Purpose: Intraoperative near-infrared fluorescence (NIRF) imaging could help stratification for the proper primary treatment for patients with pancreatic ductal adenocarcinoma (PDAC), and achieve complete resection, as it allows visualization of cancer in real time. Integrin αvβ6, a target specific for PDAC, is present in >90% of patients, and is able to differentiate between pancreatitis and PDAC. A clinically translatable αvβ6-targeting NIRF agent was developed, based on a previously developed cysteine knottin peptide for PET imaging, R01-MG, and validated in preclinical mouse models.
Experimental Design: The applicability of the agent was tested for cell and tissue binding characteristics using cell-based plate assays, subcutaneous, and orthotopic pancreatic models, and a transgenic mouse model of PDAC development (Pdx1-Cretg/+;KRasLSL G12D/+;Ink4a/Arf−/−). IRDye800CW was conjugated to R01-MG in a 1:1 ratio. R01-MG-IRDye800, was compared with a control peptide and IRDye800 alone.
Results: In subcutaneous tumor models, a significantly higher tumor-to-background ratio (TBR) was seen in BxPC-3 tumors (2.5 ± 0.1) compared with MiaPaCa-2 (1.2 ± 0.1; P < 0.001), and to the control peptide (1.6 ± 0.4; P < 0.005). In an orthotopic tumor model, tumor-specific uptake of R01-MG-IRDye800 was shown compared with IRDye800 alone (TBR 2.7 vs. 0.86). The fluorescent signal in tumors of transgenic mice was significantly higher, TBR of 3.6 ± 0.94, compared with the normal pancreas of wild-type controls, TBR of 1.0 ± 0.17 (P < 0.001).
Conclusions: R01-MG-IRDye800 shows specific targeting to αvβ6, and holds promise as a diagnostic and therapeutic tool to recognize PDAC for fluorescence-guided surgery. This agent can help improve the stratification of patients for a potentially curative, margin-negative resection. Clin Cancer Res; 24(7); 1667–76. ©2018 AACR.
Here, we report on a novel optical imaging agent that targets integrin αvβ6 in a highly specific manner to improve the treatment of pancreatic cancer patients. Pancreatic cancer has a dismal prognosis and proper selection for the most optimal treatment strategy, direct surgery, or preoperative neoadjuvant therapy, is key to provide treatment with oncologic benefit. Tumor-specific optical imaging with our agent can help with this stratification during diagnostic laparoscopy, and will improve complete resection of the primary tumor and lymph node metastases. Using a near-infrared fluorescent imaging system, even subcentimeter metastastic lesions could be identified. Clinical translation of this imaging agent in a timely fashion is promising because the separate components, the cysteine knottin peptide, and the near-infrared fluorescent dye, have shown to be safe in humans. The use of tumor-specific imaging has the potential to change the field of pancreatic cancer surgery, and optimize patient care.
Pancreatic ductal adenocarcinoma (PDAC) has a dismal prognosis (1, 2). In fact, it is the fourth leading cause of cancer-related deaths in the western world; 43,090 patients are estimated to die of PDAC in 2017 in the Unites States. The median size of a PDAC at the time of diagnosis is approximately 3.1 cm, and this has not changed in the past three decades despite major advances in both anatomic and molecular imaging technologies as well as in vitro diagnostics (3, 4). Early detection of pancreatic cancer is difficult because patients are often asymptomatic or present with nonspecific symptoms until late-stage disease (4, 5). After diagnosis, the patient selection for surgical resection is challenging. A potential curative resection can only be achieved in a minority of patients, and is dependent on tumor stage and grade. Two major predictors of long-term survival after resection are the achievement of tumor-free resection margins and the absence of systemic metastases. Unfortunately, margin-positive resections (R1) are a frequent phenomenon (up to 70%) as is the emergence of distant metastases soon after surgery, which were likely present at the time of surgery in the form of visually occult micrometastases (6–8). Failure to identify tumor-positive margins during surgery is not surprising, due to the tumor characteristics to quickly spread beyond the pancreas via perineural and perivascular pathways (9) and the inability of the surgeon to differentiate between tumor and (peritumoral) inflammatory tissue, such as associated chronic pancreatitis (10, 11).
The integration of innovative molecular imaging techniques into the operating room can lead to the transformation of current gold-standard intraoperative imaging modalities, such as intraoperative ultrasound and the surgeon's eyes, into pancreatic cancer–specific molecular imaging tools. Optical imaging in real-time, using a novel near-infrared fluorescent pancreatic cancer–targeting imaging agent, could add relevant information to the surgeon, by enhancing tumor margins and improving the detection of visually occult lesions. This would improve the selection of patients for the most suitable primary treatment.
In previous studies, out of a panel of different biomarkers, we selected integrin αvβ6 as the most promising target for pancreatic cancer detection, based on its potential to discriminate tumor and inflammation, to identify lymph node metastases, and to differentiate between viable tumor cells and necrosis in patients who received neoadjuvant therapy (12, 13).
We previously developed a cysteine knottin that binds to integrin αvβ6 with subnanomolar affinity (Kd ∼ 1 nmol/L; ref. 14). At this point, this peptide is tested in a phase I trial in humans as a targeted PET tracer, [18F]FP-R01-MG-F2, without any safety concerns (eIND #126379). Zhang and colleagues recently published data showing the development and preclinical validation of R01 conjugated to the photoacoustic dye Atto-740 (15). Zhang showed selective binding of the imaging agent to αvβ6-positive tumors. However, this agent exhibited hepatobiliary clearance marked by high uptake in the liver, spleen, and intestine. For a pancreatic cancer imaging agent, a renal clearance pathway is preferred due to the close proximity of the liver, and therefore potential interference of signal. Since the PET tracer shows renal clearance, the fluorescent dye was changed to achieve these favorable pharmacokinetics. The NIR fluorescent dye IRDye800 is known to clear via the kidneys, and has a high extinction coefficient (16). In addition, IRDye800 is already tested extensively in humans and shown safe in clinical studies (11, 17). Therefore, IRDye800 was the NIR fluorescent dye of choice in this study. The preferred imaging concept is shown in Fig. 1, highlighting the preference of an intravenously injected agent. This agent needs to bind selectively to the tumor receptors, facilitating fluorescence-guided resection, and excretion by the kidneys is favorable in pancreatic cancer surgery. Finally, the fluorescent signal needs to be correlated to tumor status on histopathology to ensure specific binding.
Here, we show the development and validation of a novel αvβ6-targeting probe for intraoperative tumor-specific fluorescent imaging of PDAC, with high potential for clinical translation.
Materials and Methods
Materials and reagents
All Fmoc amino acids and Rink Amide resin were purchased from CS Bio Co.. Peptide synthesis was carried out following the standard solid-phase Fmoc synthesis. Analysis and purification of the peptides was performed using the Dionex Summit high-performance liquid chromatography (HPLC) system (Dionex Corporation) and reverse-phase HPLC column Higins Analytical (Higins Analytical; C18, 4.6 mm × 250 mm). The mobile phase was 0.1% trifluoroacetic acid (TFA; Thermo Fisher Scientific) in water (solvent A) and 0.1% TFA in 90 % acetonitrile (CH3CN; Thermo Fisher Scientific) in water (solvent B). Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) was performed by the Canary Center proteomics facility on AB Sciex 5800 TOF/TOF System.
Development of tumor-specific fluorescent probe
Knottin peptide R01-MG-targeted integrin αvβ6 were synthesized using the described standard solid-phase Fmoc synthesis, and folded and purified as described previously (14). For labeling of the knottin peptide R01-MG, IRDye800 N-hydroxysuccinimide (NHS) ester (LI-COR Biosciences) was dissolved in anhydrous, amine-free dimethylformamide (DMF; Sigma Aldrich) to a 0.5 mg/mL solution. Knottin peptide R01-MG was dissolved in Dulbecco PBS (DPBS; Life Technologies) to a 1.5 mg/mL solution. IRDye800 NHS solution was mixed with knottin peptide R01-MG solution in 1:1 molar ratio. The reaction mixture was incubated and protected from light for up to 1 hour at room temperature. The reaction mixture was injected directly onto the HPLC column, and the separation of the product mixture followed using absorbance at 780 nm. The imaging agent had a retention time of 16.89 minutes and m/z of 4,817. The lyophilized IRDye800-peptide was dissolved in 100 μL dimethyl sulfoxide (DMSO; Thermo Fisher Scientific) and 100-μL DPBS. Maximum absorption wavelength of R01-MG-IRDye800 agent was performed using Cary50 (Varian).
Development of control cysteine knottin
The tumor-specific probe is based on the trypsin inhibitor, momordica cochinchinensis trypsin inhibitor (MCoTI-II). For the control cysteine knottin, the binding sequence of R01-MG to αvβ6 was replaced with the MCoTI-II sequence at these positions, to develop a control knottin most similar to the original R01-MG. This resulted in the following sequence; N′-GCPRILMRCRRDSDCPGACICRGNGYCG. Synthesis, folding, and purification was performed as described previously. The control cysteine knottin had a retention time of 20.60 minutes and m/z of 3033. The control peptide was labeled with IRDye800 N-hydroxysuccinimide (NHS) ester (LI-COR Biosciences) in a similar fashion as R01-MG-IRDye800. The reaction mixture was injected directly onto the HPLC column, and the separation of the product mixture followed using absorbance at 780 nm. The imaging agent had a retention time of 16.53 minutes.
One-hundred microliters of R01-MG-IRDye800 (0.5 × 10−6 M) was placed in a 96-well plate in the SpectraMax Gemini EM microplate reader (Molecular Devices) in triplicate and scanned with laser light using maximum absorption wavelength (750 nm) and emission wavelength (800 nm) for 30 minutes. Photobleaching was determined by the change in fluorescent intensity over time.
Human cancer cell lines
A total of six cancer cell lines were used for the cell-based experiments, including four pancreatic cancer cell lines (BxPC-3, BxPC-3-luc2, MiaPaCa-2, and AsPC-1), one colorectal cancer cell line (HCT-116), and one epidermoid carcinoma (A431). All cell lines, except for BxPC-3-luc2, were purchased from ATCC and expression was authenticated within 6 months before the in vitro experiments. The BxPC-3-luc2 cell line was purchased from PerkinElmer. BxPC-3 (CRL-1687TM), provided by ATCC, was used as source for the parental cell line. BxPC-3 and BxPC-3-luc2 cells were cultured in RPMI1640 medium containing 10% FBS, 1% penicillin and streptomycin. MiaPaCa-2, HCT-116, and A431 cells were cultured in DMEM containing 10% FBS, 1% penicillin and streptomycin. For MiaPaCa-2 cells, also 2.5% horse serum was added according to ATCC guidelines.
Mouse cancer cell lines
The binding of R01-MG-IRDye800 to mouse αvβ6 was assessed to determine the in vivo–specific binding to mouse αvβ6. The following two cell lines were used; a mouse breast cancer cell line (4T1), and mouse colon carcinoma cell line (CT-26). All cell lines were purchased from ATCC and expression was authenticated within 6 months before the in vitro experiments. A maximum of ten passages was performed for all cells, between thawing and uses in the described experiments. CT-26 cells were cultured in DMEM containing 10% FBS, 1% penicillin and streptomycin. 4T1 cells were cultured in RPMI1640 medium containing 10% FBS, 1% penicillin and streptomycin.
All cells were tested for αvβ6 expression using flow cytometry. The cells were grown to 90% confluence and detached with trypsin/EDTA. After evaluation of cell viability using Trypan blue, cells were adjusted to 1 × 106 per tube in ice-cold PBS with 4% formaldehyde for fixation and incubated in a binding buffer containing DMEM, 0.1% BDA, and 0.1% NaN3 with 100 μL of anti-αvβ6 mouse mAb (10D5, ab77906, Abcam) for 1 hour at room temperature. After incubation, cells were washed thrice and incubated with the secondary antibody anti-mouse IgG F(ab)2 fragment conjugated to Alexa Fluor 647 (#4410, Cell Signaling Technology, 1/800) for 30 minutes. After three washing steps with the washing buffer, cells were resuspended in 500-μL PBS. Samples were measured on the Scanford flow cytometer (Custom Stanford and Cytek upgraded FACScan Analyzer) and 1 × 105 cells were counted.
Cell binding study
Human cancer cell lines BxPC-3 (αvβ6-positive) and MiaPaCa-2 (αvβ6-negative), and murine cancer cell lines 4T1 (αvβ6-positive) and CT-26 (αvβ6-negative) were placed in a 8-well chamber slide (0.7 cm2, Nunc Lab-Tek II Chamber Slide System; Thermo Fisher Scientific) and grown to approximately 85% confluence. Cells were fixed for 20 minutes at 4°C with 1% formaldehyde. R01-MG-IRDye800 was diluted in integrin-binding buffer (IBB). The IBB was similar as previously described (18), and added to the chamber slide with concentration of 0.1 μmol/L, 0.2 μmol/L, or 0.5 μmol/L and incubated at room temperature for 2 hours. After being washed with DPBS three times, the chamber was removed and the nuclei were counter stained with ProLong Gold Antifade Mountant with DAPI (Thermo Fisher Scientific). The cellular binding of R01-MG-IRDye800 was assessed by costaining of cells with DAPI and fluorescence with the EVOS Cell Imaging Systems (EVOS FL Cell Imaging System, Thermo Fisher Scientific) with EVOS light cube Cy7, excitation 710–740, and emission: 775–746, and EVOS light cube DAPI, excitation: 357–344 and emission: 447–460. For quantitative analysis, mean fluorescent intensity (MFI), defined as total counts/region of interest (ROI) pixel area, was calculated using custom ROI generated for each specimen using the imaging software ImageJ version 1.50i and bundled with 64-bit Java for Windows (NIH, Bethesda, MD; http://rsb.info.nih.gov/nih-image). Specificity of the binding was determined by a blocking experiment with incubation of BxPC-3-luc2 cells with 0.5 μmol/L of R01-MG-IRDye800 and varying amounts of R01-MG [50× (25 μmol/L) – 150× (75 μmol/L) – 200× (100 μmol/L) excess], and by incubation of BxPC-3-luc2 cells with IRDye800 0.5 μmol/L alone. All samples were done in triplicates. For IRDye800 alone, IRDye800-NHS ester was used which is equal to the dye used in the agent. The NHS ester was reacted with water before injection, to ensure no reaction with amine groups could take place. A linear regression analysis was performed to determine whether there was a linear correlation between concentration and fluorescent signal; for BxPC-3 (Y = 118171*X + 11944) and for MiaPaCa-2 (Y = 23296*X + 9738).
To identify fluorescence intensity with increasing amount of αvβ6 receptors, an experiment was designed with both BxPC-3 and MiaPaCa-2 cells, in varying percentages; with the lowest amount of receptors in 100% MiaPaCa-2, and increasing amounts onwards; 20% BxPC-3- 80% MiaPaCa-2, 40% BxPC-3 - 60% MiaPaCa-2, 60% BxPC-3 - 40% MiaPaCa-2, 80% BxPC-3 - 20% MiaPaCa-2, and 100% BxPC-3. The cells were incubated with 0.5 μmol/L of R01-MG-IRDye800 for 2 hours. Additional cell experiments were performed comparing R01-MG-IRDye800 to the newly synthesized control peptide. BxPC-3 cells were placed on a 8-well chamber slide, and after fixation incubated with both agents separately at concentration of 0.1 μmol/L, 0.2 μmol/L, or 0.5 μmol/L.
Fluorescent imaging of αvβ6 in small-animal models
Subcutaneous mouse model of human pancreatic cancer.
The Administrative Panel on Laboratory Animal Care of Stanford University approved all procedures using laboratory animals. Eight-week-old nude female mice (n = 20; Charles River) were injected with 1 million BxPC-3-luc2, A431, or MiaPaCa-2-luc cells into the right shoulder. Mice bearing 0.4–0.8 cm tumors (n = 8 BxPC-3-luc2, n = 3 A431, n = 4 MiaPaCa-2) for R01-MG-IRDye800, and n = 5 for control cysteine knottin) were fluorescently imaged at 800 nm using a commercially fluorescent imaging system (The Pearl Impulse imaging platform (LI–COR Biosciences). For each animal we first obtained preinjection images at 800 nm. R01-MG-IRDye800 (30 μmol/L), the control cysteine knottin (30 μmol/L), or IRDye800 alone (30 μmol/L) was dissolved in 200-μL PBS (pH 7.4) and administered to mice via tail-vein injection. The data was acquired from 30 minutes up to 24 hours after the injection of the imaging agent.
Xenograft mouse model of human pancreatic cancer.
Human AsPC1 pancreatic ductal adenocarcinoma cells (αvβ6 positive; ATCC) were cultured to 70%–80% confluency. AsPC-1 cells were cultured in DMEM medium containing 10% FBS, 1% penicillin and streptomycin. AsPC1 cells were suspended in culture medium and mixed 1:1 with Matrigel (BD Biosciences) for orthotopic implantation. Sterile insulin syringes were loaded with the mixture and kept on ice until orthotopic implantation. After midline laparotomy, the pancreas of nu/nu mice was exposed, and AsPC1 cells [total of 6 × 106 cells, dissolved in 25 μL Matrigel containing EGF (0.7 ng/mL), insulin-like growth factor (16 ng/mL), and TGFβ (2.3 ng/mL); BD Biosciences] were coinjected into the body or tail of the pancreas in 8 female nude mice (6–8 weeks old; Charles River). The abdomen was then closed by layers. Orthotopic xenografts were allowed to grow for 21 days. R01-MG-IRDye800 (30 μmol/L) or IRDye800 (30 μmol/L) was dissolved in 200-μL PBS (pH 7.4) and administered to mice via tail-vein injection. The data were acquired from 30 minutes up to 24 hours after the injection of the imaging agent.
Transgenic mouse model of PDAC.
The transgenic pancreatic cancer mouse model (Pdx1-Cretg/+; KRasLSL G12D/+; Ink4a/Arf−/−) was used (n = 10), which spontaneously develops foci of pancreatic cancer within 4–7 weeks of age (19). Tumor diameters ranged between 3.0 and 4.8 mm. Age-matched littermates without KRasG12D mutation were used as normal wild-type control (n = 10). R01-MG-IRDye800 (30 μmol/L) was dissolved in 200 μL PBS (pH 7.4) and administered to mice via tail-vein injection, in both the wild-type controls and tumor-bearing mice.
Toxicity of R01-MG-IRDye800 in mice
The toxicity of R01-MG-IRDye800 in living mice was determined in Nu/Nu mice (n = 3). After injection of R01-MG-IRDye800 (30 μmol/L), the mice were followed for three weeks and weight and overall appearance was assessed. In addition, blood draws were done to assess effect on red and white blood cells, aspartate transaminase (ASAT), alanine transaminase (ALAT), and creatinine.
Ex vivo imaging and biodistribution
All the mice were sacrificed between 6 or 24 hours postinjection of the imaging agent as described above. Optical imaging of the excised organs was carried out using The Pearl Impulse imaging platform at 800 nm for biodistribution studies.
NIR-fluorescent animal imaging analysis
The Pearl Impulse small-animal imaging system (LI-COR Biosciences) was used for NIR-fluorescence measurements to measure NIR-fluorescent signals of the tumor, biodistribution analysis, and to calculate TBR (tumor-to-background ratios). The specific and control images were normalized and regions of interest were marked. The background signals were extracted from the surrounding tissue that was defined as the (normal) tissue/organ that lies around the tumor. After measuring the signal intensity from the macroscopic images, they were divided by each other using the following formula: TBR, mean signal tumor/mean signal surrounding tissue.
Formalin-fixed paraffin-embedded tumor tissues were sectioned at 5-μm thickness and fluorescence imaging was performed using the Odyssey NIR scanner (LI-COR Biosciences). All histologic sections were stained with standard hematoxylin–eosin IHC staining (H&E). To confirm the presence of αvβ6, additional sections were stained with IHC for anti-human αvβ6 (1/80, 6.2A1, Biogen Idec). For preparation of IHC staining, the slides were deparaffinized with xylene and rehydrated in serially diluted ethanol solutions (100%–50%), followed by demineralized water according to standard protocols. Endogenous peroxidase activity was blocked by incubation in 0.3% hydrogen peroxidase in PBS for 20 minutes. Antigen retrieval for αvβ6 was performed with 0.4% pepsin incubation (Dako) at 37°C for 10 minutes. Following antigen retrieval, the tissue sections were incubated overnight with the primary antibodies in 100 μL for standard tissue sections at room temperature. The slides were washed with PBS, followed by incubation with secondary antibody at room temperature according to the Vectastain ABC HRP kit (Peroxidase, Mouse IgG, Vector Laboratories Inc.). After additional washing, the staining was visualized with 3,3-diaminobenzidine tetrahydrochloride solution (DAKO) at room temperature for 5 minutes and counterstained with hematoxylin for 20 seconds. Finally, the tissue sections were dehydrated and mounted in Pertex (Histolab).
Analyses were performed in SPSS 22.0 (IBM technologies). All data were presented as the mean ± SDs. Means were compared using independent samples t test, or with one-way ANOVA with post hoc Bonferroni correction. A P value of less than 0.05 was considered to indicate a significant difference. Graphs and linear regression were performed in GraphPad Prism 7.0 (version 7.02, IBM Corp.)
IRDye800 (NIRF dye) conjugation of the cysteine knottins
R01-MG was synthesized, and labeled with IRDye800CW at 1:1 ratio (Supplementary Fig. S1A). In addition, the control cysteine knottin was synthesized (Supplementary Fig. S1B), and conjugated following the same protocol. R01-MG-IRDye800 showed high stability without clear decrease in fluorescent signal after laser exposure at 750 nm for 30 minutes (Supplementary Fig. S1C). This correlates to the stability of IRDye800 alone (20).
αvβ6 expression on cancer cell lines
The expression of αvβ6 in the cancer cell lines (listed above) was evaluated by flow cytometry using the mAb, 10D5, and shown in the Supplementary Figures (Supplementary Fig. S2). Four of the cancer cell lines showed high αvβ6 expression; BxPC-3-luc2, A431, AsPC-1, and 4T1. Three cancer cell lines showed none or low αvβ6 expression; MiaPaCa-2, HCT-116, and CT-26. On the basis of these experiments, we selected BxPC-3 and A431 as positive and MiaPaCa-2 as negative controls, respectively, for further studies.
The retained αvβ6-specificity of R01-MG-IRDye800 was confirmed using chamber slide assays. An increase in agent concentration resulted in an almost linear increase in signal intensity in the αvβ6-positive cell line (BxPC-3), while the αvβ6-negative cell line (MiaPaCa-2) showed limited signal intensity (Fig. 2A). The difference between the cell lines was observed for all the different concentrations tested; however, staining at the lowest concentration (0.1 μmol/L) was also weak in BxPC-3 cells (Fig. 2A). Competition of R01-MG-IRDye800 with unlabeled R01-MG resulted in a 77.2% reduction in NIR signal intensity, suggesting that the αvβ6-specific binding capacity of R01-MG after conjugation to IRDye800 remained. In addition, in a cell assay with different concentration of αvβ6 receptors, simulated by different percentages of BxPC-3 and MiaPaca-2 cells together, an increase in receptors led to a linear increase in fluorescent signal (R2 = 0.80; Fig. 2A).
A significant difference in fluorescent (Fig. 2B and C) signal between BxPC-3 and MiaPaCa-2 cells was detected (P < 0.001), and after 2 hours of incubation the fluorescent signal in BxPC-3 cells was 3.1-fold higher compared with the MiaPaca-2 cells. To control for nonspecific staining of the fluorescent dye, incubation with the fluorescent dye alone, IRDye800 (0.5 μmol/L), was performed. The fluorescence intensity was 2.7 times lower compared with the incubation of R01-MG-IRDye800 (0.5 μmol/L) in BxPC-3 cells (P < 0.001) and equal to the intensity in MiaPaCa-2 (P = 1.000), indicating the specific binding of R01-MG-IRDye800 to αvβ6 (Fig. 2B). A significant (P < 0.001) lower fluorescent signal was observed after incubation with the control cysteine knottin, compared with R01-MG-IRDye800 as shown in Fig. 2D. There was no fluorescent signal observed in both cell lines after incubation with PBS.
Specificity of R01-MG-IRDye800 in subcutaneous mouse models.
To validate the performance of R01-MG-IRDye800 in living subjects, three different mouse models were used: (i) a subcutaneous model with different cell lines, (ii) an orthotopic, intrapancreatic xenograft model using different cell lines, and (iii) a spontaneous transgenic pancreatic cancer model. Subcutaneous tumor models were used to compare the binding of R01-MG-IRDye800 in αvβ6-positive and -negative tumors, and to determine the TBR over time (Fig. 3A). The highest TBR for BxPC-3 mice was measured 20 hours postinjection (4.08 ± 0.95; Fig. 3B). There was a difference in TBR between BxPC-3 tumors (2.5 ± 0.1), MiaPaCa-2 tumors (1.2 ± 0.1), and A431 tumors (1.8 ± 0.5), with a linear increase in TBR for both the BxPC-3 tumors and A431 tumors over time. The TBR of MiaPaCa-2 tumors remained stable at approximately 1.0 during the investigated period of 5 hours (Fig. 3C). BxPC-3 tumors had a significantly higher TBR compared with MiaPaCa-2 tumors (P < 0.001; Fig. 3D). As controls R01-MG-IRDye800 was compared with a control peptide and fluorescent dye alone. The TBR of the mice injected with R01-MG-IRDye800 after 24 hours (3.5 ± 0.95) was significantly (P < 0.005) higher compared with control peptide (1.6 ± 0.36), and dye alone (1.4 ± 0.2; P < 0.005) suggesting specific targeting of αvβ6 by R01-MG-IRDye800 (Fig. 3E). In addition, compared with IRDye800 alone, a clear increase in mean TBR was seen over time for R01-MG-IRDye800, compared with an approximately constant mean TBR for IRDye800. This difference was significant from T = 4 hours after injection (T = 0 hour, P = 0.95; T = 1 hour, P = 0.16; T = 3 hours, P = 0.05; T = 4 hours, P < 0.001; T = 6 hours, P < 0.001; T = 24 hours, P < 0.005; Fig. 3F).
αvβ6 binding and biodistribution in orthotopic mouse models.
The efficacy of R01-MG-IRDye800 to bind and detect αvβ6 in vivo was evaluated in pancreatic cancer orthotopic xenografts implanted in nude mice. A mean TBR of 2.7 ± 1.22 at 24 hours postinjection was measured, compared with a mean TBR of 0.86 ± 0.13 for IRDye800 alone (Fig. 4A). Ex vivo NIR-fluorescence measurements showed tumor specificity of the tracer with a mean TBR of 6.96 ± 2.2. Biodistribution assays showed high fluorescence in the kidneys (MFI 1.12 AU) and urine (MFI 0.22 AU) compared with the liver (MFI 0.007 AU) indicating a renal clearance route, and some fluorescence in the bowel (MFI 0.03 AU) and skin (MFI 0.02 AU; Fig. 4B). Fluorescence in the bowel can be explained by mild αvβ6 expression in gastrointestinal tract mucosa (21, 22). At the microscopic level, high fluorescence matched well with location of the tumor on H&E (Fig. 4C).
Validation in transgenic mouse models.
In the spontaneous transgenic tumor mouse model, the tumor could be visualized, even with the skin intact, and fluorescence localized at tumor regions was clearly visible ex vivo (Fig. 5A). Even a subcentimeter metastasic lesion on the bowel (0.7 × 0.5 cm) could be visualized with fluorescence, corresponding to tumor-positive H&E (Fig. 5B). In mice with the tumor genotype, the TBR obtained with R01-MG-IRDye800 (3.6 ± 0.94) was significantly higher compared with the TBR in the normal pancreas of the WT controls after R01-MG-IRDye800 infusion (1.0 ± 0.17; P < 0.001; Fig. 5B). Fluorescent signal and tumor status corresponded well to αvβ6 expression as assessed by IHC (Fig. 5C). To resemble the human situation, fluorescent signal provided by R01-MG-IRDye800 showed successful distinction between PDAC and peritumoral inflammation when in close proximity, even on microscopic level (Supplementary Fig. S3).
Toxicity of R01-MG-IRDye800
Pilot toxicity of the agent was determined in nude mice (N = 3) by measurement of weight and hematology and chemistry blood panels. There was no significant difference observed in weight of mice preinjection of R01-MG-IRDye800, directly after injection (P = 0.054) and after the observation period of 3 weeks (P = 0.732). The liver panel slightly increased after agent injection, and red blood cells count decreased, but all returned to baseline levels after follow-up. A slight increase in white blood cell count and creatinine was also seen, but these levels remained within normal limits for the entire observation period (Supplementary Fig. S4).
To improve the dismal prognosis of patients with PDAC, focus should be on early detection, improved patient selection for treatment and better surgical outcomes. A significant number of patients with PDAC present while the disease has already spread and will therefore not benefit from surgery as primary treatment. A limitation of conventional imaging techniques, such as CT and MRI, is the low sensitivity for detection of subcentimeter lesions and therefore metastatic spread is not always detected prior to surgery (23). To overcome this limitation, optical imaging could be an additional tool to be applied during a diagnostic laparoscopy, prior to resection, to screen the abdomen for potential metastatic spread, such as liver or peritoneal metastases. For superficial lesions, fluorescent imaging is superior. Adjacent to that, for a more thorough screening of the liver, optoacoustic imaging could be added to provide sufficient depth penetration (10, 24). If no spread is found in this initial stage, the surgery can proceed using the same optical imaging techniques to guide the resection. During resection, use of optical imaging can improve radical resection rates by enhancing tumor margins and increase detection of tumor-positive lymph nodes (25).
Previously, Hutteman and colleagues showed the use of the nontargeted optical imaging agent, indocyanine green (ICG), for the detection of PDAC. It was concluded that no useful tumor demarcation could be visualized after intravenous injection of ICG. However, the common bile duct and biliary anastomoses were clearly visualized (26). Clinical trials performed in other malignancies, such as ovarian cancer, using tumor-specific imaging agents, showed successful use of this technique for the specific identification of cancer during surgery (11, 27–29).
Integrin αvβ6 has been shown to be a promising target for identification of PDAC, and differentiation with pancreatitis and normal pancreatic tissue, with retained expression in vital tumor cells after neoadjuvant therapy (12–15, 30). Nowadays, the majority of patients will receive neoadjuvant therapy prior to surgery due to late onset of symptoms and therefore, presentation with locally advanced disease, or borderline resectable tumors. Currently, the distinction after neoadjuvant therapy between necrosis, fibrosis, and vital tumor is impossible using conventional imaging techniques, and tumor-specific imaging could potentially fill this gap and be of added value in the diagnostic process.
Cysteine knottin peptides are relatively small (3–4 kDa), and have been extensively validated as alternative molecular imaging agents to antibodies (14, 31, 32). We showed that the previously engineered cysteine knottin peptide R01-MG (14) demonstrated high and specific tumor uptake and high tumor-to-normal tissue ratios in mouse models of integrin αvβ6–positive PDAC, after conjugation to the near-infrared fluorescent dye, IRDye800. Moreover, the agent did not accumulate in integrin αvβ6–negative tumors. Because of the well-defined secondary structure, cysteine knottin peptides possess improved in vivo stability, and resist both chemical and physical insult compared with linear peptides, such as “A20” (14, 33–35). R01-MG-IRDye800 targets integrin αvβ6 with high-affinity binding (Kd = 1.2 nmol/L) and in a highly specific manner (14), compared with, for example, cyclic-RGD, which mainly targets αvβ3, but also αvβ5, α5β1, α6β4, α4β1, and αvβ6 (36).
To prove that R01-MG-IRDye800 specifically targets αvβ6, and fluorescent signal in the tumor is not due to enhanced permeability and retention (EPR) effect, especially seen in subcutaneous tumor models (37), we have chosen to use a humanized transgenic mouse model. In addition, we performed several control studies using IRDye800 alone, and a newly synthesized control peptide with similar characteristics as the targeting peptide. Furthermore, to minimize nonspecific leakage of the agent to the tumor, subcutaneous tumors with BxPC-3 cells were used which are known to be poorly leaky tumors, and therefore minimize EPR effect (38). Our studies showed that the fluorescent signal of R01-MG-IRDye800 was significantly higher compared with both controls, indicating specific binding of R01-MG-IRDye800 to αvβ6.
A limitation of using optical imaging as described here, is the use of one target to image the tissue of interest compared with multiplexing, targeting of multiple markers, to increase the detection sensitivity and potentially assess aggressiveness of the tumor (39). Ideally, you would be able to deliver a cocktail of imaging agents to patients, each targeting a specific biomarker that is known to be overexpressed in PDAC and supplement each other for improved detection (13). However, at this point none of the newly developed NIR tumor-specific optical imaging agents is approved by the FDA, which is partly due to the comprehensive process and costs associated with getting regulatory approval for an optical imaging agent, and the current lack of proven patient benefit (40). Therefore, the possibility to deliver multiple agents at once can probably not be realized in the near future. A benefit of the investigated target is that αvβ6 is overexpressed in >95% of PDAC patients, and significantly higher expressed in PDAC versus pancreatitis and normal pancreatic tissue (13). Therefore, it is expected that multiplexing will not cause significant benefits over the use of R01-MG-IRDye800 as a single agent. As reported above, the differentiation between chronic pancreatitis and PDAC is one of the aspects where tumor-specific molecular imaging can be of additional benefit to the patient. A limitation of this study is that the agent is not tested in chronic pancreatitis mouse models. However, we know from previous performed research that αvβ6 expression is significantly lower in pancreatitis compared with PDAC (12, 13). In addition, we did address the difference in fluorescent signal between PDAC and peritumoral inflammation, because if the technique is able to do this, it would fill a clinically relevant gap in conventional imaging techniques. On the basis of our results, and due to high specific binding of the agent, no limitations are expected for the clinical use of this agent for tumor-specific detection.
Our developed agent has high potential for clinical translation into patients, as it consists of two parts that are already successfully translated into humans. The cysteine knot peptide, R01-MG, is currently used in the clinic as an αvβ6-targeting PET tracer for the detection of pancreatic cancer (eIND #126379, NCT02683824) and the NIR fluorescent dye, IRDye800CW, has already been used for fluorescence-guided surgery in multiple clinical trials (11, 29). We performed a pilot toxicology study that showed that the newly developed agent was well tolerated. However, some changes in hematology and chemistry blood panels were seen immediately after injection, which could potentially be due to the agent. These levels returned within the normal range after observation, but before final conclusions can be drawn, larger toxicology studies need to be performed to determine the meaning of these changes.
In conclusion, the developed optical imaging agent R01-MG-IRDye800 binds selectively to αvβ6. In cell studies, R01-MG-IRDye800 outperformed all controls, and a clear correlation between levels of receptor expression and fluorescence signal was found. In multiple preclinical mouse models, tumor-specific targeting was achieved, with favorable renal clearance. The agent exhibited high tumor-to-background ratios, and clear correlation between fluorescence signal and histopathologic evidence of PDAC. Thanks to the fact that this agent contains two elements, R01-MG and IRDye800 that were already proven to be safe for use in humans, we do not expect major hurdles for the clinical translation of this optical imaging agent.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: W.S.F.J. Tummers, R.H. Kimura, R.-J. Swijnenburg, J.K. Willmann
Development of methodology: W.S.F.J. Tummers, R.H. Kimura, L. Abou-Elkacem, C. Beinat, R.-J. Swijnenburg, J.K. Willmann, S.S. Gambhir
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W.S.F.J. Tummers, R.H. Kimura, L. Abou-Elkacem, C. Beinat
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W.S.F.J. Tummers, R.-J. Swijnenburg, J.K. Willmann
Writing, review, and/or revision of the manuscript: W.S.F.J. Tummers, R.H. Kimura, L. Abou-Elkacem, C. Beinat, A.L. Vahrmeijer, R.-J. Swijnenburg, J.K. Willmann, S.S. Gambhir
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.K. Willmann
Study supervision: R.-J. Swijnenburg, J.K. Willmann
We would like to thank James Strommer, Stanford University, for his contribution to the graphical design. We would like to acknowledge support from The Canary Foundation, Department of Energy, and NIH. W.S.F.J. Tummers' contribution to this work was supported in part by Michaël-van Vloten Fonds, Lisa Waller Hayes Foundation, Stichting De Drie Lichten, and Ketel1 Studiefonds.