Thy1-Targeted Microbubbles for Ultrasound Molecular Imaging of Pancreatic Ductal Adenocarcinoma

This article is featured in Highlights of This Issue, p. 1511

Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal form of cancer and is currently the fourth leading cause of cancer death in the United States (1). The American Cancer Society estimated 53,670 new diagnoses of PDAC and 43,090 deaths from this cancer in 2017 in the United States (1). The incidence of PDAC is on the rise, and PDAC is projected to become the second most common cause of cancer death in the United States by 2030 (2). Although patient survival is highly dependent upon tumor stage, most patients already have advanced disease at the time of diagnosis (3). Therefore, the overall median survival of patients after diagnosis is only 4 to 6 months (4). Although unpredictable and vague clinical symptoms related to the disease are factors in the delay in diagnosis, there is a lack of specific and sensitive blood biomarker or imaging tests to detect the disease early (5).

One potential strategy for earlier detection of PDAC involves the screening of moderate- and high-risk patients using molecularly-targeted contrast microbubbles that bind and amplify the signal of molecular markers differentially expressed on the neovasculature of PDAC compared with normal pancreas and chronic pancreatitis (6–8). Proteomic analysis and immunohistochemical staining of whole tissues from human PDAC and control normal pancreas and chronic pancreatitis have previously identified and validated the thymocyte differentiation antigen 1 (Thy1) as a molecular marker differentially upregulated on the neovasculature of PDAC (6, 9). A total of 81% of PDAC cases stained positive for Thy1, whereas normal pancreas and chronic pancreatitis cases were positive in 11% and 7%, respectively. Thy1 neovascular immunostaining could distinguish PDAC from normal and chronic pancreatitis tissues with 90% specificity and 81% sensitivity. Thy1, also known as cluster of differentiation 90 or CD90, is a cell-surface glycoprotein that belongs to the immunoglobulin-like supergene family. Thy1 was originally described as a mouse thymocyte differentiation marker (10, 11) and subsequently shown to be expressed in other tissues, including the surface of newly formed blood vessels in human colon cancer (12), glioblastoma (13), hepatocellular carcinoma (14), and ovarian cancer tissues (15).

Evolving protein display techniques (16) using naïve nonimmune human single-chain fragment variable (scFv) yeast-surface-display (YSD) libraries enables the development of novel ligands with high affinity, specificity, and stability that can be used to functionalize contrast microbubbles to attach to molecular markers. An antibody in scFv format (molecular weight of ∼28 kDa) consists of variable regions of heavy (VH) and light (VL) chains, joined together by a flexible peptide linker (17). ScFvs are an alternative to full-length antibodies in diagnostic and therapeutic applications and have been translated to clinical trials (18–20). In comparison with full-length antibodies, scFv have shown to be advantageous for molecular imaging due to better tumor penetration, more rapid blood clearance, shorter retention times in nontarget tissues, as well as reduced immunogenicity (21–23). Furthermore, due to their relatively small size as well as good stability and solubility, scFvs can be recombinantly produced in a cost-efficient way (24, 25). Moreover, the high stability of scFvs with a single lysine distant from the paratope/antigen-binding site eases amine bioconjugation chemistry to functionalize contrast microbubbles (26).

The purpose of this proof-of-principle study was to engineer a human scFv ligand binding to both human and murine Thy1 and to test it for in vivo ultrasound molecular imaging in two different mouse models of PDAC, in human xenografts, and transgenic PDAC mice.

Wild-type (WT) MILE SVEN 1 mouse vascular endothelial (MS1-WT) cells were directly obtained directly from the American Type Culture Collection [(CRL2279; (ATCC)] that performs standard cell line characterizations of mycoplasma, yeast, bacterial, and viral contamination. The cell line was maintained according to the recommendations of the ATCC. Cells were cultured under sterile conditions in DMEM (ATCC) with FBS at 5% and maintained in a 5% CO₂-humidified atmosphere at 37°C. Cells were transfected with human Thy1 DNA as described (6). In brief, the human Thy1 DNA sequence (gi|224589802:c119294246-119288655) was first optimized for mammalian codon usage using standard techniques (27). The transfection of MS1-WT cells with the Thy1-expression vector was performed using lipofectamine 2000 transfection reagent (Life Sciences; Invitrogen), following the recommended manufacturer's standard protocol. MS1 cells stably expressing human Thy1 (MS1_Thy1) were grown in DMEM containing 10% FBS and 0.4 mg/mL puromycin (G418, Sigma) in a 5% CO₂ humidified atmosphere and subcultured prior to confluence using trypsin.

A YSD nonimmune human scFv library was sorted and matured as described (16). For further details, please refer to the Supplementary Methods.

The dissociation constant (K_D) of the Thy1-scFv was measured using yeast as described (28). In brief, yeast cells (1 × 10⁵) transformed with pCT plasmid containing Thy1-scFv gene were incubated with 0.01 to 100 nmol/L recombinant human and murine biotinylated Fc-conjugated Thy1 (B-Thy1; Abcam), respectively, overnight at room temperature. Cells were stained with the chicken anti–c-myc antibody (eBioscience). After FACS analysis, mean fluorescence values for streptavidin-allophycocyanin (APC; Biolegend) and Alexa 555–conjugated goat anti-chicken IgY (eBioscience) in double-positive populations of each yeast sample were acquired. K_D values were analyzed by determining the ratio of the mean fluorescence intensity of [APC] and fluorescence intensity of [A555] using FACS. The ratio was plotted against the used concentration of Thy1 using Prism 5 software (Graphpad). The K_D was determined using a nonlinear least squares curve fit (16, 29, 30).

Thy1-scFv ligand was recombinantly expressed in SHuffle T7 Escherichia coli and purified by a HisTrap FF column (GE Healthcare Biosciences). For further details, please refer to the Supplementary Methods.

Thy1-coated magnetic beads were prepared as described above and analyzed with FACS. Briefly, 66 pmol of Fc-conjugated human and murine Thy1 was incubated with 10 μL of protein G-magnetic beads (ThermoFisher Scientific) in 50 μL of PBSA for 40 minutes at room temperature. As controls, naked protein G-beads and beads coated with the same amount of human and murine IgG were prepared. Subsequently, beads were washed with PBSA and incubated with 10 nmol/L purified biotinylated Thy1-scFv for 1.5 hours at room temperature. The Thy1-scFv bound to beads were stained with 5 μL of streptavidin-APC (Biolegend) and analyzed by FACS. For cell-binding assay, MS1_Thy1 was stained with APC-labeled Thy1-scFv for 1.5 hours at 4°C and analyzed with FACS. As negative control cells, both MS1-WT and MS1_CD276, a cell line overexpressing an independent biomarker CD276, were stained with Thy1-scFv. To confirm Thy1 expression on cells, primary antibody (rabbit anti-human Thy1, Sigma, 1:100) incubation was performed. After washing in PBS, secondary antibody (anti-rabbit FITC; Jackson Immunolaboratories, 1:500) was added for 1 hour at 4°C and analyzed with FACS.

To confirm human Thy1 expression on vascular endothelial MS1 cells, immunofluorescence staining of the cells was performed using standard techniques. In brief, MS1-WT and MS1_Thy1 cells were grown on cover slips under standard conditions in DMEM complete growth media for 24 hours; after the media were removed, cells were washed in PBS and fixed in 4% paraformaldehyde in PBS solution for 30 minutes at room temperature. Cells were then washed in PBS, and 1% bovine serum albumin (BSA) blocking solution was applied for 1 hour. The biotinylated Thy1-scFv (100 nmol/L) incubation was performed for 2 hours at 4°C. After washing in PBS, secondary antibody (Streptavidin-AF555; ThermoFisher Scientific) was added for 1 hour at room temperature. Cells were then washed in PBS, counterstained with 4′,6 diamidino-2-phenylindole (DAPI), mounted onto glass slides with anti-fade solution, and imaged with an Olympus IX81 system.

Commercially available perfluorocarbon-filled, lipid-shelled, streptavidin-coated microbubbles (MicroMarker, VisualSonics) were used to generate Thy1-targeted microbubbles. Microbubbles targeted to both human and murine Thy1 (MB_Thy1-scFv) using the new Thy1-scFv. Two positive control microbubbles (MB_Thy1) targeted either to murine Thy1 (for imaging of transgenic mice) or to human Thy1 (for human xenograft imaging) using commercially available antibodies as binding ligands. MB_{scFv-scrambled} and MB_Non-targeted were used as negative controls. For further details, please refer to the Supplementary Methods.

Successful coupling of both Thy1-scFv and scFv_scrambled on the microbubble shell was confirmed by FACS and microscopy. Synthesized targeted microbubbles (1 × 10⁵ each) coupled with either Thy1-scFv or scFv_Scrambled were incubated with an anti-His antibody-AF488 (Thermo Fisher) for 1 hour. The labeled molecularly-targeted microbubbles were washed 3 times by centrifugation at 300 g for 2 minutes and analyzed by FACS. FACS was used to confirm microbubble ligand coating by fluorescence intensity. Voltage, forward and side light scattering (FSC and SSC) settings were adjusted to detect solely microbubble populations. Note that 50 μL freshly synthesized microbubble solutions were diluted with 200 μL PBS prior to each measurement. Subsequent data analysis was done using FlowJo software (Stanford University, CA, USA).

FACS analysis was performed in order to assess the binding specificity of molecularly-targeted microbubbles to a soluble human and murine Thy1 protein (R&D Systems). To confirm successful ligand conjugation of scFv ligands including His-tag to the microbubble shell, MB_Thy1-scFv was prelabeled with anti-His antibody-AF488 for 30 minutes at room temperature (Fig. 1D). The microbubbles were washed twice by centrifugation at 300 g for 2 minutes. Subsequently, prelabeled microbubbles were incubated with 10 nmol/L soluble human and murine IgG-Fc–conjugated Thy1 with 900 μL mouse serum (Sigma-Aldrich) on a benchtop rotator for 40 minutes at room temperature. This was followed by washing by centrifugation at 300 g for 2 minutes and incubation with anti-human and anti-mouse IgG-Fc antibody-AF647 (eBioscience) for 30 minutes on ice with a final washing by centrifugation at 300 g for 2 minutes. The measured median fluorescence intensity data were analyzed using FlowJo software. A strong correlation of anti-His antibody-AF488–labeled microbubbles and IgG-Fc antibody-AF647–labeled Thy1 indicated positive binding. All experiments were performed in triplicate.

Binding specificity of MB_Thy1-scFv and MB_Thy1 to the target Thy1 was first assessed in cell culture experiments under flow shear stress conditions simulating flow in blood capillaries by using a flow chamber experimental set-up. Please see the Supplementary Methods for more details.

The Administrative Panel on Laboratory Animal Care of Stanford University approved all procedures using laboratory animals. An orthotopic mouse model expressing human Thy1 on its neovasculature was created as described (6). In brief, human AsPC1 PDAC cells (ATCC) were cultured to 70%–80% confluency. After midline laparotomy, the pancreas were exposed, and AsPC1 cells along with MS1_Thy1 cells at a 1:5 ratio (total of 6 × 10⁶ cells, dissolved in 25 μL Matrigel, BD Biosciences) were coinjected into the body or tail of the pancreas in 9 female nude (nu/nu) mice (6–8 weeks old; Charles River). The abdomen was then closed by layers. Orthotopic xenografts were allowed to grow between 7 and 21 days to yield a spectrum of different tumor sizes with diameters ranging between 2.5 and 8.0 mm (mean, 5.1 mm) as measured by B-mode ultrasound.

The transgenic pancreatic cancer mouse model (Pdx1-Cre^tg/+; KRas^{LSL G12D/+}; Ink4a/Arf^−/−) was used (n = 6), which spontaneously develops foci of pancreatic cancer within 4 to 7 weeks after birth (31). Tumor diameters ranged between 1.5 and 4.5 mm (mean, 3.5 mm). Age-matched littermates without KRasG12D mutation were used as normal WT control mice (n = 6).

Chronic pancreatitis was established in 6 transgenic WT mice using two subcutaneous injections of l-arginine dissolved in 0.9% saline (6 g/kg) separated by 1 hour as described (6, 32). This was repeated weekly for 6 weeks, after which the mice were allowed to recover for 14 days before imaging experiments.

All mice were kept anesthetized with 2% isoflurane in room air at 2 L/min on a heated stage for 37°C throughout the ultrasound molecular imaging sessions. In intra-animal comparison experiments, all types of microbubble (5 × 10⁷ each of MB_Thy1-scFv, MB_Thy1, MB_{scFv-scrambled}, and MB_Non-targeted) were injected intravenously through the tail vein of each of the 15 mice (9 nude mice with orthotopic human PDAC xenografts, 6 spontaneous PDAC in transgenic mice) in random order to minimize any bias from injection order. Between imaging sessions using the different contrast agents, a waiting interval of a minimum 30 minutes was used to allow for clearance of freely circulating microbubbles from the vasculature (3, 6). Also, each time at approximately 30 minutes, it was confirmed by imaging that no remaining circulating microbubbles were present from the previous injection and any remaining attached microbubbles were removed by applying a high-power destruction pulse (see below for acoustic parameters). In addition, all control littermates with normal pancreas (n = 6) and all chronic pancreatitis mice (n = 6) were scanned as Thy1-negative models after the injection of all four microbubble types. To further confirm binding specificity of MB_Thy1-scFv, an in vivo competition experiment was performed. In a subgroup of four orthotopic human PDAC-bearing mice, in vivo blocking of Thy1 by injecting 150 μg Thy-scFv via the tail vein was performed in order to block binding of MB_Thy1-scFv to its target human Thy1.

All in vivo imaging studies were performed in contrast mode using a dedicated small-animal high-resolution ultrasound imaging system (Vevo 2100; VisualSonics). Contrast mode images were acquired with a 21-MHz linear transducer (MS250, VisualSonics; lateral and axial resolution of 165 and 75 μm, respectively), and all imaging parameters (focal length, 10 mm; transmit power, 4%; mechanical index, 0.2; dynamic range, 40 dB; a center frequency, 21 MHz) were kept constant during all imaging sessions. The transducer was fixed on a railing system to maintain the acoustic focus at the center of the PDAC with the imaging plane aligned in the center of the tumor and including adjacent non-PDAC tissue of pancreata. To differentiate the acoustic signal owing to microbubbles adherent to Thy1 and the signal from freely circulating microbubbles in the bloodstream, previously described principles of destruction-replenishment techniques were used (6). In brief, the imaging signal in the field of view increases after intravenous injection of molecularly-targeted microbubbles and is composed of signal from attached and freely circulating microbubbles as well as tissue background signal. After 4 minutes following tail vein injection of microbubbles, 200 frames were acquired. This was followed by a high-pressure destructive pulse (1-second continuous high-power destructive pulse of 3.7 MPa, transmit power, 100%; mechanical index, 0.63) to destroy all bound and unbound microbubbles within the beam elevation. After 10 seconds to allow freely circulation microbubbles to replenish into the field of view, an additional set of 200 frames was acquired to measure the signal intensity from the unbound circulating microbubbles. The difference in imaging signal before and after destruction was calculated corresponding to the signal from attached microbubbles. The same imaging setting was used in all imaging sessions (Fig. 1E).

All imaging data were analyzed by one reader in random order who was blinded to microbubble types. The acoustic imaging signals were analyzed after image acquisition, averaged to compensate for breathing motion artifacts by using commercially available analysis software (VevoCQ; VisualSonics). Data analysis was accomplished by selecting a frame with the plane of interest and manually drawing a region of interest (ROI) around the PDAC in bearing pancreas, in the adjacent non-PDAC pancreatic tissue, as well as in the normal pancreas of control littermates. The magnitude of imaging signal (expressed in arbitrary units, a.u.) from attached microbubbles was assessed by calculating an average for pre- and postdestruction imaging signals and subtracting the average postdestruction signal from the average predestruction signal using Vevo2100 built-in software (VevoCQ; VisualSonics).

Ex vivo analysis was performed using standard techniques (see the Supplementary Methods).

All data were expressed as mean ± SD. For details on the statistical analysis, please refer to the Supplementary Methods.

To obtain a human Thy1-scFv ligand, a naïve scFv-yeast library with a diversity of 2.5 × 10⁹ was screened. After seven rounds of iterations, selections, and affinity maturations, the final isolated yeast clones showed higher affinity for Thy1 than those obtained from the original yeast library (Fig. 2A). The dominant clone was identified by sequence analysis (Fig. 2B). The 259-amino acid Thy1-binding scFv ligand was produced with an N-terminal 6xHis-tag and an enterokinase-cleavage site fragment. Thy1-scFv was then purified by using a His-Trap FPLC. The molecular weight of Thy1-scFv (44 kDa) was verified by using SDS-PAGE, showing high ligand purity of more than 95% (Fig. 3A).

FACS analysis was performed to analyze binding specificity of purified Thy1-scFv to a soluble human and murine Thy1 protein. In comparison with negative control beads, fluorescently-labeled Thy1-scFv showed enhanced fluorescence intensity signal, indicating solely binding to both human and murine Thy1-coated beads (Fig. 3B). The K_D of Thy1-scFv as assessed by FACS was 3.4 ± 0.36 nmol/L for human and 9.2 ± 1.7 nmol/L for murine Thy1 (Fig. 3C). Thy1-scFv also showed binding to Thy1-expressing cells, whereas there was no binding to both Thy1-negative cells (MS1-WT) and CD276-expressing control cells. Furthermore, blocking of Thy1 receptors with free Thy1-scFv resulted in significantly (P < 0.01) decreased binding of fluorescently-labeled Thy1-scFv, confirming binding specificity of Thy1-scFv to cellularly Thy1-expressing cells (Fig. 3D). This was further shown by immunofluorescence microscopy (Fig. 3E).

FACS analysis was performed to analyze binding specificity of MB_Thy1-scFv to a soluble human and murine Thy1 protein. Before FACS studies, all microbubbles were fluorescently prelabeled by anti-His antibody-AF488 and were incubated with human and murine Thy1 protein (Fig. 3F and G). In comparison with non–scFv-coated microbubbles (MB_Thy1 and MB_Non-targeted), only His-tagged scFv ligands conjugated to the surface of microbubbles (MB_Thy1-scFv and MB_{scFv-scrambled}) showed enhanced fluorescence intensity signal using anti-His antibody-AF488, indicating successful conjugation of scFv ligands to the surface of microbubbles (Fig. 3F and G). To prove binding specificity of MB_Thy1-scFv to both soluble human and murine Thy1 protein, a binding complex analysis between anti-His antibody-AF488 prelabeled MB_Thy1-scFv and AF647 postlabeled soluble human and murine Thy1 was performed; it showed a significant (P < 0.01) and strong positive correlation (human Thy1, R² = 0.97 ± 0.03; murine Thy1, R² = 0.95 ± 0.02), whereas MB_{scFv-scrambled} (R² = 0.19 ± 0.02; R² = 0.14 ± 0.01, respectively) and MB_Non-targeted (R² = 0.05 ± 0.04; R² = 0.03 ± 0.0.05, respectively) showed no correlation, indicating no binding to soluble AF647-labeled Thy1 (Fig. 3F and G). A binding complex between prelabeled Thy1 and MB_Thy1 (no His-tag) serving as positive control depicted an enhanced AF647 fluorescence intensity signal (Fig. 3F and G). These results confirmed that MB_Thy1-scFv bind specifically to both soluble human and murine Thy1 protein.

FACS analysis confirmed overexpression of human Thy1 on MS1_Thy1 cells, whereas MS1 WT cells did not express Thy1 (Fig. 4A). In flow chamber cell attachment studies simulating flow shear stress conditions in blood capillaries, the average numbers of MB_Thy1-scFv (3.0 ± 0.81 MB/cell) and MB_Thy1 (3.29 ± 0.75 MB/cell) attached per MS1_Thy1 cell were significantly higher (P < 0.01) than MB_{scFv-scrambled} (0.57 ± 0.53 MB/cell; Fig. 4B). Blocking of Thy1 receptors with free Thy1-scFv resulted in significantly (P < 0.01) decreased MB_Thy1-scFv attachment (1.14 ± 0.6 MB/cell), confirming binding specificity of MB_Thy1-scFv to Thy1-expressing MS1 cells. There was significantly lower (P < 0.01) attachment of MB_Non-targeted (0.43 ± 0.53 MB/cell) to Thy1-expressing MS1 cells compared with both MB_Thy1-scFv and MB_Thy1. Overall results suggested that MB_Thy1-scFv binds specifically to human Thy1–expressing cells under flow shear stress conditions (Fig. 4C).

To validate binding of MB_Thy1-scFv to both human and murine Thy1 in vivo, both an orthotopic human PDAC xenograft expressing human Thy1 on its vasculature as well as a transgenic mouse model of spontaneous PDAC expressing murine Thy1 on its neovasculature were used. In vivo ultrasound molecular imaging of orthotopic human PDAC xenografts showed significantly higher imaging signal following MB_Thy1-scFv (P < 0.01) and MB_Thy1 (P < 0.01) administration compared with control MB_Non-targeted (Fig. 5A and B). The imaging signals following administration of MB_Thy1-scFv (5.32 ± 1.59 a.u.) and MB_Thy1 (5.56 ± 1.57 a.u.) were not significantly different (P = 0.76). Following injection of control MB_{scFv-scrambled} (0.9 ± 0.63 a.u.), imaging signal was significantly (P < 0.01) lower compared with the signal after MB_Thy1-scFv and not significantly different (P = 0.65) compared with MB_Non-targeted (0.78 ± 0.48 a.u.; Fig. 5A and B). Furthermore, imaging signal was significantly lower (P < 0.01) in adjacent normal pancreas tissue after injection of all microbubbles types (MB_Thy1-scFv, 0.41 ± 0.23 a.u.; MB_Thy1, 0.38 ± 0.24 a.u.; MB_{scFv-scrambled}, 0.17 ± 0.2 a.u.; MB_Non-targeted, 0.3 ± 0.22 a.u.; Fig. 5A). To further confirm Thy1-binding specificity of MB_Thy1-scFv in orthotopic human PDAC xenografts, in vivo blocking of Thy1 receptors with free Thy1-scFv was performed and resulted in a significant decrease of Thy1-targeted imaging signal by 55.81% (P < 0.05, 2.38 ± 0.81 a.u.) compared with tumors without preadministration of the blocking agent (Fig. 5C and D).

Similarly, in spontaneous murine PDAC of transgenic mice, imaging signal was significantly higher following MB_Thy1-scFv (P < 0.01) and MB_Thy1 (P < 0.01) administration compared with MB_Non-targeted (Fig. 6A and B). Furthermore, the imaging signals following administration of MB_Thy1-scFv (5.68 ± 2.5 a.u.) and following MB_Thy1 (5.85 ± 2.1 a.u.) were not significantly different (P = 0.9). Also, following injection of MB_{scFv-scrambled} (1.21 ± 0.93 a.u.), imaging signal was significantly (P < 0.01) lower compared with the signal using MB_Thy1-scFv and not significantly different (P = 0.54) compared with MB_Non-targeted (0.92 ± 0.73 a.u.). Furthermore, for all microbubbles types, imaging signal was significantly lower (P < 0.01) in adjacent pancreas tissue of transgenic mice (MB_Thy1-scFv, 0.51 ± 0.3 a.u.; MB_Thy1, 0.47 ± 0.32 a.u.; MB_{scFv-scrambled}, 0.34 ± 0.26 a.u.; MB_Non-targeted, 0.26 ± 0.23 a.u.; Fig. 6A).

Also, as a negative control for nonangiogenic Thy1-negative vessels, normal pancreata in WT mice were scanned after intravenous administration of MB_Thy1-scFv, MB_Thy1, MB_{scFv-scrambled}, and MB_Non-targeted (Fig. 6C). Imaging signal in normal pancreata was significantly lower (P < 0.01; MB_Thy1-scFv, 0.67 ± 0.71 a.u.; MB_Thy1, 0.68 ± 0.29 a.u.; MB_{scFv-scrambled}, 0.32 ± 0.2 a.u.; MB_Non-targeted 0.21 ± 0.1 a.u.) for all microbubble types compared with PDAC (Fig. 6C).

Finally, all microbubble types were further tested in a chronic pancreatitis model as a model of benign disease and additional negative control. Imaging signal was also significantly lower (P < 0.01) in chronic pancreatitis tissues compared with PDAC for all microbubbles types (MB_Thy1-scFv, 0.84 ± 0.6 a.u.; MB_Thy1, 0.71 ± 0.73 a.u.; MB_{scFv-scrambled}, 0.48 ± 0.28 a.u.; MB_Non-targeted, 0.26 ± 0.18 a.u.).

Histologic analysis of tissue hematoxylin and eosin (H&E)–stained slices confirmed the presence of PDAC in orthotopic human xenografts and transgenic mice (Fig. 6D and E). Similar to in vivo ultrasound imaging results, fluorescence imaging showed accumulation of fluorescently labeled MB_Thy1-scFv within the neovasculature of PDAC in both orthotopic human xenografts and spontaneous PDAC in transgenic mice (Fig. 6F and G). Furthermore, ex vivo immunofluorescence analysis confirmed overexpression of human Thy1 on the neovasculature of orthotopic human PDAC xenografts and of murine Thy1 on the neovasculature of PDAC in transgenic mice. Quantitative immunofluorescence showed significantly (P < 0.01) increased colocalization of human (0.46 ± 0.09 a.u.) and murine (0.51± 0.08 a.u.; green) Thy1 on CD31-stained (red) tumor vessels compared with normal pancreata (0.10 ± 0.07 a.u.) and chronic pancreatitis tissues (0.13 ± 0.06 a.u.; Fig. 6H).

This study exemplifies the proof-of-principle approach from engineering a human Thy1-scFv ligand binding to both human and murine Thy1 to in vitro, cell culture, and in vivo validation of Thy1-targeted microbubbles for ultrasound molecular imaging of PDAC in two animal models, a human PDAC xenograft and a transgenic mouse model.

Ultrasound is a promising imaging technique for early cancer detection due to its inherent advantages compared with other imaging approaches, including wide availability, portability, relatively low cost, lack of ionizing radiation exposure, deep tissue penetration, and high spatial resolution (33, 34). Combined with endoscopy, ultrasound of the pancreas can also be performed in patients with limited acoustic windows. Adding contrast microbubbles that are molecularly-targeted at molecules expressed in PDAC can substantially increase the ability of ultrasound to detect and characterize small foci of cancer (35).

In this proof-of-principle study, we engineered a novel human Thy1-scFv ligand and tested whether it could be used for ultrasound molecular imaging of PDAC. Several antibodies from nonhuman species targeting Thy1 are commercially available that can be used for molecular imaging of PDAC. However, antibodies from nonhuman species are immunogenic and can cause severe allergic reactions in patients (36, 37); therefore, these ligands can only be used for preclinical imaging. Although many humanized antibodies have been successfully created and used in patients, the development and production of full-length antibodies against a specific target are challenging, costly, and time-consuming (38). Therefore, there is a critical need for an alternative platform to allow engineering of biocompatible, target specific, and high affinity binding ligands that are smaller, less expensive, and faster to produce than conventional full-length antibodies (39). Antibody-like molecules with exposed loops or surfaces that can be randomized, modified, and screened using selective phage or yeast engineered libraries are promising alternative platforms to engineer ligands that bind to molecular imaging targeted with high affinity and specificity (16). Several types of engineered binding ligands have been explored for ultrasound molecular imaging including knottins (40), nanobodies (41), fibronectins (42), and peptides (43). Engineered scFvs are an alternative platform for designing binding ligands for molecular imaging with several advantages (16, 44). Because scFvs are smaller than full-length antibodies owing to the lack of an Fc fragment, the risk for immunogenicity is minimized (45). Also, recombinant scFvs can be produced cost efficiently. Furthermore, scFvs are versatile and can be modified by adding site-specific tags for detection and bioconjugation (46). Finally, scFvs have already shown good safety profile in phase I and II clinical trials in patients (19, 47).

In this study, we have used a naïve nonimmune human scFv-yeast surface display library for engineering and screening of scFv binding to human and murine Thy1. Our results showed that the engineered human Thy1-scFv ligand from a human YSD library not only binds to soluble human and murine Thy1 but also binds to Thy1 expressed on live cells, but that it can also be used to functionalize contrast microbubbles for ultrasound imaging PDAC in vivo. For testing in vivo binding to murine Thy1, a well-established transgenic mouse model was used that resembles spontaneous development of PDAC in patients (31). This animal model has been previously shown to overexpress murine Thy1 on its neovasculature (6, 48), which was confirmed by quantitative immunofluorescence in our current study. For testing in vivo binding to human Thy1, we used a previously described orthotopic human PDAC xenograft model (6) which expresses human Thy1 on its neovasculature by coinjecting human PDAC cells with vascular endothelial cells expressing human Thy1. These human Thy1–expressing cells are integrated into the neovasculature of the xenografts during angiogenesis and tumor growth, thereby allowing microbubbles to be tested for binding to human Thy1 in a mouse model before moving into clinical trials.

We acknowledge the following limitation of our study. For this proof of concept study, biotin/streptavidin coupling was used for conjugating engineered Thy1-scFv onto microbubbles. Although this is a well-established approach for preclinical testing of new contrast microbubbles due to its flexible, quick, and site-specific conjugation chemistry, these types of microbubbles are not useful for clinical translation due to the immunogenicity of streptavidin (36). Efforts are under way to develop site-specific modifications on Thy1-scFv to enable alternative bioconjugation of the ligands to the microbubble shell for eventual clinical translation. Also, although transgenic mice may develop multiple tumor foci in the pancreas, we imaged only one focus of PDAC in this study to evaluate imaging signals obtained from different types of microbubbles in an intra-animal comparison study without assessing the whole pancreas for additional foci of cancer. Also, due to the limited number of imaged tumors, imaging signal was not stratified by tumor size. Future prospective studies are warranted to assess the minimally detectable tumor size along with sensitivity and specificity of Thy1-targeted contrast microbubbles for ultrasound molecular imaging of PDAC.

In conclusion, our results suggest that a novel Thy1-scFv ligand can be engineered for improved detection of PDAC using ultrasound molecular imaging. Ongoing developments are aimed at designing a next-generation clinically translatable Thy1-targeted contrast microbubble.

No potential conflicts of interest were disclosed.

Conception and design: L. Abou-Elkacem, R.H. Kimura, S.S. Gambhir, J.K. Willmann

Development of methodology: L. Abou-Elkacem, H. Wang, S.S. Gambhir, J.K. Willmann

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Abou-Elkacem, S.M. Chowdhury, S.V. Bachawal, J.K. Willmann

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Abou-Elkacem, H. Wang, S.V. Bachawal, S.S. Gambhir, L. Tian, J.K. Willmann

Writing, review, and/or revision of the manuscript: L. Abou-Elkacem, H. Wang, R.H. Kimura, S.V. Bachawal, S.S. Gambhir, L. Tian, J.K. Willmann

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Abou-Elkacem, H. Wang, S.S. Gambhir, J.K. Willmann

Study supervision: L. Abou-Elkacem, S.S. Gambhir, J.K. Willmann

We would like to acknowledge the SCi³ Core imaging facilities and the Canary Center at Stanford for the technical support of this study.

This research was supported by the NIH R01CA155289 grant, the NIHR21EB022770, the R41CA203090, NIH U01CA210020, and the Canary Foundation. The Mildred-Scheel Cancer Foundation from Germany provided funding for Dr. Lotfi Abou-Elkacem to study abroad at Stanford University.

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.