HER3 PET Imaging: 68Ga-Labeled Affibody Molecules Provide Superior HER3 Contrast to 89Zr-Labeled Antibody and Antibody-Fragment-Based Tracers

Simple Summary HER3 is a known driver for oncogenesis and therapy resistance in solid cancers. PET imaging could be a useful tool to non-invasively detect and monitor HER3 expression and aid in the selection of patients for HER3-targeted therapy. PET tracers based on therapeutic antibodies have thus far shown limited success in reliably imaging HER3-expressing tumors in clinical trials. Smaller-sized tracers specifically designed for imaging might be needed for higher contrast imaging and sufficient sensitivity. Our group has previously studied the use of radiolabeled affibody molecules for imaging of HER3 expression. In the present study, we compared four different types of potential PET tracers for imaging of HER3 expression in a preclinical model. We demonstrated that the affibody-based tracer, [68Ga]Ga-ZHER3, could provide overall superior imaging contrast to antibody- and antibody-fragment-based tracers shortly after injection. Our results indicate that HER3-targeting affibody molecules are promising agents for PET imaging of HER3 expression. Abstract HER3 (human epidermal growth factor receptor type 3) is a challenging target for diagnostic radionuclide molecular imaging due to the relatively modest overexpression in tumors and substantial expression in healthy organs. In this study, we compared four HER3-targeting PET tracers based on different types of targeting molecules in a preclinical model: the 89Zr-labeled therapeutic antibody seribantumab, a seribantumab-derived F(ab)2-fragment labeled with 89Zr and 68Ga, and the 68Ga-labeled affibody molecule [68Ga]Ga-ZHER3. The novel conjugates were radiolabeled and characterized in vitro using HER3-expressing BxPC-3 and DU145 human cancer cells. Biodistribution was studied using Balb/c nu/nu mice bearing BxPC-3 xenografts. HER3-negative RAMOS xenografts were used to demonstrate binding specificity in vivo. Autoradiography was conducted on the excised tumors. nanoPET/CT imaging was performed. New conjugates specifically bound to HER3 in vitro and in vivo. [68Ga]Ga-DFO-seribantumab-F(ab’)2 was considered unsuitable for imaging due to the low stability and high uptake in normal organs. The highest tumor-to-non-tumor contrast with [89Zr]Zr-DFO-seribantumab and [89Zr]Zr-DFO-seribantumab-F(ab’)2 was achieved at 96 h and 48 h pi, respectively. Despite lower tumor uptake, [68Ga]Ga-ZHER3 provided the best imaging contrast due to the fastest clearance from blood and normal organs. The results of our study suggest that affibody-based tracers are more suitable for PET imaging of HER3 expression than antibody- and antibody-fragment-based tracers.


Introduction
The human epidermal growth factor receptor type 3 (HER3) is a target of increasing interest for molecular imaging in oncology. HER3 is known to be a marker for poor prognosis and a mediator of therapeutic resistance in several cancers, for example, breast and prostate cancer [1][2][3]. Currently, several clinical studies with HER3-targeting or co-targeting therapeutic agents are ongoing (clinincaltrials.gov (last accessed 10 August 2021), [3]). PET imaging of HER3 expression could be a valuable tool for improving the treatment of patients with HER3-expressing disease, allowing for non-invasive, repetitive whole-body detection and monitoring of HER3 expression as well as assessment of therapy response. However, development of high contrast agents for HER3 imaging is challenging due to the substantial HER3 expression in several healthy organs, e.g., salivary glands, organs of the GI tract, and particularly the liver (proteinatlas.org). Moreover, the level of overexpression of HER3 in tumors is relatively low, generally not exceeding 50,000 receptors/cell [4] (for reference, breast cancer patients are considered HER2+ with expression >2.3 × 10 6 receptors/cell [5]).
Many factors need to be considered when developing potential molecular imaging agents. The choice of the targeting molecule (e.g., its size, polarity, and local charge) could appreciably influence the pharmacokinetics and biodistribution of the tracer and thereby the image contrast and sensitivity of the scan [6][7][8]. The ideal radiotracer should bind to the target receptor with high affinity, clear efficiently from the blood, and have minimal accumulation in non-targeted tissue [6,8]. Radiolabeled monoclonal antibodies were among the first tracers to be evaluated for molecular imaging of the HER-receptor family [8,9]. However, the slow extravasation and long residence time in blood are potential disadvantages for antibody-based imaging agents, because they result in an elevated background signal from blood-borne activity and increase the unspecific accumulation in non-targeted tissue [10]. Sufficient imaging contrast with antibody-based tracers is generally not achieved until several days after injection [11][12][13]. Decreasing the size of the imaging agent can provide better extravasation and faster clearance from the system, which may enable imaging with good contrast just a few hours after injection [6,14]. In recent years, a variety of smaller tracers based on different classes of targeting molecules have emerged as alternatives to antibody-based tracers and shown potential value and advantages as radionuclide-based agents for imaging of HER family members. These include peptides, antibody-fragments (e.g., Fab-and F(ab') 2 -fragments and scFvs), single-domain antibodies (sdAbs, e.g., nanobodies), and engineered scaffold proteins (ESPs) [6,8,10].
Several of the HER3-targeting therapeutic antibodies under clinical and preclinical evaluation are being explored for PET imaging. [ 89 Zr]Zr-lumretuzumab and [ 89 Zr]Zr-GSK2849330 were able to visualize HER3 expression in patients 4-7 d pi [11,15]. [ 89 Zr]Zrlumretuzumab was able to detect 67.6% of HER3-expressing lesions with a size > 10 mm, but uptake quantification was only possible in 50% of the detected lesions [11]. [ 89 Zr]Zr-GSK2849330 demonstrated dose-dependent tumor uptake in a small six-patient study [15]. With both of these 89 Zr-labeled tracers, the image contrast was limited by the accumulation in normal tissue. Particularly the high uptake in the liver (due to the hepatobiliary excretion of antibodies) was considered problematic because it limits the detection of potential liver metastases. A clinical study with [ 64 Cu]Cu-patritumab included 11 patients, but was terminated because of low uptake in HER3-expressing lesions [16]. More HER3-targeting antibodies for imaging have been studied in preclinical settings [17,18]. A few alternative HER3-imaging agents based on other targeting molecules have also been reported in preclinical studies. [ 64 Cu]Cu-DOTA-mAb105-F(ab ) 2 was able to detect changes in HER3 expression in response to therapy with AKT and PI3K inhibitors [19]. A biparatopic nanobody conjugate, [ 89 Zr]Zr-MSB0010853, showed maximum tumor uptake in xenografts 24 h pi and imaged HER3 expression up to 96 h pi. The undecapeptide [ 68 Ga]Ga-HER3P1 could visualize HER3 expression, but requires further optimization to increase the uptake in tumors [20].
More extensive preclinical work has been performed on the development of HER3targeting affibody molecules, a class of ESPs. The small size (7 kDa) of affibody molecules presents a distinct advantage in the context of radionuclide-based imaging due to the more rapid blood clearance by renal filtration compared with antibody-based radiotracers. Additional benefits include high thermal and chemical stability and straightforward bacterial production [21]. Affibody molecules have been engineered to bind with high affinity and specificity to a multitude of targets including HER3. The affibody Z HER3:08698 has been developed to bind HER3 with low picomolar affinity and with cross-reactivity to murine ErbB3 [22]. Radiolabeled HER3-binding affibody molecules for PET and SPECT imaging have been used to visualize HER3 expression as early as 1 h pi, differentiate between high-and low-expressing xenografts models, and to detect changes in HER3 expression and receptor occupancy during HER3-targeted therapy [23][24][25][26]. Additional optimization of the molecular format and labeling approach resulted in major improvement in imaging contrast, particularly the tumor-to-liver contrast [27][28][29]. Among the studied variants, [ 68 Ga]Ga-(HE) 3 -Z HER3 -NODAGA and radiocobalt-labeled (HE) 3 -Z HER3 -DOTA have shown the most promising characteristics for affibody-based imaging of HER3 expression thus far [27,30].
There might be perceived advantages in using radiolabeled versions of pre-existing therapeutic antibodies because, in addition to being used for target detection, they can also be used for dosing studies and imaging of receptor occupancy. However, for a difficult target such as HER3, smaller imaging agents, specifically designed for imaging, may provide a lower background signal, resulting in better contrast, enabling same-day scanning. Thus far, no side-by-side comparisons of different HER3-targeting diagnostic agents have been reported. The aim of this study was to compare the PET imaging properties of HER3-targeting tracers, based on three different molecules, at their respective favorable imaging time points for visualization of HER3 expression in a preclinical mouse model: a 89 Zr-labeled version of the HER3-targeting therapeutic antibody seribantumab (MM-121), which, to our knowledge, has not yet been reported; a seribantumab-derived 68 Ga and 89 Zr labeled F(ab') 2 -fragement, and the 68 Ga-labeled affibody molecule [ 68 Ga]Ga-Z HER3 . This study demonstrated that, even at their respective favorable imaging time points, the achieved image contrast varied appreciably depending on the type of targeting molecule, due to the different pharmacokinetics and interactions with healthy tissue. Due the fast clearance and low accumulation in the background, the affibody-based tracer provided the overall best imaging contrast already 3 h pi.

General
HER3-expressing human cancer cell lines BxPC-3 (pancreatic cancer) and DU145 (prostate cancer) were obtained from ATCC (via LGC Promochem, Borås, Sweden). Cells were maintained in RPMI 1640 media (with L-glutamine, Biowest, Riverside, MO, USA) supplemented with 10% fetal bovine serum (Sigma-Aldrich, St. Louis, MO, USA) and 1% penicillin-streptomycin (Biochrom, Berlin, Germany). 89 Zr (solution in 0.1 M oxalic acid) was purchased from Perkin Elmer (Waltham, MA, USA). 68 Ga was obtained by elution of a 68 Ge/Ga-generator (Cyclotron Co., Obninsk, Russia) with 0.1 M metal-free HCl. The radioactivity content in the samples was measured in an automated gamma counter with an NaI(TI) detector (1480 Wizard, Wallac, Finland). The statistical significance (p < 0.05) for in vitro and in vivo specificity experiments was assessed by a two-sided, unpaired t-test. One-way ANOVA with post-hoc t-test corrected for multiple comparisons using Tukey correction was used for statistical analysis of the biodistribution experiments. All in vivo experiments were approved by the local authorities for laboratory animal welfare (Uppsala, approval number C4/16).

Generation of Seribantumab-F(ab') 2 and DFO Conjugation
Seribantumab was purchased from Evitria (Zurich, Switzerland). Seribantumab-F(ab') 2 was generated from enzymatic digestion using the Pierce™ F(ab') 2 Preparation Kit (Thermo Scientific, Chicago, IL, USA) according to the manufacturer's instructions. Briefly, seribantumab (4.1 mg/mL) was incubated in a spin column containing immobilized pepsin equilibrated with digestion buffer at 37 • C for 6.5 h in a rotamixer. The antibody digest and immobilized pepsin were separated by centrifugation. Non-digested seribantumab and Fc fragments were separated from seribantumab-F(ab') 2 with size-exclusion chromatography (SEC) on a 1200 series HPLC system (Agilent Technologies, Santa Clara, CA, USA) using a Superdex ® 200 Increase 10/300 GL column (Cytiva, Uppsala, Sweden) and a flow rate of 0.5 mL/min of PBS. Successful removal of non-digested seribantumab and Fc fragments was evaluated with SDS-PAGE.
For biodistribution studies, mice bearing BxPC-3 xenografts were intravenously injected with equimolar protein doses corresponding to 27 2 3 h pi. At each time point, a group of n = 3-4 mice was sacrificed by heart puncture after intraperitoneal injection of a mixture of ketamine (250 mg/kg) and xylazine (25 mg/kg). Samples of blood, salivary glands, lung, liver, stomach, spleen, small intestine, kidneys, tumor, muscle and bone were collected, weighed, and measured for radioactivity content.
Mice bearing both HER3 positive BxPC-3 and HER3 negative RAMOS xenografts were used for the in vivo specificity and 96 h pi ([ 89 Zr]Zr-DFO-seribantumab) and treated according to the biodistribution protocol described above.
Hematoxylin and eosin staining of the tumor sections was performed following a standard protocol. After staining, the sections were scanned using PathScan Enabler IV (Meyer Instruments, Houston, TX, USA). Whole-body nanoPET images were acquired using a nanoScan PET/MR (Mediso Medical Imaging Systems Ltd., Budapest, Hungary). Scan times were 45-60 min for [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 and [ 89 Zr]Zr-DFO-seribantumab, and 30 min for [ 68 Ga]Ga-Z HER3 . A CT scan was performed immediately after the PET scan, using a nanoScan SPECT/CT (Mediso Medical Imaging Systems Ltd., Budapest, Hungary) with the same bed. The parameters for the CT scans were a 5 min acquisition time, an X-ray energy peak of 50 keV/670 µA, and 480 projections. Reconstruction of the PET scans was conducted using the Tera-Tomo™ 3D reconstruction engine with decay correction at the injection administration time. CT data were reconstructed using filter back projection in Nucline 2.03 Software (Mediso Medical Imaging Systems Ltd., Budapest, Hungary). PET and CT scans were fused using InterView FUSION software (Mediso Medical Imaging Systems Ltd., Budapest, Hungary).

Generation of Seribantumab-F(ab') 2 and DFO Conjugation
Close to complete digestion was achieved for seribantumab following incubation with pepsin ( Figure S1, third lane). Residual non-digested seribantumab and Fc fragments were successfully removed by SEC purification, shown by SDS-PAGE ( Figure S1, fourth lane).

Kinetic Evaluation of Seribantumab-F(ab') 2 and Seribantumab
Seribantumab-F(ab') 2 retained its binding to both human HER3 and murine ErbB3 after pepsin digestion, shown by representative sensorgrams with fitted curves ( Figure S2A,B, respectively). Average K D -values based on two surfaces with different immobilization levels are displayed in the sensorgrams. Seribantumab-F(ab') 2 and seribantumab demonstrated affinities in the low nanomolar range for both human HER3 and murine ErbB3 (mErbB3). Seribantumab-F(ab') 2 exhibited slightly higher affinity compared to seribantumab.

Radiolabeling and Stability
Labeling of all conjugates was successful. Radiochemical yields determined by radio-iTLC are shown in Table 1. [ 89 Zr]Zr-DFO-seribantumab was used without further purification. Purity of [ 89 Zr]Zr-seribantumab-DFO-F(ab') 2 after purification with NAP5-size exclusion columns was 97 ± 1%. The radiochemical purity of [ 68 Ga]Ga-Z HER3 was 97 ± 2%.  2 was stable when incubated in PBS, but after incubation in human serum for 3 h, the majority of the activity had dissociated from the protein.  Figure 1B, we did consider blocking with Z HER3 to be sufficient to demonstrate HER3 specific binding for [ 68 Ga]Ga in Figure 1C. Addition of the control antibody, VEGFR-targeting bevacizumab, did, as expected, not influence the uptake of [ 89 Zr]Zr-DFO-seribantumab and [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 . In vitro specificity of [ 68 Ga]Ga-Z HER3 was earlier reported [27].

Radiolabeling and Stability
Labeling of all conjugates was successful. Radiochemical yields determined by radio-iTLC are shown in Table 1 Ga]Ga-DFO-seribantumab-F(ab')2 was stable when incubated in PBS, but after incubation in human serum for 3 h, the majority of the activity had dissociated from the protein.

In Vitro Characterization of [ 89 Zr]Zr-DFO-Seribantumab, [ 89 Zr]Zr-DFO-Seribantumab-F(ab')2, and [ 68 Ga]Ga-DFO-Seribantumab-F(ab')2
All compounds demonstrated HER3-specific binding when tested on BxPC-3 and DU145 cells in vitro. Pre-saturation of HER3 receptors resulted in significantly lower uptake of [ 89 Zr]Zr-DFO-seribantumab, [ 89 Zr]Zr-DFO-seribantumab-F(ab')2, and [ 68 Ga]Ga-DFO-seribantumab-F(ab')2 compared with the non-blocked control group in both BxPC-3 and DU145 cells ( Figure 1). Taking into account the limitations associated with the short half-life of 68 Ga and due to the confirmed cross-blockability shown in Figure 1B, we did consider blocking with ZHER3 to be sufficient to demonstrate HER3 specific binding for [ 68 Ga]Ga in Figure 1C. Addition of the control antibody, VEGFR-targeting bevacizumab, did, as expected, not influence the uptake of [ 89 Zr]Zr-DFO-seribantumab and [ 89 Zr]Zr-DFO-seribantumab-F(ab')2. In vitro specificity of [ 68 Ga]Ga-ZHER3 was earlier reported [27].  To study the internalization of the new radioconjugates, HER3-expresisng BxPC-3 cells were continuously incubated with the compounds for up to 24 h. The total cell uptake and internalized activity of all radioconjugates continuously increased with time (Figure To study the internalization of the new radioconjugates, HER3-expresisng BxPC-3 cells were continuously incubated with the compounds for up to 24 h. The total cell uptake and internalized activity of all radioconjugates continuously increased with time (Figure 2A-F The internalized fraction of [ 68 Ga]Ga-DFO-seribantumab-F(ab') 2 was 18 ± 4% and 10 ± 2% of cell associated activity in BxPC-3 and DU145 cells after 4 h, respectively. Uptake and internalization of [ 68 Ga]Ga-Z HER3 was studied earlier [27].  Figures  3-5, and the numerical biodistribution data are available in Supplementary Materials, Tables S1 and S2.

In Vivo Biodistribution and Specificity
In vivo specificity and biodistribution of the conjugates were studied in Balb/c nu/nu mice with HER3-positive and HER3-negative RAMOS xenografts.  2 , and [ 68 Ga]Ga-DFO-seribantumab-F(ab') 2 in HER3negative RAMOS xenografts was significantly lower compared with the uptake in BxPC-3 xenografts (Figure 3), demonstrating HER3-specific binding of the radioconjugates in vivo. Binding specificity of [ 68 Ga]Ga-ZHER3 towards HER3 in vivo was confirmed previously [27].
Among all the tested tracers and time points, the uptake in BxPC-3 xenografts was the highest for [ 68 Ga]Ga-DFO-seribantumab-F(ab') 2 (11 ± 3%ID/g) and [ 89 Zr]Zr-DFOseribantumab-F(ab') 2 3 h pi (7 ± 2%ID/g) ( Figure 4). However, the uptake of [ 89 Zr]Zr-DFOseribantumab-F(ab') 2 significantly decreased from 3 h to 48 h pi. There was no significant difference in tumor uptake of [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2    For [ 89 Zr]Zr-DFO-seribantumab, the highest uptake in normal organs at both time points was observed in liver (without significant difference in liver between the time points). The elevated uptake in the stomach, small intestine and GI tract significantly decreased from 48 h to 96 h pi.
At three hours pi, the highest uptake of both DFO-seribantumab-F(ab') 2 -based tracers was observed in kidneys, blood, liver, lungs, small intestine, and spleen. The uptake of [ 68 Ga]Ga-DFO-seribantumab-F(ab') 2 in all normal organs was significantly higher than the uptake of [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 . In most normal organs, the uptake of  [ 68 Ga]Ga-ZHER3 showed the fastest clearance from blood circulation. The activity concentration of the affibody in blood 3 h pi was more than 100-fold lower compared with  [ 68 Ga]Ga-Z HER3 had a significantly higher tumor-to-blood ratio than the other radioconjugates at all time points, with the exception of [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 24 and 48 h pi ( Figure 5). The tumor-to-blood ratios of [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 significantly increased with time. There was a noticeable, however non-significant, increase in the tumor-to-blood ratio for [ 89   . Numerical data as well as detailed information on statistically significant differences between groups can be found in Supplementary Material-Tables S1 and S2.

Ex Vivo Autoradiography of BxPC-3 Xenografts
Ex vivo autoradiography showed an even distribution of activity in the tumors for all radioconjugates at all time points (Figure 6). The activity accumulation that matched with the localization of tumor cells was higher in tumor tissue than in connective tissue ( Figure 6). * Indicates a statistically significant difference (p < 0.05) with [ 68 Ga]Ga-Z HER3 . Numerical data as well as detailed information on statistically significant differences between groups can be found in Supplementary Materials, Tables S1 and S2.

Ex Vivo Autoradiography of BxPC-3 Xenografts
Ex vivo autoradiography showed an even distribution of activity in the tumors for all radioconjugates at all time points (Figure 6). The activity accumulation that matched with the localization of tumor cells was higher in tumor tissue than in connective tissue ( Figure 6).     (Figures 7 and S3).

Discussion
Upregulation of HER3 is known to be a possible mediator for therapy resistance in cancer [3]. Dynamic changes in HER3 expression and inter-tumor heterogeneity warrant the development of tools that can repeatedly and non-invasively assess the status of HER3 expression in the primary tumor and metastases. PET imaging could provide a repeatable, non-invasive whole-body assessment of HER3 expression to select patients for HER3-targeted therapy. Because of the challenges associated with HER3 as a molecular imaging target (low overexpression and substantial expression in normal tissue), potential tracers need to be carefully designed. It is important that the imaging agent binds the target with high affinity and clears efficiently from the blood and other non-targeted tissues. Several different types of targeting molecules have been explored for radionuclide imaging of HER3 expression [8]. The limited success of antibody-based tracers for PET imaging of HER3 expression suggests that the modification of preexisting therapeutic antibodies might not be sufficient for imaging of a challenging targets such as HER3. Here, we characterized the new HER3-targeting radiotracers [ 89 Zr]Zr-DFO-seribantumab and [ 68 Ga]Ga/[ 89 Zr]Zr-DFO-seribantumab-F(ab')2 in vitro and compared their PET imaging properties at their respective favorable time points with the 68 Ga-labeled affibody molecule [ 68 Ga]Ga-ZHER3 in a preclinical mouse model.
A DFO chelator was coupled to seribantumab and seribantumab-F(ab')2 to allow labeling of these proteins either with 68 Ga or 89 Zr. The exact ratio of chelate to protein was not determined in this study. However, in a similar published protocol with the same

Discussion
Upregulation of HER3 is known to be a possible mediator for therapy resistance in cancer [3]. Dynamic changes in HER3 expression and inter-tumor heterogeneity warrant the development of tools that can repeatedly and non-invasively assess the status of HER3 expression in the primary tumor and metastases. PET imaging could provide a repeatable, non-invasive whole-body assessment of HER3 expression to select patients for HER3targeted therapy. Because of the challenges associated with HER3 as a molecular imaging target (low overexpression and substantial expression in normal tissue), potential tracers need to be carefully designed. It is important that the imaging agent binds the target with high affinity and clears efficiently from the blood and other non-targeted tissues. Several different types of targeting molecules have been explored for radionuclide imaging of HER3 expression [8]. The limited success of antibody-based tracers for PET imaging of HER3 expression suggests that the modification of preexisting therapeutic antibodies might not be sufficient for imaging of a challenging targets such as HER3. Here, we characterized the new HER3-targeting radiotracers [ 89 Zr]Zr-DFO-seribantumab and [ 68 Ga]Ga/[ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 in vitro and compared their PET imaging properties at their respective favorable time points with the 68 Ga-labeled affibody molecule [ 68 Ga]Ga-Z HER3 in a preclinical mouse model.
A DFO chelator was coupled to seribantumab and seribantumab-F(ab') 2 to allow labeling of these proteins either with 68 Ga or 89 Zr. The exact ratio of chelate to protein was not determined in this study. However, in a similar published protocol with the same methodology using a DFO:mAb molar ratio of 3:1, conjugation typically resulted in 0.3-0.9 coupled DFO moieties per antibody [31]. Such modification should not dramatically influence the biological properties of the proteins, which was corroborated by our results. Labeling of DFO-seribantumab resulted in quantitative yields. Radiochemical yields for labeling of DFO-seribantumab-F(ab') 2 with either 68 Ga or 89 Zr could potentially be improved by an increase in temperature, but due to the heat sensitivity of antibodies and antibody fragment labeling, temperature would be limited to physiological temperature.
Since purification with size-exclusion columns provided sufficient purity of the product no optimization of the labeling conditions was attempted. The new radioconjugates bound specifically to HER3 in vitro and in vivo and enabled visualization of HER3 expression using nanoPET. Binding affinities of seribantumab and seribantumab-F(ab') 2 towards HER3 measured by SPR were lower compared with the affinity values published for radiolabeled Z HER3 [29,30]. SPR-analysis furthermore showed cross-reactivity of seribantumab and seribantumab-F(ab') 2 to mErbB3 providing improved value to the preclinical model and enabled excellent comparability of the new tracers with [ 68 Ga]Ga-Z HER3 (with known affinity to mErbB3 [34]).
To our knowledge, no 89 Zr-labeled version of seribantumab has yet been reported for imaging of HER3 expression. The general pattern of biodistribution, tumor uptake, tumorto-blood ratio, and hepatic uptake of [ 89 Zr]Zr-DFO-seribantumab appear to be comparable to other reported HER3-targeting 89 Zr-labeled antibodies [17,35,36]. The uptake of [ 89 Zr]Zr-DFO-seribantumab was the highest in the liver, due to the known elimination of antibodies via the hepatobiliary pathway. As expected, imaging contrast improved from 48 h to 96 h pi, mainly due to the clearance from blood. This is in agreement with the general expectation for antibody-based tracers and studies of other 89 Zr-labeled HER3-targeting antibodies that reported the most favorable imaging time points to be between 4 and 7 days pi [11,15,35] [11]. In a study by Alsaid and colleagues, the uptake of [ 89 Zr]Zr-GSK2849330 in CHL-1 xenografts was approximately 3.5%ID/g and the tumor-to-blood-ratio was approximately 3.5 48 h pi with an injected dose of 0.5 mg/kg [35]. A further increase in the dose to 1-3 mg/kg also increased the tumor uptake to 8-10%ID/g and tumor-to-blood ratio 7-8 144 h pi.
In the present study, we also investigated a newly produced seribantumab-derived F(ab') 2 -fragment labeled with 68 Ga and 89 Zr. 89 Zr could be considered an uncontroversial choice for labeling of F(ab') 2 -fragments. However, antibody fragment-based tracers are known to have faster blood clearance than full-length antibodies and, as a result, could provide the opportunity for imaging at earlier time points. Thus, we also wanted to explore the possibility of a radiolabel with a shorter half-life such as 68 Ga.
A 68 Ga-labled variant of a trastuzumab-derived DOTA-conjugated F(ab') 2 has been reported for preclinical and clinical imaging of HER2-positive breast cancer 3 h pi [37,38]. However, in this study, 68 Ga proved to be an unsuitable choice for imaging of HER3 expression with DFO-seribantumab-F(ab') 2 . The biodistribution of the tracer appeared to be affected by insufficient stability of the [ 68 Ga]Ga-DFO complex. Trans-chelation of gallium to transferrin in the blood could be the reason for the observed elevated blood concentration. Besides trans-chelation, gallium-colloids are also known to accumulate in the liver and spleen, both of which sites are where we observed high activity uptake in the biodistribution study of [ 68 Ga]Ga-DFO-seribantumab-F(ab') 2 [39,40]. The insufficient stability of the [ 68 Ga]Ga-DFO complex was also corroborated by the dramatic release of 68 Ga within a few hours of the in vitro serum challenge. This is in agreement with results from trans-chelation experiments by Brandt [43][44][45][46]. Regardless, in this study, the clearance and increasing tumor-to-non-tumor contrast, particularly tumorto-blood contrast, of [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 with time indicated that 3 h pi was an unsuitable time point for imaging with [ 68 Ga]Ga-DFO-seribantumab-F(ab') 2 . As a result, no PET images were acquired for this tracer, and it is not included in further discussions.
As opposed to [ 89 Zr]Zr-DFO-seribantumab, the tumor uptake of [ 89 Zr]Zr-DFOseribantumab-F(ab') 2 decreased with time (two-fold from 3 h to 48 h pi). This could possibly be related to the faster clearance of the F(ab') 2 -fragement, which shifts the equilibrium in the blood to dissociation. Accordingly, it can be speculated that the stable tumor uptake of [ 89 Zr]Zr-seribantumab was due to the steady supply of the tracers from the blood pool. Aside from the uptake in the kidney and tumor, the overall uptake of [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 and [ 89 Zr]Zr-DFO-seribantumab at 48 h and 96 h pi were comparable in showing uptake in organs with expression of mErbB3 due to the cross-reactivity. Interestingly, this was also the case in the liver. Even though some uptake of [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 in liver was expected because of the expression of mErbB3, we had anticipated higher hepatic accumulation of [ 89 Zr]Zr-DFO-seribantumab (due to the hepatobiliary excretion of full-length antibodies), as it was shown in a comparison of [ 111 In]In-cetuximab and [ 111 In]In-cetuximab-F(ab') 2 in the past [47]. Because of the highest tumor-to-blood ratios, we considered 96 h and 48 h pi to be the best time for PET imaging with [ 89 Zr]Zr-DFO-seribantumab and [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 , respectively. We would, furthermore, consider [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 the more suitable imaging probe compared with [ 89 Zr]Zr-DFO-seribantumab due to the significantly higher tumor-to-blood contrast. This was also reported for [ 64 Cu]Cu-anti-HER3-F(ab') 2 , a mAb105-derived fragment and the only other HER3 targeting F(ab') 2 -fragment reported in the literature, even though no full biodistribution data were published [19].
The biodistribution of [ 68 Ga]Ga-Z HER3 was in good agreement with earlier published data by our group [27,30]. Despite the lower uptake in xenograft, [ 68 Ga]Ga-Z HER3 provided equal or higher tumor-to-non-tumor ratios and visibly better PET contrast than [ 89 Zr]Zr-DFO-seribantumab and [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 because of the lower uptake in normal organs. Both observations are most likely linked to overall faster clearance of Z HER3 . In particular, we considered [ 68 Ga]Ga-Z HER3 superior to [ 89 Zr]Zr-DFOseribantumab because of the significantly higher tumor-to-blood ratio. In comparison with [ 89 Zr]Zr-DFO-seribantumab-F(ab') 2 48 h pi, the tumor-to-blood ratio of [ 68 Ga]Ga-Z HER3 was similar, but [ 68 Ga]Ga-Z HER3 provided significantly better contrast in HER3-expressing organs, particularly in liver, which is a common site for metastases in many cancers.
A 89 Zr-labeled variant of Z HER3 has been reported in the literature [25]. Comparing the available data on [ 89 Zr]Zr-DFO-Z HER3 with our data, [ 89 Zr]Zr-DFO-Z HER3 had lower tumorto-blood contrast than the [ 68 Ga]Ga-Z HER3 variant 3 h pi [27,30]. It should be mentioned, however, that [ 89 Zr]Zr-DFO-Z HER3 was injected at a lower dose (1 µg) than [ 68 Ga]Ga-Z HER3 and studied in an MCF-7 xenograft model. Our group has previously reported that next-day imaging of HER3 using a radiocobalt-labeled Z HER3 variant can considerably improve the HER3-PET contrast in mice [30]. For [ 89 Zr]Zr-DFO-Z HER3 , the tumor-to-blood ratio improved from 3 h to 24 h pi, but the tumor-to-non-tumor contrast in other tissues remained the same or decreased [25]. This suggests that [ 89 Zr]Zr-DFO-Z HER3 might not be the optimal configuration for a Z HER3 -based imaging agent. Still, the PET-image contrast of [ 89 Zr]Zr-DFO-Z HER3 could be considered favorable to [ 89 Zr]Zr-DFO-seribantumab.
In addition to the clear advantage in imaging contrast, using the 68 Ga-labeled affibody may provide several other advantages for PET imaging of HER3 expression. Release of 89 Zr from the [ 89 Zr]Zr-DFO complex in vivo is a known phenomenon, occurring most likely to the unsaturated coordination sphere of Zr(IV) by DFO [41]. Free 89 Zr is known to accumulate in the bones and particularly joints [50], which was clearly visualized on the PET images in this study. This is an obvious disadvantage for imaging of bone metastasis, which are also known to express HER3 [2,51]. A longer time between tracer injection and acquisition of the scan, as is the case for antibody-based tracers, even further exacerbates the problem of 89 Zr accumulation in normal tissues due to the continuous release of the nuclide during that period. Stability of the [ 89 Zr]Zr-DFO complex could potentially be improved by increasing the temperature during radiolabeling [46]. This, however, is not a suitable option for the labeling of antibodies and antibody-fragments, because they are not able to withstand heating above physiological temperature. Another possibility to reduce the release of 89 Zr is the use of the recently developed DFO-derivative DFO* [52].
[ 89 Zr]Zr-DFO* has shown improved stability compared with [ 89 Zr]Zr-DFO for labeling of trastuzumab, resulting in significant reduction in bone uptake [53]. The faster clearance of [ 68 Ga]Ga-Z HER3 also reduces the radiation dose to the patients. For example, for the HER3-targeting antibody [ 89 Zr]Zr-GSK2849330, the effective radiation dose to patients was reported to be 0.46-0.59 mSv/MBq [15], whereas the effective dose for the HER2targeting affibody [ 68 Ga]Ga-ABY-025 was only 0.030 ± 0.003 mSv/MBq [54]. Furthermore, the faster clearance of [ 68 Ga]Ga-Z HER3 would enable repeat scanning within a shorter time. Additionally, an image time point shortly after injection could be better suited to detect rapidly occurring HER3 expression changes or image receptor occupancy during HER3-targeted therapy as in [24].

Conclusions
In conclusion, we have shown that the choice of targeting molecule can have a profound impact on HER3-PET image contrast. While the approach of using pre-existing therapeutic monoclonal antibodies for imaging of HER3 expression might be appealing, the complicated nature of HER3 as a molecular imaging target might necessitate the design of designated probes for molecular imaging in order to provide sufficient imaging contrast. Even though all studied tracers were able to visualize HER3 expression in our preclinical model, the results of this study suggest that smaller imaging agents, such as affibody molecules, are more suitable for PET imaging of HER3 expression than full-length antibody and antibody-fragment-based agents and warrant further translational studies.