; Hercules, CA)) was run to confirm that free dye had been removed. 0.5 μg of huAR9.6-IRDye800 conjugates at increasing dye:protein ratios and free unconjugated IRDye800 control (LI-COR Biosciences, 929-08972; Lincoln, NE) were loaded onto the gel, and gel was run at 90-125V. The gel was immediately imaged on a LI-COR Odyssey^® M imaging system in the 800 nm channel.

Binding Assessment

To carry out binding studies, a MUC16 fragment corresponding to tandem repeat 5 (TR5) and flanking residues from TR4 and TR6 (Uniprot code, Q8WXI7, residues 12665-12857) was expressed in E. coli as a thioredoxin (Trx)-fusion and purified to homogeneity by immobilized metal affinity chromatography as described previously.²⁵ Binding of murine, chimeric, and humanized antibody and antibody conjugates to recombinant MUC16 were assessed with an ELISA in which a 96 well plate (Fisher Scientific, 21-377-203; Waltham, MA) was coated overnight at 4 °C with recombinant MUC16 antigen (100 ng of Trx-TEV-TR6-SEA5-TR4 in PBS). The plate was blocked with BSA (1% in PBS, 1 h, room temperature). Primary antibody was added at a starting concentration of 100 nM and serially diluted (5-fold in PBS + 0.1% tween-20) down the plate and incubated (1 h, RT). The plate was washed 5 times (PBS + 0.1% Tween-20). Secondary antibody (1:40000 dilution, anti-human IgG Kappa Horseradish Peroxidase, B7466, Novus Bio-techne; Littleton, CO) was added and incubated for 1 h at RT, and the plate was washed again. TMB Substrate Solution (Thermo Scientific, N30; Waltham, MA) was added and incubated for ~ 30-60 s and developed by the addition of an equal volume of 0.18 M Sulfuric Acid. The plate was read at 450 nm and 540 nm on a plate reader. For estimation of apparent antibody affinity, (EC50), the log concentration of antibody was plotted vs. absorbance and fit to a four-parameter logistic curve in GraphPad Prism (Graph Pad Software).

Western Blotting

Cells were lysed with radioimmunoprecipitation assay (RIPA) buffer (ThermoFisher Scientific, 89900) supplemented with Halt protease inhibitor (ThermoFisher Scientific, 78440). 20 μg of protein were separated on a 4–20% polyacrylamide gel (Bio-Rad, 4568094; Hercules, CA) and transferred onto a nitrocellulose membrane (Bio-Rad, 1620115). The membrane was blocked with 5% Blotting-Grade Blocker (Bio-Rad, 170–6404) in TBST and incubated with huAR9.6 (1:1000, 1 mg/ml stock solution) and GAPDH antibody (1:2000, Cell Signaling Technology, 2118S; Danvers, MA) overnight at 4 °C. The membrane was incubated with goat anti-rabbit HRP secondary antibody for detecting the GAPDH loading control (1:5000, Jackson Immunoresearch, 115–035-144; West Grove, PA), and goat anti-human HRP secondary antibody to detect huAR9.6 (1:5000, Jackson Immunoresearch, 109–035-003). Secondary antibodies were diluted in 5% Blotting-Grade Blocker for 1 h on a rocker and were visualized with enhanced chemiluminescent (ECL) substrate (Bio-Rad, 1705060S).

Fluorescence Microscopy

HPNE, Colo357, and OVCAR3 cells were seeded at 30,000–40,000 cells per chamber of an 8-chamber slide (ThermoFisher Scientific, 154534), and allowed to adhere overnight. Cells were washed 3x with 1X PBS and blocked with 3% BSA in TBS for 1 h at room temperature. 5 μg/ml of AR9.6-IRDye800 was incubated with cells for 1 h at room temperature in 3% BSA in TBST. Cells were washed 3x with PBS. 1 μg/ml of Hoescht 3342 stain (ThermoFisher, 62249) was added to cells, and cells were imaged in Live Cell Imaging Buffer (ThermoFisher, A14291DJ). Cells were imaged at 200X magnification on an Olympus DP80 Digital Camera and cellSens Dimension software.

Animal Models

All animal work was performed under a protocol approved by the UNMC Institution of Animal Care and Use Committee IACUC. Subcutaneous tumor models were generated by injecting 1×10⁶ T3M4 or 1.5 × 10⁶ Colo357 or Panc1 cells suspended in 100 μl of 1:1 media and Matrigel (Corning, 356234) into the left flank of 6–8 week old male NU/J mice (Jackson Laboratories, 002019; Bar Harbor, ME). T3M4 tumors were allowed to grow for ~11 days, and Colo357 and Panc1 tumors were allowed to grow for ~30 days.

A patient-derived xenograft model (J000115419, passage 4) was obtained from Jackson Laboratories. The PDX model was delivered in a female NSG mouse (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ, 005557), screened for pathogens and confirmed to be pathogen and opportunist free. The tumor was initially subcutaneously engrafted in the right flank on June 30, 2021. Tumor was resected at ~1000 mm³ in size (approximately 7 weeks after initial engraftment) and propagated in recipient 6-8 week old female NSG mice according to protocols provided by the manufacturer.²⁶

Biodistribution, Safety, and Toxicity

6-8 week old male and female CD-1 mice (Charles River, Crl:CD1(ICR)) were injected with 100 μl of saline control, 20 μg of 1:1 dye:protein huAR9.6-IRDye800 or 80 μg of 1:1 dye:protein huAR9.6-IRDye800 via tail vein. Agent was allowed to circulate for 1, 14, or 28 days after injection. Mice were weighed weekly to look for signs of toxicity for the 28-day timepoint, and weighed every three days for the 14-day timepoint. At the randomly assigned timepoint, cardiac punctures were conducted, and necropsies were performed. Resected organs were imaged on the LI-COR Pearl and quantified as previously described. Organs were fixed in 10% neutral buffered formalin for 24 hours, before paraffin embedding. Organs were stained with H&E and analyzed for signs of toxicity by a board-certified pathologist (G.A.T) who was blinded to the treatment groups. Scoring for toxicity was defined as follows: 0 = No histological changes, 1 = minimal and not significant histological changes, 2 = mild histological changes, 3 = moderate histological changes, and 4 = severe histological changes.

Optimal Dye to Antibody Ratio Assessment

Female CD-1 mice (Charles River, Crl:CD1(ICR); Wilmington, MA), 6-8 weeks old were injected via a tail vein with 50 μg of huAR9.6-IRDye800, conjugated with either 0.3, 1, 2, or 4 dyes per antibody. Each conjugate was allowed to circulate for 48 h post-injection. Cardiac punctures and necropsies were conducted, and organs were imaged on a LI-COR Pearl Trilogy imaging system. Organs were quantified in Image Studio as previously described.²¹

Dynamic Contrast Enhancement

Male NU/J mice (6-8 week old) were injected subcutaneously with T3M4 (1x10⁶), COLO357 (1.5X 10⁶), or Panc1 (1.5X 10⁶) cells in 50% Matrigel in the left flank. After tumor formation, 50 μg of huAR9.6-IRDye800 was injected via a tail vein. Mice were imaged from 4 h post-injection and then daily through 9 days. Tumor-to-Background ratio (TBR) and signal-to-noise ratio (SNR) values were calculated as previously described.^{21, 27} After 9 days, mice were euthanized, and necropsies were conducted. Necropsied organs were imaged on a LI-COR Pearl Trilogy imaging system.

Immunohistochemistry

Paraffin-embedded tissues were cut into 4 μm thick sections and placed on glass microscope slides (Fisher Scientific, 22-037-246; Waltaham, MA). H&E staining (BBC biochemical MA0101257, MA0101260, Mount Vernon, WA), and Masson’s Trichrome (Statlab, KTMRPT, Mckinney, TX) were conducted according to manufacturer’s protocols. Immunohistochemistry was conducted using an automated Ventana Discovery Ultra Staining platform. OC125 (Roche, 760-2610; Basel, Switzerland), CD31 (Abcam, ab28364, Cambridge, UK), SMA (Abcam, ab5694), and Ki67 (Abcam, ab16667) staining was conducted according to the instructions of the manufacturer. Staining of cell line xenografts was conducted with rabbit polyclonal CA125/MUC16 antibody, as both mouse and human antibodies had high background staining (Novus, 25450002; Littleton, CO). Antigen retrieval was conducted with 10 mM citrate buffer, pH 6, for 20 minutes in a rice cooker between 95-100 °C. Samples were blocked for 45 minutes with 10% goat serum in TBS, and primary antibody (1:1000) was incubated with samples overnight at 4 °C in TBS with 5% goat serum. Goat-anti rabbit HRP secondary antibody (1:1000, Jackson Immunoresearch, 115–035-144) was added for one hour in TBS with 5% goat serum at room temperature. Impact^® DAB substrate (Vector Laboratories, SK4105; Burlingame, CA) was added to tissues for ~4 min. Tissues were washed and counterstained with Hematoxylin (IHC World, IW-1400; Ellicott City, MD), and imaged on the Olympus DP80 digital microscope with cellSens Dimension software.

Fluorescence-Guided Surgery

NSG mice were implanted with the PDX tumor model subcutaneously in the left flank. Tumors were allowed to propagate for 21 days. When tumors reached 200-300 mm³, 50 μl of huAR9.6-IRDye800 (dye:protein 1:1) were diluted into PBS for a total injection volume of 100 μl and injected via a tail vein. Images were collected daily using the LI-COR Pearl Trilogy. Mice were euthanized 72 hours post-injection, and tumors were resected under image guidance using a Fluobeam 800 (Fluoptics; Cambridge, MA), and a Lab-FLARE RC2 Zoom Imaging Head FGS system controlled by an RP1 Photonic Control Unit (Curadel; Natick, MA). Images and videos were collected during surgical resection. Necropsies were subsequently conducted, and resected tumors and organs were imaged on the LI-COR Pearl Trilogy. Fluobeam images were analyzed to calculate intraoperative TBR using Image J software (version 1.52a, NIH; Bethesda, MD). Resected tumors were either frozen in OCT, or formalin-fixed and paraffin embedded for histological analysis.

Statistical Analysis

All statistical analyses were conducted using Graph Pad Prism software. Nonlinear regression was used to analyze data from ELISA assays for binding of huAR9.6 and huAR9.6 conjugates. A Two-way ANOVA was implemented, followed by Tukey’s test for multiple comparisons to compare differences in mean fluorescent signal in resected organs, and to compare TBR between Panc1, T3M4, and Colo357 tumor models. A Two-way ANOVA, followed by Tukey’s test for multiple comparisons was used to compare differences in mean fluorescent signal in resected organs for the biodistribution study. All values are reported as the mean ± standard deviation.

RESULTS

huAR9.6 and huAR9.6-IRDye800 Bind MUC16

Our previous studies developed a MUC16-targeted FGS agent that utilized a murine antibody as a targeting moiety.²¹ However, the use of murine antibodies has limited translational potential.²⁸ Patients injected with murine antibodies have a rapid human anti-mouse antibody response (HAMA), which can cause adverse allergic reactions, increase clearance, and impact tumor penetration.²³ To improve the translational potential of AR9.6 and minimize the potential immunogenicity from a murine antibody, a humanized antibody, huAR9.6, was developed using complementary determining region (CDR) grafting techniques (C.B). Humanized AR9.6 binding was compared to chimeric and murine variants and showed no significant differences in binding recombinant MUC16 (Figure 1A). This suggests that the process of humanizing this antibody did not impact the affinity of huAR9.6 for MUC16. To further confirm the retained affinity for MUC16, a Western blot was conducted, shown in Figure 1B, and demonstrated that the huAR9.6 variant bound MUC16 in MUC16-expressing pancreatic cancer cell lysates, as well as in lysates of a MUC16-expressing ovarian cancer cell line (OVCAR3) positive control. This is consistent with Western blots previously conducted with the murine AR9.6 variant.²¹ Binding was not observed in HPNE negative control cells.

Figure 1.

(A) Humanized, chimeric, and murine AR9.6 binding to recombinant MUC16 (B) Binding of huAR9.6 to human pancreatic cancer cell lines.

HuAR9.6 was conjugated to IRDye800CW NHS Ester at varying ratios of 0.3, 1, 2, and 4 dyes per antibody (Figure 2A). Analysis of samples on an SDS-PAGE gel, shown in Figure 2B, confirmed that the absence of free dye in each dye:antibody conjugate. To determine if increasing dye to protein ratios had any impact on binding, an ELISA was conducted. The results from this assay in Figure 2C showed that there were no significant differences in binding of the conjugates to recombinant MUC16 regardless of the dye to protein ratio. This is an important consideration, especially since the methodology used for conjugation herein was non-specific, and dyes could be conjugated near antigen-binding regions, thus impacting affinity. The range of dye:antibody ratios was selected based on upper and lower limits of ratios used with other antibody-based probes, both preclinically and clinically.^{14, 20, 29} Fluorescent microscopy and fluorescent Western blotting was conducted, as shown in Supplemental Figure 1, to demonstrate binding of huAR9.6-IRDye800 to pancreatic cancer cells.

Figure 2.

Characterization of huAR9.6-IRDye800 conjugates. (A) Representative absorbance spectra from huAR9.6-IRDye800 at 0.3,1,2, and 4 dyes per protein; absorbance spectra are normalized to huAR9.6. (B) SDS page of huAR9.6-IRDye800 at various dye: protein ratios. (C) Binding of huAR9.6-IRDye800 conjugates to recombinant MUC16 via ELISA.

Dye:Antibody Ratio Impacts Biodistribution

While increasing dye to protein ratios did not impact binding to recombinant MUC16 protein in vitro, the impact of conjugation ratios in vivo necessitated evaluation. Reports in the literature have suggested that dye:antibody ratios can significantly impact pharmacokinetics and distribution of fluorescently labeled IgGs.^30–32 Most importantly, these reports have shown that liver signal is significantly increased with dye:antibody ratios >1. The liver is a key background organ for imaging pancreatic cancer, as well as a site of frequent metastasis.³³ Thus, it is critical to keep background signal in the liver as low as possible in order to minimize interference with primary tumor detection, and to assist in potentially identifying metastatic lesions. To optimize the dye:antibody ratio for huAR9.6-IRDye800, the overall biodistribution of a no-injection control was compared to antibody equivalent (50 μg) injections of 0.3, 1, 2, and 4 dyes per antibody after 48 hours of circulation. The quantified mean fluorescent signal in each resected organ in Figure 3B shows a significant increase in liver signal with increasing dye to protein ratios. Representative images of the necropsied organs shown in Figure 3A highlight the increase in liver signal observed with increasing dye:antibody ratios, consistent with previous reports. However, minimizing the impact of conjugate ratio on biodistribution must be considered in conjunction with sufficient signal for tumor detection on imaging systems. Other reports have shown sufficient detection with 1:1 dye:antibody ratio.³⁴ Resected clearance organs were compared under multiple imaging systems to evaluate the variability in signal with different dye to protein ratios. Current dye:antibody ratios used in clinical trials for IRDye800 agents were also considered in the evaluation of huAR9.6-IRDye800. The majority of dye:antibody ratios for IRDye800-conjugated antibodies used in clinical trials ranged from 1 to 2.^{14, 20, 29} In order to minimize the impact of dye:antibody ratios on biodistribution, retain sufficient signal for deep seated tumors, and maintain consistency with current clinical trials, a dye:antibody ratio of 1:1 was chosen for further investigation in this study.

Figure 3.

(A) Representative images from necropsied organs of tumor-naíve mice 48 hours after injection of control (no injection), or 50 μg or 0.3, 1, 2, or 4 dye:antibody huAR9.6-IRDye800 (N=5). (B) Quantified signal from resected organs.

MUC16-Expressing Xenografts Demonstrate Dynamic Contrast Enhancement

To evaluate the tumor accumulation of huAR9.6-IRDye800 after injection, a dynamic contrast enhancement time course study was conducted. Three subcutaneous PDAC xenograft models were selected based on the in vitro expression of MUC16 as shown in Supplemental Figure 2A. Supplemental Figure 2B shows immunohistochemistry staining of MUC16 expression in each of the tumor models used. Panc1 cells do not express MUC16 in vitro and served as a biomarker negative control. This control was used to assess the impact of biomarker expression on tumor accumulation of huAR9.6-IRDye800. T3M4 and Colo357 served as the MUC16-expressing models. These cell lines were chosen in tandem because of their differential expression of MUC16. Colo357 cells expressed high levels of MUC16 in vitro, and T3M4 cells expressed moderate levels of MUC16. Mice were imaged for 9 days after the injection of huAR9.6-IRDye800 on the LI-COR Pearl Trilogy. Representative images from each of the tumor models over time are shown in Figure 4. Tumor enhancement from the MUC16-expressing tumor cell lines were consistent with the enhancement from muAR9.6-IRDye800, with increasing tumor to background ratios observed over time.²¹ The biomarker control model, Panc1 (Figure 4C), demonstrated significantly lower tumor to background ratios from 24 h after injection until 9 days after injection as compared to both MUC16-expressing tumor models (Figure 4D). Current recommendations for the development of optical imaging probes suggest that a TBR of >3.0 in preclinical studies is a sufficient metric to warrant further investigation.²² Tumor to background ratios with huAR9.6-IRDye800 were greater than 3.0 from 24 hours through 9 days, with peak TBRs reaching 6.95 ± 0.39 and 7.72 ± 1.96 for T3M4 tumors and Colo357 tumors respectively at 9 days post-injection. Comparatively, Panc1 MUC16-negative tumors had maximum TBRs of 1.93 ± 0.53 at 9 days post-injection. The significant differences in TBRs observed in MUC16-expressing compared to MUC16 -negative tumor models demonstrated the specificity of huAR9.6-IRDye800 for MUC16. Interestingly, MUC16 moderate-expressing (T3M4) and MUC16 high-expressing (Colo357) models did not have significantly different TBRs at most time points, which suggests that TBR cannot be solely attributed to differences in biomarker expression. Rather, factors such as tumor size and vascularity likely play a role in tumor to background ratio as well, as has been suggested in other FGS studies.³⁵ At 9 days post-injection, tumors and key clearance organs were resected. Representative images from each group and quantified signals from resected organs are depicted in Supplemental Figure 3. At 9 days post-injection, tumor signal in MUC16-expressing tumors (Colo357 and T3M4) was significantly higher than in all resected clearance organs (p<0.0001). Conversely, Panc1 tumors, which do not express MUC16, did not have significantly higher tumor signal compared to resected clearance organs. The optimal imaging time window for a given FGS agent should optimize high tumor to background ratios, low background signal, and sufficient overall signal for detection of lesions on multiple imaging systems.³⁶ Based on the dynamic contrast enhancement study, and current clinical trial procedures, a range of 3-6 days post-injection was selected as an optimal imaging window for further studies.

Figure 4.

Dynamic contrast enhancement in mice bearing (A) MUC16 high-expressing Colo357 subcutaneous tumors (B) MUC16 moderate-expressing T3M4 subcutaneous tumors and (C) MUC16-negative Panc1 subcutaneous tumors. N=5. (D) Tumor to background ratios in Colo357, T3M4, and Panc1 subcutaneous tumors over 9 days after injection. *p<0.05 between Panc1 either MUC16+ tumor type for Day 1 and thereafter.

Fluorescence-Guided Surgery

Translating fluorescent probes from preclinical studies into the clinic can pose many challenges. In pancreatic cancer, the depth of disease, tumor heterogeneity, and dense desmoplasia can complicate FGS and intraoperative tumor detection, but these disease characteristics are not frequently represented in preclinical models. A major weakness in preclinical testing of FGS probes is the overreliance on simplified xenograft models that have overwhelmingly high expression of the biomarker of interest.³⁷ While relying on such models is useful to show specificity of the targeting moiety, these models fail to recapitulate some of the complexities of tumor heterogeneity and surgical imaging that arise during clinical translation. Translation of FGS probes into the clinic introduces massive variation in tumors, including fluctuation in biomarker expression, degree of stroma, tumor vascularity, and disease localization. The use of patient-derived xenografts in preclinical studies may improve preclinical appropriation of clinical FGS by more closely recapitulating the tumor microenvironment as compared to single high biomarker expressing cell line xenografts. Patient-derived xenograft model J000115419 was selected based on screening expression of MUC16, diagnosis as pancreatic ductal adenocarcinoma, and specimen collection during surgical resection. Because this sample was surgically resected, this represented the key patient population in our study – patients eligible for surgical resection who could benefit from FGS. This tumor specimen was propagated in female NSG mice. Mice received injection of the contrast agent after tumors reached 200-300 mm³ in size.

At 72 h post-injection, mice were imaged on the Fluobeam, Curadel, Spectropen, and LI-COR imaging systems.³⁸ Fluorescence localization was observed in subcutaneous tumors on all imaging systems, depicted by representative Curadel images in Figure 5A (N=3). Spectral analysis showed significantly higher fluorescent signal in the tumors as compared to pancreas (p<0.0001) as shown in Figure 5B, and represented by LI-COR images in Figure 5C. Fluorescence localization was also observed microscopically in frozen tumor sections, shown in Figure 5D. To analyze the tumor microenvironment and intratumoral distribution of huAR9.6-IRDye800, tumors were sectioned, stained, and analyzed by a blinded, board-certified pathologist (G.A.T). Tumor blocks were scanned on an Odyssey M slide scanner at 800 nm, shown in Figure 6A, and depicted homogenous pockets of huAR9.6-IRDye800 localization throughout the tumor. Tumors were diagnosed as moderately differentiated invasive pancreatic ductal adenocarcinoma. Tumor differentiation was homogenous throughout, and there were no regions of necrosis, as shown by H&E stain in Figure 6B. OC125, the gold standard antibody for detection of MUC16, showed 26-50% of cells positive for MUC16 (Figure 6C, Supplemental Figure 4). High expression was observed in secretions with 3+ staining intensity, and weaker membranous staining was also observed with 1+ staining intensity. The pattern of distribution of MUC16 was consistent with pattern of fluorescence localization observed on 800 nm scan of the tumor block. Masson’s Trichrome stain was used to identify collagen within the tumor (Figure 6D). Wisps of collagen comprised approximately 5% of the total tumor mass, with homogenous localization around tumor glands. α Smooth muscle actin (SMA) stain (a myofibroblast marker), showed expression consistent with the patterns observed in Masson’s Trichrome staining, with expression observed around glands, making up about 5% of the total tumor mass, suggesting that the stroma present was largely fibroblast-derived (Figure 6E). CD31 (an endothelial marker commonly used to identify angiogenesis in tumors) showed minimal to mild angiogenesis throughout the tumor, mirroring the hypovasculature characteristic of patients with PDAC (Figure 6F). Ki67 staining showed variable expression of tumor cell proliferation with lower-expressing regions containing 50% positive cells, and higher regions with 80% positive cells (Figure 6G).

Figure 5.

Fluorescence-Guided Surgery in a patient-derived xenograft model with huAR9.6-IRDye800. (A) Representative images from Curadel-guided resection. (B) Comparative fluorescent signal from tumor and pancreas (C) Representative LI-COR images of tumor and pancreas signal. (D) Microscopic fluorescence localization of huAR9.6-IRDye800. (N=3).

Figure 6.

Histological Analysis of Patient-Derived Xenografts. (A) 800 nm signal in PDX tumor (1X magnification) (B) H&E staining (C) OC125 staining for MUC16 expression (D) Masson’s Trichrome staining to identify collagen expression (E) αSMA staining of myofibroblasts (F) CD31 staining of vascular proliferation (G) Ki67 staining of tumor proliferating cells (Images in B-G captured at 20X magnification).

Biodistribution

To assess the biodistribution of huAR9.6-IRDye800, male and female CD-1 mice were randomly assigned to a 24-hour, 14-day, or 28-day time point. 3 male and 3 female mice at each timepoint were injected with either saline (control), a low dose (20 μg) of huAR9.6-IRDye800, or a high dose (80 μg) of huAR9.6-IRDye800 via tail vein. Doses were selected based on human equivalent doses (calculated from body surface area) of ~5 mg and ~20 mg respectively.³⁹ These doses captured a range of similar antibody-based fluorescent agents used in early phase clinical trials.^{20, 40} At the assigned time point after injection of the agent, mice were euthanized, and necropsies were conducted. Organs were imaged on the LI-COR Pearl, using the white and 800 nm channels. Mean fluorescent signal (mean NIR signal per pixel) was calculated for each organ. Representative images of necropsied organs from each of group at 1, 14, and 28 days post-injection are shown in Supplemental Figure 5. Primary clearance and distribution were observed within the liver. At 1 day after injection, mean liver signal in both the low dose (0.0320 ± 0.0099) and high dose (0.1582 ± 0.0298) huAR9.6-IRDye800 groups was significantly higher (p <0.0001) than the control group (0.0014 ±0.0004) as shown in Figure 7A. While the primary clearance route for unconjugated IRDye800 is renal, it is expected to see primary clearance through the liver with antibody conjugates because the large size largely prohibits glomerular filtration.⁴¹ No significant differences in biodistribution were observed in the high dose group. By day 14, signal in the liver decreased by >10 fold in both the high and low dose huAR9.6-IRDye800 groups, with mean liver signals of 0.0102 ± 0.0043 and 0.00339 ± 0.0009 respectively (Figure 7B). Signal in all organs was diminished at 14 days post-injection. By 28 days, little to no signal remained in any organs for either the high or low dose group, demonstrating that the conjugate had been cleared. While the mean quantified signal still showed significant differences in several organs at 28 days, all values were near zero (Figure 7C).

Figure 7.

Organ biodistribution at (A) 1 day post-injection, (B) 14 days post-injection, (C) and 28 days post-injection. (N = 6) *p<0.05

Toxicology and Pathology

To monitor for signs of toxicity, animal weights were monitored for the 14-day and 28-day post-injection timepoints. No significant changes in weight were observed for male or female mice over the 14-day and 28-day time points, as depicted in Supplemental Figure 6, indicating that huAR9.6-IRDye800 was well tolerated. At the indicated timepoint of 24 hours, 14 days, or 28 days, organs resected at euthanasia were formalin-fixed and paraffin-embedded, sectioned, and stained with H&E. Slides were examined by a board-certified pathologist (G.A.T), who was blinded to treatment groups and study results, for any signs of toxicity. Results from the 1-day timepoint showed that no significant toxicity was observed in any group, as summarized in Table 1. Minimal and insignificant histological changes were observed in minor lobular inflammation in the liver, but this was also observed in the control saline group. Mild vacuolar changes and mild tubular dilation were observed in the kidneys of one subject in the saline group and high dose group respectively, but again changes were not classified as significant. At 14 days, no significant toxicity was observed in any group (Table 1). Several mice had minimal and insignificant focal lobular inflammation in the liver consistent with observations 1-day post-injection, but this was also observed in the saline control group as well, and thus is not likely to be attributed to huAR9.6-IRDye800. At 28 days, results were consistent with earlier timepoints shown in Table 1. There were several cases of minor focal lobular inflammation in the liver, but this was observed in all groups including the saline control. Several cases had minor patches of tubular dilation in kidneys. Histological changes observed at all three timepoints in the liver and kidneys were considered minimal and insignificant, and changes were consistent between the saline, low dose, and high dose groups, and are thus likely not attributed to the injection of huAR9.6-IRDye800. The absence of signs of toxicity in response to huAR9.6 at 1 day, 14 days, and 28 days post-injection provides compelling preliminary evidence that huAR9.6-IRDye800 is a safe, non-toxic probe.

Table 1.

Histological evaluation of key clearance organs necropsied 1, 14, and 28 days post-injection. Score: 0 = No histological changes, 1 = minimal and not significant histological changes, 2 = mild histological changes, 3 = moderate histological changes, 4 = severe histological changes

		DAY 1					DAY 14					DAY 28
	Score:	0	1	2	3	4	0	1	2	3	4	0	1	2	3	4
CONTROL	Heart	100%	-	-	-	-	100%	-	-	-	-	100%	-	-	-	-
	Lungs	100%	-	-	-	-	100%	-	-	-	-	100%	-	-	-	-
	Spleen	100%	-	-	-	-	100%	-	-	-	-	100%	-	-	-	-
	Kidney	83.3%	16.67%	-	-	-	83.3%	16.67%	-	-	-	100%	-	-	-	-
	Liver	83.3%	16.67%	-	-	-	66.7%	33.3%	-	-	-	50%	50%	-	-	-
LOW DOSE	Heart	100%	-	-	-	-	100%	-	-	-	-	100%	-	-	-	-
	Lungs	100%	-	-	-	-	100%	-	-	-	-	100%	-	-	-	-
	Spleen	100%	-	-	-	-	100%	-	-	-	-	100%	-	-	-	-
	Kidney	100%	-	-	-	-	100%	-	-	-	-	87.5%	12.5%	-	-	-
	Liver	100%	-	-	-	-	62.5%	37.5%	-	-	-	87.5%	12.5%	-	-	-
HIGH DOSE	Heart	100%	-	-	-	-	100%	-	-	-	-	100%	-	-	-	-
	Lungs	100%	-	-	-	-	100%	-	-	-	-	100%	-	-	-	-
	Spleen	100%	-	-	-	-	100%	-	-	-	-	100%	-	-	-	-
	Kidney	83.3%	16.67%	-	-	-	100%%	-	-	-	-	50%	50%	-	-	-
	Liver	83.3%	16.67%	-	-	-	83.3%	16.67%	-	-	-	66.7%	33.3%	-	-	-

DISCUSSION

The objective of this study was to develop a humanized MUC16-targeted fluorescent antibody conjugate, huAR9.6-IRDye800, demonstrate its specificity toward MUC16, and evaluate efficacy in detecting PDAC intraoperatively, to further progress towards clinical translation. Because of the critical and potentially curative role of surgery and the limited number of fluorescent probes under development for FGS in pancreatic cancer, the goal of this work was to focus on the development of an FGS agent towards a novel target. MUC16 is a compelling potential biomarker for FGS. It is highly expressed among primary tumors in PDAC patients, in metastatic lesions common to pancreatic cancer, and in precursor (PANIN) lesions. It has little to no expression in benign conditions such as pancreatitis, which is critical for differentiating malignant disease, and reducing potential background signal from peritumoral inflammation.⁴² Preliminary evidence in our laboratory (currently unpublished) has demonstrated that MUC16 expression is conserved after neoadjuvant therapy.

Expression of this protein on the surface of cells affords ease of access for targeting, and limited native expression provides a favorable biodistribution profile for minimized background signal for imaging. MUC16 is not expressed in the normal pancreas and is not expressed in critical background organs throughout the peritoneal cavity.⁴³ MUC16 is also expressed in many other malignancies, and thus has potential applications as a tumor agnostic biomarker for targeted FGS. It has been well established as a serum biomarker in ovarian cancer,⁴⁴ suggesting avenues for development of blood-based biomarkers for detection, patient stratification, and disease monitoring. These characteristics of MUC16 align with many of the reported criteria for development of successful FGS probes, and targeted agents.^45–48

It is necessary to address the potential challenges of targeting MUC16 as this conjugate continues preclinical development. MUC16 can be cleaved from the cell and circulate within the blood stream. While this can be advantageous as a blood biomarker, this may pose a challenge for imaging. High levels of circulating MUC16 in the serum could bind the fluorescent antibody conjugate and reduce antibody available to bind tumors. This could cause high background and reduce tumor contrast. Current mouse models of tumors tend to be poor predictors of this phenomenon, as many studies have found that these tumor models fail to recapitulate circulating antigen.⁴⁹ However, there are many preclinical and clinical studies that suggest cleaved antigen is of minimal concern. Preclinical studies have shown successful imaging in the context of many cleaved antigens.^{40, 50–52} Furthermore, the clinical success of SGM-101 as an imaging agent for FGS addresses this criticism. CEA, the target of SGM-101, can also be cleaved. However, quantification of fluorescent antibody bound to circulating antigen in patients after injection of the contrast agent showed only 3% of the total injected dose was sequestered in circulation, resulting in minimal impact on the tumor contrast.⁵³ Given the large body of work demonstrating successful imaging of cleavable antigens, we do not anticipate that this parameter will impede the efficacy of huAR9.6-IRDye800.

One of the critical aspects of FGS for PDAC not addressed within this project is the detection of metastases, including positive lymph nodes. Further studies are needed to investigate the minimum size of lesions that can be detected with FGS, since detection of occult metastases may alter the patient’s course of treatment and they can be missed during preoperative imaging. Additionally, while MUC16 has little to no expression in pancreatitis, the effects of benign inflammatory conditions on detecting malignancies should be investigated. Finally, clinically available systems should be integrated to determine the detection capabilities for huAR9.6-IRDye800.

Overall, the results of this study support huAR9.6-IRDye800 as a potential agent for further development for FGS. Data showed that huAR9.6 retained its binding properties in comparison to the murine variant. Dye to protein ratios were optimized in this study and highlighted the impact of increasing dye ratios on liver accumulation. There is currently a lack of consensus in the literature on the optimal labeling ratio for near infrared dyes, but the results of the data shown herein, as well as in previous reports, highlight the impact of the degree of labeling on biodistribution.³⁰ Dye to protein ratios of 1:1 had minimal impact on distribution. The liver is both a key background organ in pancreatic cancer, as well as a site of frequent metastasis. Thus, it is paramount to diminish liver signal, and keeping dye to protein ratios low can assist in reducing background liver signal. However, this study did not investigate the detection of metastatic lesions, and so it has yet to be determined if liver lesions are identifiable even at lower dye to protein ratios used here. Furthermore, this dye to protein ratio needs to be assessed in tumor-bearing models on clinically used imaging systems to ensure that this ratio allows for adequate detection.

In vivo studies in subcutaneous tumor models showed high tumor to background ratios in MUC16 expressing tumors, and low tumor to background ratios in tumors that did not express MUC16. The results of this study confirm that fluorescence localization is due to the specificity of the probe for MUC16, rather than accumulation due to the enhanced permeability and retention effect. However, it is important to note that these results are likely not entirely simplified to the presence or absence of MUC16. Considerations for differential vascularity and tumor size likely have a role in tumor to background ratio, as evidenced by several clinical studies.³⁵ However, these factors were not explored at length in this study, and thus should be investigated in the future.

Fluorescence-guided surgery conducted in patient-derived xenograft tumors demonstrated that huAR9.6-IRDye800 could successfully delineate a tumor model that recapitulated many characteristics of tumors observed in patients in the clinic. Cell line-derived tumor models are notoriously aggressive and poorly differentiated. Comparatively, the patient-derived xenograft model was moderately differentiated, which more accurately represented the grade of tumors frequently eligible for surgical resection. Furthermore, the patient-derived xenografts used herein exhibited hypovascularity, which mirrors the hypoxic environment characteristic of most PDACs. Stroma was present in this tumor as well. Despite the low degree of vascularity and presence of stroma, huAR9.6-IRDye800 still accumulated homogenously throughout the tumor, and illuminated the tumor for surgical resection.

However, while this PDX model more accurately recapitulates the tumor microenvironment of PDAC than cell line-derived tumor models, this model has limitations as well. Previous studies have shown that humanized antibodies exhibit anomalous biodistribution in severely immunodeficient mice, such as NSG mice used herein, which can decrease tumor uptake, and increase background signal in the liver, spleen, and bone.⁵⁴ This pattern of altered biodistribution was also observed here, with increased peritoneal signal. Tumor contrast was still very bright but may have been diminished due to altered biodistribution patterns. Previous reports have attributed altered biodistribution of IgGs in NSG mice to the lack of endogenous IgG, as well as interactions between FcγR present in myeloid cells in the spleen, liver, and bone.^{55, 56} While in this study, we were able to still demonstrate fluorescence localization in the context of these deterrents, this is an important consideration for investigation of this agent in future studies, as this may impact biodistribution and background signal in an orthotopic model. To address this, a preloading dose of antibody could be delivered to occupy FcRn receptors and minimize alterations to biodistribution, or an antibody fragment that lacks the Fc region of the antibody could be utilized. Herein, our main priority was to address localization of our conjugate in the context of PDAC microenvironment, which we were able to confirm. In addition, while our previous work with the murine variant of this conjugate demonstrated feasibility of tumor delineation in an orthotopic model,²¹ further studies need to be conducted with huAR.6-IRDye800 to evaluate the specific surgical feasibility in an orthotopic model to confirm favorable tumor to background ratios within the peritoneal cavity. Furthermore, analysis of one PDX tumor provides a snapshot of clinically replicative tumors. Assessing distribution and accretion of huAR9.6-IRDye800 across multiple PDX models with varying levels of MUC16 expression and differential microenvironments may help to strengthen the evidence for success of huAR9.6-IRDye800.

This study further demonstrated compelling evidence of a non-toxic safety profile for huAR9.6-IRDye800, and provides data to support clinical translation efforts for this fluorescent conjugate. IRDye800 alone has been well documented for its safety across vigorous preclinical and clinical studies. However, humanized AR9.6 is a new molecular entity, and as with any combination product, huAR9.6-IRDye800 must undergo rigorous evaluation before clinical translation. Among these qualifications for further investigation are in-depth analysis of the safety, toxicity, and biodistribution profiles. To that end, this study was conducted to lay the groundwork and provide preliminary evidence of the safety profile of huAR9.6-IRDye800 to support further investigation and translation.

Since the ultimate goal of this study was to produce preliminary evidence, doses for safety evaluation were chosen to directly reflect a range of doses used for similar agents in clinical trials and several preclinical studies.^{20, 40} In future studies, multiple human equivalent doses will be assessed, either 10-fold or 100-fold greater than the anticipated human dose. It is important to note that while preliminary evidence has demonstrated that AR9.6 binds to murine MUC16, further evidence is needed to support this, and to determine if there are differences in binding human as compared to murine MUC16. This is an important consideration, because interaction with native MUC16 could impact biodistribution, as could differential expression of MUC16 in mice and humans.⁵⁷ Results from the toxicity study provided strong support for the safety of huAR9.6-IRDye800, evidenced by the lack of significant histological changes in organ pathology. However, this parameter should also be investigated at higher dosing ranges and include a wider array of organs for analysis in future studies.

IRDye800 has undergone extensive analysis, and demonstrated a safe, non-toxic profile. This NIR dye is currently used in over 30 clinical trials, none of which have been withdrawn due to safety issues.⁵⁸ The majority of patients who receive antibody conjugates for FGS do not experience any serious adverse effects, regardless of the parent antibody used. However, most antibodies currently under investigation for FGS repurpose an FDA-approved antibody as a targeting moiety. huAR9.6 will likely face additional regulatory hurdles because this is not yet an approved or widely adopted antibody. This further emphasizes the critical importance of thorough safety and toxicity evaluation.

CONCLUSIONS

Herein, we synthesized a NIR fluorescently-labelled humanized MUCIN16-targeted contrast agent for FGS. We showed that humanized AR9.6 mAb has equivalent affinity as our previously reported murine AR9.6 mAb and that conjugation to the NIR dye IRDye800 does not impact binding. Moreover, the results of this study demonstrated that huAR9.6-IRDye800 is specific for MUC16, and is a safe, non-toxic probe, and can effectively illuminate tumors in translational cancer models. Data presented here demonstrates compelling evidence of the efficacy of huAR9.6-IRDye800 and necessitates further investigation for clinical translation.