Radionuclide therapy using ABD-fused ADAPT scaffold protein: Proof of Principle

Molecular recognition in targeted therapeutics is typically based on immunoglobulins. Development of engineered scaffold proteins (ESPs) has provided additional opportunities for the development of targeted therapies


Introduction
Targeted therapies are based on recognition of molecular structures that are predominantly expressed by cancer cells.One of several possible mechanisms of action for targeted therapeutics is a selective delivery of a cytotoxic payload (drugs, toxins or radionuclides) to malignant cells [1].An antibody-mediated delivery of cytotoxic radionuclides is called radioimmunotherapy [2].Radioimmunotherapy has demonstrated impressive results in treatment of hematologic malignancies, but results in solid tumors are still modest and manifested as tumor growth stabilization at the best [2].One of the issues in targeted payload delivery is the big size of antibodies, causing slow extravasation and diffusion rates [3].
To overcome this limitation, several innovative immunoglobulin-based constructs, such as minibodies or SIPs (small immunoproteins) have been engineered [4].However, the use of non-immunoglobulin-based engineered scaffold proteins (ESPs) might provide increased freedom in molecular design of small targeting agents.Several classes of ESPs have been evaluated for delivery of drugs and toxins [3,5].
ESPs have attracted an increasing interest as specific targeting vectors in radionuclide molecular imaging [6].ESPs have a number of attractive features, such as efficient and inexpensive production in prokaryotic hosts or even by peptide synthesis.Moreover, they show high stability in a broad range of temperature and pH, facilitating the use of a large variety of labeling methods.In addition, their high stability and mutational tolerance promote the ease of engineering multimeric, multispecific and multifunctional constructs.ADAPT is an ESP derived from the small (46 amino acids) albumin-binding domain (ABD) of protein G [7]. Feasibility of utilizing ADAPTs as imaging probes has been demonstrated using ADAPT6, an ADAPT that shows high-affinity binding to human epidermal growth factor receptor type 2 (HER2) [8].A clinical study (NCT03991260) demonstrated that 99m Tc-labeled ADAPT6 could image HER2 expression in breast cancer with high contrast within a few hours after injection [9].No adverse side effects were observed after injection of 1 mg of ADAPT6.However, the use of ADAPT6 for radionuclide therapy has been precluded by its high renal reabsorption after clearance via glomerular filtration.This phenomenon leads to activity accumulation in kidneys that by far exceeds accumulation in tumors, which excludes anti-tumor effect of a radionuclide therapy without an unacceptable high dose to kidney.The re-absorption cannot be reduced by co-injection of L-lysine or Gelofusine [10].Therefore, other strategies are needed to reduce the renal uptake of ADAPT6 and allow for therapeutic use.
A non-covalent binding to serum albumin is one possible strategy to extend circulatory half-life, increase the bioavailability and reduce the renal uptake of small therapeutics undergoing renal clearance [11].For the development of radiopharmaceuticals, coupling to a small albumin-binding entity has been used earlier to prolong circulation in the blood increasing bioavailability and reduce renal uptake of radiolabeled folates [12], PSMA ligands [13,14] and somatostatin analogs [15].To modulate the pharmacokinetics of another class of ESP, affibody molecule, we have previously evaluated this scaffold protein fused to the small albumin binding domain of Protein G, ABD [16].Such strategy enabled a more than 17-fold reduction of renal uptake for 177 Lu-labeled affibody molecules and by that made an affibody-mediated radionuclide therapy possible.In addition to the increase of size that dramatically decreases the renal filtration, there is also an advantage in that albumin-bound targeting proteins might be rescued from degradation by a pH-dependent interaction with the neonatal Fc receptor (FcRn) [17].Fusion with ABD has been successfully applied for extension of half-lives of therapeutic non-labeled affibody molecules binding to human epidermal growth factor receptor type 3 (HER3), interleukin-17 A (IL-17 A) and amyloid β [18,19].Interim results of a clinical trial evaluating anti-IL-17 A-affibody molecule fused to ABD for the treatment of psoriasis and psoriatic arthritis (AFFIRM-35, NCT03591887) demonstrated no toxicity or immunogenicity although dozens of patients were injected with doses up to 160 mg.Further, a fusion of ABD to an ADAPT6-PE25 targeted toxin permitted significantly increased residence time in blood and decreased renal uptake [20].
The goal of this study was to test the hypothesis that fusion with ABD would suppress renal uptake of ADAPT6 and hence permit radionuclide therapy of HER2-expressing tumors without damaging the kidneys.To test this hypothesis, a variant of ABD known as ABD 035 , engineered to bind human serum albumin with a subpicomolar affinity [21] was selected as the fusion partner.
It has to be noted that the molecular design of new types of targeting probes might have a surprisingly strong effect on their targeting properties.A fusion with another protein might make some surface area of a protein less prone to both on-target and off-target interactions with an individual cancer cell due to steric hindrances.On the other hand, some off-target interactions might be enhanced by additional off-target interactions of the fusion partner.Such cooperativity or anti-cooperativity should depend on the relative positions of the units in the construct.Thus, the order of the units might be of importance.We have observed earlier that a fusion between ABD and a HER3-binding affibody molecule with C-terminal placement of ABD (3A) has two-fold higher affinity to HER3 and 2.5-fold higher affinity to HSA compared with a variant with N-terminal placement (A3) [22].Furthermore, a fusion with an additional bulky moiety might either enhance or attenuate off-target interactions in vivo, influencing uptake and, therefore, also the absorbed doses in normal tissues.Hence, we decided to test two variants, placing ABD either on C-or N-terminus of ADAPT6.The selection of the linker length was based on the data from our previous study [23] on ADAPT6 homodimers.That study has demonstrated that the use of a single SSSG sequence as a linker prevents correct refolding of a dimer and reduces affinity compared to the monomeric ADAPT6, most likely due to steric hindrance.The use of both (SSSG) 2 and (SSSG) 3 linkers resulted in approximately a ten-fold increase of the dimers apparent affinity compared to the affinity of a monomer due to an avidity effect.There was no difference between the uptake of variants containing (SSSG) 2 and (SSSG) 3 linkers in normal tissues.The use of a longer linker in the current construct aimed at minimizing a potential steric hindrance when binding to both albumin and HER2 simultaneously.
Another essential part of the molecular design of a radiolabeled therapeutic conjugate is the labeling strategy.It might influence targeting in three ways.First, nearly all targeting proteins undergo internalization and trafficking to lysosomes after specific binding to cancer cells or unspecific binding to normal tissues.Depending on the labeling approach, proteolysis in lysosomes might lead to formation of charged/ polar or lipophilic radioactive metabolites.Charged/polar radiometabolites cannot penetrate lysosomal and cellular membranes and therefore become trapped intracellularly.The labels resulting in such metabolites are called residualizing labels [24].Lipophilic radioactive metabolites can instead diffuse through cellular membranes and leave tumors or normal tissues; such labels are called non-residualizing.The use of non-residualizing labels (typically radiohalogen-based) decreases residence of activity in normal tissues, but it also decreases residence in tumors.The use of non-residualizing labels might be considered when internalization of a targeted probe by cancer cells is slow.Further, it might be beneficial when internalization of the probe in excretory organs, e.g., kidneys is rapid [25].However, even slow internalization results in a shorter residence of activity in tumors.Second, a coupling of radionuclide (and appropriate chelator or linker) alters distribution of charge and lipophilicity on a surface of a polypeptide, which can modify its interaction with a molecular target and cause affinity changes.For example, it has been shown that small changes in the chelator charge and geometry caused by substitution of yttrium with cobalt in a somatostatin analog DOTA-TOC has increased 25-fold its affinity to somatostatin receptor 2 [26].A strong influence of the labeling approach on affinity was found for ESPs, such as e.g.99m Tc-labeled anti-EGFR affibody molecule [27] or 125 I-labeled HER2-binding DARPin G3 [28].For ADAPT6, labeling with 125 I using ((4-hydroxyphenyl)ethyl) maleimide (HPEM) at the C-terminus increased the affinity to HER2 nearly two-fold compared to labeling with 111 In using a DOTA chelator in the same position [25].Third, modification of the polypeptide surface influences off-target interaction of targeting probes, which can change their uptake in normal organs in vivo [29,30].Such impact is difficult to predict due to the complexity of the environment in vivo, which necessitates animal studies.
Two variants of conjugates were designed (Fig. 1), one with ABD 035 fused at the N-terminus of ADAPT6 (ABD 035 -ADAPT6) and another with ABD 035 fused at the C-terminus of ADAPT6 (ADAPT6-ABD 035 ).A linker containing three SSSG repeats was placed between ABD 035 and ADAPT6 moieties to enable free folding and unhindered binding of the two different domains.A leader sequence (MG-DEAVDANS), which improved protein expression and biodistribution of monomeric ADAPT6 [31] was placed at the N-terminus of all variants.Incorporation of a C-terminal cysteine enabled site-specific indirect radioiodination using ((4-hydroxyphenyl)ethyl) maleimide (HPEM) or conjugation of a maleimide-derivative of DOTA chelator for labeling with radiometals.Low-energy beta emitter 177 Lu and 125 I (as a surrogate for beta emitter 131 I) were used for radiolabeling.

Statistics
Statistical treatment was performed using GraphPad Prism software (version 4.00 for Microsoft Windows; GraphPad Software) to determine significant differences (P < 0.05).Data on cellular uptake and processing as well as biodistribution were analyzed by unpaired 2-tailed t-test when two groups were compared.Comparison of data for more than two groups was performed using ANOVA with a Bonferroni test for multiple comparisons.Biodistribution data concerning dual-label studies were analyzed using paired t-test.

Production, purification, conjugation and characterization of targeting proteins
Genes encoding ADAPT6-(SSSG) 3 -ABD 035 and ABD 035 -(SSSG) 3 -ADAPT6 were synthesized by Thermo Fisher Scientific (Waltham, MA, USA) and subcloned into expression vectors through amplification with primers containing the desired restriction sites as well as the N-terminal sequence MGDEAVDANS and the C-terminal sequence GSSC.Proteins were expressed in E. coli BL21*(DE3) cells and extracted by sonication (40% amplitude with 1.0/1.0s pulsing for 1 min and 30 s using a Vibracell (Sonics).The proteins were purified by loading the lysates on an inhouse produced affinity chromatography column coupled with Human Serum Albumin (HSA).Bound protein was washed with 1xTST (25 mM Tris-HCl, 1 mM EDTA, 0.2 M NaCl, 0.05% Tween) and 5 mM NH 4 Ac pH 5.5 prior to elution with 0.5 M HAc pH 2.8.
Proteins intended for 177 Lu labeling were conjugated with the chelator DOTA (Macrocyclics, TX, USA) through maleimide coupling at the free thiol group of the C-terminal cysteine.The thiol groups were reduced by incubation with 20 mM DTT for 30 min at 40 • C. To allow for conjugation, DTT was removed in a buffer exchange step using disposable PD-10 desalting columns (GE Healthcare, IL, USA) with 20 mM NH 4 Ac, pH 5.5 (treated with Chelex 100 resin (Bio-Rad Laboratories, CA, USA) to avoid metal-ion contamination) as elution buffer.Maleimido-DOTA was added to the sample in 3-fold excess and the samples were incubated for 60 min at 40 • C. Both the DOTA-conjugated as well as the unconjugated protein intended for 125 I labeling, were further purified through semipreparative RP-HPLC using an Agilent 1200 series system equipped with a Zorbax semi-preparative column (300SB-C18, 9.4 × 250 mm, 5 μm, Agilent).Proteins were eluted using a gradient of 40-55% B (A, 0.1% trifluoroacetic acid in water; B, 0.1% trifluoroacetic acid in acetonitrile) over 30 min using a flow rate of 3 mL/min.
Non-ABD-fused ADAPT6 (MGDEAVDANS-ADAPT6-GSSC) was used as a control in the biodistribution measurements and was produced and purified like previously described [23].
A non-target binding fusion protein was designed, denoted ADAPT-Neg -ABD 035 .This protein resembles the ABD-fused target-binding construct but contains the non-binding ADAPT Neg .ADAPT Neg was constructed by introduction of 3 point mutations in the gene of parental ADAPT6 to remove HER2 binding.The gene of this non-target-binding protein (MGDEAVDANS-ADAPT Neg -(SSSG) 3 -ABD 035 -GSSC) was synthesized by Thermo Fisher Scientific (Waltham, MA, USA) and the protein was produced as described above.This non-targeting protein was used in a control group in radionuclide therapy to evaluate the effect on tumor growth caused by a radionuclide attached to a protein that is retained in circulation but not specifically accumulated in tumors.
The purity of all variants used in the study was evaluated by RP-HPLC using an Agilent 1200 series system equipped with a Supelco analytical column (300SB-C18, 4.6 × 150 mm, 3.5 μm, Sigma Aldrich).
Elution was performed by a gradient of 40-55% B (A, 0.1% trifluoroacetic acid in water; B, 0.1% trifluoroacetic acid in acetonitrile) over min using a flow rate of 1 mL/min.Purity is described as the percentage of the area under the curve of the main peak in relation to the total area under all peaks.The molecular weights of all proteins were confirmed by liquid chromatography-electro-spray ionization-mass spectrometry (LC-ESI-MS) using an Impact II UHR QqTOF MS (Bruker Daltonics, MA, USA) and the secondary structure and thermal stability was evaluated on a Chirascan circular dichroism spectrometer (Applied Photophysics, Surrey, UK) like previously described [23].
Target binding analysis was performed on a Biacore T200 system (GE Healthcare, Stockholm, Sweden) by injecting analytes over a CM5 chip immobilized with Murine Serum Albumin (MSA), Human Serum Albumin (HSA) and Human Epidermal growth factor Receptor 2 (HER2) to a response level of ~1000 RU each.Experimental parameters have been described previously [23].In short, a dilution series of five different analyte concentrations were injected onto the chip with a flow rate of μL/min for 300 s followed by 1600 s of dissociation.All analyses were carried out in 1xPBST (PBS supplemented with 0.05% Tween, pH 7.4) at 25 • C. To confirm correct and functional refolding, an additional target binding analysis was carried out following heating of the samples to 95 • C for 1 h and then cooling them down before injection.

Radiolabeling
Yield of 125 I-HPEM was measured by radio-TLC using Silica gel F254 TLC plates (20 × 100 mm, elution path 80 mm; E. Merck, Darmstadt, Germany) eluted with acetonitrile.Radiochemical yield and purities were measured using radio-ITLC (Varian Medical Systems, Palo Alto, CA, USA).For analysis of radioiodinated proteins, a mixture of acetone:water (7:3) was used for development.In analysis of 177 Lulabeled proteins, 0.2 M citric acid, pH 2.0, was used for development.To cross-validate radio-ITLC data, radio-HPLC analysis was performed using Hitachi Chromaster HPLC systems with radioactivity detector.For analysis of 125 I-labeled ADAPTs, Vydac RP C18 column (300 Å; 3 × 150 mm; 5-μm) at room temperature was used.Solvent A was 0.1% trifluoroacetic acid (TFA) in H 2 O; solvent B was 0.1% TFA in acetonitrile.The flow rate was 1 mL/min, with a 5% B to 80% B gradient over 20 min.For analysis of 177 Lu-labeled ADAPTs, Luna C18 column (300 Å; 4.6 × 150 mm; 5-μm) at room temperature was used.Solvent A was 0.1% trifluoroacetic acid (TFA) in H 2 O; solvent B was 0.1% TFA in acetonitrile.The flow rate was 1 mL/min, with a 5% B to 70% B gradient over 15 min.
Labeling with 125 I was performed according to the method described by Ref. [32].Briefly, 125 I (14 MBq) was mixed with a solution of HPEM (10 μg, 31.5 nmol) in 5% acetic acid in methanol.Chloramine-T (10 μL, 8 mg/mL in MQ water) and 10 μL of 5% acetic acid in methanol were added, and the mixture was incubated for 5 min at room temperature.
The reaction was quenched by adding 10 μL sodium metabisulphite (12 mg/mL in water).ADAPT6 derivatives (500 μg, 38.9 nmol) were reduced by incubation in PBS with 20 mM dithiothreitol (DTT) for 60 min at 40 • C. Reduced proteins were purified using size-exclusion NAP-5 column (GE Healthcare, Uppsala, Sweden) pre-equilibrated with 0.2 M NH 4 OAc buffer, pH 6. Radioiodinated HPEM was added, and the mixture was incubated for 60 min at 40 • C. Radiolabeled proteins were purified using NAP-5 columns.
Labeling with 177 Lu was performed as previously described for DOTA-PNA labeling [33].Briefly, a DOTA-conjugated protein (80 μg, Thereafter, a 5000-fold molar excess of EDTA was added to scavenge a loosely bound nuclide, and the mixture was incubated for 15 min at 95 • C. Radiolabeled proteins were purified using NAP-5 columns.Stability of radioiodine labels was tested by 3-h incubation with 1000-fold molar excess of potassium iodide.Stability of labeling with 177 Lu was evaluated by 3-h challenge with 5000-fold molar excess of EDTA.

In Vitro evaluation
Cell lines were purchased from American Type Culture Collection (ATCC, Manassas, VA).The specificity of the binding of the radiolabeled ADAPT6 derivatives to HER2-expressing cancer cells was evaluated using ovarian carcinoma SKOV-3 and breast carcinoma BT474 cell lines using a saturation method described earlier [34].Briefly, 1,000,000 cells per cell culture dish were seeded 24 h before experiment.In a binding specificity test, radiolabeled conjugates were added to reach a concentration of 50 nM.To a set of control dishes, 5 mM of unlabeled protein was added 15 min before adding radiolabeled variant to saturate receptors.The cell-associated activity was measured after 1 h incubation at 37 • C. The cellular processing during continuous incubation was evaluated for conjugates bearing residualizing 177 Lu label.Cells were incubated at 37 • C with 50 nM of radiolabeled protein.At 1, 2, 4, 8 and 24 h after incubation start, the membrane-bound and internalized activity was determined by the acid wash method using 4 M urea in 0.2 M glycine buffer, pH 2.0, to determine the membrane-bound activity [34].The affinity of radiolabeled conjugates binding to living HER2-expressing SKOV-3 cells was determined using LigandTracer (Ridgeview Instruments AB) by an established method as described previously [35].

Animal studies
Animal studies were planned in agreement with EU Directive 2010/ 63/EU for animal experiments and Swedish national legislation concerning protection of laboratory animals and were approved by the Ethics Committee for Animal Research in Uppsala.
Biodistribution and targeting properties were evaluated in female BALB/C nu/nu mice bearing HER2-positive SKOV-3 xenografts.As a specificity control, HER2-negative Ramos xenografts were used.To establish xenografts, 10 7 cells were implanted subcutaneously.Data concerning tumors are presented using weight units, when the tumors were weighed (e.g. in biodistribution experiment) or using the volume units, when tumor volume was measured in the experiments (e.g. during experimental therapy).

Biodistribution measurements
To reduce the number of tumor-bearing mice, a dual-label approach was used for biodistribution experiments.The mice were injected into the tail vein with a mixture of 125 I-and 177 Lu-labeled ADAPT6 variants (15 kBq of 125 I and 90 kBq of 177 Lu per animal in 100 μL of PBS).The total injected protein dose was adjusted to 3 μg/mouse for non-fused ADAPT6 and 6 μg/mouse for ABD-fused variants using the corresponding unlabeled protein.After exsanguination under anesthesia, the organs and tissues of interest were excised and their activity was measured using an automated gamma-spectrometer (1480 WIZARD; Wallac Oy).Activity was measured in the energy window from 10 to keV for 125 I, and from 90 to 370 keV for 177 Lu.
To evaluate the dosimetry of the most promising variants, 125 I-HPEM-ADAPT6-ABD 035 and 177 Lu-DOTA-ADAPT6-ABD 035 , their biodistribution was measured 4, 24, 48, 72, 168 and 336 h after injection in mice bearing SKOV-3 xenografts (average mouse weight was 16.3 ± 1.3 g, and average tumor weight 0.11 ± 0.06 g).To control that the tumor uptake was HER2-specific, the biodistribution of these conjugates in mice bearing HER2-negative Ramos xenografts (average tumor weight 0.11 ± 0.06 g) was measured 48 h after injection.Four mice per data point were used in this experiment.Dosimetry was evaluated as described earlier [36].The absorbed dose to all organs, including the bone marrow and the tumor was calculated by using the absorbed dose rate factors or S-values for a 22 g mouse Moby phantom, kindly provided by Dr. Erik Larsson [37].Blood perfusion in the bone marrow was assumed to follow the theoretical marrow cellularity distribution [38].Data for 125 I-labeled variant were used to calculate absorbed dose in the case of the use of 131 I-label for therapy.

Radionuclide therapy
To evaluate the therapeutic effect of the ADAPT6 fused to ABD, 10 SKOV-3 cells per mouse were implanted subcutaneously on the abdomen.
Treatment started 7 days after tumor implantation, when the average tumor volume was 0.07 ± 0.02 cm 3 and the average mouse weight was 15.1 ± 0.4 g.The mice were randomly divided into 5 groups of 11 animals each.The first group of treated animals was injected with 6 μg (4.7 nmol)/18 MBq of 177 Lu-DOTA-ADAPT6-ABD 035 .Animals in the second treatment group received two treatments, the first at day 7 and the second three weeks after the first one (day 28).Control group 1 received vehicle, PBS, alone.To evaluate the influence of non-labeled ABD-fused ADAPT6 on tumor growth, control group 2 was injected with 6 μg (4.7 nmol) of unlabeled DOTA-ADAPT6-ABD 035 .To evaluate the effect of a non-targeting radiolabeled protein, a control group 3 was injected with 177 Lu-DOTA-ADAPT Neg -ABD 035 (6 μg/18 MBq) only.
At least twice a week, the mice were weighed and visually inspected, and tumors were measured using electronic calipers.Tumor volumes (mm 3 ) were calculated using an ellipsoid formula, as [length (mm)] × [width (mm)] 2 × 0.5.The animals were euthanized when tumors reached a size of 1000 mm 3 or became ulcerated, or if an animal's weight dropped by more than 10% during 1 week or by more than 15% since the study began.After euthanasia, tumors and kidneys were excised for subsequent histologic evaluation.

Histological evaluation
Samples of the xenografts, livers and kidneys from five mice in every group were taken at the time of sacrifice.The samples were formalin fixed, paraffin embedded and stained by hematoxylin and eosin using standard procedures.A histological examination was performed at the Department of Pathology and Wildlife Diseases, National Veterinary Institute, Uppsala, Sweden.

Production, purification, conjugation and characterization of targeting proteins
All proteins used in this study were produced in E. coli and purified to homogeneity.The purity of the variants intended for in vivo studies was determined by RP-HPLC to be above 95% (Suppl.Fig. S1) and mass spectrometry confirmed the correct mass of all proteins (Table 1).
SPR measurements of the ABD-fused constructs revealed similar affinities towards HER2 (around 5 nM), regardless of the orientation of the ABD (Suppl.Fig, S2, Table 2).Furthermore, as can be seen in Table 2, affinity measurements after heat treatment of the sample confirms that the binding is not compromised.This is comparable to the previously determined 3.5 nM affinity of the unfused control [23].The affinities towards the serum albumins were also similar between all constructs, with measured affinities around 1.8 nM towards MSA and around 50 pM for HSA (Suppl.Fig, S3-4, Table 2).Suppl.Fig 2 also demonstrates simultaneous binding of HER2 and HSA, since the fusion proteins can bind HER2 also in the presence of saturating concentrations of HSA.
SPR analysis of the non-target binding fusion protein, which was designed for use as a control in the therapy studies, confirms that it does not bind HER2 (Suppl.Fig. S5).

Radiolabeling
Radiochemical yield of labeling of HPEM with 125 I was 98%.Data concerning radiochemical yields and radiochemical purity are provided in Table 3.All labeling protocols were efficient and purification using NAP-5 provided nearly complete absence of non-conjugated radionuclide.For experimental therapy, the maximum molar activity up to 55.6 GBq/μmol was obtained for 177 Lu-DOTA-ADAPT6-ABD 035 .Identity of radiolabeled ADAPT6 derivative was confirmed by radio-HPLC (Suppl.Fig. S6 and S7).All labels were stable under challenge conditions (no measurable radionuclide release).

In Vitro evaluation
Pre-saturation of HER2 receptors in cells with unlabeled constructs resulted in a significant (p < 5 × 10 − 5 ) reduction of binding of all radiolabeled ADAPT6 variants to both SKOV-3 and BT-474 cell lines.Data concerning binding and processing of 177 Lu-labeled ADAPT6 derivatives by HER2-expressing cells are presented in Fig. 3.
The binding pattern of radiometal-labeled ABD 035 -fused variant was similar to the pattern for the non-fused variant: rapid binding during the first 2 h followed by a somewhat slower increase.The data for the nonfused variant closely resembled the data for the 111 In-labeled counterpart [40].The internalization rate for all variants was slow, and less than 20% of the cell-bound activity was internalized after 24 h.Data concerning affinity of radiolabeled ADAPT6 derivatives to living HER2 expressing SKOV-3 cells are presented in Table 4 and Supplemental Fig. S8.All variants demonstrated two interactions with HER2 on living cells, a strong one with subnanomolar apparent dissociation constant and a weaker one with a dissociation constant in the range of 20-50 nM.There was no obvious negative effect of ABD 035 fusion on the strength of ADAPT6 binding to HER2-expressing cells.

Biodistribution measurements
The effect of fusion with ABD 035 on the biodistribution of ADAPT6 is demonstrated in Fig. 4 using conjugates with a residualizing 177 Lu label.The biodistribution of the non-fused 177 Lu-DOTA-ADAPT6 was typical for radiometal-labeled derivatives of this protein [28], i.e. a reasonably high tumor uptake, rapid clearance from blood and other tissues, and a high reabsorption and retention of activity in kidneys.However, fusion with ABD changes this biodistribution pattern dramatically.More specifically, the retention in blood increased more than 200-fold for both variants.The renal accumulation of activity was reduced 10-fold for 177 Lu-DOTA-ABD 035 -ADAPT6 and 14-fold for 177 Lu-DOTA-A-DAPT6-ABD 035.At the same time, the tumor uptake of ABD-fused ADAPT6 was more than 3-fold higher than for 177 Lu-DOTA-ADAPT6.For both ABD-fused proteins, the tumor uptake was higher than the renal uptake.
The effect on the biodistribution of ABD-fused variants caused by the positioning of ABD 035 relative to the ADAPT6 moiety and the chemical nature of the label, is shown in Fig. 5.As can be observed, the biodistribution is profoundly influenced by the nature of the label; uptake in both tumor and kidneys was significantly (p < 0.005 in a paired t-test) higher for the residualizing 177 Lu label compared to 125 I (Fig. 5).At the same time, the blood-borne activity was significantly (p < 0.005 in a paired t-test) higher for radioiodinated variants.Thus, the use of 177 Lu provided higher tumor-to-blood ratios.It has to be noted, that the uptake in normal tissues (except blood) was significantly (p < 0.05 in paired t-test) higher for 177 Lu-labeled counterparts.
There was no significant difference between the biodistribution of radioiodinated 125 I-HPEM-ADAPT6-ABD 035 and 125 I-HPEM-ABD 035 -ADAPT6 (Fig. 5A).For the 177 Lu-labeled variants, however, the relative position of ABD 035 and ADAPT6 had a strong influence (Fig. 5B). 177Lu-DOTA-ADAPT6-ABD 035 demonstrated significantly higher tumor uptake and significantly lower renal uptake compared to 177 Lu-DOTA-ABD 035 -ADAPT6.Thus, the tumor-to-kidney ratio was appreciable more favorable in the case of placement of ABD at C-terminus of ADAPT6.Therefore, ADAPT6-ABD 035 was selected for further evaluations.
Data concerning biodistribution of 125 I-HPEM-ADAPT6-ABD 035 and 177 Lu-DOTA-ADAPT6-ABD 035 up to 2 weeks after injection are Fig. 2. CD Spectra of the proteins used in this study, demonstrating high alpha helical content as well as ability to refold after thermal denaturation.

Table 2
Affinity constants for the ABD-fused proteins towards HSA, MSA and HER2.The variants demonstrate similar affinities regardless of orientation of the ABD.Data for affinity of MG-DEAVDANS-ADAPT6-GSSC and ABD035 are taken from Refs.[21,40], respectively.presented in Fig. 6 and Tables 5 and 6.The kinetic profiling using onecompartment distribution model (Fig. 7) showed a 30.0 ± 2.7 %IA/g tumor uptake amplitude for 125 I with an uptake increase half-time of 6.3 ± 1.3 h and a clearance half-life of 84 ± 10 h.The 177 Lu compound showed a tumor uptake amplitude of 55 ± 17 %IA/g with an uptake increase half-time of 18 ± 7 h and a clearance half-life of 121 ± 40 h.
An in vivo specificity test demonstrated that uptake of both variants was significantly (p < 0.005) higher in HER2-positive than in HER2negative xenografts 48 h after injection (Fig. 6).The difference was 14-17 fold.
The biodistribution data (Tables 5 and 6) demonstrate slow clearance of activity from blood.The biological half-lives in blood were 29.4 (95% CI 25.7 to 33.7) and 28.4 (95% CI 24.6 to 32.8) h for 125 I-HPEM-ADAPT6-ABD 035 and 177 Lu-DOTA-ADAPT6-ABD 035 , respectively.The level of the blood-borne activity was slightly but significantly (p < 0.05 in a paired t-test) higher for the radioiodine label at all time points (Table 6).Both conjugates showed efficient targeting, as already at 24 h, the uptake in HER2-positive tumors exceeded uptake in any other organ or tissues for both variants.Starting from 72 h after injection, the tumor uptake was higher for 177 Lu label.Overall, the use of 177 Lu provided significantly higher tumor-to-blood but lower tumor-to-kidney ratios compared to 125 I label.Results of dosimetry calculations are provided in Tables 5 and 6.
Imaging performed 72 h post injection (Fig. 8) confirmed the biodistribution data showing that both 177 Lu and 125 I was accumulated in tumors to a much higher extent than in any of normal tissues.

Radionuclide therapy
The tumor growth data (Fig. 9; Supplemental Fig. S9) reflected the impact of treatment.The tumor volume doubling times were 7.0 (95% CI 6.0 to 8.3), 8.0 (95% CI 7.0 to 9.1), and 10.5 (95% CI 7.6 to 15.3) d, for groups of mice treated with PBS, unlabeled DOTA-ADAPT6-ABD 035 , and 177 Lu-DOTA-ADAPT Neg -ABD 035 , respectively.In mice treated with 177 Lu-DOTA-ADAPT6-ABD 035 , an initial tumor growth was followed by a shrinkage (at 11 d), and the tumor re-growth started only at approximately 30 d for the group treated by a single injection of 177 Lu-DOTA-ADAPT6-ABD 035 .In three mice, the tumors did not reach the critical volume of 1 cm 3 by the day of the study termination.In the group treated by two injections of 177 Lu-DOTA-ADAPT6-ABD 035 , only two tumors started re-growth, but did not reach the end-point by the study termination.Such tumor growth pattern had an apparent influence on survival (Fig. 9).Median survival in the control groups was 25, 25, and 31 d for treatment with PBS, unlabeled DOTA-ADAPT6-ABD 035 , and 177 Lu-DOTA-ADAPT Neg -ABD 035 , respectively.Treatment with 177 Lu-DOTA-ADAPT Neg -ABD 035 slightly but significantly increased the survival compared with the survival of mice treated by cold substances.Median survival in the group receiving a single injection of 177 Lu-DOTA-ADAPT6-ABD 035 was 70 days, which is significantly (p < 0.0001) longer than the median survival of mice in the group treated with 177 Lu-DOTA-ADAPT Neg -ABD 035 .The median survival in the group treated with two injections of 177 Lu-DOTA-ADAPT6-ABD 035 was not reached within the timeframe specified in the ethical permit (90 days).
The therapy was well tolerated.The appearance of the skin and eyes did not differ between treated and untreated mice, and there was no behavior indicating treatment-related pain or suffering.In the group treated with two injections of 177 Lu-DOTA-ADAPT6-ABD 035 (the second injection 21 days after the first one), one mouse had to be sacrificed at day 29 due to loss of body weight.However, the average animal weight did not differ significantly between the treated groups and the control groups (Fig. 9B).An average weight in treated groups increased gradually toward the end of the study.

Histological evaluation
In the hepatocytes in the liver samples from groups treated with 177 Lu-DOTA-ADAPT6-ABD 035 , (both double injection and single injection of 18 MBq) and 177 Lu-DOTA-ADAPT Neg -ABD 035 , rounded cytoplasmic vacuoles with increased variation in size and shape of the cell

Table 4
K D for interaction of radiolabeled ADAPT6 variants with HER2-expressing SKOV-3 cells determined using LigandTracer.Data are presented as an average ± maximum error.nucleus were found, which could indicate degenerative changes (Supplemental Fig. 10F, 10G, 10H).These changes were, however, subtle and considered as reversible because scattered mitotic figures were also detected, which indicates on-going regenerative activity.Scattered renal tubular epithelial cells in some mice from groups treated with radiolabeled 177 Lu-DOTA-ADAPT6-ABD 035 exhibited enlarged size of the cell nucleus and chromatin margination that were suggestive of degeneration (Supplemental Fig. 10I).Tubular degeneration, similar to hepatocellular changes, was also subtle and considered as a reversible change.
Tumors displayed features consistent with highly malignant carcinoma.Cells grew in solid pattern, sometimes showing in pseudolobules of varying size enclosed by thin septa of connective tissues.Mitotic activity was high or very high.Histopathological examination of tumors did not detect any differences between the group treated with ADAPT-ABD and control group.Animals treated with 177 Lu-ADAPT-ABD had areas of extensive fibrosis (3 of 3 available tumors in the group treated with double injection), centrally located areas of necrosis, hemorrhage and inflammation (1 of 3 tumors in the group treated with double injection, and 3 of 5 tumors in the group treated with single injection) (Supplemental Fig. 10J and 10H).In the group treated with unspecific 177 Lu-labeled conjugate, only 1 of 5 examined tumors had necrotic area.

Discussion
ADAPT6 has shown great promise in both pre-clinical and clinical imaging applications, mainly because of its small size and fast blood clearance.However, the small size and hence the substantial kidney uptake prevents further therapeutic use.In this study, we have utilized a half-life extension strategy where the protein of interest is fused to the Albumin Binding Domain of Protein G, to take advantage of the long half-life of the patient's own serum albumin.
Our molecular design of ADAPT-ABD fusion proteins has permitted production of probes capable of correct folding and high-affinity binding to both albumin and HER2 (Table 2).The ability of the proteins to refold after thermal denaturation (Fig. 2) allows the use of high temperatures

Table 5
Biodistribution of 177 Lu-DOTA-ADAPT6-ABD 035 in BALB/C nu/nu mice bearing SKOV-3 xenografts.Data are expressed as %IA/g and are averages from 4 animals ± SD.The absorbed dose to the bone marrow is indicated for the blood dose median value.for labeling with 177 Lu, which should ensure high coupling stability [41].Indeed, the label was stable under challenge with a large molar excess of EDTA, and binding specificity and affinity of 177 Lu-labeled fusion proteins were not compromised (Table 4).The InteractionMap analysis revealed two types of interactions of radiolabeled constructs with HER2 on living cells (Table 4, Supplemental Fig. S8).For each labeled variant, both interactions had similar association rate but different dissociation rates, which resulted in different equilibrium dissociation constants (K D 1 and K D 2, respectively).Such phenomenon has been observed earlier for several classes of targeted proteins binding to tyrosine kinase receptors of HER family [31,42,43].It has been shown that the phenomenon is associated with a dimerization of the receptors, which results in conformational changes leading to alteration of a ligand binding strength [43].Interestingly, the affinity to HER2 was the highest for 125 I-HPEM-ABD 035 -ADAPT6, which is in agreement with the data of Lindbo and co-workers [25] showing twice higher affinity of a non-fused (HE) 3 DANS-ADAPT6 labeled at C-terminus using 125 I-HPEM compared to a counterpart labeled using 111 In-DOTA in the same position.Thus, positioning of 125 I-HPEM on the C-terminus of ADAPT6 strengthens its binding towards HER2 independent of N-terminal modification.Cellular processing study with the use of residualizing 177 Lu label suggested that both 177 Lu-DOTA-ABD 035 -ADAPT6 and 177 Lu-DOTA-ADAPT6-ABD 035 internalized slowly, to the same extent as a non-fused ADAPT6 (Fig. 4).This might indicate that residualizing properties of a

Table 6
Biodistribution of 125 I-HPEM-ADAPT6-ABD 035 in BALB/C nu/nu mice bearing SKOV-3 xenografts.Data are expressed as %IA/g and are averages from 4 animals ± SD.The absorbed dose to the bone marrow in the spine is indicated for the blood dose median value.The biodistribution data were used to calculate absorbed doses in the case of use of 131 I label.Asterisk marks significant difference (p < 0.05 in a paired t-test) between uptake of 125 I-HPEM-ADAPT6-ABD 035 and 177 Lu-DOTA-ADAPT6-ABD 035 (Table 5) at the same time point.A pilot in vivo targeting experiment (Fig. 4) showed an appreciably enhanced retention of activity of ABD 035 -fused ADAPT6 variants in blood and reduction of their uptake in kidneys in comparison with nonfused ADAPT6.It has to be noted that although fusion of targeting proteins with ABD has been applied earlier [16,20,22,[44][45][46], the extent of such effect is unpredictable due to the complexity of interactions of all constituents of a construct in vivo.For example, fusing of ABD with ADAPT6-targeted toxins had a much more modest effect on retention in blood and uptake in kidneys [20].Studies with ABD-fused affibody molecules have demonstrated that the chemical nature of a chelator might have a substantial effect on biodistribution, blood retention and renal uptake [44].
In this study, the chemical nature of the radiolabel had a pronounced effect on biodistribution and targeting properties of ADAPT6 derivatives (Fig. 5).Dual-label experiments demonstrated that both 125 I-HPEM-ADAPT6-ABD 035 and 125 I-HPEM-ABD 035 -ADAPT6 had significantly (p < 0.05) higher uptake in blood and lower uptake in tumor and normal tissues compared with 177 Lu-labeled counterparts.In the case of 125 I-HPEM-label, a relative position of ABD 035 and ADAPT6 had no influence on biodistribution.On the contrary, the relative position of ABD 035 and ADAPT6 had a clear effect in the case of 177 Lu label.This resulted not only in significantly (p < 0.05) higher uptake of 177 Lu-DOTA-ADAPT6-ABD 035 in tumor but also in significantly (p < 0.001) lower uptake of this conjugate in kidneys.Taking into account that absorbed dose to kidneys might be dose limiting, these features of 177 Lu-DOTA-ADAPT6-ABD 035 make a difference between success and failure in radionuclide therapy.Therefore, radiolabeled ADAPT6-ABD 035 was selected for further evaluation.The variance in biodistribution of 177 Lu-labeled variants might be explained by a difference of handling of their adducts with albumin in kidneys.Albumin can be filtered through glomerular membranes, although to a low extent [47].The glomerular filtration depends on the properties of the protein [47].It is conceivable that filtration of 177 Lu-DOTA-ABD 035 -ADAPT6-albumin adduct is slightly more efficient than filtration of 177 Lu-DOTA-ADAPT6-ABD 035 -albumin adduct due to the difference in protein geometry.This assumption is in agreement with higher renal uptake of 177 Lu-DOTA-ABD 035 -ADAPT6.Accordingly, the bioavailability of 177 Lu-DOTA-ADAPT6-ABD 035 would be higher, which would result in its higher accumulation in tumors and in normal tissues, which is seen in Fig. 5 B. Similar effect should also take place in the case of radioiodinated variants, but leakage of radiometabolites of the non-residualizing radioiodine label from sites of catabolism complicates data interpretation.
Accumulation of both 177 Lu and 125 I-labeled ADAPT6-ABD 035 in tumors was apparently dependent on HER2 expression (Fig. 6).SPECT/ CT imaging showed that uptake in tumor was much higher than in any other tissue at 72 h after injection (Fig. 8).A comparative longitudinal dual-label biodistribution study demonstrated that 177 Lu-DOTA-ADAPT6-ABD 035 provides more efficient delivery of radionuclide to tumors than 125 I-HPEM-ADAPT6-ABD 035 (Tables 5 and 6).Although internalization of ADAPT6-ABD 035 is slow, a non-residualizing label was not capable of providing sufficient retention of activity in tumors.This should be true also for a molecule labeled with a therapeutic counterpart, 131 I. Currently, established absorbed dose limits during targeted radionuclide therapy are 2 Gy for red marrow and 28-40 Gy for kidneys [48].Our dosimetry calculation suggested that for 177 Lu-DOTA-A-DAPT6-ABD 035 the absorbed dose to tumor should be 67 Gy when keeping dose limits of 2 Gy for bone marrow and 20 Gy to kidneys, which is compatible with therapy application.In the case of 131 I-HPE-M-ADAPT6-ABD 035, the ratio of absorbed dose to tumor and kidneys (3.7) would be somewhat higher than for 177 Lu-DOTA-ADAPT6-ABD 035 (3.3).At the same time, the absorbed dose to bone marrow would be higher for 131 I due to both longer residence in blood and longer range of emitted beta-particles.In combination with shorter residence in tumor, this would result in a ratio of absorbed dose to tumor and bone marrow of 1.9.In the absorbed dose to bone marrow of 2 Gy, the absorbed dose to tumor would be only 3.8 Gy, which is too low for therapy.
Results of an experimental therapy study confirmed the dosimetry data.Control groups of mice treated with PBS and unlabeled ADAPT6-ABD 035 had a median survival of 25 days.Treatment with the nontargeting 177 Lu-DOTA-ADAPT Neg -ABD 035 caused a slight but significant extension of median survival (31 d) (Fig. 9), most likely due to activity circulating in blood or due to unspecific uptake in tumors because of the EPR effect.The effect of 177 Lu-DOTA-ADAPT6-ABD 035 on tumor growth (Supplemental Fig. S9 and Fig. 9) was strikingly different.The tumor growth was suppressed in a dose-dependent manner and median survival was enhanced more than two-fold even in the case of a single injection of 18 MBq of 177 Lu-DOTA-ADAPT6-ABD 035 per mouse.In addition, the double injection had a much more pronounced effect.Importantly, the treatment was not associated with any observable toxicity.For example, the average animal weight did not differ significantly between treated and control groups (Fig. 9).Histology evaluation discovered only minor and transient alterations in liver and kidneys (Supplemental Fig. 10).
Currently, there are several treatment options for patients with disseminated HER2-expressing breast cancers, e.g.treatment with monoclonal antibodies trastuzumab and pertuzumab or antibody-drug conjugated trastuzumab-DM1.These therapeutics significantly improve the survival of breast cancer patients.Unfortunately, tumors eventually develop resistance to such therapies.Remarkably, the resistant tumors preserve a high level of HER2 expression.Consequently, therapy using 177 Lu-DOTA-ADAPT6-ABD 035 might be an additional option for such patients.
In conclusion, half-life extension of ADAPT6 through fusion with ABD is a promising strategy to increase the potential of the ADAPT6 for radionuclide therapy.However, the effect of ABD-fusion is critically dependent on molecular design of the ADAPT6-based probe as well as selection of a suitable therapeutic radionuclide and labeling strategy.In the case of an optimal design, radiolabeled ADAPT6-ABD 035 fusion is capable of tumor suppression with minimal side-effects.

Fig. 1 .
Fig. 1.The two targeting agents evaluated in this study, consisting of ADAPT6 (black) and ABD 035 (grey with ABD binding surface marked in red) as well as a flexible linker (SSSG) 3 and a leader sequence (MG-DEAVDANS) to improve protein expression and biodistribution.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

6. 2
nmol) in 1 M ascorbate buffer, pH 5.5, was mixed with 345 MBq 177 Lu in 0.1 M HCl.The mixture was incubated for 45 min at 95 • C.

4 Fig. 4 .
Fig. 4. Effect of ABD 035 fusion on biodistribution of 177 Lu-labeled ADAPT variants.The biodistribution in BALB/C nu/nu mice bearing SKOV-3 xenografts was measured at 48 h after injection.Data are expressed as %IA/g and are averages from 5 animals ± SD.

Fig. 5 .
Fig. 5. Effect of ABD 035 position on biodistribution of ABD-fused ADAPT6 variants labeled with 125 I (A) or 177 Lu (B).The biodistribution in BALB/C nu/nu mice bearing SKOV-3 xenografts was measured at 48 h after injection.The data for 177 Lu-DOTA-ABD 035 -ADAPT6 and 177 Lu-DOTA-ADAPT6-ABD0 35 are the same as in Fig. 4, but presented in the same scale as for the radioiodinated counterparts to simplify comparison.Data are expressed as %IA/g and are averages from 5 animals ± SD. *marks significant difference (p < 0.05 in unpaired t-test) and ** marks highly significant difference (p≪0.001 in unpaired t-test) in uptake.

Fig. 7 .
Fig. 7. Modelling of tumor uptake kinetics.The values are presented as an average and standard deviation and the dashed lines indicate the 95% confidence interval of the fit; Tmax is reached at 25.5 ± 4.3 h with 125 I (A) and at 57.3 ± 15.4 h with 177 Lu (B).

Table 1
Theoretical and experimental molecular weights of all proteins used in this study.

Table 3
Radiochemical yields and purity of radiolabeled ADAPT6 variants.