Introduction

Driven by the established broad clinical applicability of 18F-FDG scans (Sheikhbahaei et al. 2017), positron emission tomography (PET) has become a leading technique for imaging of aberrant metabolic processes in the human body (Unterrainer et al. 2020). Diagnostic PET imaging aims to visualize disease-specific processes and/or markers using an appropriate selective radiotracer (Kilbourn et al. 2021). Typically, a PET radiotracer is a small target-seeking molecule equipped with a short-lived β+-emitting radioisotope. Amongst suitable isotopes for PET imaging, 18F is routinely used due of its convenient half-life (109.8 min) and low positron energy (Emean = 0.250 MeV), which ensures high image resolution (Conti and Eriksson 2016). Considering the half-life of radionuclides, PET-tracers carrying them are required to possess compatible pharmacokinetics properties, i.e. fast biodistribution and accumulation in the tissue of interest, to enable optimal selective imaging after minutes up to few hours post-injection. Therefore, large target-seeking molecules, such as antibodies, constitute far from ideal candidates for the construction of PET tracers. Typically, antibodies, while offering selective, efficient, rapid, and sustained target binding, have impractically long circulation times of up to days, which limits their use in immuno-PET tumor imaging, particularly with respect to difficult-to-penetrate solid tumours. While the long half-life of 89Zr (78.4 h) has been successfully used for direct labelling of antibodies (Heskamp et al. 2017; McKnight and Viola-Vilegas et al. 2018), this strategy suffers from major impediments related to the long circulating time of the 89Zr-PET probe, notably the high irradiation dose received by healthy tissues (Jauw et al. 2016; Dewulf et al. 1868).

To overcome the limitations of using conventional antibodies for immuno-PET imaging, promising pretargeting approaches have been developed (Altai et al. 2017; Bailly et al. 2017; van de Watering et al. 2014; Marquez and Lapi 2016). Typically, pretargeting involves the injection into the circulation of the patient with an unlabelled, tumour-seeking, bi-functional antibody, which is given time (typically 24–48 h) to reach its optimal target binding at acceptable off-target background levels. Subsequently, a low molecular weight tracer is injected that is labelled with a positron-emitting radioligand for PET imaging, equipped with an inert binding tag (hapten) that can be efficiently captured by bifunctional antibodies that have accumulated at the site(s) of the tumour(s), whereas any unbound tracer rapidly clears from the circulation. Importantly from the application point of view, it is conceivable to administer the bispecific antibody outside of the specialized imaging department, after which the 18F-labeled agent is administered at the hospital site, shortly before the imaging procedure.

Pretargeting strategies notably exploit the strong biotin-streptavidin interaction (Weber et al. 1989; Hnatowich et al. 1987; Rusckowski et al. 1997; Goldenberg et al. 2006) and, in more recent examples, take advantage of the biorthogonal reaction of tetrazines with a trans-cyclooctenes (Rossin et al. 2010; Rossin and Robillard 2014; Rondon and Degoul 2020). Alternatively, pretargeting can effectively be achieved in vivo using so-called bispecific antibodies (bsAbs) (Molema et al. 2000; Lütje et al. 2014; Yu et al. 2017). BsAbs represent a unique class of rapidly emerging therapeutics that combine two target functionalities into one recombinant antibody-based molecule. In this respect, bsAbs have been engineered in which the primary binding domains are directed towards a preselected cancer-associated cell membrane marker and the secondary binding domains towards an inert small molecule (e.g. hapten). Upon selective binding and accumulation of the bsAb at the tumour site(s), the secondary binding domains are available for capturing a haptenylated compound of choice. Previously, pretargeting approaches have been described in which the tumour-directed bsAbs showed second specificity towards a chelator incorporated with a radiometal (Le Doussal et al. 1990), or a histamine-succinyl-glycine (HSG)-labelled tracer (Janevik-Ivanovska et al. 1997; Sharkey et al. 2012; Schoffelen et al. 2013).

Results and discussion

Design and synthesis

Here, we present a proof-of-concept for the use of a novel pretargeting platform, for PET imaging, combining bispecific antibodies (bsAbs) and a fluorescein-based 18F-PET tracer (Fig. 1). We selected fluorescein as a particularly suitable hapten for this approach, since it is an established, non-toxic, FDA-approved agent that is routinely used in angiography and that has confirmed low immunogenicity (Lumbroso et al. 2014). To construct a versatile platform for fluorescein-based pretargeted immuno-PET, we equipped bsAb with affinity-maturated scFv antibody fragments (KD = 1.8 × 10–10 M) that potently bind to fluorescein and several derivatives thereof, including carboxy-fluorescein, fluorescein isothiocyanate (FITC), and Oregon green, and retain the binding potency even when these are conjugated to other compounds (Schwesinger et al. 2000; He et al. 2017). For the design of the fluorescein-based, 18F-bearing tracer, an alkyne moiety was introduced to the fluorescein molecule, enabling rapid labelling using copper(I) azide-alkyne cycloaddition (CuAAC) reaction (Meyer et al. 2016; Gill et al. 2011), to link the modified fluorescein with a fluorinated polyethylene glycol (PEG) chain as shown in Scheme 1 (Sirion et al. 2007; Li et al. 2007). The choice of 18F for the PET-tracer is noteworthy as the majority of previously mentioned strategies utilize higher energy radiometals combined with bispecific antibodies (Le Doussal et al. 1990; Janevik-Ivanovska et al. 1997; Sharkey et al. 2012; Schoffelen et al. 2013). The reference compound, triazole-pegylated-fluorescein (TPF), was prepared first (for details of the synthesis and characterization, see Additional file 1).

Fig. 1
figure 1

Immuno-PET pretargeting approach using fluorescein-based tracers. A bispecific antibody (containing two different scFv antibody fragments) is administered and allowed to circulate for 1–2 days for optimal biodistribution. Then, a labelled fluorescein-based 18F PET-tracer is administered, followed by PET data acquisition

Full size image
Scheme 1
scheme 1

Synthesis of the TPF reference compound, employing CuAAC as the key labelling step. a 4-pentynoic acid, oxalyl chloride, DMAP, NEt3, DCM, rt, overnight (62%). b NaOH in MeOH, rt, 2 h (82%). c 1-azido-2-(2-(2-fluoroethoxy)ethoxy)ethane (3), CuSO4.5H2O, sodium ascorbate, TBTA, DMSO, 75 °C, 1 h (64%)

Full size image

In vitro binding assessment of TPF with bsAb EpCAMxFITC

Preliminary binding studies were performed between the non-radiolabeled TPF and the bsAb EpCAMxFITC that was engineered to have dual high-affinity binding activity for the pan-carcinoma cell surface marker EpCAM and fluorescein. Upon capture by bsAb EpCAMxFITC, the fluorescence of TPF is quenched, which enables the monitoring of the binding in real-time (Saunders et al. 2014). Titration of a TPF solution with bsAb EpCAMxFITC resulted in a significant reduction of the fluorescence signal (see Additional file 1: Figure S2). This effect was also observed when the bsAb EpCAMxFITC was first bound to EpCAM-expressing OVCAR3 cancer cells (see Additional file 1: Figure S3), indicating that TPF binds to bsAb EpCAMxFITC even at lower concentrations of bsAbs, relevant in pretargeting approaches, in this case limited by EpCAM expression. The selective binding of bsAb EpCAMxFITC to the cell surface of EpCAM-expressing OVCAR3 cancer cells was confirmed by flow cytometry (see Additional file 1: Figure S1).

Radiolabelling of [ 18 F]TPF

Next, the radiosynthesis of [18F]TPF was developed to allow for a rapid and efficient labelling procedure (Scheme 2). [18F]F− (3–5 GBq) was azeotropically dried, before the addition of the tosylate precursor 4 (3 mg, 9 µmol) in MeCN. The 18F-fluorination was carried out for 10 min at 110 ºC. The labelled azide [18F]3 was isolated by HPLC in 56 ± 8% radiochemical yield (RCY, n = 7, see Additional file 1: Figure S4 for radio-TLC analysis and Additional file 1: Figure S4 for the HPLC analysis), concentrated in an HLB Plus Short cartridge, eluted with 1.2 mL of DMSO, and further reacted with the fluorescein alkyne 2 (1.0–1.5 mg, 2.3–3.5 µmol) in the presence of a premixed catalytic mixture containing CuSO4, sodium ascorbate and TBTA. The CuAAC reaction allowed for the synthesis of [18F]TPF in high yields (78 ± 7% RCY, n = 8) with a molar activity of Am = 27.6 ± 0.8 TBq/mmol (n = 3), and high radiochemical purity (> 99%). The detailed labelling procedure is available in Additional file 1, together with the HPLC analysis of the reaction (Additional file 1: Figure S6) and product purity (Additional file 1: Figure S7). The in vitro stability of [18F]TPF was evaluated indicating no measurable degradation after incubation in human plasma at 37 °C for 2 h (see Additional file 1: Figure S8).

Scheme 2
scheme 2

Radiosynthesis of [18F]TPF. The fluorescein-based tracer for PET pretargeting is obtained by radiolabeling of precursor 4 towards prosthetic group [18F]3 and its subsequent CuAAC reaction with alkyne 2

Full size image

In this study we focused exclusively on employing [18F]TPF in combination with FITC-containing bsAb for pretargeting purposes. Nonetheless, capitalizing on the diverse uses of fluorescein derivatives, [18F]TPF can prove to be relevant beyond this application. For example, taking advantage of the fluorescent properties of [18F]TPF offers opportunities in dual-modality imaging, combining a PET signal with optical imaging (Jennings et al. 2009). Alternative applications for [18F]TPF also includes the sensitive detection of cerebrospinal fluid leaks by PET imaging (Kommidi et al. 2017; Guo et al. 2019), where similar fluorescein-based 18F-PET tracers have emerged in the recent years.

In vitro pretargeting assay with [ 18 F]TPF on EpCAM-expressing OVCAR3 cells

The radiotracer [18F]TPF was used in a radio-assay to confirm binding to bsAb EpCAMxFITC. Immobilizing of bsAb EpCAMxFITC on protein A agarose beads and comparing the uptake of [18F]TPF with a similarly constructed bsAb EpCAMxMock enabled the evaluation of binding specificity. A nine-fold increase in binding was observed for the bsAb EpCAMxFITC compared with EpCAM-binding bsAb with an irrelevant second binding specificity (bsAb EpCAMxMock) (Additional file 1: Figure S9). Next, a pretargeting experiment was executed using EpCAM-expressing OVCAR3 cancer cells pretargeted (or not) with bsAb EpCAMxFITC, washed to remove any unbound antibody, and then incubated with [18F]TPF. After a final washing step to remove any unbound [18F]TPF, the radiotracer uptake was evaluated. To demonstrate EpCAM-selectivity of this approach, control experiments were performed using OVCAR3 cancer cells in which EpCAM expression was knocked out using CRISPR-Cas9 gene editing technology. Visualization and quantification of bsAb-mediated uptake of [18F]TPF was performed by autoradiography of the cancer cells (Fig. 2). The low, and similar, [18F]TPF uptake observed for both EpCAMneg and EpCAMpos cells that where not incubated with bsAb, confirms that [18F]TPF does not bind directly to EpCAM and is not subject to unspecific uptake by cells, both being important criteria for a suitable PET-tracer. The selectivity of the EpCAMxFITC bsAb for the EpCAM and the efficiency of the washing procedure is demonstrated by the low [18F]TPF uptake observed for EpCAMneg cells incubated with bsAb that did not show a significant difference compared with the uptake observed for EpCAMneg cells not incubated with this bsAb. Finally, the results of the autoradiography indicated enhanced uptake of [18F]TPF by EpCAM-positive OVCAR3 cancer cells when these cells were pretargeted with bsAb EpCAMxFITC, compared to all three control groups (n = 12, P < 0.001). This result indicates that [18F]TPF retains its high affinity for the EpCAMxFITC bsAb, even when the bsAb is already engaged in binding with EpCAM, an essential indication that this approach is suitable for pretargeting of EpCAM-expressing cancer cells. These results suggest that EpCAM-selective PET tumour imaging may be achieved in a two-step pretargeting approach using bsAb EpCAMxFITC and [18F]TPF.

Fig. 2
figure 2

Autoradiography assessment of the capacity of cancer cell-bound bsAb EpCAMxFITC to capture [18F]TPF. See Additional file 1: Section 11 for experimental details

Full size image

[ 18 F]TPF uptake by a variety of bsAbs

To demonstrate the versatility of our bsAb-based pretargeted immuno-PET platform, we extended our study with additional FITC-binding bsAbs, each directed at clinically relevant cancer-associated target molecules, including melanoma marker MCSP (aka CSPG4), epidermal growth factor receptor (EGFR) and immune checkpoint PD-L1. In particular, bsAb MCSPxFITC was engineered to have high affinity binding capacity for the  cell surface molecule MCSP, a promising target for malignant melanoma (Kageshita et al. 1991; Jordaan et al. 2017). Analogously, bsAb EGFRxFITC was engineered to have high affinity binding capacity for EGFR, a clinically relevant cancer-associated cell surface marker that is selectively overexpressed on various difficult-to-treat carcinomas and glioma (Sun et al. 2018). Finally, we evaluated the applicability of our pretargeting strategy using bsAb PD-L1xFITC that was engineered to have high-affinity blocking capacity for the inhibitory immune checkpoint molecule PD-L1, which is frequently overexpressed in various cancer types. Imaging of PD-L1-expressing tumors may be of therapeutic value to identify cancer patients who may profit from cancer immunotherapy based on PD-L1 immune checkpoint blockade (Koopmans et al. 2019; van de Donk et al. 2020; Koopmans et al. 2018). Gratifyingly, target antigen-selective binding was observed for all the studied bsAbs. Specifically, incubation of each of these bead-immobilized bsAbs with the [18F]TPF radiotracer resulted in an up to four-fold increase in binding of [18F]TPF by the corresponding bsAb compared with control bsAb EpCAMxMock, which was not equipped with the anti-fluorescein scFv fragment responsible for [18F]TPF binding (Fig. 3). Noticing no significant difference in [18F]TPF uptake between the bsAb EpCAMxMock and the protein A agarose beads only, we can confidently exclude binding of the [18F]TPF independent of the anti-fluorescein scFv fragment. Importantly, there was no significant difference in [18F]TPF uptake for all bsAbs containing this fluorescein-binding scFv fragment, suggesting that modifying the bsAb format to direct it towards different cell surface tumor markers does not influence [18F]TPF binding by the anti-fluorescein scFv fragment. These results indicate the versatility of our bsAb-based two-step PET imaging approach for various tumour types using one and the same [18F]TPF tracer.

Fig. 3
figure 3

Capture of [18F]TPF by bsAbs immobilized on Protein A agarose beads. See Additional file 1: Section 10 for experimental details

Full size image

Conclusion

In conclusion, the study of a novel pretargeted immuno-PET platform based on a combination of bispecific antibodies and a 18F-fluorescein-based PET-probe showed promising results in vitro. An efficient radiolabeling protocol, exploiting copper-catalyzed click chemistry, was developed for the radiosynthesis of [18F]TPF. The 18F-modified-Fluorescein tracer retained its affinity for FITC bsAbs and the specific uptake of [18F]TPF in EpCAM-overexpressing OVCAR3 cells, pretargeted with EpCAMxFITC bsAb, was confirmed by in vitro autoradiographic assays. These results should encourage further development and optimization of [18F]TPF for in vivo ovarian cancer imaging applications. Lastly, the versatility of the pretargeting approach was demonstrated by [18F]TPF capture by a series of cancer-selective bsAbs, targeting clinically relevant cancer-associated cell surface markers MSCP, PD-L1 and EGFR. We envision that this novel and versatile bsAb-based approach may prove to be useful to broaden the clinical efficacy of immuno-PET tumour imaging.

Methods

General and synthetic methods

For the general description of the materials, cell lines, methods and synthetic procedures, please see Sections 1–3 of the Additional file 1.

Antibody production and binding assessment

For the general description of the antibody construction, eucaryotic production and the assessment of binding and florescence quenching capacity, see Sections 4–7 of the Additional file 1.

Radiolabeling of [18F]TPF

A QMA cartridge was pre-conditioned by passing 10 mL of 1.4% aq. sol. NaHCO3 followed by 15 mL of deionized H2O to reach pH = 7. The cartridge was then dried under argon flow. The [18F]F− in enriched [18O]H2O obtained from the cyclotron (typically 3–5 GBq in 1.2 mL) was passed through the dried cartridge, followed by 5 mL of air to elute most of the H2O. The [18F]F− remaining on the QMA was eluted, into a 4 mL conical vial, with a 1 mL solution containing 15 mg of Kryptofix K222 in 800 μL of MeCN and 1 mg of K2CO3 in 200 μL H2O. Solvents were removed at 115 °C under magnetic stirring and argon flow, 1 mL of anhydrous MeCN was added and evaporated to dryness. This process was repeated 3 times with 0.5 mL of anhydrous MeCN. To the azeotropically dried [18F]F−/K222 mixture was added 3 mg (9 μmol) of 2-(2-(2-azidoethoxy)ethoxy)ethyl 4-methylbenzenesulfonate (5) in 0.5 mL of anhydrous MeCN. (NB.: The tosylate 5 was previously azeotropically dried at 100 °C by repeated addition of anhydrous MeCN and evaporation (3 × 0.5 mL)). The reaction mixture was then stirred at 110 °C for 10 min and, at the end of the reaction, was diluted with 0.5 mL of H2O. The reaction mixture was then injected on HPLC (see Additional file 1: Section 8 for HPLC conditions) and the product ([18F]3) was collected (tR = 8.2 min., 56 ± 8% RCY from dry [18F]F−). The collected fraction was diluted with 80 mL of H2O and concentrated in an Oasis PRiME HLB Plus Short cartridge, which was pre-conditioned with 10 mL of H2O. In the meantime, a catalytic mixture was prepared for the CuAAC reaction. First, 4 mg (20 μmol) of sodium ascorbate in 200 μL of H2O were added to 2.5 mg (10 μmol) Cu2SO4.5H2O dissolved in 100 μL of H2O. The resulting solution was pre-mixed for 5 min to obtain a dark brown solution. 5.6 mg (10.5 μmol) of TBTA in 200 μL of DMSO was added and mixed to afford a light orange solution, from which 200 μL was taken and added to a 4 mL conical vial containing 1.5 mL of fluorescein alkyne (4). [18F]3 was eluted from the HLB cartridge into the same 4 mL conical vial by passing 1.2 mL of DMSO through the HLB cartridge. The reaction mixture was stirred at 80 °C for 20 min. At the end of the reaction, 0.8 mL of H2O (containing 0.1% of formic acid) was added and the solution was purified by HPLC (see Additional file 1: Section 8 for HPLC conditions) and the [18F]TPF product was collected (tR = 23 min, 78 ± 7% RCY from eluted [18F]3). The collected fraction was diluted with 70 mL of H2O and concentrated in a pre-conditioned (10 mL of EtOH followed by 20 mL of H2O) Waters Sep-Pak Plus Light C18 cartridge. [18F]TPF was eluted from cartridge with 1 mL of EtOH and diluted with 9 mL of PBS affording typically 300–700 MBq of product. A sample was taken and submitted to UPLC measurement for QC to assess the purity and molar activity of the product. All the radio-TLC and HPLC analyses can be found in the Additional file 1: Section 8.

Stability [18F]TPF in human serum

In an Eppendorf tube, containing 200 μL of human serum, was added 5 μL of [18F]TPF (50 kBq) in PBS/EtOH (9/1). The mixture was shaken at 37 °C for 0, 60 and 120 min. 200 μL of MeCN was the added to induce protein precipitation. The sample was then vortexed and the supernatant was analyzed on an iTLC and compared with the initial [18F]TPF (Rf = 0.45; 10% MeOH in DCM as eluent). After exposure to a phosphor plate, the iTLC was read on an Amersham Typhoon Imager (see Additional file 1: Figure S8).

Assessment of capacity of bead-bound bsAb EpCAMxFITC to capture [18F]TPF

In short, 10 μg of bsAb EpCAMxFITC (or bsAb EpCAMxMock) was mixed with a slurry of 50 µl Protein A agarose beads (SinoBiological) in final volume of 500 µl PBS, and the mixture incubated at rt for 30 min. Next, the beads were pelleted by centrifugation (2500 rpm, 5 min) and washed 2 times with PBS to remove unbound bsAb. The washed beads were resuspended in 100 μL PBS, mixed with 25 μL of a [18F]TPF (0.5–1.0 MBq) solution in PBS/EtOH (9:1 vol/vol), and incubated under continuous gentle shaking at rt for 15 min. Then, PBS was added to a final volume of 500 µl, after which the beads were pelleted by centrifugation (2500 rpm, 5 min). Next, 480 μL of the supernatant solution was collected and replaced by the same volume of PBS after which the beads were pelleted again. This washing procedure was repeated 2 more times after which radioactivity associated with beads and that present in the combined supernatants were then measured using a gamma counter.

Assessment of capacity of cancer cell-bound bsAb EpCAMxFITC to capture [18F]TPF

In short, 10 × 103 OVCAR-3 cells per well were seeded in an 8 wells permanox chamber slide (LAB-TEK, Nagle Nunc Int.), allowed to adhere, and then incubated (or not) with 10 μg of bsAb EpCAMxFITC (or bsAb EpCAMxMock) at 37 °C for 1 h. Next, the cells were washed 2 times with PBS to remove unbound bsAb and then incubated (or not) with 200 μL of 0.5–1 MBq of [18F]TPF in PBS (< 2% EtOH) at rt for 30 min. Subsequently, the cells were washed two times with PBS to remove unbound [18F]TPF. Finally, the growth chamber was carefully removed, after which the glass slide with adhered cells was carefully washed with PBS and then air-dried. The dried glass slide was wrapped in aluminium foil and exposed to a phosphor plate for up to 1 h. The phosphor plate (see Section 11 of the Additional file 1) was imaged using a Amersham Typhoon Imager to record the autoradiographic images.