Pretargeting of internalizing trastuzumab and cetuximab with a ¹⁸F-tetrazine tracer in xenograft models

During recent years, the radiolabelling of clinically relevant monoclonal antibodies (mAbs), like cetuximab and trastuzumab, has been under intense investigation for the generation of diagnostic and therapeutic radiopharmaceuticals that can identify and treat tumours expressing the target antigen [1]. Traditional radiolabelling methods of mAbs for PET (positron emission tomography) and SPECT (single-photon emission computed tomography) usually involve the use of long-lived radioisotopes, such as zirconium-89 (t _½ = 3.3 days), indium-111 (t _½ = 2.8 days), and iodine-124 (t _½ = 4.2 days); the half-lives of which are matched to the biological half-lives of mAbs. Although the target-to-background ratios obtained with directly radiolabelled mAbs are excellent, typically, several hours to days are required before non-target radioactivity has decreased to sufficiently low levels necessitating the use of long-lived isotopes. An additional caveat is the increased radiation burden to the subject, especially to the bone marrow resulting from long circulation time of mAbs. In pretargeting, the antibody is radiolabelled in vivo after it has accumulated to the tumour and most of the unbound antibody has cleared from the blood flow (Fig. 1), enabling the use of short-lived radioisotopes with dosimetric benefits [2‐4]. Earlier approaches to pretargeting included streptavidin-biotin system [5], bispecific antibodies targeting the antigen with In-DTPA [6] or Y-DOTA [7] complexes, and antibody-modified oligonucleotide conjugates [8]. In the recent years, bioorthogonal chemical reactions have emerged as efficient tools for pretargeted nuclear imaging of mAbs [2, 4, 9‐15]. Bioorthogonal reactions occur under physiological conditions without interfering with or being interfered by any of the native biochemical processes of the system [16, 17]. The most studied bioorthogonal reactions for pretargeting are the strain-promoted alkyne azide cycloaddition (SPAAC) and the inverse electron-demand Diels-Alder (IEDDA) cycloaddition between a tetrazine and a dienophile.

The radioisotopes used in previous in vivo pretargeting studies have included indium-111 (t _½ = 2.8 days) [15, 18], copper-64 (t _½ = 12.7 h) [2, 13, 19], technetium-99 m (t _½ = 6.0 h) [12, 20], fluorine-18 (t _½ = 109.8 min) [4, 21, 22], gallium-68 (t _½ = 67.6 min) [11], and carbon-11 (t _½ = 20.4 min) [23]. The half-life dictates the maximum time allowed after tracer injection for imaging. With indium-111, copper-64, and technetium-99 m, the time needed for the clearance of the tracer from non-target tissues plays a lesser role because the imaging can be performed several hours or even days later than the tracer injection. While the advantage in the use of the latter short-lived radioisotopes is that the effective dose to the subject is decreased, the challenge is to modify the pharmacokinetics of the tracer to allow for rapid elimination within the imaging time frame dictated by the shorter half-life. High molar activity ¹⁸F is nowadays widely available in clinical PET centres, and the half-life of 109.8 min of ¹⁸F allows for multistep synthesis, the preparation of multiple imaging doses from one batch of the radiotracer, and transportation of the tracer to remote sites if needed. Therefore, the choice of a fluorine-18-labelled tracer balances between the dosimetric limitations in the use of the long-lived radiometal-labelled tracers and the synthetic challenges of short-lived ⁶⁸Ga- and ¹¹C-labelled ones. In this work, we used a previously developed ¹⁸F-radiolabelled tetrazine tracer, [¹⁸F]TAF [24] (Fig. 1). The radiosynthesis of [¹⁸F]TAF is carried out using [¹⁸F]-5-fluoro-5-deoxy-ribose ([¹⁸F]FDR) as a prosthetic group to radiolabel aminooxy-functionalized tetrazine via oxime formation. The fluoroglycosylation with [¹⁸F]FDR is a very attractive choice for radiolabelling tetrazines with fluorine-18 since it is a high-yield reaction with mild reaction conditions compatible with the inherent instability of tetrazines. In addition, by using different linkers between the tetrazine and the aminooxy group, the structure of the tetrazine radiotracer can be altered depending on the application. We have previously used [¹⁸F]TAF successfully for pretargeting of porous silicon (PSi) nanoparticles after intravenous administration [25]. Although radiotracers labelled with longer-lived radioisotopes including ⁶⁴Cu and ¹¹¹In have been successfully used for pretargeting antibodies previously [9, 15], the time between the injection of the radiotracer and the start of the imaging acquisition have been rather long, and this lag time as well as the radioburden to the subject could be decreased further with radiotracers labelled with shorter-lived isotopes.

We have previously shown that the higher the molar activity of the tetrazine radiotracer, the more efficient the bioorthogonal reaction is in vivo [25], while others view that diluting the tetrazine radiotracer with additional non-radiolabelled compound is necessary for maintaining high reaction rates in vivo [4, 9, 26]. Since our earlier experiments were done with nanoparticles that were cleared from the blood rapidly, we selected an amount of 1:1 ratio of cold TAF tetrazine to the amount of trans-cyclooctene (TCO) in the antibody because this has been shown to be successful in the literature specifically in antibody pretargeting. This approach also provides both blocking of the excess TCO in the blood as well as driving of the equilibrium of the bioorthogonal reaction towards the product.

Herein, we have investigated the efficacy of two-step pretargeting strategy based on IEDDA reaction for tracing two clinically relevant antibodies, cetuximab and trastuzumab, with [¹⁸F]TAF (Fig. 1) in epidermal growth factor receptor (EGFR)-positive (A431 epidermoid carcinoma) and HER2-positive (BT-474 ductal carcinoma) xenografts in vivo. Cetuximab binds to the extracellular domain of the EGFR that is over-expressed and/or activated in many types of cancer. Trastuzumab is an anti-HER2 mAb that is used for the treatment of HER2/neu-positive breast cancers. Both EGFR (ErbB1/HER1) and HER2/neu (ErbB2) belong to the HER family of tyrosine kinase receptors. Both cetuximab and trastuzumab exhibit slow pharmacokinetics and are known to internalize over time [27‐31]. In principle, the cell membrane-bound antibody is the only fraction that is available for pretargeting with a non-internalizing radiotracer. Notably, the antibodies previously used successfully for pretargeted imaging either accumulate to the tumour at exceptionally high concentrations such as 5B1 targeting the carbohydrate antigen CA 19.9 in pancreatic cancer [32, 33] or show high persistence on the cell surface even when bound to the antigen, like huA33 targeting a transmembrane glycoprotein in colorectal cancer [9, 34]. In this work, the known internalization of the selected mAbs brings an additional challenge for the pretargeted system together with the residual unbound antibody remaining in the blood flow at the time of tracer injection. Despite these complications, we could localize the tumour in the PET images, showing that pretargeted imaging strategies could be extended to the imaging of a wider selection of antibodies with careful tailoring of the radiotracer pharmacokinetics.

For both antibodies, the tumour could be visualized from the pretargeted PET images, albeit the tumour uptake of radioactivity was considerably lower than that for the ⁸⁹Zr-labelled mAbs. This is in good agreement with the cell uptake studies (see Additional file 1: Figure S1) where we showed that the hydrophilicity of [¹⁸F]TAF prevents cellular internalization of the tracer alone in EGFR and HER2-positive cell lines, but in vitro pretargeting with cetuximab and trastuzumab allows for the internalization of the tracer as a result of the IEDDA reaction occurring with the mAb on the cell surface (Additional file 1: Figure S3). The degree of cell-associated radioactivity, however, is considerably lower than what is seen either with the DFO-conjugated ⁸⁹Zr-labelled mAbs or when [¹⁸F]TAF is allowed to click with the TCO-modified mAb before it is applied to the cells, as illustrated by the results of the immunoreactivity assay, corroborating that the internalized fraction of the antibody remains beyond reach for this tracer.

With cetuximab, the highest tumour-to-blood ratios were achieved when the lag time between mAb and tracer injections was 72 h. This is most likely due to the lower levels of circulating unbound TCO-cetuximab in the blood flow at the later time point. An additional reason might be the downregulation of the EGF receptor induced by cetuximab [41], as the internalization rate slows down at the later time points compared to the earlier time points. This might lead to higher fraction of surface-bound cetuximab accessible for [¹⁸F]TAF at the later time points. With trastuzumab, no difference was seen between the lag times of 48 h and 72 h. With trastuzumab, the tumour-to-blood ratios have been reported to improve between 48 and 72 h due to a higher amount of the mAb at the tumour while the blood levels of unbound trastuzumab remain unchanged between these time points [38, 39]. This explains why we did not observe any difference in the radioactivity concentration in the blood between different antibody administration lag times with pretargeted trastuzumab. Furthermore, although there is more trastuzumab at the tumour at 72 h, the amount of non-internalized trastuzumab available for the reaction with [¹⁸F]TAF might in fact be the same. Hence, we did not observe increased tumour uptake at 72 h compared to 48 h.

Houghton et al. and Sharkey et al. have both showed that it is possible to pretarget an antibody that internalizes with a tracer that does not internalize [2, 42]. Our work is in accordance with this finding, but naturally, a more interesting approach would be to use a tracer that would internalize. In this case, it would be important that the unreacted radiotracer would not remain inside the cells, which could be challenging to achieve by tuning the lipophilicity of the tracer alone. This might restrict the use of internalizing radiotracers for pretargeted imaging due to the high background radioactivity in non-target organs.

Development of a “reverse” approach to Tz/TCO ligation started with the development of ¹⁸F-labelled TCO by Li et al. [43]. In this reverse approach, the tetrazine moiety is attached to the desired platform, e.g., antibody or nanoparticle, and the radiolabel is attached to the TCO counterpart. This approach is very appealing because the higher hydrophilicity of tetrazine compounds compared to TCO compounds may prevent hydrophobic burying of the tags to the antibody structure as reported by Rahim et al. [44] resulting in a higher effective functional loading on the surface of the platform. The reverse approach has been successfully used in pretargeted fluorescence live cell imaging, where tracer pharmacokinetics and metabolism are understandably less of a concern than in vivo [45]. However, utility of ¹⁸F-TCO for in vivo pretargeting is very limited [46]. ¹⁸F-TCO enters the brain indicating that this tracer could also internalize cells, but unfortunately, it was quickly metabolized in vivo precluding its use for pretargeting. A more stable conformationally strained TCO (s-TCO) has been used in live cell imaging and radiolabelling of biomacromolecules, but not in pretargeted experiments in live animals [47‐50]. Darko et al. have developed even more stable conformationally strained TCO derivatives (d-TCO) and studied their stability and reaction kinetics [51]. The in vivo properties of the new d-TCO are very promising when considering possible internalizing tracers for pretargeting [22].

A crucial factor for the development of a successful pretargeted system is the elimination properties of the radiotracer. Renal clearance should be favoured over the hepatobiliary excretion in order to avoid unwanted, long-residing background radioactivity in the gastrointestinal tract, as high radioactivity levels in the intestines hamper the delineation of nearby organs. The pharmacokinetic and elimination profile of a tetrazine can be altered, for example by the addition of a hydrophilic linker and/or changing the chelator [10]. Comparison of the structures and biodistribution of the published tetrazines [2, 4, 9‐13, 19, 20, 22] for pretargeted PET imaging shows that there are multiple factors that contribute to the pharmacokinetic profile. However, there might be one highly influencing factor, namely the introduction of positive charge to a radiopharmaceutical, which can increase the renal clearance [52, 53]. Despite the non-optimal elimination kinetics of [¹⁸F]TAF, our data shows that it can be used for the pretargeted PET imaging of antibodies. Given that this approach for pretargeting cetuximab and trastuzumab is intended for the imaging of tumours outside the abdominal area, the hepatobiliary excretion would not interfere with the image analysis as much as it would in orthotopic colorectal tumour models. The three hydroxyl groups in the [¹⁸F]TAF precursor, [¹⁸F]-5-fluoro-5-deoxy-ribose ([¹⁸F]FDR), lower the lipophilicity of the radiolabelled oxime product, which is desired for fast urinary elimination. However, [¹⁸F]TAF has some uptake in the gastrointestinal tract and further optimization of the pharmacokinetics of [¹⁸F]TAF is needed to improve the present system. The structure of [¹⁸F]TAF, however, can be modified by attaching almost any linker between the tetrazine and the aminooxy groups. The radiolabelling could still be done by using [¹⁸F]FDR as a prosthetic group, making the oxime radiolabelling approach very appealing to be used with tetrazines due to the mild reaction conditions of the glycosylation. The oxime formation with [¹⁸F]FDR is a high-yield reaction that can be performed in room temperature in 10 min. The optimization of the structure of [¹⁸F]TAF towards renal clearance, however, fell beyond the scope of this work.

Recently Knight et al. presented a new pretargeted approach called HaloTag enzyme for the pretargeting of two targets: internalizing HER2 and non-internalizing TAG-72 [18]. Although the HaloTag enzyme could be used for pretargeting TAG-72 in vivo and was capable of detecting HER2 expression in vitro, it lacked the sensitivity for in vivo pretargeting of trastuzumab. The tumour uptake in the pretargeted trastuzumab experiments did not reach values significantly higher than that in pretargeted IgG. In our experiments, while the tetrazine tracer accumulated to the tumour in higher degree in the pretargeted trastuzumab experiments than in the pretargeted IgG experiments, the radioactivity accumulation in the tumour was not nearly as high as with directly radiolabelled [⁸⁹Zr]Zr-DFO-trastuzumab. Both pretargeting approaches clearly faced the same challenge of an internalizing target antigen rapidly localizing in the cellular compartment beyond the reach of the radiotracer, prompting the future development of internalizing bioorthogonal reagents for the pretargeted imaging of trastuzumab.

Evans et al. have pretargeted cetuximab with a ⁶⁸Ga tracer with 23-h lag time [11]. They observed high background levels in non-target organs. In our study, we observed some unbound antibody circulating even after a lag time of 72 h between the antibody and radiotracer injections. A lucrative strategy for the optimization of pretargeted systems involving long-circulating antibodies might be the use of antibody-clearing agents [54, 55]. With these agents, it might be possible to reduce the amount of circulating TCO-antibody in the blood and improve tumour delineation. This could be beneficial in two ways: (i) to improve the in vivo IEDDA reaction yield at the tumour site and (ii) to reduce the lag time between the antibody and tracer injection so that more antibody is still on the surface of the tumour cells and available for reacting with the non-internalizing tracer. However, the synthetic development of efficient TCO-scavenging clearing agents is still in its infancy limiting wider adaptation of the technique to pretargeted PET studies at present.

We have pretargeted two clinically relevant monoclonal antibodies by using ¹⁸F-tracer under challenging conditions, wherein the mAb is internalized and unbound mAb remains in circulation, and demonstrated that a specific, bioorthogonal IEDDA reaction takes place at the tumour despite these challenges. This demonstrates the potential of the pretargeted approach for targeting of rapidly internalizing antigens expanding the scope of bioorthogonal chemistry for the development of molecular imaging tools for a wider variety of antibodies and intracellular targets. In our case, some further optimization of the pharmacokinetics of the tracer is still needed for reduction of the background radioactivity in the excretory organs. Besides the pharmacokinetic optimization of the tetrazine tracer structure, the use of an antibody clearing agent could be beneficial in the models used in order to reduce the observed high radioactivity levels in the blood remaining even after 72 h of antibody administration. It might also make it possible to shorten the time between the antibody and tracer injections, when a higher concentration of the TCO-modified antibody still resides in the cell surface available for the in vivo IEDDA reaction.