Specific Targeting of Somatostatin Receptor Subtype-2 for Fluorescence-Guided Surgery Free

Neuroendocrine tumors (NET) are slow-growing tumors that produce significant morbidity and eventual mortality. The diagnosis of NETs has increased nearly 5-fold over the last three decades (1) and is predicted to continue rising at a faster rate than other common cancers such as breast and lung cancer (2). Similar to most cancers, surgery remains the only curative treatment option for patients with NETs. Unique to this patient population, however, is that surgery is indicated not only for localized lesions, but also for advanced tumors to control excessive hormone production by debulking the metastatic disease burden (3). Thus, surgery can improve overall survival in both NET patients with localized disease and those with metastatic disease who comprise more than 50% of clinical cases. Although surgery can be curative if all cancer cells are removed, patients with NETs have a 5-year recurrence rate of 94% (4). Improved identification and resection of these lesions would have a major impact on surgical efficacy and outcomes for NET patients.

In recent years, intraoperative tools that enable real-time tumor visualization have emerged. One approach, known as fluorescence-guided surgery (FGS), can help surgeons identify tumor margins with higher accuracy, which is crucial for achieving complete tumor resection and minimizing local recurrence (5). This may also reduce the need for additional surgeries and can limit removal of healthy tissues. In NETs, intraoperative imaging could also increase tumor detection rates, enhance primary tumor localization for enucleation, and identify small lesions and locoregional disease with lymph node involvement to improve surgical outcomes. Despite the growing number of fluorescent contrast agents entering clinical trials, validated intraoperative imaging agents are currently lacking for this patient population. Conversely, radiolabeled somatostatin analogs have a long history of use in nuclear medicine and PET, and are considered to be the gold standard for noninvasive imaging of NETs (6–11). Moreover, radiolabeled somatostatin analogs are currently used preoperatively for surgical planning and postoperatively for surveillance. In order to extend the utility of PET into the operating room, our group developed the first bioactive fluorescent analog of a clinical radiotracer using 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid-Tyr³-octreotide (⁶⁸Ga-DOTA-TOC) as a model (12). We introduced a synthesis scheme that enabled conversion of ⁶⁸Ga-DOTA-TOC into a dual-labeled counterpart, ⁶⁸Ga-MMC(IR800)-TOC, where MMC represents a multimodality chelator that replaces the standard chelating agent, DOTA (Fig. 1). In vitro studies demonstrated excellent retention of somatostatin receptor subtype-2 (SSTR2) binding and specificity after dye conjugation, and biodistribution in healthy mice identified the kidneys as the major excretion organ.

In the present study, we examine the tumor-targeting properties of Ga-MMC(IR800)-TOC in SSTR2-expressing xenografts to determine effectiveness for FGS. Near-infrared fluorescence (NIRF) imaging was performed to determine the imaging time point that produces optimal tumor contrast, and radiolabeled analogs were prepared to quantitatively measure tracer biodistribution. Tumor specificity was examined by in vivo NIRF imaging, followed by ex vivo NIRF imaging of tissues and frozen sections. To determine the translational utility of Ga-MMC(IR800)-TOC, binding was evaluated in surgical biospecimens from patients with pancreatic NETs. Specificity for tumor tissue and SSTR2 were determined by histology and confocal microscopy, respectively.

All chemicals were purchased from Sigma-Aldrich unless otherwise noted. Reversed-phase high-performance liquid chromatography (HPLC) was performed on an analytical Hitachi LaChrom system using a Kinetex C18 column (2.6 μm; Phenomenex) with a mobile phase of A = 0.1% TFA in H₂O, B = 0.1% TFA in CH₃CN (gradient: 0 minute, 10% B; 12 minutes, 90% B); flow rate, 1 mL/min. Radiochemical purities of ≥95% were assessed by radio-HPLC using an in-line radioactive detector (Berthold Technologies). Electrospray ionization mass spectra were acquired on a LCQ FLEET instrument (Thermo Scientific).

Radiolabeling was performed using a cation exchange cartridge as previously described (12). Following Sep-Pak Light C18 (Waters) purification, the product was diluted with PBS and analyzed by radio-HPLC. ⁶⁷Ga-citrate was purchased from a radiopharmacy (Cardinal Health) and was added to an equal volume of 0.1 mol/L HCl to produce ⁶⁷GaCl₃. The radioactive solution was then processed identically to generator-produced ⁶⁸Ga with the exception of using 0.5 mol/L ammonium acetate (pH 4.5) as the reaction buffer. Cold Ga labeling was performed according to methods established above for the radiolabeled compounds. MMC(IR800)-TOC (60 nmol) was mixed with a 4-fold molar excess of GaCl₃ and heated at 95°C for 15 minutes. The crude mixture was purified by ultrafiltration, and the pure product was characterized by HPLC and mass spectrometry.

Athymic female nu/nu mice (Charles River Laboratories) were housed under standards of the Institutional Animal Care and Use Committee of the University of Texas Health Science Center at Houston and maintained on normal rodent chow. HCT116-SSTR2 (courtesy of Dr. Buck Rogers, Washington University in St. Louis, and Dr. Carolyn J. Anderson, University of Pittsburgh), HCT116-WT (ATCC), and NCI-H69 (ATCC) cell lines were cultured in Roswell Park Memorial Institute 1640 medium with 10% (v/v) FBS and maintained at 37°C with 95% humidity and 5% CO₂ atmosphere. HCT116-SSTR2 cells were additionally supplemented with 100 μg/mL Zeocin (Gibco). For all procedures, mice were anesthetized with 1% to 2% isoflurane. For xenografting, 6- to 8-week-old mice were s.c. injected with 1 × 10⁶ HCT116-SSTR2, 1 × 10⁶ HCT116-WT, or 5 × 10⁶ NCI-H69 cells in Matrigel (Corning):PBS (1:1) in the shoulder. Studies were conducted 3 to 4 weeks after implantation when tumor size reached approximately 5 to 10 mm maximum diameter. Overdose of anesthesia followed by cervical dislocation was the method of euthanasia for mice in terminal studies.

Radioactive uptake studies were performed as previously described (12). Briefly, HCT116-SSTR2, NCI-H69, and non–SSTR2-expressing HCT116-WT cells (200,000 cells/well) were incubated with a 10 nmol/L solution of ⁶⁸Ga-MMC(IR800)-TOC, ⁶⁸Ga-DOTA-TOC, or a mixture of radiotracer with a 100-fold excess of octreotide at 37°C for 1 hour. After incubation and washing, cells were collected and radioactivity was quantified in a Wizard² automated γ counter (Perkin Elmer) to determine uptake as percentage of total radioactivity added.

Fluorescence imaging of Ga-MMC(IR800)-TOC cellular uptake was performed as previously described (12). Briefly, HCT116-SSTR2 and HCT116-WT cells were plated (50,000 cells/well), allowed to attach for 48 hours, and incubated with 5 μmol/L Ga-MMC(IR800)-TOC for 1 hour at 37°C. NCI-H69 cells were identically processed with the exception of being in suspension. After the incubation period, cells were washed with media 2 times for 5 minutes. Subsequently, cells were fixed. NCI-H69 cells were immobilized onto Shi-fix coverslips (Everest Biotech) following the manufacturer's instructions. Cells were mounted with Mowiol mounting medium containing Nucspot Live 488 nuclear stain (diluted 1:1,000 in Mowiol), and microscopic images were acquired on a confocal microscope (SP8, Leica). IR800 was detected using a 730 nm laser, and NucSpot Live 488 was detected using a 488 nm laser with appropriate filter settings.

Athymic female nu/nu mice were i.v. injected with 15 nmol (42.5 μg) of ⁶⁷Ga-MMC(IR800)-TOC. Blood samples were obtained by postmortem exsanguination of the heart 24 hours after injection and centrifuged to isolate the plasma. Ice-cold acetonitrile was then added to an equal volume of the plasma to precipitate plasma proteins. Finally, the supernatant was collected and analyzed by HPLC based on emission in the fluorescent channel with conditions as stated in the General Methods section.

Mice with HCT116-SSTR2 tumors (n = 5) were i.v. injected with Ga-MMC(IR800)-TOC (2 nmol, 5.67 μg), and imaging was performed at 3 and 24 hours after injection. In vivo NIRF images were acquired for 200 ms without background subtraction using a custom-built electron-multiplying charge-coupled device (EMCCD) fluorescence imaging system at λ_ex/λ_em = 785/830 nm (13), and image analysis was performed with the ImageJ software package (NIH). At the conclusion of the imaging studies, the mice were euthanized and selected organs were excised and underwent ex vivo NIRF imaging using an IVIS Lumina II (Perkin Elmer). To demonstrate feasibility of imaging with a clinically approved system, mice bearing HCT116-SSTR2 tumors were i.v. injected with Ga-MMC(IR800)-TOC (5 nmol, 14.2 μg) and imaged using the Firefly fluorescence imaging system on the da Vinci surgical robot (λ_ex = 805 nm; Intuitive). NIRF images were acquired under ambient room lighting at 3 and 48 hours after injection.

Mice with HCT116-SSTR2 tumors were i.v. injected with 370 kBq (10 μCi, 2 nmol) of ⁶⁸Ga-MMC(IR800)-TOC or ⁶⁷Ga-MMC(IR800)-TOC and euthanized at 3, 24, or 48 hours after injection. Selected tissues were excised and underwent ex vivo optical imaging on the IVIS with the following settings: lamp level (high), excitation (745 nm), emission (ICG), epi-illumination; binning (S); FOV (C, 10); f-stop (2); acquisition time (1 second). Region-of-interest (ROI) analysis was performed with the vendor software package (Living Image) to obtain tumor-to-background ratios (TBR). Parameters were the same for all acquired images. At the completion of the optical imaging studies, tissues were weighed and counted for radioactivity using the γ counter. The total injected activity per mouse was determined from an aliquot of the injected solutions. The results are expressed as a percentage of the injected activity per gram of tissue (%IA/g) and represent the mean ± SD of n = 5 mice/time point. To demonstrate SSTR2 binding in an animal model with endogenous receptor expression, xenografts were prepared using NCI-H69 human small cell lung cancer cells, and biodistribution was performed at 48 hours after injection.

For specificity studies, Ga-MMC(IR800)-TOC (2 nmol) uptake in HCT116-SSTR2 was compared with SSTR2-negative HCT116-WT xenografts by in vivo and ex vivo optical imaging at 24 hours after injection. Tumors and key organs (muscle, pancreas, and small intestine) were cryoconserved in optimal cutting temperature and used to prepare frozen sections (10 μm) for mesoscopic and microscopic analysis to localize the IR800 signal within the tissue. Using an Odyssey slide scanner (LiCOR), sections were scanned and fluorescence intensities at 800 nm were quantified on 16-bit images using ImageJ. ROIs were drawn around the outline of each organ, and means ± SD were calculated using GraphPad Prism (n = 3 mice/group). Adjacent sections underwent hematoxylin and eosin (H&E) staining to permit morphologic analysis of the tissue.

To analyze Ga-MMC(IR800)-TOC distribution and receptor specificity at the cellular level, the frozen sections were fixed in 4% cold paraformaldehyde for 10 minutes and embedded in Mowiol mounting medium. For counterstaining, we added NucSpot Live 488 nuclear stain (Biotium) to the mounting medium (1:1,000 dilution directly in Mowiol), and NIRF confocal microscopy was performed.

In order to evaluate the binding and specificity of Ga-MMC(IR800)-TOC in human tumor tissues that express SSTR2, surgical biospecimens of pancreatic NETs and surrounding normal tissue were obtained from the Elkins Pancreas Center at Baylor College of Medicine. The use of tissues was approved by the Institutional Review Boards of Baylor College of Medicine and The University of Texas Health Science Center. All patients gave written-informed consent. Biospecimens (n = 5) were fresh frozen and used to prepare 4 μm frozen sections for ex vivo staining with Ga-MMC(IR800)-TOC. Frozen sections were thawed at room temperature for 15 minutes and briefly wetted with PBS. A 5 μmol/L solution of Ga-MMC(IR800)-TOC was added onto the slide and incubated for 45 minutes at 37°C. Slides were then washed twice for 3 minutes with PBS. Fixation was performed in 4% cold paraformaldehyde for 10 minutes, and slides were mounted with Mowiol mounting medium containing Nucspot Live 488 nuclear stain (diluted 1:1,000 in Mowiol). Subsequently, NIRF confocal microscopy was performed. Sections stained with H&E were used for correlation of tissue morphology.

To detect SSTR2 expression, IHC was carried out on frozen sections of xenograft tumor tissue (HCT116-SSTR2, HCT116-WT) and human NET biospecimens using the Discovery XT processor (Ventana Medical Systems) at the Molecular Cytology Core Facility of Memorial Sloan Kettering Cancer Center. After thawing, sections were baked at 50°C for 1 hour, followed by a 30-minute incubation with Background Buster solution (Innovex). Sections were then incubated with the anti-SSTR2 rabbit monoclonal antibody (Abcam) at 2.2 μg/mL for 5 hours, followed by a 1-hour incubation with biotinylated goat anti-rabbit IgG (Vector Labs). For detection, a DAB detection kit (Ventana Medical Systems) was used according to the manufacturer's instructions. Sections were counterstained with H&E and coverslipped with Permount (Fisher Scientific). Slides were digitalized using a MIRAX Slide Scanner (3DHISTECH).

Curve fitting and linear regression calculations were performed with GraphPad Prism (v 5.01). All data are presented as mean ± SD, and group comparisons were performed with two-tailed t tests.

Supplementary Figure S1 shows a representative HPLC trace for ⁶⁸Ga-MMC(IR800)-TOC and ⁶⁷Ga-MMC(IR800)-TOC. Radiochemical yields were 78.2 ± 8.1% and 72.7 ± 4.1% for ⁶⁸Ga-MMC(IR800)-TOC and ⁶⁷Ga-MMC(IR800)-TOC, respectively, and purity of >99% was achieved after Sep-Pak purification. Serum stability studies showed >60% intact radiotracer at 24 hours with minor metabolite formation (Supplementary Fig. S2).

Longitudinal NIRF imaging of Ga-MMC(IR800)-TOC was performed in HCT116-SSTR2 xenografts to determine the pharmacokinetic profile of the agent and subsequent effects on contrast at early (3 hours) and delayed (24 hours) time points. Ga-MMC(IR800)-TOC uptake was observed in the tumor region at 3 hours but was also accompanied by notable background fluorescence in the thoracic/abdominal regions and high kidney signal (Fig. 2A; Supplementary Fig. S3). At 24 hours, a more focal pattern of agent uptake was evident in all tumors, along with lower background signal. ROI analysis revealed similar tumor fluorescence at 3 and 24 hours (5.6 ± 0.9 × 10⁵ and 4.2 ± 1.4 × 10⁵ counts/pixel², respectively, P > 0.05) but showed a 2-fold reduction in background fluorescence at 24 hours (4.8 ± 1.0 × 10⁵ to 2.3 ± 1.0 × 10⁵ counts/pixel², P < 0.005; Fig. 2B), thus identifying clearance from normal tissues as the primary reason for improved tumor visualization with delayed imaging. These findings were in agreement with images acquired from the Firefly system (Supplementary Fig. S4).

MMC(IR800)-TOC was radiolabeled with ⁶⁸Ga or ⁶⁷Ga and injected into HCT116-SSTR2 xenografts for early (3 hours) and delayed (24 and 48 hours) analysis of agent uptake, respectively, by NIRF imaging and gamma counting. Macroscopic evaluation by ex vivo IVIS imaging revealed similar tumor fluorescence at 3, 24, and 48 hours, along with high uptake in the kidneys and moderate uptake in the liver (Fig. 3A). Notable fluorescence was also seen in the lungs at 3 hours, but was progressively reduced to background levels at 48 hours. Signal in other tissues was minimal. Image analysis was performed to obtain semiquantitative measurements of fluorescence intensity (Fig. 3B; Supplementary Table S1). Tumor signal was higher than all nonclearance organs and showed no significant differences between early and delayed time points. Clearance was primarily through the kidneys, which showed high fluorescence throughout. Conversely, fluorescence in all other tissues decreased as a function of time. From 3 to 48 hours, the highest percentage of signal washout in normal tissues occurred in the muscle (77.0%, P < 0.001), small intestine (81.5%, P < 0.001), and lungs (86.8%, P < 0.001). This, in turn, led to a 3.4-, 4.1-, and a 5.4-fold increase in tumor-to-muscle, tumor-to-small intestine, and tumor-to-lung ratios, respectively (P < 0.01; Fig. 3C). Importantly, ratios in key sites of NET formation with endogenous SSTR2 expression (pancreas, small intestine, and lungs; ref. 14) ranged from 4.3 to 17.2, which is critical since a TBR of at least 2 is generally suitable for tumor delineation in the operating room.

Radioactive biodistribution results are summarized in Fig. 3D (Supplementary Table S2). At 3 hours, administration of ⁶⁸Ga-MMC(IR800)-TOC resulted in high renal clearance (45.6 ± 3.8 %IA/g), along with prominent accumulation in the lungs (6.7 ± 0.9 %IA/g), liver (5.0 ± 0.9 %IA/g), and stomach (4.5 ± 0.8 %IA/g). Tumor uptake was notable (3.54 ± 0.85 %IA/g) and higher than pancreas (1.7 ± 0.2 %IA/g) and small intestine (1.3 ± 0.1 %IA/g). At 24 and 48 hours, ⁶⁷Ga-MMC(IR800)-TOC biodistribution showed excellent agent retention in tumors with %IA/g values of 4.3 ± 1.1 and 6.7 ± 0.9, respectively, whereas nearly all nonclearance organs had significantly lower uptake. At 48 hours, signal had decreased by 61.2% in muscle (P < 0.01) and 95.9% in blood (P < 0.001), giving tumor-to-muscle and tumor-to-blood ratios of 34.2 ± 14.0 and 81.0 ± 35.3, respectively, which translated into a 5.7-fold (muscle) and 50.3-fold (blood) increase from values obtained at 3 hours (Fig. 3E). Notably, TBRs improved for the lungs, pancreas, and small intestine to final values of 4.5 ± 0.6 (8.6-fold increase), 5.8 ± 1.3 (2.7-fold increase), and 8.2 ± 1.1 (2.9-fold increase), respectively (P < 0.001).

A major limitation in NET research is the lack of robust orthotopic or patient-derived xenograft animal models that accurately represent human disease. For further agent evaluation, we selected NCI-H69 cells, which are known to endogenously express moderate to low levels of SSTR2 (15) and have been widely used for assessment of SSTR2-targeted agents (16–20). As anticipated, in vitro results demonstrated specific binding and internalization in NCI-H69 cells (Supplementary Figs. S5 and S6) at lower levels than HCT116-SSTR2 cells, but higher than non–SSTR2-expressing HCT116-WT cells. In line with these findings, we observed modest in vivo tumor uptake in NCI-H69 (1.2 ± 0.4 %IA/g; Supplementary Fig. S7 and Supplementary Table S3) compared with HCT116-SSTR2 cells. Nevertheless, low background signal still produced optical contrast ratios that were >2 in the muscle, small intestine, and lung (Supplementary Fig. S7B).

In order to demonstrate SSTR2-mediated targeting in vivo, Ga-MMC(IR800)-TOC uptake was compared in mice bearing HCT116-SSTR2 or HCT116-WT tumors. In vivo, high NIRF signal intensity was observed in SSTR2-overexpressing tumors, whereas wild-type tumors were poorly delineated (Fig. 4A). Examination of resected tissue by ex vivo imaging (IVIS) confirmed in vivo results (Fig. 4B) and showed higher fluorescence in HCT116-SSTR2 tumors (P < 0.01) compared with the wild-type counterparts (Fig. 4C). This, in turn, resulted in TBRs that were 1.8-, 1.8-, 2.3-, and 2.4-fold higher in the muscle, pancreas, small intestine, and lungs of HCT116-SSTR2 mice, respectively (Fig. 4D).

To enable evaluation of receptor specificity at the mesoscopic level, frozen sections of tumors that were injected with 2 nmol Ga-MMC(IR800)-TOC 24 hours prior to necropsy were scanned on a slide scanner and showed localized Ga-MMC(IR800)-TOC uptake only in HCT116-SSTR2 tumors (Fig. 5A). Comparison to H&E staining revealed that uptake was confined to viable areas of the HCT116-SSTR2 tumors. Excellent colocalization was demonstrated between the Ga-MMC(IR800)-TOC signal and SSTR2 expression as determined by anti-SSTR2 IHC staining (Supplementary Fig. S8). Quantification of the Ga-MMC(IR800)-TOC signal showed, on average, 8.6-fold higher fluorescence signal in HCT116-SSTR2 tumors compared with HCT116-WT tumors, whereas signals in the pancreas, muscle, and small intestine were low and similar between the groups (Fig. 5B). Further validation of receptor specificity was shown by confocal microscopy, which revealed Ga-MMC(IR800)-TOC fluorescence in HCT116-SSTR2 tumors but not in HCT116-WT tumors (Fig. 5C). NIRF signal was observed at the cell membrane and intracellularly, in accordance with the SSTR2 staining, and indicates that Ga-MMC(IR800)-TOC was internalized with the receptor upon binding.

A proof-of-principle approach was used to show Ga-MMC(IR800)-TOC binding and specificity in human NETs. To simulate in vivo binding in patients, we developed an ex vivo staining protocol on fresh-frozen sections of biospecimens (n = 4) from pancreatic NET patients that consisted of tumor and surrounding nontumor, pancreatic tissue for each patient. We first confirmed receptor expression via SSTR2 IHC, which clearly delineated tumor from normal tissue in accordance with H&E staining (Fig. 6A). The receptor was most abundant at the cell membranes of tumor cells, whereas the staining was at background levels in nontumor tissue. Staining of unfixed frozen sections with Ga-MMC(IR800)-TOC resulted in active binding to tumor tissue. This produced NIR fluorescence that was localized to the cell membrane and was very low in stromal areas and nontumor tissue, in excellent agreement with SSTR2 IHC (Fig. 6B; Supplementary Fig. S9A). We were also able to confirm binding using an NIRF slide scanner, which showed that the emitted NIRF signal from Ga-MMC(IR800)-TOC was within the tumor boundaries identified by H&E staining and absent in normal tissues, indicating excellent specificity of the agent for human NETs and potential tumor delineation capability (Supplementary Fig. S9B).

Given the importance of achieving maximal fluorescent contrast for surgical applications, the relationship between injected dose and contrast was studied by comparing the effects of low (0.5 nmol) and high (2 nmol) dose administration of Ga-MMC(IR800)-TOC in SSTR2-expressing tumors at 24 hours (Supplementary Fig. S10A). Although ex vivo imaging showed that absolute tumor fluorescence was 3.8-fold higher (P < 0.001) following injection of the higher dose (Supplementary Fig. S10B), fluorescence in nontumor tissues was also elevated and, importantly, resulted in a >6-fold increase in background signal in the pancreas, small intestine, and lungs (P < 0.001). Because signal in background tissues was essentially absent with 0.5 nmol dose, the corresponding contrast ratios were higher for this group and increased by approximately 40% for each tissue compared with the high-dose group (Supplementary Fig. S10C).

Our findings show that a fluorescent somatostatin analog can be used to differentiate tumor from normal tissues based on highly specific targeting of SSTR2. Importantly, we also demonstrate that the MMC is an ideal linker because it produces a bioactive fluorescent analog of ⁶⁸Ga-DOTA-TOC and enables quantitative analysis. FGS agent design approaches from our group and others have benefited immensely from the successful first-in-human application of a targeted fluorescent agent by Van Dam and colleagues (21), which represented a major milestone in the field of FGS and played an important role in the development of new dyes and targeted probes (22), optimization of optical imaging systems, and identification of appropriate metrics and validation methods (23, 24). Translational initiatives for FGS can be further complemented by combining the strengths of nuclear and optical imaging. Clinically, the addition of a radiolabel to a fluorescent agent has the potential to enhance the detection of deep-seated tumors by adding intraoperative gamma probe guidance to existing surgical imaging tools, such as intraoperative ultrasound (25, 26). However, radiolabeling may be even more significant in the preclinical setting, where quantitative validation is critical for developing an expansive translational pipeline of targeted fluorescent contrast agents (27, 28). Instead of applying our strategy to a novel targeting agent or biomarker, we exploited the established clinical effectiveness of radiolabeled somatostatin analogs to develop an FGS counterpart. The major advantage of this approach arises from the well-characterized clinical performance of TOC and IR800, which, when combined with analysis of recent FGS clinical trials, provided an experimental strategy that examined key clinical endpoints such as optimal imaging time points and injection dose, and efficacy for tumor demarcation.

During surgery, differential uptake of an FGS agent in tumor and normal tissues is critical to produce the necessary contrast to identify tumors, margins, residual disease in the wound bed, and regional lymph nodes metastases. One determinant for obtaining maximal contrast is identification of the optimal imaging time point. We previously showed that ⁶⁸Ga-MMC(IR800)-TOC undergoes time-dependent elimination from circulation and normal tissues up to 3 hours after injection (12) and selected this time point for the initial in vivo NIRF imaging studies. Because tumors were not consistently delineated at 3 hours, we hypothesized that low agent accumulation, slow clearance from nontarget sites, or a combination of the two were contributing factors. Therefore, NIRF imaging was repeated at 24 hours and showed improved tumor contrast that was attributed to lower background signal (i.e., washout). Because optical imaging is inherently semiquantitative (29), we applied the radiolabeling utility of the MMC to produce dual-labeled analogs for early (⁶⁸Ga) and delayed (⁶⁷Ga) biodistribution studies that would serve as the gold standard for quantification of tissue uptake. The biodistribution profiles showed that the improved contrast at 24 hours was attributable to persistent tumor signal and clearance from background organs. Biodistribution analysis at 48 hours supported this observation and produced the highest TBRs. Importantly, the quantitative data obtained by gamma counting were identical to the ex vivo optical data and demonstrate the utility of dual-labeling for agent validation.

Image contrast is also dependent upon identifying the optimal injection dose of an FGS agent. Because the final formulation of a radiotracer contains a mixture of unlabeled precursor and the radiolabeled product, administering higher doses could result in competition for binding sites and reduce radioactive signal in tumors. Fluorescent agents, however, generally consist of a single chemical species and do not experience such effects. Thus, FGS agents may benefit from injecting higher doses. This has indeed been shown in a dose-escalation study in patients with IR800-conjugated cetuximab where higher TBRs were obtained with the highest injection dose (30), as well as with a fluorescent diabody targeting an antiprostate stem cell antigen (31). However, numerous studies have reported better TBRs with lower doses because increasing the injection amount also leads to more nontarget uptake that reduces contrast, increases the number of false-positives, and negatively affects specificity (32–35). In order to examine this effect with Ga-MMC(IR800)-TOC, xenografts were injected with 0.5 or 2 nmol and showed a direct correlation between injected dose and absolute fluorescence in tumor and normal tissues. Although the 2 nmol dose produced a nearly 4-fold increase in tumor fluorescence, the TBRs were all higher in the low-dose group. This is likely attributable to nonspecific binding of Ga-MMC(IR800)-TOC that is amplified at higher doses. Given the excellent tumor targeting of Ga-MMC(IR800)-TOC and the availability of NIRF imaging systems that can detect microdose levels of dye, further dose optimization studies could identify maximal TBRs that reduce false-positives and enhance diagnostic accuracy.

The efficacy of an FGS agent has been defined by FDA guidance documents as the ability to locate and outline normal structures or distinguish between normal and abnormal anatomy (24). Therefore, efficacy is dependent upon the specificity of the agent for its target and its overexpression in tumor tissue. We previously showed SSTR2-mediated uptake of ⁶⁴Cu-MMC(IR800)-TOC in vitro and in vivo, where competition with octreotide strongly reduced uptake in SSTR2-expressing tumors (36). In a subsequent study with ⁶⁸Ga-MMC(IR800)-TOC, we demonstrated selective in vitro uptake by SSTR2-expressing cells that was 20-fold higher than cells that lacked the receptor, as well as a 16-fold reduction in uptake when coincubated with octreotide (12). The present study confirms SSTR2-mediated uptake of Ga-MMC(IR800)-TOC in vivo as shown by clear differences in tumor fluorescence between HCT116-SSTR2 and HCT116-WT tumors macroscopically (in vivo, ex vivo) and microscopically.

Nonspecific binding of an FGS agent must also be examined as it will affect image contrast in the surgical field-of-view and, thus, efficacy. Surgical resection of primary NETs is most commonly performed in the pancreas, small intestine, or lungs. Therefore, it is imperative that Ga-MMC(IR800)-TOC exhibits low uptake in these tissues in order to generate sufficient contrast for identification of surgical margins or lymph node metastases. Despite endogenous SSTR2 expression in these tissues (94% homology between mouse and human SSTR2 isoforms; ref. 37), ex vivo optical imaging in mice at 48 hours produced TBRs > 4. The highest TBRs were found in the small intestine (17.2 ± 3.5), muscle (15.3 ± 5.0), and lungs (7.6 ± 2.0), suggesting even higher detection sensitivity for subclinical disease in these sites. Given the use of numerous drug-device combinations in preclinical and clinical studies, a definitive TBR threshold that indicates a positive fluorescent signal is difficult to define. However, successful intraoperative visualization of tumors has been demonstrated with several targeted agents that produce TBRs > 2 (30, 38, 39) and suggests that contrast ratios for Ga-MMC(IR800)-TOC may be sufficient for tumor identification in a clinical setting.

Intraoperative frozen section analysis is widely used for assessment of margin status and can show the tumor edge by demarcation of tumor from normal tissue by H&E staining. When combined with IHC, this method can identify SSTR2 expression and distribution within tissues and confirm the specificity of Ga-MMC(IR800)-TOC binding. Using frozen sections prepared from mouse xenografts injected with Ga-MMC(IR800)-TOC, we obtained fluorescence images that delineated HCT116-SSTR2 tumors from normal tissue in a manner that was consistent with tumor morphology shown by H&E staining and IHC. HCT116-WT tumors had negative IHC staining and minimal fluorescence. As shown by ex vivo tissue imaging, minimal signal was found in normal tissues and shows the potential effectiveness of Ga-MMC(IR800)-TOC for identifying tissues based on overexpression of SSTR2.

Clinically, ex vivo macroscopic imaging has emerged as an effective validation method with implications for fluorescence-guided pathology. A recent clinical study used an ex vivo fluorescence imaging platform to assess the microdistribution, tumor specificity, and margin identification ability of bevacizumab-IR800 and showed excellent sensitivity and negative predictive value for margin detection (40). The authors also proposed this approach as a first step in the development of novel FGS agents to show tumor-specific binding prior to more specific studies for dose-finding, time point optimization, and assessment of diagnostic accuracy. Similarly, we used Ga-MMC(IR800)-TOC for ex vivo staining of fresh biospecimens obtained from patients with pancreatic NETs to examine the ability of our agent to bind to human tumors. In addition to showing intense and preferential uptake in tumor regions, Ga-MMC(IR800)-TOC staining accurately maintained the tumor boundaries depicted by H&E and may enable margin detection in situ. These results compare favorably with ex vivo clinical data from several FGS agents (39–41) and could potentially lead to image-based prioritization of margin samples to support current surgical pathology practices. The most impactful application of Ga-MMC(IR800)-TOC, however, may be for real-time surgical imaging. For patients with pancreatic NETs (e.g., gastrinomas, VIPomas, somatostatinomas, and insulinomas), intraoperative tumor localization with Ga-MMC(IR800)-TOC could provide better specificity and stronger NIRF signal than currently used methylene-blue imaging (42, 43). Ga-MMC(IR800)-TOC may also enhance the ability of the surgeon to achieve greater debulking of metastatic burden, which can be beneficial both from a symptom management and overall survival standpoint (44). Thus, delineating the tumor with fluorescence might permit more complete cytoreduction and improve outcomes for patients with metastatic NETs.

In summary, we demonstrated for the first time that an SSTR2-targeted FGS agent can be used for highly specific tumor targeting. Our data showed the effectiveness of radiolabeling in the preclinical characterization of an FGS agent by providing definitive measurements of agent biodistribution that support imaging data. Imaging at the macro-, meso-, and microscopic levels provided extensive validation of SSTR2 specificity, which was extended to human NETs and indicates significant promise for in vivo application of Ga-MMC(IR800)-TOC. Our approach can be used in combination with SSTR2-targeted radiotracers (i.e., ⁶⁸Ga-DOTA-TOC and ⁶⁸Ga-DOTA-TATE) to identify patients that are candidates for FGS, thereby improving surgical care for patients with NETs.

A. Azhdarinia holds ownership interest (including patents) in RadioMedix Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Hernandez Vargas, S.C. Ghosh, W.E. Fisher, A. Azhdarinia

Development of methodology: S. Hernandez Vargas, S. Kossatz, J. Voss, S.C. Ghosh, H.S. Tran Cao, A. Azhdarinia

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Hernandez Vargas, S. Kossatz, J. Voss, J. Simien, S. Dhingra, A. Azhdarinia

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Hernandez Vargas, S. Kossatz, J. Voss, S.C. Ghosh, H.S. Tran Cao, W.E. Fisher, A. Azhdarinia

Writing, review, and/or revision of the manuscript: S. Hernandez Vargas, S. Kossatz, J. Voss, H.S. Tran Cao, T. Reiner, S. Dhingra, W.E. Fisher, A. Azhdarinia

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Voss

Study supervision: A. Azhdarinia

This work was supported by the National Institute of Biomedical Imaging and Bioengineering (R01 EB017279), NCI (P30 CA008748, K99 CA218875-01A1), the Tow Foundation, and Memorial Sloan Kettering Cancer Center's Center for Molecular Imaging & Nanotechnology. The authors thank Martina Cagigas and Amy McElhany for assistance with surgical biospecimens, and Dr. Erik Wilson and Armando Garcia for access to the Memorial Hermann Surgical Innovation and Robotics Institute. The authors also acknowledge the support of the Imaging Core Facility at The University of Texas Health Science Center–Center for Molecular Imaging, and the Radiochemistry and Molecular Imaging Probes Core Facility and Molecular Cytology Core Facility at Memorial Sloan Kettering Cancer Center.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.