Targeted In vivo Imaging of Integrin α_vβ₆ with an Improved Radiotracer and Its Relevance in a Pancreatic Tumor Model

In recent years, positron emission tomography (PET) imaging, using 2-[¹⁸F]fluoro-2-deoxy-d-glucose ([¹⁸F]FDG), has become routine for detection of tumors and monitoring of treatment response (1–3). [¹⁸F]FDG, a radiolabeled analogue of glucose, is taken up and trapped primarily in cells with high metabolism. As glucose hypermetabolism is commonly found in inflammation, as well as in many malignancies, distinguishing cancer from other hypermetabolic states remains a challenge (1, 4, 5). Although PET has been combined with other imaging modalities, most notably computed tomography (CT), which relies on subtle differences in soft tissue density to distinguish between normal and malignant tissues (5–7), [¹⁸F]FDG-PET/CT still leaves considerable room for improvement in diagnosis and clinical management of many cancers (8). The availability of more disease-specific imaging agents (7) could help to mitigate the limited specificity of [¹⁸F]FDG for malignant tissue. Rather than relying on slight changes in tissue density or metabolic rate alone, targeting disease-specific tissue markers could considerably aid in tumor detection and localization. The feasibility of this approach has been shown with [¹⁸F]galacto-RGD, a small radiolabeled peptide used for PET imaging of the cancer-related integrin α_vβ₃ (9). In contrast to α_vβ₃, other members of the integrin family have received relatively little attention. However, newly emerging evidence indicates that α_vβ₆ may also be an important target for diagnosis and treatment of cancer (10–14). The integrin α_vβ₆ was first identified in a human pancreatic carcinoma cell line (15). Subsequent studies established that although α_vβ₆ is epithelial specific, its expression is low or undetectable in healthy adult tissues. Up-regulation has been recognized during tissue remodeling, including inflammation and wound healing (16–18). Significantly, expression has also been shown to be increased in many different cancers (13, 19). Among gastroenteropancreatic adenocarcinomas, α_vβ₆ expression was found to be strongest in pancreatic ductal adenocarcinomas (PDAC; ref. 20). Moreover, α_vβ₆ has recently been described as a prognostic indicator, with high levels of expression correlating with poor prognosis for cancer of the colon, cervix, lung, and stomach (10–12, 21). Antibody-blockade of α_vβ₆ was shown to inhibit tumor progression in vivo in animal models (13). Thus, mounting evidence implicates α_vβ₆ as an important and biologically relevant target for molecular imaging of cancer.

Recently, we showed the ability to selectively image α_vβ₆-expressing tumors in vivo with microPET using the α_vβ₆-specific [¹⁸F]fluorobenzoyl-labeled peptide [¹⁸F]FBA-A20FMDV2 (22). Whereas targeted imaging was achieved, low uptake and poor retention in the target tissue limited its general utility. For this current study, we conjugated small, monodisperse polyethylene glycol (PEG) polymers to A20FMDV2 to improve pharmacokinetics, generating the two new radiotracers [¹⁸F]FBA-PEG₂₈-A20FMDV2 and [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2. We report that both PEGylated A20FMDV2 variants showed significantly improved retention in two different α_vβ₆-expressing human tumor xenograft models: DX3puroβ6, a human melanoma cell line transduced to express α_vβ₆, and BxPC-3, a human pancreatic carcinoma cell line that endogenously expresses α_vβ₆. In the BxPC-3 model, retention of [¹⁸F]FBA-PEG₂₈-A20FMDV2 in tumors was 12-fold greater than retention of the non-PEGylated [¹⁸F]FBA-A20FMDV2. Tumor/pancreas and tumor/blood ratios of >23:1 and >47:1, respectively, were achieved with [¹⁸F]FBA-PEG₂₈-A20FMDV2 at 4 hours. The ability to noninvasively detect regional tumor invasion or occult metastatic disease using α_vβ₆ PET imaging would have a large effect on patient management, reducing the number of unnecessary surgeries of unresectable disease (23).

Reagents and cell lines. Reagents, materials, and DX3puro/DX3puroβ6 cell lines have been described previously (22). Pancreatic cancer cell lines were obtained from American Type Culture Collection. Transforming growth factor β1 human latency-associated peptide (TGFβ1–LAP) was purchased from Sigma. A Leica CM1850 cryostat (Leica Microsystems) was used for tissue sectioning.

Competitive binding ELISA. The in vitro efficacies of the peptides toward integrins α_vβ₆, α_vβ₃, α_vβ₅, α₅β₁, and α_IIbβ₃ were analyzed by competitive binding ELISA as described previously (22).

Flow cytometry. Integrin expression on the cell lines was determined as previously described (24).

Cell binding and internalization. A cell suspension (3.75 × 10⁶ cells) was incubated with the radiotracer (0.2 μCi) in serum-free medium (pH 7.2, 100 μL). For each radiotracer, the cell lines were assayed concurrently (n = 4 samples per cell line per time point). The fraction of bound radioactivity was determined with a γ-counter (cell pellet versus supernatant). To determine the fraction of internalized radioactivity, the cells were then treated with acidic wash buffer [0.2 mol/L sodium acetate, 0.5 mol/L sodium chloride (pH 2.5), 4°C] to release surface-bound activity (25). The internalized fraction was determined with a γ-counter (cell pellet versus radioactivity released into supernatant).

Xenograft cell lines and animal model. All animal studies were carried out using male athymic nude mice (nu/nu; Charles River Laboratories) following approved protocols. DX3puroβ6 (α_vβ₆ positive) and DX3puro (control) cells were injected s.c. (n = 20 mice; 3 × 10⁶ cells/cell line in 100 μL serum-free DMEM; Invitrogen) on opposite flanks in the shoulder region, as described previously (22). BxPC-3 cells were injected s.c. (n = 30 mice; 1.5 × 10⁶ cells in 100 μL 1:1 serum-free DMEM/Matrigel HC; BD Biosciences) in the left shoulder region. Studies commenced once tumors had reached a diameter of ∼0.5 cm.

Biodistribution studies. Radiotracers (∼15–30 μCi) in saline/PBS (150–200 μL) were injected, and data were analyzed as previously described (n = 3 per time point; ref. 22).

MicroPET imaging studies. The [¹⁸F]FBA peptides or [¹⁸F]FDG (∼120–230 μCi) in saline/PBS (100–200 μL) were injected as previously described (n = 3 per time point; ref. 22). Two animals were imaged side-by-side in a feet-first prone position. For blocking studies, the [¹⁹F]FBA peptide (30 mg/kg as 10 mg/mL solution in saline) was injected 10 min before the radiotracer. Imaging data were collected and analyzed as previously described (22, 26).

Standard uptake values (SUV) were computed by dividing the activity concentration in each voxel by the injected dose and then multiplying by the weight of the animal (26). Maximum intensity projection (MIP) images were created by collapsing across one dimension with the maximum value in each column of voxels.

Autoradiography studies. Frozen tissue slices (tumor and surrounding tissue, 50 μm) in freezing medium (Tissue-Tek, Sakura Finetek) were exposed to a storage phosphor screen (GE Healthcare) overnight. The screen was read at a 50-μm resolution using a Storm 860 phosphorimager (GE Healthcare).

Statistical analysis. The data are reported as mean ± SD. Two-tailed Student's t tests were performed to evaluate statistical significance. P < 0.05 was considered significant. In plots, the SD is smaller than the size of the symbol for some data points.

PEGylated peptide ligands for α_vβ₆. PEGylated A20FMDV2 variants were compared by ELISA to the established α_vβ₆ peptide ligand FBA-A20FMDV2 (Fig. 1). They exhibited IC₅₀ of 3 to 6 nmol/L, similar to FBA-A20FMDV2. Inhibition of binding by the natural ligands to integrins α_vβ₃, α_vβ₅, α₅β₁, and α_IIbβ₃ required at least >1,000-fold higher concentrations of peptide (IC₅₀ >10 μmol/L), indicating that introduction of the PEG moieties did not have a deleterious effect on α_vβ₆ affinity and selectivity of A20FMDV2.

Integrin α_vβ₆ expression in human pancreatic cancer cell lines. Seven cell lines were screened by flow cytometry for α_vβ₆ expression (Fig. 2). For reference, the melanoma cell line DX3puroβ6, stably transduced to express α_vβ₆, and its paired control line DX3puro, lacking expression of α_vβ₆, were analyzed as positive and negative controls, respectively; both of these cell lines express similar levels of other integrins, including α_vβ₃, α_vβ₅, and α₅β₁ (22). BxPC-3 and Capan-2 showed high α_vβ₆ expression levels comparable with that of the positive control DX3puroβ6. Capan-1 and L3.3 showed intermediate expression of α_vβ₆, whereas MIA-PaCa2, Panc-1, and BCL-2cl9 showed expression levels equivalent to that of the negative control DX3puro.

Cell binding and internalization. Binding of the radiotracers [¹⁸F]FBA-(PEG₂₈)_n-A20FMDV2 (n = 0, 1, 2; Supplementary Fig. S1 and S2; ref. 22) to DX3puroβ6, DX3puro, and BxPC-3 cell lines was strongly dependent on the expression of α_vβ₆ by the cells and on the presence of a PEG unit in the radiotracer (Fig. 3). Whereas <2% of any radiotracer bound to the α_vβ₆-negative DX3puro within 1 hour (Fig. 3A), 5.3 ± 0.8% and 9.7 ± 1.2% of [¹⁸F]FBA-A20FMDV2 bound to the α_vβ₆-expressing DX3puroβ6 and BxPC-3 cell lines, respectively. For [¹⁸F]FBA-PEG₂₈-A20FMDV2, the binding levels increased significantly to 73.5 ± 1.3% and 68.9 ± 0.2%, respectively. Similarly, [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2 showed markedly elevated binding levels of 38.5 ± 1.7% and 25.8 ± 0.6%, respectively. Approximately two-thirds of the bound radioactivity was internalized in the α_vβ₆-expressing cell lines in the case of the two PEGylated radiotracers, whereas less than half of the bound [¹⁸F]FBA-A20FMDV2 was internalized. Incubation in the presence of excess nonradioactive [¹⁹F]FBA-PEG₂₈-A20FMDV2 (≥1μmol/L) completely abrogated binding of the radiotracer to DX3puroβ6 cells (Supplementary Fig. S3).

Figure 3B illustrates the rapid cell binding of radiotracer. Within 1 minute, approximately one quarter of [¹⁸F]FBA-PEG₂₈-A20FMDV2 bound to DX3puroβ6 and BxPC-3 (30.5 ± 2.7% and 24.5 ± 1.2%, respectively). Binding levels approached their maximum within 15 to 30 minutes and then remained steady. When the two cell lines were incubated with [¹⁸F]FBA-PEG₂₈-A20FMDV2 for 1 hour in the presence of LAP, the radiotracer was able to efficiently compete for binding; binding levels in presence of 5 nmol/L LAP remained at ≥75% of those observed under LAP-free conditions (Supplementary Fig. S4).

[¹⁸F]FBA-(PEG₂₈)_n-A20FMDV2 (n = 0, 1, 2) biodistribution in the DX3puro/DX3puroβ6 model. Initial in vivo evaluation of [¹⁸F]FBA-PEG₂₈-A20FMDV2 and [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2 was performed using our previously described mouse model bearing the paired human xenografts of DX3puroβ6 (α_vβ₆ positive) and DX3puro (α_vβ₆ negative; Supplementary Table S1). Whereas uptake of both PEGylated radiotracers in the target tissue at 1 hour postinjection was slightly below the level previously seen for the non-PEGylated [¹⁸F]FBA-A20FMDV (22), the introduction of the PEG unit(s) resulted in excellent retention of radioactivity in the α_vβ₆-positive tumor ([¹⁸F]FBA-PEG₂₈-A20FMDV2, 1 hour: 0.49 ± 0.12% ID/g, 4 hours: 0.49 ± 0.04% ID/g; [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2, 1 hour: 0.52 ± 0.09% ID/g, 4 hours: 0.54 ± 0.08% ID/g; compared with [¹⁸F]FBA-A20FMDV2, 1 hour: 0.66 ± 0.09% ID/g, 4 hours: 0.06 ± 0.00% ID/g). A comparison of the uptake in the two xenografts consistently showed preferential uptake in the α_vβ₆-positive xenograft over the α_vβ₆-negative xenograft, particularly for the PEGylated radiotracers. Whereas [¹⁸F]FBA-A20FMDV resulted in a 4.3:1 maximal α_vβ₆-positive/α_vβ₆-negative ratio, ratios of 8.0:1 and 9.2:1 were seen for [¹⁸F]FBA-PEG₂₈-A20FMDV2 and [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2, respectively (for P values, see Supplementary Table S1).

[¹⁸F]FBA-(PEG₂₈)_n-A20FMDV2 (n = 0, 1, 2) biodistribution in the α_vβ₆-endogenous BxPC-3 model. The feasibility of α_vβ₆-targeted pancreatic tumor detection in vivo was evaluated in the BxPC-3 mouse model (Fig. 4; Supplementary Table S1). For [¹⁸F]FBA-A20FMDV2 initial uptake in and washout from the BxPC-3 xenograft (1 hour: 0.69 ± 0.19% ID/g, 4 hours: 0.12 ± 0.03% ID/g) were similar to those seen in the DX3puroβ6 xenograft (1 hour: 0.66 ± 0.09% ID/g, 4 hours: 0.06 ± 0.00% ID/g). By contrast, for both of the PEGylated radiotracers uptake in the BxPC-3 xenograft 1 hour after injection was three to four times higher than in the DX3puroβ6 xenograft, combined with excellent retention ([¹⁸F]FBA-PEG₂₈-A20FMDV2, 1 hour: 1.85 ± 0.44% ID/g, 4 hours: 1.48 ± 0.04% ID/g; [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2, 1 hour: 1.57 ± 0.25% ID/g, 4 hours: 2.08 ± 0.37% ID/g). Autoradiography images obtained for the PEGylated radiotracers of BxPC-3 xenografts 1 hour after injection also showed preferential, tumor-specific uptake (Fig. 5).

Renal clearance was the major route of elimination for all three radiotracers (Fig. 4 and data not shown). High-performance liquid chromatography (HPLC) analyses of urine samples showed three metabolites for [¹⁸F]FBA-A20FMDV2 and one major metabolite for each of the PEGylated radiotracers. Distribution of the radiotracers in healthy tissues generally depended on the number of PEG units present (Fig. 4; Supplementary Table S1). Non-PEGylated [¹⁸F]FBA-A20FMDV2 was washed out efficiently from all tissues examined, with gall bladder and kidneys containing the highest levels of radioactivity. By contrast, the di-PEGylated [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2 was generally retained throughout, most notably in the kidneys (∼42% ID/g); only blood showed clearing of radioactivity. Pharmacokinetics of the mono-PEGylated [¹⁸F]FBA-PEG₂₈-A20FMDV2 in healthy tissues compared favorably to the di-PEGylated radiotracer, resulting in a desirable clearing behavior resembling that observed for the non-PEGylated compound in healthy tissues; highest levels of radioactivity were detected in the kidneys and the gall bladder.

The BxPC-3/blood ratio remained relatively low for [¹⁸F]FBA-A20FMDV2 (1 hour: 3.3:1, 4 hours: 7.0:1) and the BxPC-3/muscle ratio only reached statistically significant levels at the 4-hour time point (1 hour: 1.3:1, P = 0.38; 4 hours: 2.5:1, P = 0.02) because of more rapid washout from healthy tissues than from the tumor (Fig. 4, Supplementary Table S1). By comparison, improved BxPC-3/muscle ratios were seen for [¹⁸F]FBA-PEG₂₈-A20FMDV2 (1 hour: 2.5:1, 4 hours: 3.6:1) and [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2 (1 hour: 4.1:1, 4 hours: 4.4:1), along with significantly higher BxPC-3/blood ratios ([¹⁸F]FBA-PEG₂₈-A20FMDV2, 1 hour: 11.8:1, 4 hours: 47.9:1; [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2, 1 hour: 9.0:1, 4 hours: 150:1), reflecting the improved retention of these radiotracers in the BxPC-3 tumor.

Importantly, levels of radioactivity in the BxPC-3 tumor were at least 4-fold higher than those for the pancreas itself for all radiotracers at all time points measured (Supplementary Table S1). For [¹⁸F]FBA-A20FMDV2, the BxPC-3/pancreas ratio rose to 9.5:1 at 4 hours postinjection due to the more efficient washout from the pancreas than from the tumor. On the other hand, because of retention of [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2 in healthy pancreas, the BxPC-3/pancreas ratio for this radiotracer remained nearly unchanged throughout (1 hour: 5.6:1, 4 hours: 4.4:1). [¹⁸F]FBA-PEG₂₈-A20FMDV2 exhibited good retention in the tumor together with efficient washout from the pancreas (and other healthy tissues), resulting in a steadily increasing BxPC-3/pancreas ratio (1 hour: 4.2:1, 4 hours: 23.6:1). In summary, the mono-PEGylated [¹⁸F]FBA-PEG₂₈-A20FMDV2 had the most favorable pharmacokinetic characteristics.

MicroPET imaging of the α_vβ₆-endogenous BxPC-3 pancreatic cancer model. Figure 5A depicts representative coronal and transaxial MIPs obtained with the three [¹⁸F]FBA peptide tracers and with [¹⁸F]FDG (1 hour postinjection). Tumor uptake was compared based on maximum SUVs (SUV_max; n = 3 animals).

The observations made for the peptide radiotracers during microPET imaging paralleled the results seen in the biodistribution study with preferential uptake noted in the tumor along with major uptake in kidneys, bladder (urine), and, for [¹⁸F]FBA-A20FMDV2, gall bladder. [¹⁸F]FBA-A20FMDV2 showed moderate initial tumor uptake and significant washout over time (SUV_max, 15 minutes: 0.50 ± 0.22, 1 hour: 0.22 ± 0.11; 4 hours: < 0.05). [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2 showed improved uptake and retention in the tumor (SUV_max, 15 minutes: 0.99 ± 0.18, 1 hour: 0.79 ± 0.24, 4 hours: 0.59 ± 0.20), along with significant renal retention. [¹⁸F]FBA-PEG₂₈-A20FMDV2 combined favorable tumor uptake characteristics (SUV_max, 15 minutes: 1.07 ± 0.29, 1 hour: 0.91 ± 0.25, 4 hours: 0.63 ± 0.12) with good renal clearance.

Administration of nonradioactive [¹⁹F]FBA-PEG₂₈-A20FMDV2 (30 mg/kg) 10 minutes before the corresponding [¹⁸F]FBA peptide greatly reduced BxPC-3 tumor uptake [Δ(SUV_max) = −77%; Fig. 5B]. Similar results were obtained for [¹⁹F]FBA-(PEG₂₈)₂-A20FMDV2 (30 mg/kg)/[¹⁸F]FBA-peptide [Δ(SUV_max) = −60%; Fig. 5C]. Interestingly, [¹⁸F]FBA-PEG₂₈-A20FMDV2 and [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2 were retained in the mouse oral cavity in scans without prior administration of nonradioactive FBA-peptide, suggesting that there may be limited expression of α_vβ₆ (27).

MicroPET scans obtained with [¹⁸F]FDG revealed an SUV_max of 0.62 ± 0.13 in the BxPC-3 tumor (1 hour postinjection, n = 3; Fig. 5A). This level of tumor uptake is relatively low for [¹⁸F]FDG, and it was within the range observed for the surrounding tissue (∼0.4–0.8). Thus, whereas the [¹⁸F]FBA peptide tracers clearly revealed the BxPC-3 tumor in vivo, [¹⁸F]FDG did not.

The integrin α_vβ₆ has been identified as a prognostic biomarker for cancer of the colon, cervix, lung, and stomach (10–12, 21). In addition, α_vβ₆ is reported to be expressed by many other types of carcinoma (13, 19) whereas corresponding normal tissue is weak or negative. Thus, α_vβ₆ represents an excellent target for imaging and therapy of carcinoma. We previously showed the ability to selectively image α_vβ₆-expressing tumors in vivo using the PET radiotracer [¹⁸F]FBA-A20FMDV2 (22). While targeted imaging was achieved, low uptake and poor retention in the target tissue, together with metabolic instability, limited its general utility, leading us to explore strategies to improve pharmacokinetics. As introduction of PEG moieties (PEGylation) generally increases biological half-life, metabolic stability, and tumor-uptake (28, 29), we generated two variants of A20FMDV2 that included PEG moieties on the N-terminus of the peptide. These PEGylated variants of [¹⁸F]FBA-A20FMDV2 were examined and compared with the parent compound. Monodisperse PEG polymers of moderate molecular weight [Fw(PEG₂₈) = 1.3 kDa] are short enough to be compatible with solid-phase chemistries yet large enough to be expected to approach a range with some beneficial in vivo (pharmacokinetic) properties. Additionally, owing to the monodisperse nature of the polymer, the final product is obtained as a single compound, allowing for precise characterization and batch-to-batch reproducibility, in contrast to mixtures containing polymer chains of different lengths encountered with the larger PEG products available.

As presented here, initial in vitro evaluation by ELISA indicated unchanged, high α_vβ₆-specificity for the PEGylated compounds (Fig. 1). Further binding experiments with the DX3puroβ6 and DX3puro cell lines showed strong, α_vβ₆-specific and PEG-dependent binding to the receptor expressed on cells in the presence of other related integrins (including α_vβ₃, α_vβ₅, and α₅β₁; Fig. 3). In vivo studies also showed improved, α_vβ₆-targeted PEG-dependent tumor retention in the DX3puroβ6/DX3puro model (Supplementary Table S1). Taken together, these data confirmed our initial expectations for the two new PEGylated radiotracers, [¹⁸F]FBA-PEG₂₈-A20FMDV2 and [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2.

The DX3puroβ6/DX3puro melanoma cancer model was generated by viral transduction of human β6 into human melanoma cells. To examine a model more closely resembling the actual human disease, we sought a cancer wherein the expression of α_vβ₆ was relevant to the disease and for which cell lines that had endogenous expression of α_vβ₆ existed; we chose to investigate pancreatic cancer. Sipos and colleagues (20) reported, as part of a study of gastroenteropancreatic adenocarcinomas, that α_vβ₆ expression was strongest in PDACs. Thus, of the 34 human PDAC samples analyzed, 32 (94%) received the maximum score (the mean score was 2.8/3.0, with none of the other carcinomas receiving a score over 1.5). These observations make α_vβ₆ a very appealing target for selective in vivo detection of pancreatic cancer. The possibility of detecting primary tumors and metastases earlier by imaging α_vβ₆ and, thus, directing therapy sooner may improve upon the very poor prognosis of this disease. In addition, recent studies associate strong expression of α_vβ₆ with poor prognosis (for colon, lung, cervical, and stomach cancers; refs. 10–12, 21), so the relevance of our data is not restricted only to pancreatic cancer.

As shown here by flow cytometry (Fig. 2), many common pancreatic carcinoma cell lines exhibit an α_vβ₆ profile that reflects expression seen for PDAC in the clinic (20). In vitro cell-binding assays showed that [¹⁸F]FBA-PEG₂₈-A20FMDV2 performed equally well with the endogenously α_vβ₆-expressing pancreatic BxPC-3 cell line and the transduced α_vβ₆-expressing DX3puroβ6 cell line, showing rapid and high binding together with significant internalization (Fig. 3). Contrary to conditions in the in vivo environment, these assays were performed in the absence of natural ligands competing for binding to α_vβ₆. We, therefore, sought to further show their relevance as predictors for in vivo performance by carrying out a binding assay in the presence of a biologically relevant α_vβ₆ ligand. LAP was chosen as it binds to α_vβ₆ with high affinity and has significance in the context of tumor biology (13, 18, 30). In association with TGFβ, particularly TGFβ1, LAP binds to α_vβ₆, thereby affecting signaling pathways understood to play significant roles in tumor development (18, 30). We found that, even in the presence of 5 nmol/L LAP, the radiotracer, present at subpicomolar levels, was still able to efficiently bind to the cells (at ≥75% of binding observed under LAP-free conditions; Supplementary Fig. S4). The feasibility of evaluating α_vβ₆-specific in vivo imaging radiotracers for pancreatic carcinoma was shown in a mouse model bearing BxPC-3 xenografts. In microPET images, the BxPC-3 tumors were clearly identifiable throughout. These data confirm that selective in vivo imaging of α_vβ₆-expressing cancers is possible with cancers that endogenously express α_vβ₆ and not only those engineered to express it. Importantly for pancreatic imaging, the best pharmacokinetic behavior was observed for [¹⁸F]FBA-PEG₂₈-A20FMDV2, resulting in the highest tumor/pancreas ratio (>23:1; 4 hours), making this radiotracer a promising lead compound for in vivo detection.

Close examination of the biodistribution data revealed that introduction of the second PEG unit was not beneficial. Simply modifying from [¹⁸F]FBA-PEG₂₈-A20FMDV2 to [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2 resulted only in a slight increase in tumor uptake, whereas, concurrently, washout of radioactivity from healthy tissue (except blood) was suppressed over the time span examined; particularly notable are the high and constant levels of radioactivity in the kidneys (Fig. 4). PET images also showed that [¹⁸F]FBA-PEG₂₈-A20FMDV2 outperformed [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2; it showed slightly higher uptake (SUV_max) together with better BxPC-3/background ratios, largely due to preferential clearing of nonspecifically bound activity (Fig. 5). Thus, in this study, [¹⁸F]FBA-PEG₂₈-A20FMDV2 seemed to have the best balance of α_vβ₆-dependent tumor uptake and a suitable rate of clearance from healthy tissues, including the pancreas. The high retention of [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2 in healthy tissues, caused by introduction of the second PEG unit, had not been predicted. These observations seem to stand in contrast to the common belief that renal clearance remains highly efficient for compounds bearing small PEG units (under ∼30 kDa; refs. 28, 29). Despite the comparatively small PEG units used in this study, the PEGylation of A20FMDV2 seemed to result in an “enhanced permeation and retention” effect toward the BxPC-3 tumors along with some improved radiotracer stability (29). Positive effects of PEGylation on the in vivo stability of pharmaceuticals are generally recognized (28). Likewise, in our study, only one major radioactive metabolite was found in HPLC samples of urine collected 1 hour after injection of [¹⁸F]FBA-PEG₂₈-A20FMDV2 or [¹⁸F]FBA-(PEG₂₈)₂-A20FMDV2, whereas the non-PEGylated analogue had yielded three equally large metabolite signals (22). However, because the cell-based assays showed very rapid binding (∼25% of [¹⁸F]FBA-PEG₂₈-A20FMDV2 bound to BxPC-3 within 1 minute; Fig. 3B) and significant, PEG-dependent internalization (approximately two thirds of bound activity were internalized at 1 hour; Fig. 3A), delivery efficiency to the target likely was of principal importance for these radiotracers, provided that radioactive metabolites were rapidly cleared from the body, as observed particularly for [¹⁸F]FBA-PEG₂₈-A20FMDV2. Taken together, these data suggest that rapid binding and internalization of the PEGylated radiotracers in cells are key to increased uptake and retention in the tumors. That being said, a detailed characterization of metabolites and further modification of the radiotracers remain of interest. They include selective deletion of redundant amino acids and replacement with unnatural amino acids in conjunction with multimeric radiotracers and incorporation of branched PEG. These approaches have already been successful in improving the pharmacokinetics of tracers for integrin α_vβ₃ (28, 31, 32).

Paralleling the in vitro blocking studies with the α_vβ₆-expressing cell lines, in vivo blocking experiments using the nonradioactive analogues showed a significant reduction of uptake in the BxPC-3 tumors. Before the blocking experiments, the animals were also evaluated by [¹⁸F]FDG-PET, the standard radiotracer for PET imaging. Significantly, [¹⁸F]FDG failed in all animals to detect the BxPC-3 tumors. In the clinical setting, [¹⁸F]FDG is by far the most widely used PET radiotracer for the detection of many cancers, including pancreatic cancer, despite the risk of false positive (e.g., pancreatitis) or false negative (e.g., low GLUT-1 glucose transporter expression in tumor; hyperglycemic patient) diagnosis. Thus, our observation parallels findings from the clinical setting wherein [¹⁸F]FDG is estimated to miss over one third of all pancretic malignancies (and about half of the lesions smaller than 1 cm; refs. 6, 8, 33).

Better understanding of the biology of pancreatic cancer (34) and improved early, more accurate diagnosis are sorely needed. By combining detailed anatomical information with functional metabolic information, [¹⁸F]FDG-PET/CT has shown some promise (5, 6, 8, 33). However, the use of [¹⁸F]FDG as radiotracer does not take advantage of the molecular biological differences of tumors beyond their increased (glucose) metabolism. The integrin α_vβ₆ is expressed strongly by most human pancreatic cancers (20) and is likely to be a prognostic indicator for this cancer. Therefore, it is clear that targeted in vivo imaging of α_vβ₆ will be a major improvement for preoperative staging and monitoring response to treatment for this lethal malignancy in humans. Using a metabolic probe ([¹⁸F]FDG) in conjunction with receptor-specific radiotracers in PET/CT could provide crucial additional information, particularly for occult disease. In a step toward making this a clinical reality, we have shown that [¹⁸F]FBA-PEG₂₈-A20FMDV2 holds promise for α_vβ₆-specific tumor imaging.

No potential conflicts of interest were disclosed.

Grant support: NIH grant R21 CA107792.

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.

We thank J. Choi, D.L. Kukis, L. Planutyte, C.D. Griesemer, J.Y. Fung, and S.V. Rendig for excellent technical support.