Identification of a Novel ITGα_vβ₆-Binding Peptide Using Protein Separation and Phage Display

This article is featured in Highlights of This Issue, p. 3975

Head and neck squamous cell carcinoma (HNSCC) represent more than 90% of all head and neck cancers and account for more than 5% of all malignant tumors worldwide (1). Because of a high prevalence for the development of lymph node metastases, high recurrence rates, and an increased occurrence of secondary tumors the 5 years survival rate is less than 30% (2). There is a lot of hope that patient outcome might be improved by better imaging tools for more precise tumor detection as well as by the help of targeted therapies allowing for a personalized approach depending on the individual tumor phenotype. Accordingly, the antiangiogenic peptide RGD cilengitide has been shown to specifically target the ITGα_vβ₃ receptor, which is highly expressed on HNSCC (3). Not only HNSCC-radiolabeled RGD peptides have been used to image neoangiogenesis not only in HNSCC, but also in a variety of other tumor types (4, 5). However, binding of these peptides occurs predominantly on endothelial cells of neovessels rather than on tumor cells mostly resulting in weak signal intensity. Therefore, identification of tumor cell-binding tracers is an urgent need especially in the context of metastasis and micrometastasis.

In this regard, phage display has proven to be a powerful tool to identify novel peptides with high specificity for a variety of molecules on tumor cells or on tumor neovasculature. Even without knowing the exact target, phage display on viable cells allowed for the identification of tumor cell-specific peptides that have been successfully used as carriers of chemotherapeutic drugs and toxic molecules (6).

To improve the likelihood to identify high affine and stable peptides binding to tumor cell–specific target molecules of HNSCC, we used a novel approach by combining the two-dimensional protein fractionation system ProteomeLab PF2D (7–9) with phage display. To enrich for tumor cell binders before phage display isolated tumor cell membranes were subjected to protein fractionation. The phage display SFTI8Ph library used is based on the molecular scaffold of the sunflower trypsin inhibitor (SFTI)-1, a 14-residue cyclic miniprotein isolated from sunflower seeds which is stabilized by a single disulfide bridge (10). Employing the SFTI8Ph library for alternate biopanning against the HNSCC cell line HNO97 and the respective separated tumor cell-specific PF2D membrane protein fractions we identified a SFTI-1 derivate containing the amino acid sequence FRGDKMQL with specificity for different HNSSC and squamous carcinoma cell lines of other tumor entities. The presence of the arginine-glycine-aspartate (RGD) motif flanked by the KXXL sequence within the amino acid sequence implicates that the peptide represents a ligand for ITGα_vβ₆ (11, 12). Integrins comprise a large family of cellular adhesion receptors that play an important role in development, immune response and cancer (13). The RGD motif occurs in many extracellular matrix ligands of integrins; however, the motif DLXXL was identified by phage display within 7- and 12-residue peptides as a ITGα_vβ₆-specific binding motif (14, 15). Because ITGα_vβ₆ is overexpressed in many carcinomas associated with poor prognosis including lung, pancreatic, ovarian, colorectal and cervical cancers, but only at low or undetectable levels in normal tissues (16) it might represent an important target for imaging and anticancer therapies (17–23).

The data presented here emphasize the clinical value of the identified ITGα_vβ₆-specific binding peptide as a diagnostic tool and possibly useful ligand for endoradiotherapy of HNSCC and other ITGα_vβ₆-expressing tumors.

To harvest membrane proteins HNO97 and HPV16GM cells (for further information regarding cell lines and patient samples see Supplementary Methods) were grown to 90% confluence in multilayer flasks (HYPERFlask, Corning) and detached by pre-incubation with PBS/0.5% EDTA and subsequent 0.025% trypsin treatment. Membrane proteins were extracted according to an ultracentrifugation-based protocol. Briefly, cells were washed, pelleted and lysed by mechanical dissociation using a dounce tissue grinder (Wheaton) with 4 mL of lysis buffer (50 nmol/L TRIS (pH 7.3), 250 mmol/L sucrose, 2 mmol/L EDTA, 2 mmol/L protease inhibitor). Lysates were centrifuged (2,800 × g, 4°C, 20 minutes) and supernatants, including cytosolic and membrane proteins were collected. Next, ultracentrifugation (>100,000 × g/1 hour/4°C) (Sorvall Dicovery 90SE, Hitachi) was used to separate cytosolic and membrane proteins. Membrane fraction purity was verified by Western blot (Supplementary Methods) assessing EGFR expression (Supplementary Fig. S1). Pellets containing the membrane fractions were resuspended in PF2D start buffer (Beckman Coulter). Protein concentrations were determined applying the Micro BCA Protein Assay Kit (Thermo Fisher Scientific) according to the manufacturer's protocol.

Fractionation of membrane proteins was conducted by the liquid chromatography system ProteomeLab as described before (7, 24). Briefly, a total of 2.5 mg protein was loaded on the first dimension (1D) chromatofocusing column according to manufacturer's protocol (Beckman Coulter). Proteins were separated by isoelectric focusing and fractions were collected at 0.3 pH intervals during the pH gradient. In the second dimension (2D) 200 μL of each 1D fraction was loaded onto the 2D reversed phase column (heated to 50°C) for further fractionation. Column-bound proteins were resolved by a 30 minutes linear gradient from solvent A [0.1% aqueous trifluoracetic acid (TFA)] to solvent B (0.08% TFA acid in acetonitrile), respectively. Eluted proteins were detected at 214 nm and collected. UV absorbance data were further analyzed using the ProteoVue and DeltaVue software (Beckman Coulter). Protein profiles of tumor cells and human keratinocytes (HPV16GM) were compared regarding the appearance of distinct tumor-specific peaks. Corresponding fractions were selected and pooled (8) and depending on the protein concentration concentrated using a vacuum concentrator (Bachhofer Savant).

For the selection of HNSCC-specific peptides an alternating biopanning using the SFTI8Ph library (Supplementary Methods) against HNO97 cells and the corresponding protein fraction was performed. Initially, 1×10⁹ phages were incubated for 1 hour with HNO97 grown to 90% confluence. Unbound phages were removed by washing steps and the cells were lysed with 1% Triton X-100 solution. Phages isolated from cell lysate were amplified and packaged in XL1 blue bacteria overnight and precipitated in polyethylenglycole solution (Supplementary Methods). Subsequently, the phages were exposed alternatingly to HNO97 cells and HNO97 protein fractions (100 nmol/L) in 96-well plates for 1 hour. After 5 PBS washing steps phages were eluted in 100 μL Glycin/HCl (pH 2.2) per well, neutralized by 15 μL Tris-HCl (pH 9.1) and amplified in XL1 blue bacteria. For titration, the phages were diluted (10⁻², 10⁻⁴, 10⁻⁶) and grown on agar plates. Twelve selection rounds were performed followed by single-stranded DNA isolation of 16 clones (QIApreo Spin M13 Kit; Qiagen). DNA sequencing (GATC Biotech) allowed for the identification of the corresponding peptides which were synthesized using standard Fmoc/tBu chemistry (Supplementary Methods).

For binding studies, 2.5–4 × 10⁵ cells were seeded in 6-well plates and cultivated for 48 hours. Cells were incubated for 60 min with 1 mL serum-free medium containing the ¹²⁵I-labeled peptide (supplementary methods). For competition experiments, cells were simultaneously exposed to unlabeled (10⁻⁴ to 10⁻¹⁰ mol/L) and ¹²⁵I-labeled peptides. After three washing steps with PBS (pH 7.4) the cells were lysed with 0.5 mL of 0.3 mol/L NaOH. Experiments concerning the internalization, kinetics and efflux were additionally performed with ¹⁷⁷Lu-DOTA-SFITGv6. To evaluate kinetics and efflux HNO97 cells were incubated for different time intervals (10–480 minutes) and 60 minutes, respectively, with ¹⁷⁷Lu-DOTA-SFITGv6 and lysed as described. To continue with the efflux experiment radioactive medium was replaced by non-radioactive medium and cells were incubated again for 60 to 240 minutes before measuring radioactivity of the cell lysates and the medium. For internalization experiments the cells were exposed to ¹⁷⁷Lu-DOTA-SFITGv6 for different time intervals (10 to 240 minutes) at 37°C and 4°C. After three washing steps with PBS the cells were incubated with 1 mL of glycine-HCl 50 mmol/L in PBS (pH 2.8) for 10 minutes at room temperature to remove the surface bound activity. Then, cells were washed with 3 mL of ice-cold PBS and lysed as described. Radioactivity was determined in a gamma counter and calculated as percentage of the applied dose per 1×10⁶ cells. Experiments were performed three times, and three repetitions per independent experiment were acquired.

Assessment of the binding affinity was performed on a BiaCore ×100 (GE Healthcare). To avoid denaturation of the heterodimer during amide activation ITGα_vβ₆ and ITGα_vβ₃, respectively, were used as analytes and SFITGv6 was immobilized at its N-terminus on a C1 sensor Chip (BR-1005-35, GE Healthcare) using a manual amine coupling protocol. Briefly, after activation of the sensor chip surface with a solution of EDC 0.4 mol/L/NHS 0.1 mol/L for 5 minutes, a 100 μmol/L peptide dissolved in HBS-EP running buffer (0.1 mol/L HEPES, 30 mmol/L EDTA, 1.5 mol/L NaCl, 0.5% surfactant P20, pH 7.2) was immobilized with a contact time of 30 seconds until a loading level of 18 response units (RU) was obtained. After saturation of the sensor chip surface for 5 minutes with 1 mol/L ethanolamine/HCl solution (pH 8.4) the binding affinity of the analyte dissolved in HBS-EP running buffer in appropriate concentrations to the ligand was measured (flow rate: 30 μL/min). The surface plasmon resonance assay (SPR)-sensogram data were evaluated with the BiaCore evaluation software. The dissociation constant (K_D) was determined by a 1:1 Langmuir model fit of the SPR-sensograms.

Five MBq of the purified ¹³¹I-radiolabeled peptide was incubated in 300 μL human serum at 37°C. After different time intervals (15 minutes to 24 hours), 20 μL serum was precipitated with 40 μL acetonitrile. The stability of the labeled peptide in the supernatant was monitored by radio-HPLC at selected time points using a chromolith performance RP18ec column (3 mm × 100 mm) equipped with a gamma detector (Packard COBRA Auto-Gamma, GMI). Separation condition was a gradient of 0% to 60% aqueous acetonitrile supplemented with 0.1% TFA over 10 minutes with a flow rate of 2 mL/min.

All experiments were conducted in compliance with the German animal protection laws. Eight-week-old Balb c/c nude mice (Charles River Laboratories) were inoculated subcutaneously at the right shoulder with 5×10⁶ HNO97 cells in BD Matrigel (BD Biosciences). Xenografts were grown to a tumor diameter of 10–15 mm. For small animal-PET imaging mice were anesthetized using isoflurane inhalation and injected via tail-vein with 50 MBq (2 nmol) of the ⁶⁸Ga-labeled DOTA-SFITGv6 peptide solution (see also Supplementary Methods) in 100 μL PBS. Images were recorded on an Inveon small-animal PET scanner (Siemens) using a 60-minute emission scan in list mode and a 10-minute transmission scan. Images were taken in 3-dimensional (3D) mode and reconstructed iteratively with a fully 3D algorithm from a 256 × 256 matrix for viewing transaxial, coronal, and sagittal slices of 0.9 mm thickness. Pixel size was 0.38 × 0.38 × 0.79 mm³, and transaxial resolution obtained was 0.9 mm. For blocking experiments 100 μL of a 1 mmol/L aqueous solution of SFITGv6 was pre-administered intraperitoneally 30 minutes before injection of the radiolabeled peptide. Biodistribution studies were performed after administration of 100 μL of a 20 nmol/L ¹⁷⁷Lu-DOTA-SFITGv6 solution (1 MBq) as an intravenous bolus injection into the tail vein of the mice. After different time points (30 minutes to 6 hours) three animals, respectively, were sacrificed. Peripheral blood, heart, lung, spleen, liver, kidneys, muscle, brain, intestine, and injection site (tail, after intravenous injection only) were collected and weighted. Tissue-associated radioactivity was measured in a gamma counter (Berthold LB951G) and expressed as the percentage of the injected dose per gram tissue (% ID/g).

Staining with the biotinylated PEG(12)-SFITGv6 peptide was performed on acetone-fixed cryosections (5 μm) of tumor tissues after blocking of unspecific binding using the Avidin/Biotin Blocking Kit (SP-2001, Vector Laboratories). A stock solution of the lyophilized peptide was prepared by dilution in 5% aqueous DMSO. Slices were incubated overnight at 4°C with 10⁻⁵ mol/L peptide concentration in antibody diluent (DAKO). Detection of bound peptide was carried out with the Vectastain Elite ABC Kit (PK-6100, Vector Laboratories) according to the manufacturer's protocol. Peptide specificity was ensured by a scrambled (GRD) PEG(12)-SFITGv6 derivate, the somatostatin receptor ligand DOTATOC and negative controls (without peptide or primary antibody). Staining results were assessed by bright field microscopy (BX50) with the SC30 camera and the cell Sense software (all Olympus).

The PET/CT scan was performed 1 and 3 hours after tracer administration with a Biograph mCT Flow PET/CT-Scanner (Siemens Medical Solution) using the following parameters: slice thickness of 5 mm, increments of 3 to 4 mm, soft-tissue reconstruction kernel, care dose. Immediately after CT scanning, a whole-body PET was acquired in 3D (matrix 200 × 200) in FlowMotion with 0.7 cm/min. The emission data were corrected for random, scatter and decay. Reconstruction was conducted with an ordered subset expectation maximization (OSEM) algorithm with two iterations/21 subsets and Gauss-filtered to a transaxial resolution of 5 mm at full-width half-maximum (FWHM). Attenuation correction was performed using the low-dose non-enhanced CT data. The quantitative assessment of standardized uptake values (SUV) was done using a region of interest technique.

Employing the SFTI8Ph library for alternate selection rounds on the HNSCC cell line HNO97 and respective PF2D membrane protein fractions 7 out of 16 sequenced phages displaying the peptide sequence FRGDKMQL (SFPF-10) were selected for further analysis (Supplementary Table S1). The sequence comprises a RGD and KXXL motif indicating ITGα_vβ₆-specificity of the peptide. ITGα_vβ6-expression could be confirmed by flow cytometry analysis on several squamous cell carcinoma cell lines, including HNSCC (e.g., HNO97; up to 99.8%), bladder cancer (UM-UC-5; up to 88.7%), lung cancer (LUDLU-1; up to 90.9%), and breast cancer (MCF-2; up to 36.1%; Supplementary Fig. S2A). In contrast, the liposarcoma cell line SW872 was completely ITGα_vβ₆-negative. In addition, ITGα_vβ₆ expression was assessed in situ applying immunohistochemistry in HNSCC, dysplasia-free normal mucosa tissue as well as in breast and lung cancer-derived brain metastases (Supplementary Fig. S2B–S2G). All tumors showed a strong tumor cell-specific staining, whereas the tumor-surrounding stromal cells and the epithelial cells of the dysplasia-free mucosa were negative, indicating a tumor-associated ITGα_vβ₆ expression in squamous cell carcinomas of different origins. Accordingly, in vitro¹²⁵I-SFPF-10 displayed binding to ITGα_vβ₆-expressing HNSCC cell lines HNO97 (7.2%) and HNO223 (11%) and to the bladder cancer cell line UM-UC-5 (7.5%) but less binding to the HNSCC cell lines HNO210 (5%), HNO199 (3.1%), HNO258 and (2.6%), and MCF-7 (1.6%), (Fig. 1A). In all cancer cell lines tested co-incubation of ¹²⁵I-labeled and unlabeled SFPF-10 (10⁻⁶ mol/L) decreased the binding to values below 1% of the applied dose (Fig. 1A; see also Supplementary Results and Supplementary Fig. S3A and S3B).

To specify the amino acids contributing to the target-specific binding mutations were introduced either into the RGD motif or the adjacent KXXL sequence. As shown in Fig. 1B, both the mutation of RGD to DRG (FDRGKMQL) and the mutation of KXXL to AXXA (FRGDAMQA) almost completely abolished (< 0.1%) the binding of ¹²⁵I-SFPF-10 to HNO97 and to UM-UC-5 cells, respectively. In contrast, the substitution of lysine (K) to leucine (L) within the initially identified sequence (FRGDKMQL) substantially increased the binding to both cell lines (Fig. 1B). Subsequently, experiments concerning binding and affinity as well as the in vivo application was performed employing the modified peptide SFITGv6 containing the FRGDLMQL sequence (Fig. 2).

Compared with the originally identified peptide ¹²⁵I-SFITGv6 displayed higher binding to the HNSCC cell lines HNO97 (24.5%), HNO210 (7.1%), HNO199 (6.6%), and HNO258 (3.5%) and to UM-UC-5 cells (10.4%). In addition, binding of ¹²⁵I-SFITGv6 to the carcinoma-derived cell lines LUDLU-1 (4.3%) and to the adenocarcinoma cell line HT29 (2.8%) was measured (Fig. 1C). The specific binding to these cells was reduced to values below 1% of the applied dose by addition of 10⁻⁶ mol/L unlabeled peptide as competitor (Fig. 1C). However, in accordance with lower expression levels of ITGα_vβ₆ (Supplementary Fig. S2A) less than 2% binding of ¹²⁵I-SFITGv6 to the breast carcinoma cell lines MCF-7 and T47D as well as to the liposarcoma cell line SW872 was measured (Supplementary Fig. S4A).

Because time-dependent deionization of ¹²⁵I was expected experiments concerning the kinetics, internalization, and efflux of SFITGv6 were additionally performed with the ¹⁷⁷Lu-DOTA-labeled SFITGv6 (Fig. 2). In fact, binding of the ¹⁷⁷Lu-DOTA-SFITGv6 to HNO97 cells continuously increased to 57.3% within 480 minutes, whereas the maximal uptake of ¹²⁵I-SFITGv6 (37.3%) was measured after exposure for 60 minutes followed by a decrease to 14.5% (Fig. 1D). Furthermore, a fast and continuously increasing internalization of ¹⁷⁷Lu-DOTA-SFITGv6 up to 37.5% (Fig. 1E) but maximal internalization of up to 24.1% of ¹²⁵I-labeled SFITGv6 within 60 minutes was noticed followed by a decrease to 12.6% (Supplementary Fig. S4B). Finally, the efflux experiment revealed retention of more the 50% of the originally accumulated ¹⁷⁷Lu-DOTA-SFITGv6 (Fig. 1F), but less than 5% of ¹²⁵I-SFITGv6 (Supplementary Fig. S4C) was measured 240 minutes after the termination of the uptake. These data point to time-dependent deionization of ¹²⁵I-SFITGv6.

ITGα_vβ₆-specificity of SFITGv6 was further demonstrated by competition of SFITGv6 binding by already known ITGα_vβ₆-binding molecules TP H2009.1 (12), A20FMDV2 (25) and HBP-1 (ref. 15; Fig. 3A). Like the unlabeled SFITGv6 these peptides competed for the binding of the ¹²⁵I-SFITGv6 to HNO97 cells with IC₅₀ values of 16.15 nmol/L (SFITGv6), 3.2 nmol/L (A20FMDV2), 41 nmol/L (TP H2009.1), and 45.1 nmol/L (HBP-1), respectively (Fig. 3A). In accordance with the IC₅₀ value of SFITGv6, we measured a high affinity (K_D = 14.8 ± 26.0 nmol/L) of the peptide for ITGα_vβ₆ (Fig. 3B) using SPR spectroscopy whereas ITGα_vβ₃ bound to immobilized SFITGv6 with a 10-fold lower affinity (K_D = 185 ± 0.8 nmol/L; ref. Fig. 3C). The proteolytic stability of SFITGv6 was determined by radio-HPLC analysis of ¹²⁵I-labeled SFITGv6 after incubation in heparinized human serum (Fig. 3D). Serum aliquots were taken after 0 minutes, 15 minutes, 1, 2, 4 and 24 hours, respectively, and ¹²⁵I-SFITGv6 revealed high stability with no degradation over a 24 hours' time period demonstrating the suitability of SFITGv6 for in vivo experiments.

Small-animal PET imaging of Balb/c mice bearing ITGα_vβ₆-expressing HNO97 xenografts is shown in Fig. 4. Within 20 minutes after injection of ⁶⁸Ga-DOTA-SFITGv6 radioactivity accumulated in the tumor and was maintained for at least 140 minutes (Fig. 4A and C). Nonspecific activity cleared quickly from the blood within 60 minutes after injection resulting in a low background and images with good tumor-to-background ratios (Fig. 4A and C). In contrast, no intratumoral accumulation of the peptide was visible after intraperitoneal administration of unlabeled SFITGv6 as competitor 30 minutes before the injection of the radiolabeled compound (Fig. 4B).

To expand on the biodistribution of SFITGv6 the ¹⁷⁷Lu-DOTA-linked peptide was injected intravenously into HNO97 tumor-bearing mice. Radioactivity in individual organs was measured after different time points following injection of the peptide and calculated as the percentage of injected dose (ID)/g (Fig. 4D). In the tumor, an activity of more than 6% ID/g was measured 30 minutes after injection followed by a washout to 2.4% ID/g after 6 hours. A significantly higher uptake of almost 42% ID/g with hardly any clearance was observed in the kidneys, whereas less than 1% ID/g was measured in the blood. Except for the kidneys, the tumor-to-tissue ratios were above one (supplementary table S2) and in accordance with the PET results.

In a next step, tumor cell affinity and intratumoral distribution of SFITGv6 was further assessed by histochemical peptide staining of different carcinomas (HNSCC, NSCLC, breast cancer) using biotin-labeled SFITGv6 (Fig. 5). We observed a strong and homogenous tumor cell–specific binding of SFITGv6, whereas the surrounding tumor stroma was negative (Fig. 5A, D). Moreover, for the negative control peptides containing either the GRD sequence (Fig. 5B and E) or DOTATOC (Fig. 5C and F) we could not observe any specific binding. In addition, SFITGv6 specificity was assessed on brain metastases derived from breast and NSCLC (Supplementary Fig. S5). We detected a moderate but distinct tumor cell staining in breast cancer BM (NCH640k; Supplementary Fig. S5A), but a very strong tumor-specific staining pattern in lung BM (NCH2407) (Supplementary Fig. S5D). Inflammation-associated binding was excluded by staining of tumor-free lymph nodes, which did not show any SFITGv6-specificy (Supplementary Fig. S6A), further corroborating tumor cell specificity of SFITGv6 peptide for HNSCC as well as for further carcinoma.

PET/CT scans in a compassionate use setting were performed in two tumor patients after application of ⁶⁸Ga-DOTA-SFITGv6 and ¹⁸F-FDG, respectively (Fig. 6). The first patient suffered from recurrent hypopharynx carcinoma before radiotherapy. One and 3 hours after application of the tracer increasing ¹⁸F-FDG accumulation with time was seen in the recurrent tumor but also in the elbow (Fig. 6A), in several axillary lymph nodes (LN), and the left hilus. In contrast, the PET/CT scan performed after application ⁶⁸Ga-DOTA-SFITGv6 revealed a stable uptake of the tracer only in the tumor (Fig. 6B). We, therefore, considered inflammatory reactions to account for the ¹⁸F-FDG uptake in the axillary LN of this patient. In fact, histological examinations of the lymph nodes excluded the presence of tumor cells in these lesions (data not shown). These results correspond to the negative histochemical staining of carcinoma-free lymph nodes (Supplementary Fig. S6A). The second patient suffering from NSCLC revealed a time-dependent increase of ¹⁸F-FDG accumulation in the tumor but also uptake of the tracer in the left shoulder and two mediastinal LN which might be due to inflammation (Fig. 6C). ⁶⁸Ga-DOTA-SFITGv6 instead accumulated specifically in the tumor (Fig. 6D) and decreased moderately with time (data not shown). Moreover, the PET/CT scans of both patients revealed a high uptake in the kidneys, the stomach and the bowel and moderate uptake in the thyroid whereas a rather low background was determined in other organs and the blood pool (Supplementary Tables S3 and S4). Because the bowel activity showed a shift in the localization when comparing the early and the late image a more detailed analysis was done for jejunum, terminal ileum and cecum. In both patients, a time-dependent decrease of the radioactivity in the jejunum and a time-dependent increase of the radioactivity in terminal ileum and cecum were noticed (Supplementary Tables S3 and S4) suggesting that the tracer is secreted into the lumen of the stomach or duodenum/jejunum and is transported intraluminally to the cecum.

Radiolabeled peptides for diagnosis and endoradiotherapy of tumors are characterized by efficient transport to the tumor cells and a fast clearance. Because linear peptides display poor in vivo stability, peptides embedded in disulfide-stabilized miniproteins, which are endowed with an excellent proteolytic stability, and beneficial pharmacokinetic profile are of growing interest for the development of tumor-affine ligands. In this context, a couple of cysteine knot peptides have been identified by display technologies or engineering that specifically target the angiogenesis marker Delta-like ligand 4 (DLL4; ref. 26) or ITGα_vβ₆ on pancreatic tumor cells (16, 27).

Using a phage display library based on the molecular scaffold of SFTI-1 for alternate biopanning on HNO97 cells or selected membrane protein fractions of this cell line we identified a peptide containing the RGDKXXL motif. The amino acid substitution of K⁴ to L⁴ in the original peptide improved the binding and affinity of the RGDLXXL containing SFTI-1 derivate (SFITGv6) for a variety of HNSCC and other tumor cell lines of epithelial origin. The RGDLXXL motif, in particular the DLXXL sequence, has been found to be responsible for ITGα_vβ₆ specificity and ITGα_vβ₆-dependent Foot-and-Mouth Disease Virus (FMDV) infectivity (28) while having only minimal interactions with other integrin heterodimers, such as α_vβ₃, α_vβ₅, and α_IIbβ₃ (29). Recently, an ITGα_vβ₆-specific linear peptide HBP-1 containing a DLXXL motif has been enriched by phage display on HNO223 cells using a commercially available library. HBP-1 as well as the ITGα_vβ₆-specific 20 amino acid long peptides TP H2009.1 and A20FMDV2 almost completely competed for SFITGv6 binding to HNO97 cells. In contrast, SFTI-1 derivates containing the mutated motifs DGRLXXL and RGDAXXA, respectively, completely failed to bind to these cells indicating that SFITGv6 targets ITGα_vβ₆.

ITGα_vβ₆ has been shown to be highly expressed on HNSCC as well as lung (12), colon (30), breast and pancreas carcinoma (31) and is often associated with poor prognosis (32). Thus, ITGα_vβ₆ represents an excellent target for imaging and/or therapy for a variety of epithelial malignancies. Given that ITGα_vβ₆ is overexpressed in up to 100% of HNSCC as shown by immunohistochemical studies (14, 32) it is not surprising that phage display performed on living HNO97 and HNO223 cells with either a commercially available or the SFTIPh8 library led to enrichment of the very similar RGDK/LXXL motives.

As expected from stability analysis of the natural occurring SFTI-I (10) and the SFTI-I derivate–targeting DLL4 (26, 33) SFITGv6 demonstrated an outstanding stability in human serum over a period of 24 hours and high affinity (K_D =14.8nmol/L) for ITGα_vβ₆. In correlation to the ITGα_vβ₆ expression of different cell lines as measured by FACS analysis the ¹²⁵I-labeled peptide displayed specific binding to HNO97 cells and other HNSCC cells as well as to further carcinoma cell lines of different origin including lung, bladder and colon with an average of 7% which was competed to more than 90% by addition of 10⁻⁶ mol/L unlabeled analog. The experiments concerning the binding kinetics and internalization revealed a maximal binding of ¹²⁵I-SFITGv6 to HNO97 cells after exposure for 60 minutes followed by a decrease to less than 15%. On the contrary, an increasing uptake and internalization of ¹⁷⁷Lu-DOTA-SFITGv6 to values of up to 57.3% and up to 37.4%, respectively, as well as retention of more than 50% of the applied dose for at least 4 hours in HNO97 cells was measured. This indicates a time-dependent intracellular deionization of ¹²⁵I-labeled peptide followed by the efflux of free radioiodine. The long-lasting accumulation of the ¹⁷⁷Lu-DOTA labeled peptide in tumor cells, however, allows for late imaging and possibly for a therapeutic application.

Using ⁶⁸Ga-DOTA-SFITGv6, we were able to successfully and selectively image the ITGα_vβ₆-expressing HNO97 tumor of Balb/c mice within 20 minutes in a small animal PET. The rapid clearance of unbound and unspecifically bound activity from the blood and the surrounding tissues resulted in an excellent tumor-to-background ratio 40 minutes after the injection remaining for at least 140 minutes. In parallel to the in vitro competition experiments a complete inhibition of ⁶⁸Ga-DOTA-SFITGv6 accumulation in the tumor was achieved by injecting a nonradioactive analog before. This is evidence for the selective in vivo imaging of a tumor that endogenously expresses ITGα_vβ₆. The biodistribution data revealed a significantly higher accumulation of ¹⁷⁷Lu-DOTA-SFITGv6 in the HNO97 tumor as compared with healthy tissues even after 4 and 6 hours (supplementary table S1). However, in contrast with the binding kinetics in culture a washout of the radioactivity from the HNO97 tumor with time was noticed, which might be due to a fast clearance of the tracer from the blood within the first 30 minutes. Of particular interest is the high and constant level of radioactivity of almost 42% ID/g in the kidneys in HNO97 tumor bearing Balb/c nude mice with hardly any clearance. Hausner and colleagues (31) also observed a high retention of (¹⁸F)FBA-(PEG₂₈)₂-A20FMDV2 of more than 40% in the kidneys of mice bearing ITGα_vβ₆-expressing BxPC-3 xenografts obviously due to the introduction of the second PEG unit in the peptide.

Peptide-based histochemical staining of HNO97-xenotransplanted mouse tumors and different human HNSCC tumor tissues revealed a specific, strong and homogeneous binding of biotinylated SFITGv6. As expected from the binding to different squamous and adenocarcinoma cell lines in vitro, the peptide displayed a moderate to strong binding in brain metastases derived from breast cancer and NSCLC, respectively. Because tumor-free lymph nodes did not show any staining inflammation-associated binding of the peptide can be excluded. Accordingly, tumor-specific but not inflammation-associated binding of SFITGv6 was observed in PET/CT performed in two patients suffering from either recurrent hypopharynx carcinoma or NSCLC after application of ⁶⁸Ga-DOTA-SFITGv6 and ¹⁸F-FDG, respectively. In line with the small animal PET images of HNO97 xenografts the SUV values of SFITGv6 in these patients persisted from 1 to 3 hours after injection. Interestingly, in contrast to ⁶⁸Ga-DOTA-SFITGv6 ¹⁸F-FDG accumulation was detected in inflammatory lesions and reactive lymph nodes of both patients. FDG accumulation in inflammatory or infectious processes is well-known and may occasionally lead to false-positive evaluation and incorrect up-staging of tumor patients with a tremendous impact on therapy management. Thus, in ITGα_vβ₆-positive tumors the peptide-based tracer provided a clear advantage over FDG with respect to false-positive lesions. However, not all tumors are ITGα_vβ₆-positive. Several immunohistochemistry studies revealed a high expression in ovarian cancer (100% of the cases), pancreatic cancer (100%), cervical cancer (92%), skin (84%), and tumors of the oral cavity (90-100%). For these tumor entities, our ITGα_vβ₆ ligand may add useful information for staging and treatment planning, in ITGα_vβ₆-negative tumors FDG is superior and also more widely applicable. Also, the tracer uptake in the tumor lesion is lower for the ITGα_vβ₆ ligand than for FDG, possibly leading to a lower sensitivity of the former. However, also the background signal is lower for the ITGα_vβ₆ ligand. This fact and the more specific nature of tracer uptake are expected to compensate the lower uptake of the ITGα_vβ₆ ligand.

SFITGv6 also accumulated in the kidneys, the bowel, the stomach and in the thyroid of both patients. Interestingly, the localization of the radioactivity in the bowel changed with time indicating the excretion of the tracer. In fact, regions of interest drawn in jejunum, terminal ileum, and cecum revealed a decrease of the activity in the jejunum and a time-dependent increase in terminal ileum and cecum. This is evidence for a secretion of SFITGv6 into the stomach or duodenum/jejunum followed by an intraluminal transport to the terminal ileum and the cecum. With regard to the identification of small tumor lesions in the abdomen or a peritonitis carcinomatosa the administration of laxatives might be necessary to reduce intrabowel activity and avoid a high background. The high uptake in the kidney may represent an obstacle to endoradiotherapy. However, the quantitative analysis in these two patients and also other patients (data not shown) revealed an SUV_max between 10.4 and 29.2 at 1 hour postinjection and between 8.3 and 10.5 at 3 hours postinjection. This is in the range of DOTATOC and the PSMA ligand PSMA617 and should, therefore, nor present a principal problem. However, future attempts should address the mechanism of kidney uptake and, based on this knowledge engage in the displacement of renal radioactivity.

In conclusion, SFITGv6 is a novel stable ITGα_vβ₆-specific peptide with high affinity for a variety of HNSCC and other tumors. Because of the accumulation of the peptide in different tumors but not in inflammatory lesions and normal tissues of tumor patients SFITGv6 represents a promising tracer for imaging and endoradiotherapy of ITGα_vβ₆-positive carcinoma. The theranostic use of the tracer for ITGα_vβ₆-positive tumor lesions below the diaphragm, including pancreatic and ovarian cancer (14); however, requires the modulation and improvement of the pharmacokinetic properties of the peptide.

No potential conflicts of interest were disclosed.

Conception and design: A. Altmann, M. Sauter, S. Roesch, W. Mier, R. Warta, J. Debus, C. Herold-Mende, U. Haberkorn

Development of methodology: A. Altmann, M. Sauter, S. Roesch, R. Warta, C. Herold-Mende, U. Haberkorn

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Sauter, R. Warta, J. Debus, G. Dyckhoff, C. Herold-Mende, U. Haberkorn

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Altmann, M. Sauter, S. Roesch, R. Warta, J. Debus, C. Herold-Mende, U. Haberkorn

Writing, review, and/or revision of the manuscript: A. Altmann, M. Sauter, S. Roesch, W. Mier, J. Debus, G. Dyckhoff, C. Herold-Mende, U. Haberkorn

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

Study supervision: A. Altmann, W. Mier, J. Debus, C. Herold-Mende, U. Haberkorn

The authors thank Jennifer Poloczek, Vanessa Kohl, Marlene Tesch, Melanie Greibich, and Farzaneh Kashfi for technical assistance and also thank Ursula Schierbaum and Karin Leotta for performing the animal experiments. The authors are grateful to Dr. Volker Eckstein, Medical Clinic V, University Hospital Heidelberg, for providing the FACS Core Facility and to Dr. Esther Herpel tissue bank of the National Center for Tumor Diseases (NCT) Heidelberg, Germany, and Dr. Carolin Mogler for the procurement and histopathologic examination of tissue specimens.

This work was supported by: Deutsche Forschungsgemeinschaft (DFG) HA 2901/12-1.

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.