Rapid optical imaging of EGF receptor expression with a single-chain antibody SNAP-tag fusion protein
- First Online:
- Cite this article as:
- Kampmeier, F., Niesen, J., Koers, A. et al. Eur J Nucl Med Mol Imaging (2010) 37: 1926. doi:10.1007/s00259-010-1482-5
The epidermal growth factor receptor (EGFR) is overexpressed in several types of cancer and its inhibition can effectively inhibit tumour progression. The purpose of this study was to design an EGFR-specific imaging probe that combines efficient tumour targeting with rapid systemic clearance to facilitate non-invasive assessment of EGFR expression.
Genetic fusion of a single-chain antibody fragment with the SNAP-tag produced a 48-kDa antibody derivative that can be covalently and site-specifically labelled with substrates containing 06-benzylguanine. The EGFR-specific single-chain variable fragment (scFv) fusion protein 425(scFv)SNAP was labelled with the near infrared (NIR) dye BG-747, and its accumulation, specificity and kinetics were monitored using NIR fluorescence imaging in a subcutaneous pancreatic carcinoma xenograft model.
The 425(scFv)SNAP fusion protein accumulates rapidly and specifically at the tumour site. Its small size allows efficient renal clearance and a high tumour to background ratio (TBR) of 33.2 ± 6.3 (n = 4) 10 h after injection. Binding of the labelled antibody was efficiently competed with a 20-fold excess of unlabelled probe, resulting in an average TBR of 6 ± 1.35 (n = 4), which is similar to that obtained with a non-tumour-specific probe (5.44 ± 1.92, n = 4). When compared with a full-length antibody against EGFR (cetuximab), 425(scFv)SNAP-747 showed significantly higher TBRs and complete clearance 72 h post-injection.
The 425(scFv)SNAP fusion protein combines rapid and specific targeting of EGFR-positive tumours with a versatile and robust labelling technique that facilitates the attachment of fluorophores for use in optical imaging. The same approach could be used to couple a chelating agent for use in nuclear imaging.
KeywordsSingle-chain antibodyhAGT (SNAP)-tagNIR optical imagingSite-specific labellingMolecular imagingTumour targeting
Human O6-alkyl-DNA alkyltransferase
Single-chain variable fragment
Epidermal growth factor receptor 1
Approximately one third of all epithelial cancers show elevated levels of epidermal growth factor receptor (EGFR) expression and EGFR activity is often correlated with poor prognosis [1, 2]. The receptor is therefore an attractive therapeutic target and antibodies blocking the binding of EGF and small molecule inhibitors, blocking intracellular tyrosine kinase activity, are frequently used to treat different types of cancer including breast, head and neck, pancreas and colorectal cancers . The non-invasive evaluation of EGFR status using molecular imaging could have significant impact on the choice of treatment and follow-up strategies [4–6].
Fluorescence imaging is rapidly developing as a sensitive and cost-effective tool to visualize biomarker expression in vivo . It can be used to assess the response to therapy in pre-clinical research and to pre-select agents for further clinical development. Although the direct clinical applications of fluorescence imaging may be limited to superficial tissues such as breast and skin, fluorescent probes could be used for detection or visualization during endoscopy and in “guided surgery” procedures. A number of optical imaging probes based on EGF, EGFR-specific antibodies and antibody fragments have been analysed for their tumour targeting properties [8–10]. Probes used for imaging should balance tumour accumulation and systemic clearance to provide high contrast images for a short period after administration.
We have previously reported that single-chain variable fragment (scFv) antibodies and other protein ligands fused to an engineered version of the human DNA repair enzyme O6-alkyl-DNA alkyltransferase (SNAP-tag) can be modified with fluorescent dyes or covalently coupled to nanoparticles in a site-specific manner without affecting the ligand’s binding activity . In this study we analysed the capability of an scFvSNAP fusion protein directed against EGFR to probe EGFR expression by in vivo optical imaging. The 425(scFv)SNAP fusion protein was labelled with the near infrared (NIR) dye BG-747, and distribution and tumour accumulation were analysed in mice bearing subcutaneous tumour xenografts. The intermediate size of the protein (48 kDa) results in circulation times that permit efficient tumour binding combined with clearance rates that allow measurement with low background 8–10 h after administration.
Materials and methods
Cell lines and targeting antibodies
L3.6pl pancreatic carcinoma cells  and L540 Hodgkin lymphoma cells were grown in RPMI 1640 medium supplemented with 10% (v/v) heat inactivated fetal calf serum (FCS), 50 µg/ml penicillin, 100 µg/ml streptomycin and 2 mM l-glutamine. L3.6pl cells express a moderate number of EGFRs (EGFR+), whereas L540 cells express high levels of CD30 but do not express EGFR (EGFR-). For ease of detection, L3.6pl cells were stably transfected with copGFP enabling co-localization of the targeted signal with the subcutaneous tumours. The single-chain fragment 425(scFv) is derived from the monoclonal antibody 425, which binds to extracellular domain III of human EGFR . C225 (cetuximab, ImClone Systems) is a chimaeric mouse/human IgG1 antibody with high affinity to a different epitope on EGFR domain III. Both antibodies block EGF binding. The 425(scFv) molecule has previously been used to target L3.6pl tumours in vivo . The CD30-specific Ki4 single-chain antibody SNAP fusion has been described .
Construction and expression of scFv-hAGT fusion proteins
Construction and expression of the scFv-hAGT fusion proteins have been described previously . Briefly, scFv genes were cloned as N-terminal fusions with an engineered version of the human O6-alkylguanine-DNA alkyltransferase known as SNAP-tag (Covalys AG). SNAP antibodies were expressed in 293T cells and purified from culture supernatant by immobilized metal ion chromatography (IMAC) to capture the His-tag. The activity of the purified fusion proteins was determined by flow cytometry.
Labelling of scFvSNAP and C225
Labelling of the fusion proteins with BG fluorophores was performed as described . A ratio of 1.5:2 of dye to protein was used in the labelling reactions and coupling was performed for 1 h at room temperature followed by overnight incubation at 4°C. Unbound dye was removed using PD-10 columns (GE Healthcare) and the labelled samples were concentrated to up to 4 mg/ml using 10 kDa MWCO columns (VivaSpin6, Sartorius). The benzylguanine-dye derivatives BG-Alexa Fluor 647 or BG-488 were used for in vitro analysis and BG-747 was used for detection in vivo. Probes were prepared freshly for each experiment, passed through 0.22-µm sterile centrifugal filter units (Ultrafree-MC, Millipore) and stored for no longer than 3 days at 4°C before use in the imaging studies.
The C225 antibody was labelled with the 747 dye as an NHS ester derivative (DY-747-NHS, Dyomics); 1.4 mg C225 in 1 ml phosphate-buffered saline (PBS) were reacted with six equivalents of DY-747-NHS in dimethyl sulfoxide (DMSO) (5 nmol/µl) for 2 h at room temperature. Residual dye was removed with a PD-10 column and the labelled antibody was concentrated to 6 mg/ml. The dye to protein ratio for this probe was 2.7:1.
In vitro binding analysis
Binding of the scFvSNAP antibodies was analysed by flow cytometry. Antibodies were labelled with BG-Alexa 647 or BG-488 and binding was detected directly in FL-4 or FL-1. Binding of antibodies labelled with BG-747 for in vivo applications was detected with a secondary antibody (anti-Penta-His Alexa 488, Qiagen). C225-747 full-size antibody was detected with anti-human heavy and light chain F(ab′)2-fluorescein isothiocyanate (FITC) (Chemicon). Approximately 4 × 105 cells were incubated with 50–500 ng of scFvSNAP antibody in 300 µl PBS for 20 min, washed and analysed with a FACSCalibur and CellQuest software (Becton Dickinson).
To confirm binding at the cellular level, cells were stained as described for FACS analysis and images were captured with a Leica confocal microscope (TCS SP5). DRAQ5 (Biostatus, Shepshed, UK) was added to counterstain the nucleus immediately before image acquisition.
Female BALB/c nu/nu mice, 6–8 weeks old (Charles River, Germany), were used for animal experiments, which were approved by the Ethics Committee for Animal Experiments. Approximately 2.5 × 105 L3.6pl-GFP cells in 25 µl PBS were injected subcutaneously in the left upper leg. Cells were harvested from sub-confluent cultures by short trypsination. For the injection of cells and probes and before image acquisition, mice were anaesthetized by intramuscular administration of 75 mg/kg ketamine and 10 mg/kg xylazine. Imaging studies were performed after 6 days when tumours had reached an approximate diameter of 3–4 mm. Mice were placed on a purified, chlorophyll free diet (AIN93G, SSNIFF GmbH) 7 days before the imaging experiments were started.
In vivo fluorescence imaging
We administered 0.5 nmol scFvSNAP probe (corresponding to 25 µg protein) or 0.25 nmol of C225-747 per mouse (n = 4) by retro-orbital injection in 30 µl PBS. Mice were imaged at time points 0, 10, 24, 48 and 72 h. For the C225-747 probe, mice were monitored for up to 144 h. Imaging was performed with a CRi Maestro imaging system (CRi Inc., Woburn, MA, USA) using the blue and deep red filter sets in the multi-filter acquisition mode and the predefined settings for these filters (blue 500–720 nm plus deep red 700–950 nm; 10 nm increment). Exposure times were 300 ms for the blue filter set and 5,000 ms (1,000 ms for C225–747) for deep red.
Competitive binding in vivo
To compete for binding of the labelled 452(scFv)SNAP-747 fusion protein, we injected a 20-fold excess of unlabelled probe 3–5 min before injection of the labelled sample (n = 4).
Unmixing of spectral images
To separate the single dye spectra in the acquired cube images, the background and the tumour GFP spectrum was defined for each mouse in the images before injection of the probe. The DY-747 signal was defined after subcutaneous injection of 5 μg labelled 425(scFv)SNAP in a control mouse. All images were unmixed with these previously defined spectra. To quantify the targeting efficacy, tumour to background ratios (TBRs) were determined for each time point by building the ratio of the average fluorescence intensity/pixel at the tumour site and the signal/pixel of a larger area on the back representing average background intensity. All data are presented as means ± SD of n independent measurements. Statistical analysis was performed with a Mann-Whitney test using GraphPad Prism (GraphPad Software). Statistical significance was assigned for p values < 0.05.
Expression and labelling of imaging probes
In vitro binding analysis
In vivo fluorescence imaging
At the 10-h time point, DY-747 fluorescence was found in the bladder and in urine, as validated after micturition. After 24–48 h a strong DY-747 signal was observed in parts of the digestive tract in 9 of the 12 mice treated with scFvSNAP. This signal was not observed in mice treated with the C225 full-length antibody (Fig. 5). Some animals showed a weak signal in the leg opposite to the tumour resulting from the intramuscular injection of the anaesthetic, which brings about mild bruising of the tissue and the accumulation of circulating probe.
TBRs were calculated for each probe and time point to quantify the targeting efficacy. The 425(scFv)SNAP-747 probe gave an average peak TBR of 33.2 ± 6.3 10 h post-injection. The TBR in the Ki4(scFv)SNAP-747 control group was significantly lower at 5.44 ± 1.92 (p < 0.05, n = 4).
A competition experiment was performed to verify the specificity of accumulation in the tumour. Four mice were injected with a 20-fold excess of unlabelled probe prior to the injection of labelled probe. Under these competitive binding conditions, the DY-747 fluorescence in the tumour decreased to values comparable with those obtained using the non-specific probe Ki4(scFv)SNAP, resulting in a significantly lower TBR of 6 ± 1.35 (p < 0.05, n = 4) (Fig. 4, Fig. 7b).
Comparison of C225 full-length antibody and 425(scFv)SNAP
To compare the distribution and kinetics of the probe, an additional group was injected with the anti-EGFR chimaeric full-length antibody C225 (cetuximab) labelled with DY-747-NHS. As shown in Fig. 5, strong accumulation in the tumour was observed after 10 h with average signal intensities approximately 10 times higher than those achieved with the single-chain construct. TBRs were significantly lower at 4.8 ± 0.4 (p < 0.05, n = 4, Fig. 7c, d). No renal uptake or excretion via the bile was observed for the full-length antibody. The TBR increased slowly from 4.8 to 8.6 ± 0.4 after 72 h. In contrast to the single-chain probe, the average fluorescence in the tumour increased during the first 24 h of measurement before reaching a plateau at around 48 h (Fig. 5, Fig. 7c, d).
In this study we evaluated an EGFR-specific single-chain antibody fragment fused to the SNAP-tag as a targeting probe for in vivo optical imaging. The SNAP-tag has been applied in a variety of experimental set-ups to mediate site-specific covalent modification or immobilization of recombinant proteins [15, 16]. The scFv comprising only the variable domains of an antibody connected via a peptide linker is the smallest antibody format to preserve full antigen-binding properties. However, most scFvs do not tolerate random chemical modification, so the combination with SNAP-tag technology in the form of a fusion protein allows easy modification, e.g. for imaging purposes. We have observed no limitations thus far regarding the substrate to be coupled, and the technology can be used as a platform for the directed modification of many different ligands.
The specificity of the parental antibodies used to generate the scFvSNAP-tag fusion proteins (425 and Ki4) has been demonstrated before [13, 17]. Flow cytometry and confocal microscopy confirmed this specificity for the scFvSNAP probes irrespective of the dye used for labelling.
The goal of this study was to design an EGFR-directed imaging probe that accumulates rapidly in tumours but is also cleared quickly from the body. We observed specific accumulation of 425(scFv)SNAP-747 in subcutaneous xenograft tumours. The DY-747 signal co-localized with the tumour-derived GFP fluorescence and an approximately sixfold higher TBR was observed compared with the non-specific probe Ki4(scFv)SNAP-747. The weak signal observed with the control antibody could be a result of the “enhanced permeability and retention” effect , given that L3.6pl tumours are well vascularized due to over-expression of several pro-angiogenic factors . The EGFR-specific binding of 425(scFv)SNAP in the tumours was further confirmed by the ability of an excess of unlabelled probe to outcompete the labelled probe efficiently.
Next to target specificity, size and functional affinity are key parameters for a targeted imaging probe. Size affects a probes’s capacity to penetrate tissues as well as the clearance rate from the circulation and non-target tissues by renal filtration [19, 20]. Tumour retention of a probe largely depends on its affinity to the target molecule and on the extent of internalization by the targeted cells [21, 22]. ScFvs show rapid but limited accumulation in tumour tissue because their small size ensures they are cleared from the circulation within minutes or hours [23, 24]. We observed a similar clearance profile for both scFvSNAP fusion proteins. Both probes accumulated in the kidneys immediately after injection and were subsequently detected in the bladder, which indicates clearance through renal filtration. However, despite this rapid clearance, tumour accumulation and retention of the 425(scFv)SNAP-747 probe is evidently sufficient to yield very high TBRs 10 h post-injection, allowing for tumour detection and monitoring. The size increase from 28 to 48 kDa, caused by adding the SNAP-tag, may contribute to accumulation at the tumour site due to the prolonged circulation time, as renal filtration efficiency is considered inversely proportional to protein size . The rapid internalization of 425(svFv)SNAP, as demonstrated in vitro, may also contribute to the tumour retention and accumulation observed in vivo.
All mice were placed on a purified diet to avoid the background fluorescence in the digestive tract seen with conventional mouse feed. Nevertheless, a prominent signal was observed in parts of the digestive tract after 24–48 h, although it is possible this reflects the accumulation and breakdown of the probes in the liver and excretion via the bile. We observed limited accumulation of scFvSNAP probes in liver in earlier experiments. No such signal is observed with the C225 full-length antibody, which may indicate an scFv-specific phenomenon.
Full-length antibodies usually display excellent tumour retention. Their large hydrodynamic diameter (∼10 nm) prevents glomerular filtration and together with their high serum stability this leads to prolonged circulation. Although this effect is desirable in therapeutic applications, it results in high background signals in radio or optical imaging. The EGFR-specific C225 has been conjugated to different NIR dyes and has been used successfully as an optical imaging probe [10, 26]. Also in our experiments, targeting of L3.6pl tumours with C225 was very efficient, but the TBR was significantly lower than that of 425(scFv)SNAP because of a high systemic background signal.
In general, scFvs are considered less favourable as the basis for imaging probes due to their lower affinity and rapid clearance from blood [24, 25]. Several reports describe the use of bivalent and multivalent constructs to prolong circulation times and to enhance tumour retention by increasing functional affinity [21, 22, 27]. Fusion to serum albumin has also been shown to increase circulation time and the overall tumour uptake of an scFv . However, for monitoring applications (e.g. EGFR expression during treatment), a high TBR combined with a brief signal, as described herein for the scFvSNAP construct, is beneficial as it allows repeated measurements within a short time frame. A recently reported 111In-labelled affibody molecule directed against EGFR was found superior for imaging purposes in a monovalent form compared to the bivalent form because of the faster systemic clearance . SNAP-tag technology allows the generation of bivalent and trivalent scFvSNAP probes by the use of cross-linkers containing two or more benzylguanine moieties. We are currently investigating this as a tool to customize the pharmacokinetics of scFvSNAP-based probes in combination with more accurate quantification of signals using fluorescence molecular tomography.
In summary, we show the efficient targeting and visualization of EGFR on a subcutaneous xenograft tumour with an NIR-labelled scFvSNAP-tag fusion protein. The tumour accumulation, internalization and rapid clearance of this probe, resulting in high TBRs, allow short-term monitoring of EGFR over-expression using optical imaging. In addition, the SNAP-tag labelling technique could be extended to nuclear imaging modalities, which are more frequently used in clinical settings.
We would like to thank Dr. Richard Twyman for the critical reading of the manuscript and Agnieszka Dreier (Institute for Neuropathology, University Hospital, RWTH Aachen University) for supplying the cetuximab antibody.
Conflicts of interest