Abstract
In vitro properties of antibody–drug conjugates (ADCs) such as binding, internalization, and cytotoxicity are often well characterized before in vivo studies. Interpretation of in vivo studies might be significantly enhanced by molecular imaging tools. We present here a dual-isotope cryoimaging quantitative autoradiography (CIQA) methodology combined with advanced 3-dimensional imaging and analysis allowing for the simultaneous study of both antibody and payload distribution in tissues of interest in a preclinical setting. Methods: TAK-264, an investigational ADC targeting anti–guanylyl cyclase C (GCC), was synthesized using tritiated monomethyl auristatin E. The tritiated ADC was then conjugated to diethylenetriaminepentaacetic acid, labeled with 111In, and evaluated in vivo in animals bearing GCC-positive and GCC-negative tumors. Results: CIQA revealed the time course of drug release from ADC and its distribution into various tumor regions that are less accessible to the antibody. For GCC-positive tumors, a representative section obtained 96 h after tracer injection showed only 0.8% of the voxels to have colocalized signal, versus over 15% of the voxels for a GCC-negative tumor section, suggesting successful and specific cleaving of the toxin in the GCC-positive lesions. Conclusion: The combination of a veteran established autoradiography technology with advanced image analysis methodologies affords an experimental tool that can support detailed characterization of ADC tumor penetration and pharmacokinetics.
See an invited perspective on this article on page 1459.
Antibody–drug conjugates (ADCs) are targeted pharmaceuticals that comprise a tumor-targeting monoclonal antibody conjugated to a highly potent cytotoxic drug. This approach aims to assess the safety profile of highly potent drugs with limited tolerability (1,2). Conjugation to monoclonal antibodies exploits the specific binding of the antibody as a selective delivery mechanism. This process relies on a series of events including target binding, receptor-mediated cellular internalization, lysosomal processing of the ADC to release active drug, and drug action on cellular machinery (e.g., DNA or microtubule) (3–5). In addition to exerting cytotoxic effects on GCC-positive cells, the processed cytotoxic drug may also diffuse out and damage adjacent tumor cells, resulting in a bystander effect.
The ADC approach causes the payload (active drug) to take on the pharmacokinetic properties of the antibody. As such, ADCs encounter one of the major challenges of antibodies, heterogeneous distribution (6). As ADC technologies have advanced, the ability of a payload to exert bystander effects has been an area of active research. Linker designs aimed to release metabolites that are potent and cell-permeable have been developed with the goal of enabling cytotoxicity in proximal cells and greater efficacy in tumors with heterogeneous antigen expression or heterogeneous penetration of the ADC.
Traditionally, mechanism and efficacy are characterized in vitro for most new ADCs (7–9), with subsequent in vivo characterizations focused on pharmacokinetic and pharmacodynamic profiling, efficacy studies, and gross and destructive measurements of payload concentrations using mass spectrometry ex vivo. These techniques, although informative, do not allow for quantitative visualization of the extent of tumor penetration achieved by the antibody and processed drug.
The use of in vivo molecular imaging modalities such as PET combined with long-lived isotopes such as 89Zr has allowed researchers to noninvasively evaluate parameters such as specific and nonspecific tumor uptake, target engagement, and off-target accumulation in both preclinical and clinical settings (10). PET imaging, however, is limited to a spatial resolution on the order of millimeters, precluding evaluation of intratumoral distribution of antibody and drug. Matrix-assisted laser desorption/ionization mass spectrometry is an emerging methodology for imaging the 3-dimensional (3D) distribution of drugs, peptides, and proteins ex vivo. Despite the high resolution and versatility of the method, it is hampered by the need for arduous optimization of sample preparation and acquisition for different molecules and tissues (11,12) and by its lower sensitivity toward larger proteins. Autoradiography is a longstanding method for evaluating the distribution of target proteins using radiolabeled ligands ex vivo. The main uses of autoradiography have been the study of changes in neurotransmitter receptors in brain tissue (13–15) and the study of absorption, distribution, metabolism, and excretion (16,17).
Cryoimaging quantitative autoradiography (CIQA) is an imaging methodology that involves the autoradiographic analysis of sections throughout a tissue of interest to determine the intratissue distribution of a radiolabeled molecule of interest. The radiolabeling of molecules with isotopes of significantly different half-lives enables dual-isotope imaging, assisting in understanding how the molecules interact with or affect each other.
This paper describes the development of CIQA combined with 3D modeling to evaluate the intratumoral distribution of both antibody and drug by using a dual-labeled ADC (111In-diethylenetriaminepentaacetic acid [DTPA]-3H-TAK-264). This technology opens a new door to understanding how an ADC’s dosing and linker structure together with tumor vasculature and antigen expression affect drug distribution and efficacy.
MATERIALS AND METHODS
Radiochemistry
3H-MMAE (185 MBq, 1 TBq/mmol; Moravek Biochemicals) was conjugated to the fully human anti-GCC monoclonal antibody, 5F9, via a protease-cleavable linker (MMAE and linker technology licensed from Seattle Genetics). The conjugate 3H-TAK-264 was purified and formulated in 10 mM histidine, 7.5% sucrose, 0.08% PS20, pH 5.2, to obtain 70.3 MBq of radiopharmaceutical with a radiochemical purity of 98.9%, specific activity of 3.9 MBq/mg, and a drug-to-antibody ratio of 4.3:1.
3H-TAK-264 (17.6 MBq/4.5 mg) was buffer-exchanged into HEPES buffer (0.1 M, pH 8.5) using an ultrafiltration tube (Amicon) and concentrated to approximately 5 mg/mL. Diethylenetriaminepentaacetic acid (DTPA) dianhydride in dimethyl sulfoxide (2 mg/mL) was added to the protein solution to obtain a DTPA-to-antibody molar ratio of 20:1 and the reaction incubated at 4°C overnight. The reaction was buffer-exchanged into ammonium acetate buffer (0.1 M, pH 6) and concentrated to approximately 5 mg/mL.
A solution of 111InCl3 (8 μL, 300 MBq) was mixed with an equal volume of sodium acetate buffer (0.1 M, pH 6). DTPA-3H-TAK-264 (1.89 mg, 4.5 mg/mL) was added to the reaction mixture and the reaction diluted to 1 mL with additional ammonium acetate buffer (0.1 M, pH 6). The reaction mixture was left at room temperature for 30 min, was stopped by addition of a 10% (v/v) solution of ammonium acetate buffer (0.1 M, pH 6) containing 50 mM ethylenediaminetetraacetic acid, and was incubated for a further 5 min at room temperature. The labeling solution was formulated by dilution with saline to a final volume of 2 mL. The final formulation was found have a radiochemical purity of 96.5% and specific activities of 296 MBq (111In) and 7.4 MBq (3H) per 1.89 mg of TAK-264 in 2 mL.
111In-DTPA-5F9 and 111In-DTPA-TAK-264 were prepared from 5F9 and TAK-264 using similar strategies for DTPA conjugation and 111In labeling as described for the dual-labeled TAK-264. 111In-DTPA-5F9 and 111In-DTPA-TAK-264 were obtained in radiochemical purities of 98% and 93% and specific activities of 2.46 and 2.51 MBq/μg, respectively.
The affinities of the DTPA-conjugated antibodies to GCC was compared with that of the parent antibodies using enzyme-linked immunosorbent assay and was found to be similar (data not shown).
Animal Model
All animal studies were performed at the University of Massachusetts and at Invicro with the approval of their respective Institutional Animal Care and Use Committees. The animals were kept at a temperature of 18°C–26°C and a relative humidity of 50% ± 20%, with intermittent light and dark cycles of 12 h and with food and water available ad libitum. Female SCID mice, 5–8 wk old, were implanted with HEK-293 GCC2 or HEK-293 cells (5 × 106 cells in 100% Matrigel [Corning]) subcutaneously. Tumor xenografts were allowed to grow to 200–1,500 mm3 for imaging studies and 180–520 mm3 for CIQA studies.
SPECT/CT Imaging
The animals were imaged on a NanoSPECT/CT system (Bioscan, Inc.) using a 9-pinhole aperture (1.4-mm diameter) at 2 and 24 h after injection (20-min scans, each with 50 s/projection), at 48 and 96 h after injection (24-min scans, each with 60 s/projection), and at 144 h after injection (a 24- to 32-min scan with 60–80 s/projection). Scanning was performed with dual energy windows (154–188 and 221–270 keV). Each SPECT scan was followed by a CT scan for anatomic reference (180 projections per rotation and 0.5 s/projection). Anesthesia during imaging was induced with 4%–5% isoflurane and maintained with 1%–2% isoflurane (0.5–1 L/min).
Animals bearing HEK-293 GCC2 tumors (n = 8) were anesthetized, injected with either 111In-DTPA-5F9 (19.8 ± 0.55 MBq, n = 4) or 111In-TAK-264 (19.4 ± 0.85 MBq, n = 4), and imaged at the time points described above to evaluate biodistribution profiles.
Biodistribution and Autoradiography
Tumor-bearing animals (HEK-293 [n = 8] and HEK-293 GCC2 [n = 8]) were anesthetized and injected with 15.5 ± 0.6 and 14.1 ± 1.4 MBq, respectively, of 111In-DTPA-3H-TAK-264. Two animals from each group were euthanized at each time point (1, 8, 24 and 96 h after injection), and their tumors were resected. On resection, the tumors were immediately blocked in 5% carboxymethyl-cellulose on crushed dry ice. 14C-2-doxyglucose–spiked India ink fiducial markers were inserted into the blocks, which were then stored in a −20°C freezer. High-resolution optical images of each 30-μm section were acquired using a Canon EOS 70D camera, and each tenth section was stored on 3M tape (Scotch Label Protection Tape 821), dehydrated, and exposed to phosphor image screens (BAS-SR; GE Healthcare) for 2–3 h, which were scanned using a BAS5000 image acquisition system to determine 111In content.
3H signal was overshadowed by 111In signal and by the signal of the long-lived radioactive contaminants in the 111InCl3 used for labeling. 111In signal was removed by exploiting the fact that the physical decay of 111In is faster than that of 3H. In this study, 3H signal was acquired after a delay of more than 10 111In half-lives. Due to the presence of long-lived contaminants following 111In decay, the 3H signal was extrapolated from the total signal as described in the “Image Analysis” section.
Image Analysis
SPECT/CT
SPECT images were reconstructed using a scanner manufacturer–provided multipinhole ordered-subsets expectation maximization algorithm. Regions of interest were delineated and quantified by manual fitting of constant volumes of interest to the heart (243 mm3), liver (143 mm3), spleen (75 mm3), left and right kidneys (500 mm3), and muscle (50 mm3) using VivoQuant (Invicro). Tumor regions of interest (130–1,600 mm3) were drawn manually and maximized to the size of the lesion. Nontumor volumes of interest were fitted on the basis of the CT data and centered on the peak-activity profile.
Autoradiography
Autoradiography image analysis was performed using VivoQuant. Individual white-light images were aligned into a 3D dataset using a fiducial marker–based registration approach. First, the 14C-ink fiducial centers within each slice were detected by applying a normalized cross correlation with a disk image. Then, a rigid-body transformation was estimated to align these images to a reference image. Least-squares distance error was used as the optimization parameter for estimating the registration parameters. After registration, variation in intensity throughout the image stack was corrected by a histogram-based color-matching algorithm (18) to complete the final 3D white-light dataset.
111In and 3H autoradiography images were intensity-calibrated using a set of standards included in each plate. Linear regression yielded a plate-specific calibration parameter to convert intensity values into units of kBq/g. Reconstruction of 111In tumor 3D volumes started by splitting the calibrated plates containing multiple tumor section into individual ones. Next, the 14C-ink fiducial markers were detected using the same algorithm as used for the white-light images. The fiducial markers were used to coregister the plates into a 2-dimensional (2D) stack. Because autoradiography images were captured every 10 white-light slices, linear interpolation was performed to produce the final 3D autoradiography dataset with isotropic voxel size. Fiducial centers from the white-light stack and the 111In stack were used to register the white-light volume to its corresponding 111In volume using a Procrustes transformation that allows translation, rotation, and scaling. Both volumes were resampled to 0.025 × 0.025 × 0.025 mm isotropic voxels.
The original strategy for dual-isotope imaging was to quantify the higher-energy and shorter-lived 111In signal, allowing for sufficient 111In decay (30 d, 11 half-lives), followed by quantification of the 3H signal. The presence of small quantities (<0.1%) of long-lived, γ-emitting isotope contaminants (114mIn and 65Zn with half-lives of 50 and 244 d, respectively) required an alternate quantification strategy. After allowing 2 mo for 111In decay to the background level, the tissues were quantified twice for radioactivity signal. First, the total radioactive content (γ-emitting contaminants plus tritium) was quantified by exposing tissues to standard phosphor imaging screens. Exposures were repeated with the screens covered in foil to block the tritium signal, effectively quantifying contaminant signal only. The relative intensity of the 111In standards (containing only contaminants; no measurable 111In signal and no 3H signal) in images on phosphor imaging screens with and without aluminum foil was used to estimate the contribution of the 111In contaminants to 3H signal in tumors. Images of the standards acquired with and without foil were corrected for differences in exposure time, and a correction factor based on the relative intensity of the standards was calculated. This correction factor was used to scale the foil-covered (contaminant) image. After the 2 plates had been registered, a smoothing step (2D gaussian kernel with a full width at half maximum of 0.150 mm) was applied to the scaled, foil-covered image to preserve structure in the contaminant image while reducing the impact of noise. The scaled, smoothed contaminant image was subtracted from the image without foil to remove the signal contribution from the contaminants and yield the tritium tissue signal. To improve voxel-level comparison between the 3H and 111In autoradioluminograms, a second smoothing step (3D gaussian kernel with a full width half maximum of 0.100 mm) was applied to the 3H images.
RESULTS
Evaluation of Intratumoral Accumulation of 111In-DTPA-5F9 and 111In-DTPA-TAK-264 Using SPECT/CT
On SPECT/CT images, 111In-DTPA-5F9 showed a continuous accumulation in tumors, averaging 16–18 %ID/g at 144 h (Supplemental Figs. 1 and 2; supplemental materials are available at http://jnm.snmjournals.org). 111In-TAK-264 peak tumor uptake was reached at 96 h after injection (11–13 %ID/g) and remained constant at 144 h after injection (Fig. 1). The differences in tumor uptake between 111In-DTPA-5F9 and 111In-TAK-264 were statistically significant at the first and last time points (2 h and 144 h, with P = 0.015 and P = 0.007, respectively). Significant differences in plasma clearance (calculation based on the heart region of interest, which approximates blood pool radioactivity) were apparent (P = 0.03).
Distribution of 111In-DTPA-5F9 (left) and 111In-DTPA-TAK-264 (right) in mice bearing HEK-293 GCC2 subcutaneous tumors. Data are mean ± SD (n = 3–4 for all time points).
On the basis of these data, the CIQA experiment was conducted on tumors resected at 1, 8, 24 and 96 h after injection using the HEK-293 GCC2 tumor model with the HEK-293 tumor as the GCC-negative control.
Evaluation of Intratumoral Distribution of 5F9 and Its MMAE Payload Using CIQA Plus 3D Image Reconstruction
Figure 2 presents representative autoradiography sections. Maximum-intensity projections were used to visualize tumors in 3 dimensions. Supplemental Fig. 3 is a diagram of the acquisition and reconstruction process, and Supplemental Video 1 shows the 3D reconstructed maximum-intensity projections of GCC-positive and -negative tumors at 96 h after injection.
Representative HEK-293 GCC2 tumor sections showing subtraction technique used to compensate for presence of long-lived contaminants in 111In-labeled ADC. Voxels are 0.025 × 0.025 × 0.025 mm. Images are displayed for qualitative review in units of photostimulated luminescence and include white-light image (A) and images showing 111In signal (B), total (3H + contaminants) signal (C), contaminant-only signal (D), and isolated 3H signal (E).
Demonstration of Differential Distribution of MMAE in GCC-Negative and -Positive Tumors on CIQA
Representative tumor sections at 1, 24, and 96 h after injection are shown in Figure 3. At 1 and 8 h (data not shown) after injection, 3H and 111In signals were very colocalized in both GCC-positive and GCC-negative tumors (Fig. 3A). At 24 h, 3H and 111In signals were still very colocalized in the GCC-negative tumors but were starting to diverge in the GCC-positive tumors (Fig. 3B). By 96 h, 3H and 111In showed a visible drop in signal intensity in the GCC-negative tumors and exhibited some divergence (Fig. 3C). In the GCC-positive tumors, the 3H and 111In signals were more divergent at 96 h than at 24 h, with the 3H signal penetrating more deeply into the tumor (Fig. 3C). Both tumor types demonstrated significant heterogeneity in the distribution of the antibody signal (111In), as expected for compounds of high molecular weight (19,20).
Representative sections of HEK-293 GCC2 tumors (top) and HEK-293 tumors (bottom) excised at 1 h (A), 24 h (B), and 96 h (C) after tracer injection. Voxels are 0.025 × 0.025 × 0.025 mm. 3H signal is red, 111In signal is green, and both signals coregistered is yellow. Strikingly, image of HEK-293 GCC2 at 24 h shows initial diffusion of drug away from antibody accumulation site and deeper into tumor.
The 111In signal in HEK-293 GCC2 tumors steadily increased over the 96-h period, whereas the 111In signal in HEK-293 tumors decreased as the nonspecifically bound tracer was slowly washed out of the tumor (Fig. 4). A similar trend was clearly visible in the MMAE-affiliated 3H signal, with the main difference being that the tritium signal spread into the center of the HEK-293 GCC2 tumors whereas the 111In signal remained localized at the tumor edges.
Distribution of voxel %ID/g for 3H-HEK-293 GCC2 (A), 111In-HEK-293 GCC2 (B), 3H-HEK-293 (C), and 111In-HEK-293 (D) at 1, 8, 24, and 96 h. Each group contains 2 animals; hence, there are 2 lines per color.
Demonstration of Bystander Effect via Colocalization Analysis of 3H and 111In in GCC-Positive Tumors
Differences in background and photon scatter characteristics in autoradioluminograms of the 2 isotopes can affect assessment of colocalization of the 111In and 3H signals (21). Analysis of signal overlap focused on, first, differences in overall signal intensity between tumor lines and, second, the extent of colocalization of the 111In and 3H signals in the highest-intensity voxels for each tumor line. The fraction of the highest-intensity voxels that are common to both isotopes provides a measure of colocalization. Figure 5 shows a 2D scatterplot of the threshold-constrained voxels for tumors from each cell line at 96 h. The thresholds in the HEK-293 GCC2 tumors were 2- to 3-fold higher than those in the HEK-293 tumors for both 111In and 3H. Additionally, these high-intensity voxels were almost completely disparate between the 2 isotopes for the HEK-293 GCC2 tumors (0.8% overlap) whereas a colocalization fraction of 15.5% was observed between the highest-intensity voxels in the HEK-293 tumors (Supplemental Fig. 4 shows the same data without the thresholding). Taken together, these data demonstrate greater uptake of TAK-264 in HEK-293 GCC2 tumors than in GCC-negative HEK-293 tumors. In addition, the disparate signal observed between the 2 isotopes in the GCC-positive tumors is indicative of binding, internalization, metabolism, and release of MMAE.
Colocalization analysis of top 3% of highest-intensity voxels for GCC-positive tumors (left) and GCC-negative tumors (right). Voxels with high 3H signal (MMAE), high 111In signal (antibody), or both are shown.
DISCUSSION
Preliminary SPECT/CT imaging studies were performed on 111In-labeled parent antibody and ADC to examine the tissue distribution of the compounds over time. With the relevant distribution information at hand, the ADC was dual-labeled with both 111In (antibody) and 3H (MMAE) and CIQA was performed in both HEK-293 GCC2 (GCC-positive) and HEK-293 (GCC-negative) tumor-bearing animals at different times after injection.
Analysis of the highest-intensity voxels allowed us to investigate the colocalization of the 2 signals. For GCC-positive tumors, a representative section obtained at 96 h after tracer injection showed that only 0.8% of the voxels had colocalized signal, versus over 15% of the voxels for a GCC-negative tumor section. These results are in line with our expectation that GCC-positive tumor cells would internalize and metabolize the ADC, allowing the free drug to diffuse across the tumor and affect untargeted cells in what is known as the bystander effect (22). In GCC-negative tumors, there is no target-mediated internalization; thus, a high percentage of colocalized signal is expected. Macropinocytosis has been shown to occur in response to growth factor stimulation and constitutively in some cell types and may cause the decoupling of 111In and 3H signals observed in the GCC-negative tumors.
The indium contaminants made the CIQA process more difficult, time-consuming, and expensive than desired. A good companion isotope for 3H/14C would have a half-life of several days (matching well the biologic half-lives of antibodies yet enabling a reasonable decay period). The isotope of choice should residualize in tumors to minimize the redistribution of metabolites, emit low-energy γ-energies to retain autoradiography resolution, have a high specific activity, and have short-lived contaminants. 67Ga is a residualizing isotope with established radiochemistry (23–25), a half-life of 3.3 d, specific activity of 148 TBq/mmol, low-energy γ-emissions, and a minimal presence (<0.6%) of short-lived contaminants (66Ga and 68Ga, with half-lives of 9.5 h and 68 min, respectively). The use of 67Ga as a surrogate for 111In has been validated internally to ensure that future CIQA studies will be more streamlined and cost-effective.
To avoid potential translocation and loss of the MMAE due to the tissue handling, the animals in this study were not perfused before tumor harvesting (26). In the interpretation of the results, it is important to consider that the tumor signal includes contributions from both tumor tissue and blood pool.
Despite some limitations such as the use of radioactive compounds, an inability to distinguish between the different radioactive metabolites, and the need to allow extended periods for radioisotope decay, CIQA is well positioned when compared with imaging methodologies of similar scope. Fluorescence imaging is an inexpensive and straightforward methodology that allows for depth-limited in vivo imaging combined with superior ex vivo resolution. However, ADC conjugation to fluorescent dyes affects pharmacokinetics significantly, even in low conjugate-to-antibody ratios (27), and the dyes themselves may undergo quenching after cellular uptake, making analysis challenging. Mass spectrometry imaging is an attractive modality but has a low sensitivity for detecting large proteins over a tissue background, and because different metabolites will have different masses and may be conjugated to DNA or proteins, the signal may become more diluted over time and specific standards would be required for the expected metabolites. A final possibility is immunohistochemical identification of both drug and antibody, but this methodology requires specialized antibody pairs that may not have affinity to the formed metabolites and may produce signals that are more difficult to quantify.
CONCLUSION
CIQA is a unique and powerful method to evaluate the distribution and efficacy of ADCs in vivo. CIQA allows the evaluation of ADC metabolism in tumors of varying antigen expression and organs (e.g., liver) known to have off-target toxicities. The results of this study are a proof of concept demonstrating that the CIQA technology can help explain the spatiotemporal dynamics of different antibodies, linkers, and drug conjugates.
DISCLOSURE
Ohad Ilovich, Mohammed Qutaish, Jacob Hesterman, Kelly Orcutt, Jack Hoppin, Ildiko Polyak, and Marc Seaman were employed by Invicro LLC at the time the experiments were conducted. Adnan Abu-Yousif, Donna Cvet, and Daniel Bradley were employed by Takeda Pharmaceuticals International at the time the experiments were conducted. No other potential conflict of interest relevant to this article was reported.
Acknowledgments
We are extremely grateful to Vijay Gottumukkala, Paige Czarnecki, Mihaela Plesescu, Ozlem Yardibi, Rick Coelho, Merryl Lobo, and Chris Graul for the execution and operationalizing of the study and the highly useful advice they shared.
Footnotes
Published online May 4, 2018.
- © 2018 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- Received for publication January 3, 2018.
- Accepted for publication April 19, 2018.