Comparative evaluation of affibody- and antibody fragments-based CAIX imaging probes in mice bearing renal cell carcinoma xenografts

Carbonic anhydrase IX (CAIX) is a cancer-associated molecular target for several classes of therapeutics. CAIX is overexpressed in a large fraction of renal cell carcinomas (RCC). Radionuclide molecular imaging of CAIX-expression might offer a non-invasive methodology for stratification of patients with disseminated RCC for CAIX-targeting therapeutics. Radiolabeled monoclonal antibodies and their fragments are actively investigated for imaging of CAIX expression. Promising alternatives are small non-immunoglobulin scaffold proteins, such as affibody molecules. A CAIX-targeting affibody ZCAIX:2 was re-designed with the aim to decrease off-target interactions and increase imaging contrast. The new tracer, DOTA-HE3-ZCAIX:2, was labeled with 111In and characterized in vitro. Tumor-targeting properties of [111In]In-DOTA-HE3-ZCAIX:2 were compared head-to-head with properties of the parental variant, [99mTc]Tc(CO)3-HE3-ZCAIX:2, and the most promising antibody fragment-based tracer, [111In]In-DTPA-G250(Fab’)2, in the same batch of nude mice bearing CAIX-expressing RCC xenografts. Compared to the 99mTc-labeled parental variant, [111In]In-DOTA-HE3-ZCAIX:2 provides significantly higher tumor-to-lung, tumor-to-bone and tumor-to-liver ratios, which is essential for imaging of CAIX expression in the major metastatic sites of RCC. [111In]In-DOTA-HE3-ZCAIX:2 offers significantly higher tumor-to-organ ratios compared with [111In]In-G250(Fab’)2. In conclusion, [111In]In-DOTA-HE3-ZCAIX:2 can be considered as a highly promising tracer for imaging of CAIX expression in RCC metastases based on our results and literature data.

Animal studies. Accumulation of [ 111 In]In-DOTA-HE 3 -ZCAIX:2 in CAIX-expressing xenografts in mice was highly specific (Fig. 5). Pre-saturation of CAIX with a large amount of unlabeled HE 3 -ZCAIX:2 resulted in        3 -HE 3 -ZCAIX:2 was lower in kidneys and in muscle. At 4 h after injection, the uptake of [ 111 In]In-G250(Fab') 2 in majority of organs and tissues was significantly higher than the uptake of radiolabeled affibody molecules. By 24 h after injection of [ 111 In]In-G250(Fab') 2 , the activity was reduced significantly in blood, lungs, muscles and bones. The most pronounced was the reduction of the blood-borne activity.
The biodistribution features were translated into differences in tumor-to-organ ratios (

Discussion
Genetic interpatient and intratumour heterogeneity is a serious issue for targeted anticancer therapy 43 . In the case of disseminated disease, biopsy sampling of all metastases is often unrealistic, which prompts for alternative approaches. Radionuclide molecular imaging might offer a solution in this case. However, the sensitivity of molecular imaging diagnostics must be sufficiently high. The major issue is imaging of small metastases, with a size comparable or smaller than the spatial resolution of an imaging camera. Detection of such small metastases requires a high absolute tumour uptake and contrast with normal tissue, especially at the main metastatic sites 44 . For RCC, metastases are most frequently observed in lungs and bone, and somewhat less frequently in liver and brain 45,46 . Accordingly, a probe for visualization of CAIX expression in RCC metastases should have maximal tumor-to-lung, tumor-to-bone, and tumor-to-liver ratios. High tumor-to-blood ratio is always desirable to reduce the background that is due to the blood-borne activity. Thus, development of very good binders with high affinity and high tumor uptake is not enough; we have to minimize the uptake in critical tissues by reduction of off-target interactions. We took into account these considerations when planning further improvement of an affibody-based tracer for imaging of CAIX in disseminated RCC. Furthermore, a simple and robust labeling procedure is essential to ensure successful clinical translation.
To simplify the labeling, we intended to replace a multistep procedure involving the use of [ 99m Tc]Tc(CO) 3 + by more straightforward labeling with 111 In. Our experience with affibody molecules for imaging of HER2 expression suggests that increasing the overall hydrophilicity is generally an efficient approach to reduce off-target interactions 41,47 . The use of the aminocarboxylate chelator DOTA increases the overall hydrophilicity of the affibody molecule. Besides, DOTA is a very versatile chelator, permitting labeling with a variety of nuclides suitable for radionuclide imaging. Still, a modification of a small targeting protein is always associated with a risk of undesirable effects, such as decreased affinity or poor refolding after labeling in non-physiologic conditions.
In the case of DOTA-HE 3 -ZCAIX:2, the site-specific conjugation of maleimide-DOTA did not negatively affect refolding (Fig. 2). Labeling of DOTA-HE 3 -ZCAIX:2 with 111 In was more straightforward and two-fold quicker compared with [ 99m Tc]Tc(CO) 3 -HE 3 -ZCAIX:2, and provided a stable coupling of the radionuclide. After labeling, [ 111 In]In-DOTA-HE 3 -ZCAIX:2 retained specific binding to CAIX-expressing cells in vitro (Fig. 4A). Affinity of [ 111 In]In-DOTA-HE 3 -ZCAIX:2 to living SK-RC-52 cells was high, 1.2 ± 0.5 nM, which is higher than for [ 99m Tc]Tc(CO) 3 -HE 3 -ZCAIX:2 (6.13 ± 0.03 nM) measured using the same technique. It has to be noted that the affinity of [ 111 In]In-G250(Fab') 2 was even higher (0.12 ± 0.05 nM), which is most likely due to avidity effects from bivalent binding. However, affinity in the single digit nanomolar range is sufficient for successful imaging of targets with high expression, such as CAIX in RCC 36 . www.nature.com/scientificreports www.nature.com/scientificreports/ To compare imaging properties of the newly designed [ 111 In]In-DOTA-HE 3 -ZCAIX:2, the previous best affibody-based tracer [ 99m Tc]Tc(CO) 3 -HE 3 -ZCAIX:2 and the best antibody-based tracer [ 111 In]In-G250(Fab') 2 were selected. Their biodistribution was measured in the same batch of mice bearing SK-RC-52 xenografts. This was done to minimize batch-to-batch variability in mice physiology and xenograft quality. Accumulation of [ 111 In]In-DOTA-HE 3 -ZCAIX:2 in SK-RC-52 was highly specific (Fig. 5). During this head-to-head comparison, the tumor uptake of [ 111 In]In-DOTA-HE 3 -ZCAIX:2 at 4 h after injection was two-fold higher than the tumor uptake of [ 99m Tc]Tc(CO) 3 -HE 3 -ZCAIX:2 (Table 1). Most likely, the difference in tumor uptake could be explained by higher affinity of the 111 In-labeled variant. Internalization of both [ 111 In]In-DOTA-HE 3 -ZCAIX:2 (Fig. 4B) and [ 99m Tc]Tc(CO) 3 -HE 3 -ZCAIX:2 36 after binding to cancer cells is slow. Therefore, the high affinity permits better retention of the labeled probe on the surface of cells in the tumor while the tracer is cleared from non-specific compartments. In addition, the new tracer design provided more rapid clearance from blood and lower uptake in liver compared to [ 99m Tc]Tc(CO) 3 -HE 3 -ZCAIX:2 (Table 1). Taken together with the higher tumor uptake, the new tracer provided approximately two-fold higher tumor-to-blood, tumor-to-lung, tumor-to-bones, and tumor-to-liver ratios, which has potential for increasing the sensitivity of imaging of CAIX expression in RCC metastases.
The use of [ 111 In]In-DTPA-G250(Fab') 2 demonstrated earlier excellent potential for imaging of hypoxia-induced CAIX-expression in head-and-neck cancer 32,48 . This construct is appreciably larger (110 kDa) than the affibody molecules (7-8 kDa), and its clearance from blood and other tissues is slower ( Table 1). The accumulation in tumor is lower compared with [ 111 In]In-DOTA-HE 3 -ZCAIX:2 despite higher affinity. This might be because the rates of extravasation and diffusion in the tumor interstitium are lower for larger proteins. Both affibody-based constructs showed higher tumor-to-organ ratios than [ 111 In]In-DTPA-G250(Fab') 2 at 4 h after injection (Table 2). At 24 h after injection, [ 111 In]In-DTPA-G250(Fab') 2 provided a tumor-to-blood ratio of 67 ± 12. This is as good as the tumor-to-blood ratio provided by [ 111 In]In-DOTA-HE 3 -ZCAIX:2 and by an order of magnitude better than the radiometal-labeled full-length G250 provided at 7 days after injection 31,49 . Still, the ratios between [ 111 In]In-DTPA-G250(Fab') 2 uptake in tumor and in lung, bone or liver were lower than ratios for [ 111 In]In-DOTA-HE 3 -ZCAIX:2. Thus, [ 111 In]In-DOTA-HE 3 -ZCAIX:2 would be a better tracer for imaging of CAIX expression in RCC metastases.
During the last years, several small molecule sulfonamide derivatives were labeled with 18 F 50,51 and 68 Ga 52 for in vivo imaging of CAIX. During preclinical evaluation, these tracers demonstrated appreciably lower tumor uptake and tumor-to-organ ratios than the tracers evaluated in this study. To be fair, we have to mention that different tumor models were used in these studies, which complicates the comparison. A tracer composed from facetazolamide, a spacer, and a peptidic chelator was recently developed 53 . Upon labeling with 99m Tc, this tracer demonstrated an excellent tumor uptake (22% ID/g at 3 h after injection in SK-RC-52 xenografts) and very good tumor-to-blood ratio (approximately 70 and 100 at 3 and 6 h after injection, respectively). Unfortunately, tumor-to-liver and tumor-to-lung ratios peaked only at 4.7, and 2.  55 , have also demonstrated excellent targeting of SK-RC-52 xenografts (tumor uptake at 4 h of 20.8 ± 6.3 and 19.3 ± 4.5%ID/g, respectively) and high tumor-to-blood ratios at the day of injection. However, tumor-to-lung ratios were below 5 for both tracers.
It has to be noted that the use of DOTA permits labeling with the generator-produced positron-emitting radionuclide 68 Ga. The use of this radionuclide would permit the use of PET for imaging and benefit from advantages such as better spatial resolution and quantification accuracy compared to SPECT.
It has to be noted that both G250 antibody and ZCAIX:2 affibody molecule do not cross-react with murine CAIX (Car9). This means that the difference in biodistribution and targeting properties of evaluated imaging probes was caused by their size (influencing their rates of clearance from blood, extravasation, and diffusion in extracellular space of tumor), their affinity and/or avidity to the target expressed in human tumor xenografts and their off-target interactions with normal tissues. However, their interaction with CAIX expressed in normal tissues cannot be evaluated in this model. Normal human tissue expression of CAIX was reported to be restricted to the upper gastrointestinal mucosa, bile ducts and pancreas 5 . Nevertheless, four clinical imaging studies in more than 100 patients, using radiolabeled G250 monoclonal antibodies or G250(Fab') 2 , have not demonstrated noticeable accumulation of the tracer in these tissues 16,[26][27][28] . Furthermore, clinical radionuclide therapy with doses stabilizing previously progressing tumors or causing some tumor shrinkage, have not caused any gastrointestinal toxicity 16,17,26 . This lead us to a conclusion that the expression in normal tissue is much lower than in tumors and would not influence imaging contrast. Thus, the murine model is adequate for assessment of CAIX targeting probes in human xenografts, although it does not permit evaluation of the effect of non-cancerous CAIX expression.
Both [ 111 In]In-DTPA-G250(Fab') 2 and [ 111 In]In-DOTA-HE 3 -ZCAIX:2 have high uptake in kidneys. In the case of [ 111 In]In-DOTA-HE 3 -ZCAIX:2, the uptake is unspecific since it is not blocked be unlabeled tracer. This would not prevent affibody-mediated imaging of metastases 38 , but makes radiometal-labeled affibody molecules unsuitable for therapeutic targeting. Earlier studies with HER2-targeting affibody molecules have demonstrated that the renal uptake of affibody molecules is not mediated by megalin, and it could not be blocked by co-or preinjection of lysine or Gelofusine 56 . We have shown that the use of affibody-based bioorthogonal chemistry-or peptide nucleic acid-medicated pretargeting prevents high kidney accumulation in the case of HER2-targeting affibody molecules 57,58 and enables successful radionuclide therapy in mice 59 . Application of this approach might also enable pretargeting radionuclide therapy of CAIX-expressing tumors.
A recent clinical study demonstrated that a combination of [ 89 Zr]Zr-DFO-girentuximab-immunoPET/CT and CT detected more lesions than CT alone (91% vs 56%) and more than combination of CT and [ 18 F]FDG-PET/CT (84%) in the case of clear cell RCC 60 . Thus, in principle imaging with radiolabeled DOTA-HE 3 -ZCAIX:2 might be www.nature.com/scientificreports www.nature.com/scientificreports/ used for staging of clear cell RCC, where expression of CAIX is high. It has to be stressed that we propose to use [ 111 In]In-DOTA-HE 3 -ZCAIX:2 first and foremost for detection of CAIX expression in known RCC metastases for selection of patients for CAIX-targeting treatment, but not for detection of metastatic RCC. Since a substantial fraction of granular cell and mixed cell RCCs is CAIX-negative 18 , the use of [ 111 In]In-DOTA-HE 3 -ZCAIX:2 might result in false-negative findings.

Conclusion
Modification of labeling strategy permitted appreciable improvement of tumor uptake and tumor-to-organ ratios of a tracer based on the ZCAIX:2 affibody molecule. Based on our results and literature data, [ 111 In] In-DOTA-HE 3 -ZCAIX:2 can be considered a highly promising tracer for imaging of CAIX expression in RCC metastases.

Methods
Reagents, equipment and statistics. [  − was obtained by elution of UltraTechneKow generator (Mallinckrodt Pharmaceuticals, Dublin, Ireland) with sterile 0.9% NaCl. The CRS kits for production of technetium tricarbonyl were purchased from Center for Radiopharmaceutical sciences (Villigen, Switzerland). An automatic gamma-spectrometer with a NaI (Tl) detector (1480 WIZARDWallac Oy, Turku, Finland) was used for activity measurement activity in cell binding and biodistribution experiments. Formulation of injection solutions was performed using VDC-405 ionization chamber (Veenstra Instruments BV, The Netherlands).
For purification, a FPLC system (GE Healthcare AKTA purifier 10 system) and for polishing a 1200 series HPLC (Agilent Technologies, Santa Clara, CA) were used. ESI-MS with a 6520 Accurate-Mass Q-TOF LC/ MS (Agilent Technologies) was used for confirmation of molecular masses of the affibody molecules. Circular dichroism spectroscopy was performed using a Chirascan spectropolarimeter (Applied Photophysics, United Kingdom).
Instant thin layer chromatography (ITLC) was performed using silica gel-impregnated glass microfiber sheets (ITLC-SG strips, Varian, Lake Forest, CA). Distribution of the radioactivity on the strips was measured on Cyclone Phosphor Storage Screen using OptiQuant software for data processing (both Packard Instrument Company, Meriden, CT, US).
Statistical analysis of data was performed using GraphPad Prism (version 8.00 for Windows GraphPad Software, San Diego CA) to find any significant differences (p < 0.05). An unpaired two-tailed t-test was used for analysis if it was not stated otherwise. One-way ANOVA analysis with Bonferroni's multiple comparison test was used to evaluate differences between more than two data sets.
Proteins production and characterization. Anti-CAIX HE 3 -ZCAIX:2 affibody molecule was produced as described earlier 36 . A C-terminal cysteine-containing variant was produced as described earlier 61 . Briefly, the protein was produced in E. coli BL21*(DE3) (Thermo Fisher Scientific) in an overnight culture at 25 °C after induced expression with 100 μM Isopropyl β-D-1-thiogalactopyranoside (IPTG) at an OD600 of 0.8. Following cell lysis with French press, the supernatant was heated to 95 °C for 10 min with subsequent incubation on ice for 20 min, followed by centrifugation to remove precipitated proteins. Supernatant, containing affibody molecule, was purified by immobilized metal affinity chromatography (IMAC), using nickel-nitriloacetic acid (Ni-NTA) column on an ÄKTA FPLC system (GE Healthcare, Uppsala, Sweden). IMAC purification was done by running 20 mM tris-hydrochloride; 500 mM sodium chloride, pH 8 (buffer A) and 300 mM imidazole (buffer B) onto the loaded column with the filtered cell lysate. The column was washed with 30 mM imidazole followed by elution using a 30-300 mM imidazole gradient. The buffer of the eluate was changed to 20 mM ammonium acetate, pH 5.5, and the proteins were freeze-dried.
The conjugation was performed using the method described earlier 56 . The protein was dissolved in 20 mM ammonium acetate, pH 5.5, and reduced with an equimolar concentration of tris(2-carboxyethyl)phosphine (TCEP) for 30 min at 37 °C. The proteins were incubated at 37 °C for 90 min with ten-fold molar excess of maleimide derivative of DOTA for site-specific conjugation to the C-terminal cysteine. Metal ion contaminations were removed from all buffers with Chelex 100 resin (Bio-Rad Laboratories). The conjugate was purified by reverse-phase high performance liquid chromatography (RP-HPLC) using a Zorbax 300SB-C18 semi-preparative column (Agilent Technologies, Santa Clara, CA). Water with 0.1% trifluoroacetic acid was used as running buffer and an acetonitrile gradient was used for elution.
Molecular masses of both HE 3 -ZCAIX:2 and DOTA-HE 3 -ZCAIX:2 were determined using LC/MS. Circular dichroism spectroscopy was performed using a spectropolarimeter with an optical path length of 1 mm, to analyse the alpha-helical content, thermal stability and refolding capacity of DOTA-HE3-ZCAIX:2 at a concentration of 0.25 mg/mL. The thermal stability was evaluated by measuring the change in ellipticity at 221 nm during heating (5 °C/min) from 20 to 90 °C. The melting temperature (T m ) was approximated from the data acquired from variable temperature measurements (VTM) by curve fitting using a Boltzmann Sigmoidal model (GraphPad Prism, version 7). The refolding capacity was assessed by comparing spectra obtained from measurements at wavelengths in the range 195-260 nm at 20 °C, before and after thermal denaturation.
To test stability of the label, [ 111 In]In-DOTA-HE 3 -ZCAIX:2 was incubated with 5000-fold excess of Na 4 EDTA for 2 hours at room temperature. The mixture was analyzed using ITLC as described above.
Labeling of HE 3 -ZCAIX:2 for comparative biodistribution experiment with [ 99m Tc(CO) 3 ] + was performed as described earlier by Garousi and co-workers 36 . Briefly, a generator eluate (400-500 μL) containing ca. 3 GBq of 99m Tc was added to a CRS kit and the mixture was incubated at 100 °C for 30 min. After incubation, 40 μL of mixture was transferred to a vial containing 50 μg of Affibody molecule in 40 μL of PBS. The mixture was incubated for 100 min at 50 °C. Thereafter, a 5000-fold molar excess of histidine was added to the reaction mixture, and it was incubated at 50 °C for 20 min. The radiolabeled HE 3 -ZCAIX:2 were purified using NAP-5 columns pre-equilibrated and eluted with PBS. The radiochemical yield and purity of each preparation was measured using radio-ITLC eluted with PBS. Affinity of binding was determined using LigandTracer Yellow (Ridgeview Instruments, Vänge, Sweden) as described previously 36,62 and analyzed using InteractionMap software 63 . This device records in real time kinetic binding to and dissociation of radiolabeled tracers from living cells. The TraceDrawer Software (Ridgeview Instruments, Vänge, Sweden) was used to calculate the affinity based on the association and dissociation rates determined by adding increasing concentrations of each radioconjugate to the cell cultures followed by monitoring the retention at zero concentration. For [ 111 In]In-DOTA-HE 3 -ZCAIX:2 and [ 99m Tc]Tc(CO) 3 -HE 3 -ZCAIX:2, the association was measured at concentration of 2 and 6 nM. For [ 111 In]In-G250(Fab') 2 , the association was measured at 1.5 and 4.5 nM.

In vitro characterization of [ 111 In]In
To evaluate specificity and cellular processing, SK-RC-52 cells (~1 × 10 6 cells/dish) were used. For the determining of specificity, a set of six cell dishes was used. [ 111 In]In-DOTA-HE 3 -ZCAIX:2 was added at a concentration of 10 nM. Fifteen min before adding of [ 111 In]In-DOTA-HE 3 -ZCAIX:2, CAIX on cells in a set of three dished was pre-saturated by non-labeled HE 3 -ZCAIX:2 to a total concentration of 1 mM. All cells were incubated for 1 h at 37 °C in a humidified incubator equilibrated with 5% CO 2 . Then, the medium was collected, the cells were washed with cold serum-free medium and detached by treatment with trypsin-EDTA solution for 10 min at 37 °C. The detached cells were collected, and the radioactivity of cells and media was measured. Binding specificity of [ 99m Tc]Tc(CO) 3 -HE 3 -ZCAIX:2 and [ 111 In]In-G250(Fab') 2 has been confirmed earlier in a similar way 32,36 .
For evaluation of cellular processing, cells were incubated in a humidified incubator (5% CO 2 , 37 °C) with [ 111 In]In-DOTA-HE 3 -ZCAIX:2 (at concentration of 10 nM). A set of three dishes was taken from the incubator at 1, 2, 4, 8 and 24 h after the incubation initiation, and membrane-bound and internalized radionuclides were discriminated using a modified acid wash method, as describe earlier 64 . Animal studies. Animal studies were performed according to national legislation on laboratory animal protection and were approved by the Ethical Committee for Animal Research in Uppsala.
For tumor implantation, 10 7 of CAIX-expressing SK-RC-52 cells in 100 µL of RPMI 1640 medium were subcutaneously injected in the right hind leg of 7-weeks old female BALB/c nu/nu mice (Scanbur A/S, Karlslunde, Denmark). The experiments were performed two weeks after tumor cell implantation. The average animal weight was 16.6 ± 1.3, and the average tumor weight at dissection was 0.22 ± 0.11 g.
In the biodistribution study, four mice per data point were used. Two groups of mice were intravenously injected with 30 kBq [ 111 In]In-DOTA-HE 3 -ZCAIX:2 in 100 µL PBS. The injected protein dose was adjusted to 5 µg per mouse with nonlabeled conjugate. To test in vivo targeting specificity, animals in one group were pre-injected with 500 µg of nonlabeled ZCAIX:2 30 min before injection of [ 111 In]In-DOTA-HE 3 -ZCAIX:2 to saturate tumors. One group was injected with 30 kBq (5 µg) of [ 99m Tc]Tc(CO) 3 -HE 3 -ZCAIX:2. Two groups were injected with 30 kBq (10 µg) of [ 111 In]In-G250(Fab') 2 . Biodistribution was measured at 4 h after injection of radiolabeled affibody molecules and 4 and 24 h after injection of [ 111 In]In-G250(Fab') 2 . The mice were anesthetized by an intraperitoneal injection of a lethal dose ketamine and xylazine solution and exsanguinated by heart puncture. Blood was collected with a heparinized syringe, organs were collected, weighed and activity was measured using a gamma spectrometer. The percent of injected dose per gram of sample (%ID/g) was calculated, except for gastrointestinal tract and carcass, where %ID per whole sample was calculated.
Small animal SPECT/CT (nanoScan SC equipped with 16-pinholes collimators, Mediso Medical Imaging Systems, Hungary) imaging was performed to obtain a visual confirmation of the biodistribution data. Two mice per data point were used to check a reproducibility of the data. One set of two mice was injected with 5 MBq