Gadolinium-labeled Affibody-XTEN Recombinant Vector for Detection of HER2+ Lesions of Ovarian Cancer Lung Metastasis Using Quantitative MRI

Abstract

Ovarian cancer has the highest mortality rate among all gynecologic malignancies. HER2⁺ ovarian cancer is a subtype that is aggressive and associated with metastasis to distant sites such as the lungs. Therefore, accurate biological characterization of metastatic lesions is vital as it helps physicians select the most effective treatment strategy. Functional imaging of ovarian cancer with PET/CT is routinely used in the clinic to detect metastatic disease and evaluate treatment response. However, this imaging method does not provide information regarding the presence or absence of cancer-specific cell surface biomarkers such as HER2. As a result, this method does not help physicians decide whether to choose immunotherapy to treat metastasis. To differentiate the HER2⁺ from HER2⁻ lesions in ovarian cancer lung metastasis, AbX₅₀C_4:Gd vector composed of a HER2 targeting affibody and XTEN peptide was genetically engineered. It was then labeled with gadolinium (Gd) via a stable linker. The vector was characterized physicochemically and biologically to determine its purity, molecular weight, hydrodynamic size and surface charge, stability in serum, endotoxin levels, relaxivity and ability to target the HER2 antigen. Then, SCID mice were implanted with SKOV-3 (HER2⁺) and OVASC-1 (HER2⁻) tumors in the lungs and injected with the Gd-labeled HER2 targeted AbX₅₀C₄:Gd vector. The mice were imaged using dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), followed by R1-mapping and quantitative analysis of the images. Our data demonstrate that the developed HER2-targeted vector can differentiate HER2⁺ lung metastasis from HER2⁻ lesions using DCE-MRI. The developed vector could potentially be used in conjunction with other imaging modalities to prescreen patients and identify candidates for immunotherapy while triaging those who may not be considered responsive.

Introduction

Human epidermal growth factor receptor-2 (HER2) is overexpressed in different cancers such as ovarian and breast cancers and is typically associated with more aggressive tumors with metastatic potential. Determining the presence of HER2 in potentially responsive cancer patients is becoming increasingly important, given the growing arsenal of HER2-targeted therapies. Therefore, proper assessment of HER2 expression ensures that eligible patients receive HER2-targeted therapy, while others are not exposed to unnecessary treatment [1]. This emphasizes the need for a noninvasive diagnostic imaging method with high specificity for detecting HER2⁺ metastatic lesions that are not amenable to biopsy and monitoring of response to therapy. This is especially true for epithelial HER2⁺ ovarian cancer, which is an invasive phenotype. It has been shown that as the level of HER2 expression increases, the rate of invasiveness and cancer recurrence increases [2]. As lung metastasis is the second most prevalent metastatic site in ovarian cancer, it is clinically valuable to have a reliable diagnostic approach for the biological characterization of metastatic sites originating from primary ovarian tumors.

The current clinical diagnosis of ovarian cancer includes computed tomography (CT) and positron emission tomography (PET) of the peritoneal cavity and thorax applied as a single technique or combination of the two methods [3, 4]. Yet, these diagnostic methods do not provide any information on the presence or absence of HER2 expression. To address this medical need, antibodies such as Trastuzumab can be directly labeled with imaging probes such as Gd or radiotracers; however, the use of antibodies for cancer imaging has some limitations. For example, the imaging probe is covalently linked to the antibody at very low molar ratios to prevent interference with the antigen-binding site [5]. As a result, large doses of the antibody need to be injected to generate a sufficient signal-to-noise ratio suitable for imaging. This, in turn, increases the possibility of observing toxicity to non-target tissues mainly due to the presence of Fc region in antibodies. Considering that magnetic resonance imaging (MRI) is a powerful diagnostic imaging technique for tumors with a demonstrated safety record in clinics [6, 7], the objective of this study was to develop a gadolinium (Gd)-labeled targeted system with high sensitivity for the detection of HER2⁺ lesions in ovarian cancer lung metastasis using MRI. To achieve this objective, we genetically engineered a recombinant fusion vector (lacks Fc region) comprising two functional motifs. The first motif is an affibody (Ab, ~7 kDa) with a picomolar affinity toward HER2 to facilitate recognition of HER2⁺ cancer cells. The application of this HER2 targeting affibody in cancer imaging and therapy has been previously demonstrated [8–11]. The second motif is XTEN (~50 kDa) which is an unstructured, highly soluble, and negatively charged peptide [12]. The ability of XTEN to extend the blood circulation time of other proteins such as the human growth hormone and glucagon has been previously shown [13]. Along the XTEN sequence, we designed four cysteine residues (C₄) to facilitate Gd conjugation via a stable linker. For simplicity, this Gd-labeled targeted fusion vector is referred to as AbX₅₀C₄:Gd (Figure 1). We also engineered Gd-labeled XTEN without an affibody (non-targeted) and used it as a control (indicated as X₅₀C₄:Gd). Both the AbX₅₀C₄ and X₅₀C₄ vectors were genetically engineered and expressed in E. coli, followed by purification. The vectors were characterized in terms of purity, size, charge, molecular weight, endotoxin levels, Gd content, stability in serum, and Gd relaxivity. The affinity of the targeted AbX₅₀C₄ and non-targeted X₅₀C₄ toward the HER2 antigen was examined using ELISA, whereas the binding to HER2 on the surface of SKOV-3 (HER2⁺) and OVASC-1 (HER2⁻) ovarian cancer cells was analyzed using flow cytometry and confocal microscopy. Then, SKOV-3 and OVASC-1 cells were implanted in the lungs of SCID mice and evaluated using dynamic contrast-enhanced MRI (DCE-MRI). A MATLAB-based image analysis program was also developed for a 1 Tesla (T) MRI scanner to rapidly scan the mice without significant loss of resolution and perform voxelby-voxel analysis of the scans.

Figure 1:

Schematic representation of AbX₅₀C₄:Gd composed of the HER2 affibody, XTEN, and four cysteine residues (S) along the XTEN sequence. DOTA linkers were conjugated to the cysteine residues followed by chelation of gadolinium (Gd) molecules to the DOTA. The structure of AbX₅₀C₄ was predicted using the RaptorX protein structure and function prediction program (http://raptorx.uchicago.edu).

Materials and Methods

Cloning, expression, and purification of AbX₅₀C₄ and X₅₀C₄ recombinant fusion vectors

Using standard cloning techniques, the DNA sequences encoding AbX₅₀C₄ and X₅₀C₄ were designed, codon-optimized for bacterial expression, and cloned into a pET21b+ bacterial expression vector with a C-terminal 6xHis-tag (GenScript, US). The fidelity of each gene sequence to the original design was verified by DNA sequencing followed by the translation into the corresponding peptide sequences (Genewiz, US) (Table S1). The cloned vectors were transformed into LOBSTR-BL21(DE3)-RIL E.coli expression system (Kerafast, Inc) using the recommended transformation protocol [14]. E.coli was grown at 37°C in TB media, and the peptide expression was induced by IPTG (0.4 mM) when the OD₆₀₀ reached 3. Bacterial cells were grown for three hours post-induction followed by centrifugation (6000x g, 10min). The bacterial pellet was lysed by using the lysis buffer (Table S2). For every gram of bacterial pellet, 5 ml of lysis buffer was used. The slurry was vigorously stirred at room temperature for one hour, centrifuged at 20,000g (4°C), followed by the removing the supernatant. Then, Ni-NTA resin (1 ml resin per 15 g pellet) was washed and equilibrated with 10 ml of the lysis buffer. The supernatant was diluted three times its volume with the lysis buffer and incubated with the Ni-NTA resin for one hour on ice while rocking gently. The mixture was passed through a 10-ml polypropylene filter column (Bio-Rad, US) using a vacuum-driven filtration. The column was washed with 100 ml of the lysis buffer, followed by different wash buffers (Table S2). Finally, the peptide was eluted by 5 ml of the elution buffer (Table S2), filtered through a 0.2 μm syringe filter, mixed with EDTA (5 mM), and concentrated to 2 mg/ml by an Amicon® Ultra-15 centrifugal fFilter unit with the molecular weight cut off (MWCO) of 3 kDa (MilliporeSigma) at 10,000g (4 °C). The concentrations of the peptides were measured by the BCA assay (Thermo Fisher Scientific, US). The expression of the peptides was confirmed by western blot using anti-histag antibody, whereas the purity of the peptides was determined by SDS-PAGE (12% gel) and coomassie blue staining.

Measurement of the endotoxin levels in AbX₅₀C₄ and X₅₀C₄ vectors by LAL assay

Limulus Amebocyte Lysate (LAL) chromogenic endotoxin assay (Thermo Fisher Scientific, US) was used to quantify the endotoxin contents of the purified vectors. Briefly, 50 μl of the peptide samples (2 mg/ml) and different concentrations of the endotoxin standards were prepared in 1.5 ml microfuge tubes. The temperature of the tubes was equilibrated to 37 °C using a heating block. Next, 50 μl of the LAL reagent was added to each tube, gently mixed, and incubated for ten minutes. Then, 100 μl of the pre-warmed chromogenic substrate was added to each tube, mixed, and incubated for six minutes at 37 °C. To stop the reaction, 100 μl of 25% acetic acid was added to the tubes. Finally, 200 μl of each sample was transferred to a transparent 96-well plate and the absorbance was measured at 410 nm using a Tecan Infinite M200 plate reader (Tecan Trading AG, Switzerland). Samples and standards were prepared in triplicates. The endotoxin levels are presented as the mean endotoxin unit (EU) per ml ± standard deviation (s.d.).

Gd and FITC labeling of AbX₅₀C₄ and X₅₀C₄ vectors

First, 1,4,7,10-Tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA) molecules were conjugated to the cysteine residues in the XTEN sequence. For this purpose, maleimidomonoamide-DOTA (Mal-DOTA) (Macrocyclics, US) was used to bind to the thiol group in cysteine. Mal-DOTA (1 mg, 0.65 mM) was dissolved in 500 μl of elution water (without TCEP) supplemented with EDTA and then added to 2 ml of AbX₅₀C₄ or X₅₀C₄ (17 μM) in the elution buffer (pH 7.0). The reaction mixture was stirred for 2.5 hrs at room temperature under nitrogen and then passed through a PD-10 column (GE Healthcare, US) to exchange the buffer with 0.2 M ammonium acetate buffer (pH 5.5). The eluted fractions were collected, loaded onto an ultrafilter centrifugal unit with MWCO of 3 kDa, and washed three times with 0.2 M ammonium acetate buffer. Next, AbX₅₀C₄-DOTA and X₅₀C₄-DOTA were chelated with Gd using the GdCl₃ solution. A solution of 2 mg/ml GdCl₃ (100 μl) was prepared and added to the synthesized vector-DOTA solution (2 ml of 1 mg/ml in 0.2 M ammonium acetate buffer, pH 5.5). Then, the mixture was stirred for 45 min at 37 °C under nitrogen. The Gd consumption was tracked by Arsenazo III reagent using a calibration curve at the absorbance of λ660 nm. To remove the free Gd, EDTA (10 mM) was added to the mixture and then passed through a PD-10 column conditioned with 150 mM NaCl. The fractions were collected, concentrated by using an ultrafilter unit (MWCO of 3 kDa), and then washed three times by 2 ml of 150 mM NaCl. Finally, AbX₅₀C₄:Gd and X₅₀C₄:Gd vector solutions were concentrated to ~20 mg/ml. The concentration of the final product was measured by BCA assay using bovine serum albumin (BSA) as a standard.

To synthesize AbX₅₀C₄-FITC and X₅₀C₄-FITC, 2 mg of each vector (17 μM) in elution buffer (pH 7.0) was reacted with 1 mg fluorescein-5-maleimide (1.2 mM, in 100 μl DMF) (Thermo Fisher Scientific, US) for 2.5 hours at room temperature under nitrogen while protected from light. Next, the buffer was exchanged to 150 mM NaCl by passing through a PD-10 column (GE Healthcare, US), and concentrated using an ultrafilter centrifugal unit (MWCO of 3 kDa). The final volume was adjusted to 150 μl and kept at 4 °C in the dark until further use.

Measurement of the molecular weights of the vectors by MALDI-TOF

A matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDITOF MS) was used to measure the molecular weights of the vectors and vector conjugates. The stock samples were prepared at the concentrations of ~5 mg/ml of the vector or vector conjugates in 150 mM NaCl. Vectors were then mixed at ratios of 1:1 to 1:125 with the matrix solution containing 10 mg/ml sinapinic acid in 50% acetonitrile and 0.1% TFA. 1 μl of the mixture was spotted on the target plate and allowed to evaporate at room temperature. Next, the data were acquired at the linear mid-mass positive mode from 25 kDa to 90 kDa using the Applied Biosystems/MDS SCIEX 4800 MALDI-TOF/TOF mass spectrometer at the Center for Integrative Proteomics Research, Rutgers University. The data were acquired and processed by Applied Biosystems/MDS SCIEX 4000 Series Data Explorer software. BSA (5 pM) was used as a standard for equipment calibration.

Measurement of the stability of the AbX₅₀C₄ in serum

The stability of the AbX₅₀C₄ vector in serum was evaluated by western blot analysis. A 2 mg/ml solution of the vector was passed through a PD-10 column to remove TCEP. Then, 50 μg of the vector was mixed with 1 ml of the off-the-clot pooled serum (Innovative Research, US) and incubated at 37 °C for 24 hours while keeping in the dark. Then, 50 μl samples were taken at times 0, 1, 3, 5, 24 hours and analyzed by western blot using an anti-histag antibody. The intensity of the bands was measured by importing the grayscale images into the ImageJ software (NIH, US).

Measurement of the hydrodynamic size and zeta potential of the vectors

To measure the hydrodynamic size of the vectors and conjugates, 30 μl of 1 mg/ml samples were added to a low-volume quartz batch cuvette (ZEN2112, Malvern Inc, UK) and measured by Malvern Zetasizer ZS (Malvern Inc., UK). To measure the zeta potential of the vectors, 50 μl of the samples (1 mg/ml) was mixed with 600 μl of 5 mM NaCl. The diluted samples were then measured by the Zetasizer using the Universal Dip Cell for zeta potential measurements (ZEN1002, Malvern Inc., UK). At least six independent batches for each sample were prepared. The results are presented as mean ± s.d.

Measurement of the Gd content in vectors and stability by ICP-MS

The gadolinium contents of AbX₅₀C₄:Gd and X₅₀C₄:Gd were measured by inductively coupled plasma mass spectrometry (ICP-MS) using Thermo Scientific^™ iCAP^™ Q ICP-MS system. For this, samples were diluted to approximate concentrations of less than 10 ng/ml in 2% nitric acid. A calibration curve of GdCl₃ in nitric acid was also prepared in the range of 1 to 20 ppb. The samples and standards were analyzed by iCAP^™ Q ICP-MS system at Rutgers Department of Earth and Planetary Sciences Core Facility. The data are reported as mean Gd concentrations ± s.d. (n=9).

To measure the stability of the chelated Gd, 100 μl (5 mg/ml stock solution) of Gd-labeled vector was mixed with 900 μl of deionized water, placed in Slide-A-Lyzer (MWCO 3.5 kDa), and dialyzed against 100 ml of the deionized water for 24 hours at 4 °C in the dark. Next, 50 μl of each dialysis medium was collected and diluted with deionized water 10³ to 10⁴ folds. The Gd content was quantified by iCAP^™ Q ICP-MS. As a control, 100 μl of GadoSpin^™ P (Miltenyi Biotech, Germany) containing equivalent to 2.5 μmole Gd was treated similar to the Gd-labeled vectors and then evaluated for its Gd stability. GadoSpin^™ P is a polymeric gadolinium-based contrast agent of high molecular weight (~200 kD) formulated for preclinical MRI. The results of the ICP-MS measurements are shown as the mean Gd concentrations ± s.d. (n=9).

Measurement of the binding of the vectors to HER2 antigen by ELISA

For evaluation of the vector’s ability to bind to HER2 antigen, an enzyme-linked immunosorbent assay (ELISA) was performed. First, 100 μl of 1 μg/ml HER2 antigen (Abcam, ab168896) was coated onto a maxi absorbance 96-well plate and incubated overnight at 4 °C. Each well was washed three times by phosphate buffered saline (PBS) plus 0.1% Tween (PBST) and blocked by 2% BSA in PBS-T for two hours at room temperature. Then, different concentrations of AbX₅₀C₄ or X₅₀C₄ ranging from 8.5 to 0.0085 nM were added to each HER2-coated well and incubated for 30 min at room temperature. Each well was washed three times by PBS-T and then incubated with 100 μl of the HRP-conjugated anti-His tag antibody (Abcam, ab1187) for 60 min at room temperature. Samples were washed six times by PBS-T, followed by the addition of 50 μl 3,3′,5,5′-tetramethylbenzidine (TMB) and incubation for 15 min at room temperature. Finally, the reaction was stopped by 50 μl of 2 M sulfuric acid and the absorbance was measured at 450 nm using a Tecan microplate reader. Trastuzumab (Abcam) at 6.9 nM concentration was used as control.

Additionally, the dissociation constant (K_d) of AbX₅₀C₄ was estimated by using competitive ELISA as previously described [15]. Briefly, 100 μl of two concentrations of HER2 antigen (i.e., 0.1 μg/ml and 10 μg/ml) were coated onto a maxi absorbance 96-well plate, washed three times by PBS-T, and blocked by 2% BSA-PBST, as mentioned above. Then, nine serial dilutions of HER2 antigen ranging from 0 to 100 nM were prepared in microfuge tubes. 150 μl of the AbX₅₀C₄ solution (0.1 μg/ml, 2 nM) was added to each tube and incubated for 60 min at room temperature. Next, 100 μl of each sample was added to the HER2-coated wells and incubated for 30 min at room temperature. Each well was washed and incubated with an HRP-conjugated anti-His tag antibody (Abcam, ab1187). Wells were rewashed six times by PBS-T, and then 50 μl of TMB was added and incubated for 15 min. Finally, the absorbance of each well was measured at 450 nm using a Tecan microplate reader. To calculate the K_d of AbX₅₀C₄ for the HER2 antigen, the following equation was used [16]:

$$A=\left(A_{max}-A_0\right)\times \frac{-\left(nx-a+K_d\right)+\sqrt{{\left(nx-a+K_d\right)}^2+4a{K}_d}}{2a}+A_0$$

where, A is the measured sample absorbance at 450 nm, A_max and A₀ are the maximal and minimal ELISA absorbance obtained at 0 nM and 100 nM HER2 antigen, K_d is the dissociation constant, a is the total AbX₅₀C₄ concentration, n is the number of binding sites per HER2 antigen, and x is the total HER2 antigen concentration. At A_max, all AbX₅₀C₄ are considered attached to the HER2-coated plate surfaces, and the upper solution completely devoid of the free AbX₅₀C₄. As both HER2 antigen and AbX₅₀C₄ are monomers, thus n was considered as 1. The affinity is presented as the mean value of 18 independent measurements ± s.d.

Evaluation of the ability of the vectors to bind to HER2 on SKOV-3 cells

First, the expression levels of HER2 on the surfaces of two ovarian cell lines (i.e., SKOV-3 and OVASC-1) were evaluated using flow cytometry (CytoFLEX, Beckman Coulter). OVASC-1 is a patient-derived ovarian cancer cell line obtained from a patient with advanced epithelial ovarian cancer [17, 18]. SKOV-3 and OVASC-1 cells were incubated with primary anti-HER2 antibody (Abcam) for one hour at 4 °C. Next, samples were washed, incubated with FITC-labeled secondary goat anti-rabbit polyclonal antibody (Abcam), and analyzed by flow cytometry (CytoFLEX, Beckman Coulter) to measure the expression levels of HER2. Untreated cells and IgG isotype were used as controls. The results are presented as the mean fluorescent intensity of 10⁵ events per sample.

To examine the ability of AbX₅₀C₄ and X₅₀C₄ to bind to HER2 on the surface of HER2⁺ SKOV-3 cells, 1.5 × 10⁴ SKOV-3 cells were seeded in a 96-well plate. Cells were incubated with AbX₅₀C₄ or X₅₀C₄ at two concentrations of 1 nM and 50 nM for 2 hours. Trastuzumab at 50 nM concentration was incubated with SKOV-3 cells at 37°C for two hours and used as a control. To differentiate the receptor-mediated active uptake from non-specific uptake, the plates were incubated at 37 °C and 4 °C. After 2 hours of incubation, each well was washed three times by PBS and evaluated visually by a fluorescent microscope. Then, cells were trypsinized, fixed with 3% formaldehyde, and evaluated by flow cytometry (CytoFLEX, Beckman Coulter). The data are reported as the geometric mean fluorescent intensities ± s.d.

The internalization of the vectors into SKOV-3 cells was also examined by confocal microscopy. SKOV-3 cells were seeded at the density of 2.5 × 10⁴ cells on a Nunc Lab-Tek^™ chamber slide (Thermo Fisher Scientific, US). The next day, cells were treated with 20 nM of FITC-labeled AbX₅₀C₄ or X₅₀C₄ and incubated at 37 °C. At different time points ranging from 10 min to 2 hours, cells were washed three times by cold PBS, immediately fixed by 4% paraformaldehyde for 15 min, and then stained by 300 nM DAPI solution for 1 min. Slides were gently washed three times by cold PBS and then observed under a Leica TCS SP8 Confocal Microscope (Leica Microsystems GmbH) with the 63× objective lens using immersion oil. Z-stack images were taken each at 500 nm apart. The images of different z-stacks were processed by Leica software.

Genetic engineering of luciferase-expressing SKOV-3 and OVASC-1 cell lines

The gene encoding nanoluciferase (NanoLuc®, Promega, Madison, WI) was designed and then synthesized for mammalian cell expression using the PiggyBac transposon system (VectorBuilder Inc., USA). The plasmid also contained the hygromycin B resistant gene to facilitate colony selection. Cells were seeded in a 6-well plate at the density of 3 × 10⁵ cells/well and then co-transfected with the constructed transposon plasmids along with the helper plasmid encoding hyperactive PiggyBac transposase system by using Lipofectamine 3000 (Thermo Fisher Scientific, US). The transfected cells were maintained in growth media supplemented with 10% FBS and 400 μg/ml of hygromycin B (Clontech, US) for four weeks. The media was refreshed every three days to obtain stable clones that express nanoluciferase (nLuc). The stable expression of the nLuc was tested by the IVIS imaging system using the NanoGlo substrate (Promega, WI). A stable clone was selected and stored under the liquid nitrogen for future use.

Measurement of the longitudinal relaxivity (r1) of the Gd-labeled vectors

A series of phantoms composed of different dilutions of vectors were prepared in 150 mM NaCl. Phantoms were scanned by multiple 3-dimensional spoiled gradient echo (mGRE-SP) sequences using 1T ASPECT M2 compact High-Performance MRI system (Aspect Imaging, Israel). The acquired images were exported in DICOM format and then imported into an in-house developed MATLAB script. The scans at different flip angles were combined in MATLAB and the R1-maps of the phantoms were created at the voxel-by-voxel scale [19]. From the generated R1 maps, the mean R1 and then ΔR1 of each Gd concentration were calculated by subtracting the R1 of the phantom lacking Gd. The following equation is valid at different concentrations of Gd for each vector:

$$\Delta \text{R}1=1/\Delta \text{T}1=[\text{Gd}].\text{r}1$$

where, [Gd] is the Gd concentration in mM, and r1 is the longitudinal relaxivity in mM.S⁻¹. By plotting [Gd] vs. ΔR1, the r1 of each vector was calculated from the slope of the best-fit line. Various phantoms of Magnevist® (gadopentetate dimeglumine) were prepared and used as controls.

Ovarian xenograft tumor implantation in the lungs of SCID mice and imaging

All the animal handling procedures and maintenance were conducted according to the guidelines of the Rutgers University Institutional Animal Care and Use Committee (IACUC). The SHO SCID hairless female mice (4–6 weeks old) were purchased from Charles River (Fairfield, NJ). To develop ovarian cancer lung metastasis tumor model, 1 × 10⁶ SKOV-3-Luc or OVASC-1-Luc cells were suspended in PBS and then injected retroorbitally using a 27-gauge needle.

For bioluminescent imaging of the ovarian cancer lung metastasis, the bioluminescent signal intensity in mice was monitored weekly using an IVIS imaging system (PerkinElmer Inc., US) until a stable signal in the lungs was detectable. For the detection of the nanoluciferase signal, 100 μl of NanoGlo® (Promega, US) reagent was diluted in PBS (100-fold dilution) and injected systemically before each imaging. Mice were scanned immediately after injection of NanoGlo® by IVIS with an exposure time of 60 s.

For MRI, a 35-mm small animal body coil and a 1T ASPECT M2 MRI System were used. Mice were first scanned at Scout mode to find the target location. A T2-weighted fast spin echo (FSE) scanning mode was used to visualize mouse anatomical organs. The FSE scanning parameters were as follows: coronal view, field of view (FOV) 100 × 100 mm, slice thickness 1 mm with no gap, sampling 256, encoding 250, time of echo (TE) 74.6 millisecond, repetition time (TR) 4593.7 millisecond, excitation 2, external average 3, flip angle of 180 degrees, and with the scan time of 10.05 min. A 3-dimensional gradient echo with the external average (GRE-EXT) was used to acquire the high-resolution T1-weighted images for qualitative evaluations. The GRE-EXT scan mode visually confirms the co-localization of the vectors with the lung metastasis lesions at a high spatial resolution. For that, mice were scanned by GRE-EXT at several timepoints from pre-injection to 24 hours post-injection. The acquired DICOM files were imported into the VivoQuant software (Invicro, MA), and then the enhancement of signals in lung lesions was monitored over time. The GRE-EXT sequence parameters were as follows: coronal view, FOV of 100 × 100 mm, slice thickness 1 mm, sampling 256, encoding 256, TE 3.8 ms, TR 15.0 ms, excitation 8, flip angle of 25 degrees, and with the total scan time of 10.54 min.

Quantitative analysis of MRI scans

Mice were injected with the AbX₅₀C₄:Gd, X₅₀C₄:Gd, or Magnevist equivalent to 1 μmole Gd. Then, the mGRE-SP sequence was used for the R1-mapping and quantitative analysis at several flip angles, including 5, 8, 13, 25, 60, and 160 degrees. The mGRE-SP scan with the varied flip angles generated a lower spatial resolution, but it resulted in higher temporal resolution scans (a resolution of less than 2 min per scan) than the GRE-EXT enabling us to quantify Gd concentration in lung tumors. Mice were scanned by mGRE-SP sequence using the following parameters: coronal view, FOV 100 × 100 mm, slice thickness 1 mm, sampling 128, encoding 128, TE 2.3 ms, TR 16.0 ms, excitation 2, flip angles of 5, 8, 13, 25, 60, and 160 degrees, and the total scan time of 1.40 min. sec per flip angle. Then, the mGRE-SP scans were imported as DICOM files into and analyzed by an in-house developed MATLAB script to create the R1 maps of the pre- and post-injection scans. Using the voxel-by-voxel-based R1 values at pre/post-injections, the ΔR1 data (ΔR1= R1_post – R1_pre) were calculated and from that the concentration maps of the MRI scans were created. Ultimately, the mean concentrations of Gd over time at the selected regions of interest (ROIs) were reported. The reported data represent the repeated measure analysis of the tumors from 3 to 4 slices. To compare the Gd kinetics data, the area under the Gd concentration-time curves (AUCs) were calculated by GraphPad Prism Software and statistically compared using ANOVA followed by Tukey’s post hoc test. A similar mGRE-SP method assisted by MATLAB-based R1-mapping was used to quantify the distribution of the vectors in mice organs. The organ distribution data are presented as the mean Gd concentration in the selected ROI ± s.d. (n=3).

Results

Biosynthesis, bioconjugation, and physicochemical characterization of the vectors

Using genetic engineering techniques along with a specialized purification protocol, AbX₅₀C₄ and X₅₀C₄ were biosynthesized in E. coli and then purified. Western blot analysis confirmed the expression of both vectors (Figure S1), while the SDS-PAGE images indicated the purity of the vectors to be above 99% (Figure 2A). The yields of the expression and purification processes were ~3 and ~1 mg per liter of the bacterial culture for AbX₅₀C₄ and X₅₀C₄, respectively.

Figure 2.

Physicochemical characterizations of the genetically engineered vectors. A) SDS-PAGE of the purified vectors and their conjugates. For AbX₅₀C₄ and its conjugates, equivalent of 5 μg peptide was loaded onto each well, whereas for X₅₀C₄ and its conjugates, equivalent of 10 μg peptide was loaded. B–C) MALDI-TOF of the vectors and their conjugates. D–E) Hydrodynamic size and zeta potential of the peptides (pep) and their conjugates (*t-test, p<0.05). F) Serum stability of AbX₅₀C₄ after a 24-h incubation period. G) The concentration of Gd released from the vectors after a 24-h dialysis against distilled water. The Gd concentration was measured by ICP-MS for AbX₅₀C₄ and compared to the commercially available GadoSpin^™ P. H) Levels of endotoxin in the purified vectors. The red dashed line shows the maximum allowable concentration as per FDA regulations. The data are presented as mean±s.d.

In the next step, Gd was used to label AbX₅₀C₄ and X₅₀C₄ via DOTA. The yield of Gd-labeling for both AbX₅₀C₄ and X₅₀C₄ was 65±10%. The MALDI-TOF data indicated that one mole of AbX₅₀C₄ on average reacted with 3.96 moles of DOTA, whereas one mole of X₅₀C₄ reacted with 3.09 moles of DOTA (Figures 2B and 2C, Table 1). In the FITC-labeled vectors, these values were 2.4 and 1.6 moles of FITC per mole of AbX₅₀C₄ and X₅₀C₄, respectively. Overall, the theoretical molecular weights of the vectors and their conjugates were in agreement with the MALDI-TOF results.

Table 1.

Theoretical and observed molecular weights of the vectors and their conjugates as determined by MALDI-TOF.

Vector	AbX50C4	AbX50C4-(DOTA)4	X50C4	X50C4-(DOTA)4
Theoretical Mw (Da)	58678	60782	51690	53794
Observed Mw (Da)	58446	60534 (n=3.96*)	51172	52800 (n=3.09*)

^*n: average number of conjugated DOTA per vector calculated based on the MALDI-TOF data.

The vectors were also characterized in terms of their hydrodynamic size and surface charge as these two factors affect the blood circulation times. The results of these experiments showed that AbX₅₀C₄ and X₅₀C₄ had sizes of 16.5±1.8 nm and 12.6±0.4 nm, respectively (Figure 2D). Statistical analysis of data showed that in general, the presence of the HER2 affibody slightly increased the vector size (t-test, p<0.05). Analysis of the surface charges of the vectors showed that both vectors were negatively charged, with no significant difference between the AbX₅₀C₄ and X₅₀C₄ vectors (t-test, p>0.05) (Figure 2E).

The stability of AbX₅₀C₄ in serum was studied by incubating the vector with serum for 24 h followed by western blot analysis. The results of this experiment revealed that the vector was stable in serum, showing no significant increase in protein degradation during the tested period (Figure 2F, and Figure S2).

To assess the stability of the chelated Gd in the vectors, ICP-MS was used. Here, the release of Gd was measured after 24 h of dialysis against distilled water and compared to GadoSpin^™ P. The ICP-MS data showed that the amount of Gd released from AbX₅₀C₄:Gd was negligible and statistically the same as GadoSpin^™ P (Figure 2G). This indicated that Gd in the vectors was stable, and the free unconjugated Gd was effectively removed during the synthesis process.

As both vectors are intended for in vivo use, we then measured the endotoxin levels using the LAL assay. The results of this experiment showed that the endotoxin levels in AbX₅₀C₄ and X₅₀C₄ were 0.133±0.003 and 0.070±0.020 EU/ml, respectively (Figure 2H). These values are below the US Food and Drug Administration recommended maximum endotoxin limit for the injectable solutions (i.e., < 0.25 EU/ml) [20].

Evaluation of the ability of the vectors to bind to the HER2 antigen and internalize

The ability of the AbX₅₀C₄ and X₅₀C₄ vectors to bind to HER2 antigen was first evaluated by ELISA. The results of ELISA showed that AbX₅₀C₄ was able to bind to HER2 at levels comparable to trastuzumab, whereas X₅₀C₄ failed to bind to the HER2 antigen (Figure 3A). In the next step, we estimated the dissociation constant (K_d) of AbX₅₀C₄ by competitive ELISA. The results of this experiment showed that the estimated K_d value for AbX₅₀C₄ was 3.71±0.62 nM (Figure 3B).

Figure 3.

Evaluation of the ability of AbX₅₀C₄ to bind to the HER2 antigen. A) Binding of AbX₅₀C₄ and X₅₀C₄ to the HER2 antigen as compared to trastuzumab (Tzmab). B) Measurement of the dissociation constant (K_d) of AbX₅₀C₄ by competitive ELISA at two different concentrations of the HER2 antigen; i.e., 0.1 and 10 μg/ml. The data are presented as mean±s.d.

To evaluate the ability of AbX₅₀C₄ to bind to HER2 on the surface of cancer cells, SKOV-3 cells (HER2⁺) were used as the positive control and OVASC-1 cells (HER2⁻) as the negative control (Figure 4A–4B). In this experiment, AbX₅₀C₄ and X₅₀C₄ were labeled with FITC and then incubated with the SKOV-3 and OVASC-1 cell lines at 37 °C. Fluorescence microscopy and the corresponding flow cytometry data showed that the FITC-labeled AbX₅₀C₄ was selectively attached to the SKOV-3 cells but not the OVASC-1 cells. The binding of AbX₅₀C₄ to the SKOV-3 cells was detectable when the concentration of the vector was as low as 1 nM. In contrast, the FITC-labeled X₅₀C₄ could not bind to any of the two cell lines at 37 °C (Figure 4C–4D).

Figure 4:

Evaluation of the expression levels of HER2 in ovarian cancer cells and cell binding of the AbX₅₀C₄-FITC and X₅₀C₄-FITC vectors using flow cytometry. A) Flow cytometry dot plots of the SKOV-3 and OVASC-1 cells after incubation with trastuzumab and IgG isotype antibodies. B) Bar chart summarizing the mean fluorescent intensity (MFI) of the cells. C) Cell association of AbX₅₀C₄- FITC and X₅₀C₄-FITC vectors at 37 °C. The representative phase images (upper row), fluorescent microscopy images (middle row), and flow cytometry histograms (lower row) are shown for SKOV-3 and OVASC-1 cell lines treated with the vectors. D) Bar chart summarizing the mean fluorescent intensity (MFI) of three independent experiments (mean± s.d.). This figure shows that AbX₅₀C₄ was able to bind to SKOV-3 cells but not OVASC-1, whereas X₅₀C₄ failed to bind to any of the two cell lines. Neg.Ctrl., negative control.

To examine whether the vectors attached to the surface of the SKOV-3 cells were internalized, we performed two experiments. First, we performed the same experiment mentioned above, but at 4 °C (instead of 37 °C) to determine whether the binding of AbX₅₀C₄ to the cells was energy dependent. The fluorescence microscopy and flow cytometry data showed that the cell uptake was significantly reduced under this condition, indicating that the binding of the vector to the SKOV-3 cells was energy dependent (Figure S3).

Furthermore, we used confocal microscopy to examine whether the FITC-labeled AbX₅₀C₄ vectors were internalized or immobilized on the cell surfaces. Microscopy images of the SKOV-3 cells incubated with FITC-labeled AbX₅₀C₄ complexes showed that the vectors were internalized as early as 10 min. The highest levels of green fluorescence in the midsection slices (z-stacks), as well as the presence of green fluorescence in the same planar view (edge) as the cell nucleus, validated the internalization of the vectors into the cells (Figure 5 and Figure S4).

Figure 5:

Confocal microscopy images of the SKOV-3 cells treated with AbX₅₀C₄-FITC and imaged at different time points. The cell nucleus is labeled with DAPI (blue) and vector with FITC (green). The overlay images from the top view show time-dependent binding to cells, whereas the side view images (edge) show the internalization of the fluorescent-labeled vector.

Magnetic longitudinal relaxivity (r1) of the vectors

To determine the magnetic longitudinal relaxivity of the Gd-labeled vectors, a varied flip angle R1-mapping approach was employed. Various concentrations of Gd from Magnevist, AbX₅₀C₄:Gd, and X₅₀C₄:Gd were scanned using a 1T MRI system. The acquired scans were analyzed using our in-house MATLAB code, which resulted in an R1-map for each vector at different concentrations (Figure 6A). Using this method, the mean ΔR1 of each vector was extracted and plotted against the corresponding Gd concentration (Figure 6B). The data showed that AbX₅₀C₄:Gd had the highest r1 value of 3.99 s⁻¹mM⁻¹ as compared with 3.10 s⁻¹mM⁻¹ and 2.25 s⁻¹mM⁻¹ for Magnevist and X₅₀C₄:Gd, respectively (Figure 6C). The data suggest that both AbX₅₀C₄:Gd and X₅₀C₄:Gd have a suitable r1 value and can be used as a T1-weighted contrast agent for MRI.

Figure 6:

MRI longitudinal relaxivity (r1) of the Gd-labeled vectors. A) Voxel-by-voxel R1 maps of different concentrations of Magnevist, AbX₅₀C₄:Gd, and X₅₀C₄:Gd which were constructed from the exported mGRE-SP MRI scans. A standard Gd phantom was also used to normalize the scans at different flip angles for R1-mapping. B) Measurement of the r1 relaxivities from the linear regression slope of the mean ΔR1 versus [Gd] curve. C) The calculated r1 values for the vectors.

Detection of HER2⁺ metastatic lesions in mouse lungs by the vectors

To examine the ability of the AbX₅₀C₄:Gd vector to differentiate HER2⁺ lesions from the HER2⁻ lesions in the lungs, stable clones of the SKOV-3-Luc and OVASC-1-Luc cells were genetically engineered and then injected retroorbitally into SCID mice to implant tumors in the lungs. The establishment of small cancer lesions in mouse lungs was confirmed by BLI ~3 weeks post implantation (Figure 7A).

Figure 7.

BLI and GRE-EXT scans of mice with SKOV-3 and OVASC-1 lung lesions. A) Bioluminescent images of mice with SKOV-3 and OVASC-1 lung lesions immediately after administration of the NanoGlo® substrate. The presence and location of the lesions in the lungs were confirmed post mortem. B) GRE-EXT MRI scans after administration of the vectors showing the presence of the SKOV-3 and OVASC-1 lesions in the lungs. The yellow and red arrows indicate the locations of the lesions. Magnevist was used as a control contrast agent. C) The mean signal intensity enhancement (Δ Mean) in the lung tumors was measured from the GRE-EXT scans using the VivoQuant software.

A T1-weighted GRE-EXT sequence was then used to visualize the position of the lung metastasis using MRI (Figure 7B). Once the presence of metastasis was confirmed, the mouse was injected sequentially with AbX₅₀C₄:Gd, X₅₀C₄:Gd, and Magnevist every 48 hours to semi-quantitatively evaluate the lung tumors. After each injection the mouse was scanned for 24 h at different time points using the GRE-EXT and mGRE-SP sequences. While the GRE-EXT sequence was employed for qualitative evaluation of the lesions, the mGRE-SP sequence was used to generate R1-maps and calculate the Gd concentrations. The acquired T1-weighted GRE-EXT images showed that among the three injected agents (dose normalized to equal the Gd concentration), AbX₅₀C₄:Gd had the highest and sustained signal intensity post injection in the SKOV-3 lung tumors but not OVASC-1 (Figure 7B, and Figure S5–S6). In contrast, the signal intensities of X₅₀C₄:Gd and Magnevist rapidly diminished and came close to the baseline 120 min post-injection (Figure 7C).

To quantify the concentration of Gd in lung tumors, mGRE-SP scans were performed and voxel-by-voxel R1 maps were generated using our in-house image analysis program. The mGRE-SP images of the lung tumors with their corresponding R1-maps before and after the injections are shown in Figures 8A. Using this information, Gd concentration maps were generated and the concentration of Gd in the lung tumors over time was calculated (Figure 8B). Statistical analysis of data from the area under the curve (AUC) showed that the lung tumors had the highest accumulation and retention of Gd after the administration of AbX₅₀C₄:Gd over the first three hours post injection. In contrast, the non-targeted X₅₀C₄:Gd and Magnevist accumulation was significantly lower (Figure 8C) (ANOVA/ Tukey’s post hoc test, p< 0.0001). As a result, a considerably higher signal-to-noise ratio was generated by AbX₅₀C₄:Gd, providing a reliable method for detecting HER2⁺ metastasis in the lungs.

Figure 8:

Tumor targeting and kinetics of the vectors in mice with SKOV-3 and OVASC-1 lung tumors. A) A slice of the GRE-SP (flip angle 25°) MRI scan showing the SKOV-3 and OVASC-1 lesions in the mouse lungs. The R1 maps of pre/post-injection and the corresponding Gd concentration maps are also shown. B) The Gd concentration in the lung tumors over time after injection of the vectors. C) Statistical analysis of the Gd concentrations in different groups by ANOVA followed by Tukey’s post hoc test. ###, p< 0.0001 compared to the SKOV-3 group treated with AbX₅₀C₄:Gd.

Using the same mGRE-SP MRI sequence method, the biodistribution of the vectors in the blood, liver, and kidneys was also evaluated. The blood concentration for each agent was estimated from the R1-mapping of the aorta by considering the hematocrit effect as described previously [21]. The analysis of the biodistribution and elimination of the vectors showed that AbX₅₀C₄:Gd and X₅₀C₄:Gd had a significantly longer blood circulation time than Magnevist. It was also found that the kidneys were the major route of Magnevist elimination, whereas AbX₅₀C₄:Gd and X₅₀C₄:Gd accumulated in the liver followed by sustained excretion through the kidneys for more than 7 h post-injection (Figure S7).

Discussion

HER2⁺ ovarian cancer is a subtype that is aggressive and associated with metastasis to distant sites such as the lungs. Therefore, accurate biological characterization of metastatic lesions is vital as it would help physicians select the most effective treatment strategies. At present, the HER2 expression levels are determined by ex vivo analysis of tissue biopsy using fluorescence in situ hybridization, immunohistochemistry, and other methods with the score range of 0 to 3+ [22]. These methods, although commonly used in clinical practice, have several limitations. Most notably, they require tissue removal from the body, which restricts their analysis only to the sampled parts and may not properly represent the overall tumor characteristics (tumor heterogeneity). Variability in scoring between these techniques, whether a result of true heterogeneity or artifacts in preparation, has led to decreased reliability of the final HER2 status determination [23]. Introducing MRI-assisted methodology for in vivo mapping and quantification of HER2 receptors in tumors would present a noninvasive option to obtain real-time information that could not only facilitate selection of patients for HER2-targeted therapy, but also provide information on the immediate response to therapeutic interventions, especially in distant metastases that is not amenable to biopsy. Functional imaging of ovarian cancer with PET/CT in combination with ¹⁸F-FDG is routinely used in the clinic to detect metastatic disease and evaluate the treatment response [24]. However, this imaging approach only shows increased glucose metabolism and does not provide information regarding the presence or absence of cancer-specific cell surface biomarkers such as HER2. To overcome this challenge, antibodies have been labeled with radioactive elements to enable the detection of specific cell surface biomarkers. For example, radiolabeled HER2 affibody has been used as a PET imaging probe to determine the levels of HER2 expression and to monitor the response to HER2-targeted therapy in breast cancer patients [11]. While effective, this methodology exposes patients to high concentrations of radiotracers to generate a meaningful signal. Compared to PET, MRI is a safer alternative with minimum hazards to the patients and provides greater accuracy and resolution for detecting tumor lesions at the primary ovarian sites and locations of metastasis. For example, Magnevist (Gd-DTPA) has been routinely used in cancer diagnosis using MRI. However, Magnevist does not provide any information about cancer surface biomarkers.

To address the aforementioned limitations, we developed a Gd-labeled HER2-targeted vector to detect HER2⁺ metastatic lesions in the lungs using DCE-MRI. The vector is composed of an anti-HER2 affibody (K_d= 22 picomolar) with a significantly higher affinity toward HER2 than trastuzumab (K_d=120 picomolar) [8]. As the affibody rapidly clears through the kidneys in less than 10 min due to its small size (~7 kDa) [11], we fused the affibody with a ~50 kDa XTEN to generate the AbX₅₀C₄ vector (Figure 1) [13]. The size of XTEN was carefully selected to be above 40 kDa to minimize rapid renal filtration of the vectors [25]. In addition, we designed the sequences of the vectors so that their molecular weights remained less than 60 kDa to provide an estimated whole body half-life of less than 5 h [25]. This allowed us to administer different vectors and perform multiple sequential MRI scans without interference between the signals.

To construct the vectors, the genes encoding AbX₅₀C₄ and X₅₀C₄ were cloned, transformed into a bacterial expression system, expressed, and purified. The SDS-PAGE images showed vectors with high purity; however, the observed molecular weights of the vectors appeared to be approximately twice their theoretical expected values (Figure 2A). The reason for this discrepancy is that XTEN lacks hydrophobic amino acids, which results in loose binding to sodium dodecyl sulfate (SDS). Consequently, the peptide is not sufficiently negatively charged to migrate through the gel and toward the anode. A closer look at the literature revealed that the same phenomenon was observed by other groups [12, 26]. Nevertheless, our MALDI-TOF data resolved this discrepancy and confirmed a close agreement between the theoretical and observed molecular weights of both vectors (Figures 2B–2C). We also characterized the vectors in terms of size and charge, as these parameters affect their blood circulation time. The data showed that both the AbX₅₀C₄ and X₅₀C₄ vectors had hydrodynamic sizes of bigger than 5 nm, which is the renal filtration cut-off size (Figure 2D) [25]. The analysis of surface charge showed that both vectors were negatively charged (Figure 2E). The negative charge of the vectors diminishes their opsonization and clearance by the reticuloendothelial system [27].

Next, we evaluated the serum stability of the vectors and the stability of the chelated Gd. These two factors play important roles in MRI investigations as an unstable Gd-labeled vector may not only be rapidly degraded and eliminated from the body, but also release Gd which would interfere with accurate imaging of the HER2⁺ lesions. Incubation of the vectors with serum showed that both vectors were stable for at least 24 h (Figure 2F). To attach Gd molecules to the vectors, we used DOTA as a linker because it has been shown that macrocyclic chelating agents such as DOTA and NOTA provide effective chelating forces [28]. The Gd stability study also showed that the chelated Gd was stable in the vector structure and the free Gd concentration was negligible and similar to that of GadoSpin^™ P (Figure 2G). In addition to stability, we also measured the endotoxin levels in the purified vectors as they were intended for in vivo use. Our data showed that the purification protocol effectively removed the endotoxins reaching levels lower than the FDA-recommended limit of endotoxin in “Water for Injection” (i.e., < 0.25 EU/ml) [20]. The effective removal of endotoxins from the vectors during the purification process could be due to their negative surface charges, which may have minimized their interactions with the negatively charged endotoxins allowing for their effective removal.

After physicochemical characterization of the vectors, we characterized them biologically. Here, we examined whether the fusion of XTEN to the affibody hindered its ability to bind to the HER2 antigen. The results of ELISA showed that AbX₅₀C₄ was able to bind to the HER2 antigen at sub-nanomolar levels, similar to trastuzumab (Figure 3A–B). We then used confocal microscopy to determine whether the AbX₅₀C₄ bound to the HER2 antigen on the surface of the SKOV-3 cells could be internalized. The results of this experiment showed energy-dependent internalization of AbX₅₀C₄ and the presence of FITC-labeled vector molecules in the cytoplasm of the SKOV-3 cells, as evidenced by the z-stack images (Figure 4). Overall, ELISA showed that the AbX₅₀C₄ vector could effectively bind to the HER2 antigen and confocal microscopy demonstrated selective internalization of AbX₅₀C₄ into HER2⁺ but not HER2⁻ cells.

Once we confirmed that the AbX₅₀C₄ vector meets our physicochemical and biological requirements for targeting HER2⁺ cancer cells, we measured its longitudinal relaxivity prior to our in vivo studies. Relaxivity is a measure of the sensitivity of the contrast agent and is the degree to which the agent can enhance the longitudinal or transverse water relaxation rate constant normalized to the concentration of the contrast agent. These experiments showed a higher relaxivity value for AbX₅₀C₄:Gd than for Magnevist (Figure 6). This higher r1 value indicates that in comparison to Magnevist, AbX₅₀C₄:Gd can enhance the visibility of the tumor lesions with less Gd. It was also found that X₅₀C₄:Gd had a lower r1 value (i.e., 2.25 mM⁻¹s⁻¹) than Magnevist. This lower r1 value for X₄₀C₄:Gd is due to the lower number of conjugated DOTA-Gd per molecule of X₅₀C₄, as observed by MALDI-TOF. The lower number of conjugated DOTA-Gd in X₅₀C₄ could be attributed to a high probability of inter-chain disulfide bond formation due to the extended structure of the XTEN molecule, rendering some of the cysteine residues unavailable for DOTA conjugation. In contrast, we postulate that the presence of affibody in the AbX₅₀C₄ sequence may have induced physical distance between the AbX₅₀C₄ molecules, thus lowering the probability of inter-chain disulfide bond formation resulting in a higher number of DOTA-Gd conjugations. Learning that the AbX₅₀C₄ vector has a suitable relaxivity, we designed our in vivo studies to examine the efficacy of the vector in labeling HER2⁺ lung tumors.

DCE-MRI, a T1-weighted scan protocol, combined with a Gd-based contrast agent, is becoming a key part of ovarian cancer diagnosis and therapy in the clinic because of its simplicity, accuracy, and safety [29, 30]. This powerful method can provide information not for only tumor staging [31–33], but also for measuring the tumor response to therapy [34, 35]. Using a targeted approach, it is possible to not only expose the patient to lower doses of Gd, but also personalize the therapy regimen according to the expressed target antigens on the surface of the cancer cells. To examine the application of the AbX₅₀C₄:Gd vector in vivo, we implanted tumors in the lungs of immunocompromised mice. Once the establishment of the tumors in the lungs was confirmed by BLI (Figure 7A), the mouse was injected with AbX₅₀C₄:Gd, X₅₀C₄:Gd, and Magnevist sequentially 48 h apart. The injection dose was normalized to the Gd concentration to ensure that equal amounts of Gd were administered. Overall, the MRI scans showed that the targeted AbX₅₀C₄:Gd accumulated in the SKOV-3 (HER2⁺) lung tumors but not in the OVASC-1 (HER2⁻) tumors (Figure 7B–7C). However, Magnevist was rapidly cleared from the lungs due to its low molecular weight and was not detectable after one hour. The signal associated with the untargeted X₅₀C₄:Gd vector was detectable for almost two hours, although at significantly lower intensities in both the tumor models. These clear differences among the signal intensities of the groups encouraged us to further analyze the MRI scans and generate voxel-based contrast agent concentration maps. This technique allowed us to calculate the approximate Gd organ distribution kinetics [36, 37]. As shown in Figure 8, the SKOV-3 lung tumors treated with AbX₅₀C₄:Gd generated a substantially brighter signal than X₅₀C₄:Gd and Magnevist. Therefore, to examine whether the presence of the affibody in AbX₅₀C₄:Gd could differentiate between the HER2⁺ and HER2⁻ tumors, we compared the concentrations of AbX₅₀C₄:Gd with X₅₀C₄:Gd in both the SKOV-3 and OVASC-1 lung tumors. If accumulation in the tumors was merely due to the large sizes of the vectors and the affibody did not play any role in tumor targeting, we would expect to observe similar Gd concentrations in the SKOV-3 lung tumors for both AbX₅₀C₄:Gd and X₅₀C₄:Gd. However, the statistical analysis of the MRI scans showed significantly higher amounts of Gd in the SKOV-3 lung tumors after treatment with AbX₅₀C₄:Gd in comparison with X₅₀C₄:Gd. In addition, data analysis showed significantly higher signal intensity in the SKOV-3 tumors that were treated with AbX₅₀C₄:Gd than the OVASC-1 tumors. Overall, the data show the ability of the AbX₅₀C₄:Gd vector to differentiate HER2⁺ tumors from HER2⁻ tumors in the lungs.

Quantitative analysis of the DCE-MRI scans also showed that Magnevist was rapidly eliminated through the kidneys leading to a rise in Gd concentration in the kidneys and a sharp decrease in its blood concentration. However, AbX₄₀C₄:Gd and X₅₀C₄:Gd appeared to clear more through the liver with higher blood circulation time.

Conclusions

DCE-MRI is a safe diagnostic method that provides high accuracy and resolution for detecting tumor lesions at not only primary sites, but also metastatic locations. At present, tumor samples are taken from patients and scored using the 0 to 3+ scale for HER2 expression. Since tissue biopsy from lung metastasis is not recommended in many cases, MR imaging could help determine HER2 expression levels based on the average intensity of the signal per voxel. The scoring criteria that would correlate average signal intensity to the HER2 expression levels is the next logical step for our studies. Our data demonstrated that the developed HER2 targeted AbX₄₀C₄:Gd fusion vector can be used to differentiate HER2⁺ lung metastasis from HER2⁻ lesions using the low-powered but widely available 1T MRI. The developed method could potentially be used in conjunction with other imaging modalities to prescreen patients and identify those who are the best candidates for immunotherapy while triaging those who may not be considered responsive. While we used ovarian cancer lung metastasis as a model, this methodology could be used for detecting other types of metastatic HER2⁺ cancers.

Highlights:

• A Gd-labeled HER2-targeted vector was engineered for quantitative DCE-MRI.

• The HER2-targeted vector showed high affinity toward HER2⁺ cancer cells.

• The Gd-labeled vector showed high r1 value of 3.99 s⁻¹mM⁻¹@1T

• A MATLAB-based image analysis code was created for R1 mapping of the MRI scans.

• Using DCE-MRI, the developed vector could differentiate HER2+ from HER2⁻ tumors.