Engineered antibodies: new possibilities for brain PET?
Almost 50 million people worldwide are affected by Alzheimer’s disease (AD), the most common neurodegenerative disorder. Development of disease-modifying therapies would benefit from reliable, non-invasive positron emission tomography (PET) biomarkers for early diagnosis, monitoring of disease progression, and assessment of therapeutic effects. Traditionally, PET ligands have been based on small molecules that, with the right properties, can penetrate the blood–brain barrier (BBB) and visualize targets in the brain. Recently a new class of PET ligands based on antibodies have emerged, mainly in applications related to cancer. While antibodies have advantages such as high specificity and affinity, their passage across the BBB is limited. Thus, to be used as brain PET ligands, antibodies need to be modified for active transport into the brain. Here, we review the development of radioligands based on antibodies for visualization of intrabrain targets. We focus on antibodies modified into a bispecific format, with the capacity to undergo transferrin receptor 1 (TfR1)-mediated transcytosis to enter the brain and access pathological proteins, e.g. amyloid-beta. A number of such antibody ligands have been developed, displaying differences in brain uptake, pharmacokinetics, and ability to bind and visualize the target in the brain of transgenic mice. Potential pathological changes related to neurodegeneration, e.g. misfolded proteins and neuroinflammation, are suggested as future targets for this novel type of radioligand. Challenges are also discussed, such as the temporal match of radionuclide half-life with the ligand’s pharmacokinetic profile and translation to human use. In conclusion, brain PET imaging using bispecific antibodies, modified for receptor-mediated transcytosis across the BBB, is a promising method for specifically visualizing molecules in the brain that are difficult to target with traditional small molecule ligands.
KeywordsTransferrin receptor 1 (TfR1)-mediated transcytosis Alzheimer’s disease (AD) Amyloid-β (Aβ) Antibody Blood–brain barrier (BBB) Positron emission tomography (PET)
Positron emission tomography (PET) is a non-invasive, quantitative, functional imaging method. Clinically, PET is used to aid diagnosis, especially in cancer, where the radioactive glucose analogue [18F]FDG is used to localize primary tumours and metastases. PET has also become an important tool for diagnosis of brain disorders, since naturally it is difficult to obtain biosamples from the brain. Further, PET is an attractive method in translational research and drug development, as the same experiments can be performed in vivo in both animals and humans, and it allows for repeated investigations in one subject.
The main hurdle for the delivery of drugs (and radioligands) to the brain, irrespective of their size, is the blood–brain barrier (BBB), comprising tightly connected endothelial cells. Traditionally, PET radioligands for the central nervous system (CNS) have been based on small “drug-like” molecules preferably labelled with clinically compatible positron-emitting radionuclides such as carbon-11 (11C) or fluorine-18 (18F). Radioligands for brain imaging have to be fairly lipophilic to be able to pass through the BBB into the brain parenchyma. Unfortunately, increased lipophilicity also increases nonspecific distribution into the lipophilic brain tissue. This may lead to a poor specific-to-nonspecific PET signal. Further, and especially relevant in proteopathies such as Alzheimer’s disease (AD) and Parkinson’s disease (PD), it is unlikely that small-molecule radioligands could discriminate between different aggregation forms of a protein or proteins with similar fibrillary structures. Thus, in line with the shift in therapeutic focus from small-molecule drugs to biologics, antibodies or fragments thereof could turn out to be a completely novel class of neuroPET radioligands and could be used for highly specific PET imaging in the CNS, including imaging of target proteins for which radioligands are lacking today.
Antibody transport across the blood–brain barrier
Radioligands based on antibodies or other proteins have already been introduced for peripheral targets related to cancer diagnostics and theranostics, including some applications in clinical use as well [1, 2]. However, antibodies are large molecules, displaying highly restrictive BBB transcytosis. It has been reported that only 0.1% of peripherally administered antibody reaches the brain [3, 4], and it has even been questioned whether antibodies penetrate the brain parenchyma at all, or whether antibody concentrations measured in the brain rather reflect transport from the blood into the cerebrospinal fluid (CSF) . Thus, antibodies and other proteins will most likely have to be specifically engineered for facilitated transport across the BBB to enable their use as PET radioligands within the CNS.
The transferrin receptor
The TfR is found in two different isoforms: TfR1 and TfR2. TfR1 (also known as cluster of differentiation 71, or CD71) is expressed in brain endothelial cells, hepatocytes, and erythrocyte precursors, especially in bone marrow, lung, and other rapidly dividing cells. TfR2 is only expressed in hepatocytes, enterocytes of the small intestine, and erythroid cells. Both forms are expressed as transmembrane glycoproteins composed of two disulfide-linked monomers joined by two disulfide bonds. Each monomer binds one holo-transferrin molecule, creating an iron-Tf-TfR complex which enters the cell by endocytosis. In the endosome, the lower pH of around 5.5 will cause Tf to release its iron ions, which can subsequently be used by the cell. The TfR-Tf complex will then be recycled to the cell surface.
A number of antibodies have been generated against TfR1, among them the monoclonal mouse antibody OX-26  that binds to the rat TfR1, and rat antibody 8D3  that binds to the murine TfR1. Initially, OX-26 and 8D3 were developed for immunohistochemical visualisation of brain capillaries. However, it was later discovered that the antibodies were also found in high concentrations in the brain after in vivo systemic administration, and further, that when fused to a protein cargo, they were able to carry its cargo across the BBB [15, 17, 18].
despite binding to TfR1 . It is debated which factors govern the ability to induce TfR1 transcytosis. Some studies have suggested that low/moderate affinity promotes transcytosis, while high affinity reduces release from the receptor and leads to lysosomal degradation . Other studies have suggested that the binding valency to the TfR1 is of importance, monovalent binding being more efficient than bivalent binding, which may cause TfR clustering at the cell surface and intracellular degradation . These theories are not necessarily conflicting, as monovalent binding in general results in lower affinity (avidity). Yet another feature that may govern transcytosis efficiency is pH-dependent affinity to the receptor. Ideally, the antibody should have moderate TfR affinity at neutral pH for optimal engagement at the cell surface, and low affinity at low pH for release in the acidic endosome and further transport across the cell, into the brain. While it has not been systematically studied and reported, the location of the TfR binding epitope may also be of importance for an antibody’s ability to undergo TfR-mediated transcytosis.
Alzheimer’s disease and amyloid-β
Antibody-based radioligands for imaging of Aβ
Antibody vs PIB PET
Additional evidence for the sensitivity of antibody PET is provided by a study where 10-month-old tg-ArcSwe mice were treated for 3 months with a BACE-1 inhibitor to decrease Aβ production. PET imaging with the recombinant [124I]RmAb158-scfv8D3 readily detected reduced Aβ levels in treated compared to non-treated animals, at an age when [11C]PIB can hardly detect Aβ .
Other pathological changes in need of novel radioligands
Similar to AD, protein misfolding and aggregation is also a pathological feature in PD. In PD, the presynaptic protein α-syn initially forms oligomers and later insoluble aggregates. There are presently no PET ligands available for imaging of either soluble or insoluble α-syn, and several research programmes are aimed at developing small-molecule PET ligands for α-syn . The lack of in vivo biomarkers for α-syn is a major limitation for the development of disease-modifying treatments for PD, and the Michael J. Fox Foundation (MJFF) recently announced a $2 million prize to the first team to develop a viable selective α-syn PET radioligand. A number of initiatives to develop a radioligand for α-syn are ongoing, and most of these are based on small molecules. Although some promising compounds have been presented, one major hurdle is the cross-reactivity and binding to Aβ. Even if the affinity of a specific compound is much higher for α-syn than Aβ, the abundance and availability of Aβ may be higher. Thus, it might be difficult to differentiate between an α-syn-containing brain and an Aβ-containing brain, even if Aβ levels are low. This may be especially a problem when different protein pathologies coexist, which is often the case in older individuals. In this respect, highly specific antibodies for α-syn may be a new option for developing radioligands truly specific for α-syn. One potential challenge is that the majority of aggregated α-syn in the brain appears to be intracellular, and thus an additional barrier has to be conquered by the bispecific antibody in order to reach its primary target.
Microglia and astrocytes
PET imaging of activated microglia and reactive astrocytes has been used as an indication of neuroinflammation. For example, a number of PET radioligands have been developed for the 18-kDa translocator protein (TSPO), which is highly expressed on activated microglia. However, quantitative interpretation of the PET signal with the second- and third-generation TSPO PET radioligands is confounded by large interindividual variability in binding affinity due to a genetic polymorphism leading to a trimodal distribution, reflecting high-affinity binders (HABs), low-affinity binders (LABs), and mixed-affinity binders (MABs) . In addition, TSPO is also expressed on astrocytes, and hence the TSPO ligands are not specific for microglia. Reactive astrocytes, also indicative of neuroinflammation, can be imaged with the PET radioligand deuterium-L-depreneyl ([11C]DED). [11C]DED binds to monoamine oxidase-B, primarily found in activated astrocytes, and although studies indicate that the radioligand indeed visualizes astrocytosis, its binding differs from that of other astrocytic markers often used in immunohistochemical analysis in post-mortem neuropathological studies, such as glial fibrillary acidic protein (GFAP) [48, 49]. Attempts have been made to image GFAP with antibody fragments , and myriad well-characterized antibodies for proteins expressed on microglia and astrocytes have been described in the literature. Hence, the possibility of engineering these into bispecific brain-penetrating radioligands is tempting and could potentially allow for “in vivo immunohistochemistry” of classical ex vivo immunohistochemical targets.
TfR transport capacity
One central paradigm of PET is the use of tracer doses that do not elicit a pharmacological response or occupy a significant fraction of potential binding sites. The use of doses above true tracer doses may have an impact on the PET signal, and hence the interpretation of the study. It is therefore important to estimate the capacity of the TfR. For example, a study using [125I]RmAb158-scfv8D3 showed that, compared with unmodified [125I]RmAb158, the transport of bispecific radioligand into the brain was increased almost 100-fold at tracer doses (0.05 mg/kg), while a tenfold increase was observed at a dose of 10 mg/kg . The blood pharmacokinetics were linear, i.e. the half-life in blood was the same for the tracer and the pharmacological dose, and did therefore not contribute to the changed BBB transcytosis efficacy. Moreover, the study showed that co-administration of full-sized 8D3, also at a dose of 10 mg/kg, even more efficiently inhibited transcytosis by reducing the brain uptake of [125I]RmAb158-scfv8D3 to threefold more than [125I]RmAb158. The lower inhibition capacities of RmAb158-scfv8D3 compared with 8D3 can most likely be explained by the former compound’s monovalent TfR binding.
Protein- and antibody-based PET radioligands have traditionally been radiolabelled with radionuclides such as iodine-124 (124I; half-life 4.2 days), zirconium-89 (89Zr; half-life 3.3 days), or gallium-68 (68Ga; half-life 68 min). Except for 124I, these radionuclides require the attachment of a chelator on the antibody/protein backbone before the introduction of the radionuclide. The instability of the chelators was initially a major challenge, but new and more stable versions that may be better suited for clinical use have been introduced . On the other hand, 124I has not been a preferred alternative due to its accumulation in the thyroid, resulting in high local exposure. Further, the above-mentioned radionuclides are not ideal for clinical use due to low fraction of positron decay (26% for 124I and 23% for 89Zr—only this fraction of radioactivity is detected by PET). Also, the high energy of 124I positrons, which allows them to travel a longer distance in the tissue before electron annihilation, causes low-resolution PET images. In contrast, the low-energy positrons from 18F decay generate PET images with high resolution.
For small-molecule radioligands used clinically, fluorine-18 (18F; half-life 110 min) is the first choice for radiolabelling. However, radiochemistry methods for introducing 18F on antibody-based ligands have been lacking. With the increased interest in protein-based radioligands, increasing efforts have also been observed when it comes to 18F radiolabelling. Different strategies involving” click-chemistry”, e.g. Diels-Alder reaction, have been described ; the radiolabelling is somewhat similar to the above-mentioned chelator methods, i.e. a two-step process. First, the antibody/protein is modified by introducing a small-molecule group that in a second step can react with another small molecule that carries the 18F. Some methods for direct radiolabelling of amino acids, e.g. amine groups, with 18F have also been described [53, 54].
The radioactivity measured with PET in the brain (or any tissue of interest) comprises radioactivity in the tissue itself as well as radioactivity in the blood pool of the tissue. For example, around 3% of the brain volume is blood [55, 56]. Thus, PET radioligands should be cleared from the blood fairly rapidly to minimize the radioactivity contribution from the blood pool of the tissue. This is especially important for the brain and for neuroPET radioligands with limited brain distribution. Full-sized antibodies are generally associated with long systemic half-life, while a more rapid elimination can be achieved by modifying antibodies into fragments (Fab, F(ab’)2, scFv). However, even fragments may display half-lives that are not compatible with clinically preferred radionuclides 11C and 18F. In addition, specific and nonspecific binding to peripheral targets may contribute to the observed half-life. For example, binding to soluble circulating and erythrocyte-expressed TfR1 may either prolong or shorten the circulation time. Although there are limited published data, it appears that smaller size and lower TfR1 affinity leads to faster elimination from blood [40, 41]. Hence, studying the systemic pharmacokinetics of antibody-based ligands is essential before deciding on a labelling strategy. Moreover, to achieve a high signal-to-noise ratio, unbound ligand must also be eliminated from the brain within a time frame that matches the half-life of the radionuclide. Knowledge about clearance rates and routes of bispecific antibodies from the brain is sparse, and more research will be required to elucidate whether brain clearance is passive or mediated by active transport mechanisms, and whether it is size-dependent.
Translation from preclinical imaging to clinical imaging is challenging from many perspectives. In addition to dosimetry and pharmacokinetics that may differ between different species, the actual target under investigation may also be different although its biological “purpose” may be the same in different species. In PET, species differences have been observed for both neuroreceptors and transporters at the BBB [57, 58]. The differences include different density of the target, different function and different amino acid composition. Limited knowledge about species-related differences in BBB receptors that mediate transcytosis, e.g. the TfR1 and the insulin receptor, makes it difficult to predict the brain delivery in humans based on preclinical data. The murine TfR1 and the human TfR1 differ in the amino acid sequence at the domain where the 8D3 antibody binds, resulting in no binding of 8D3 to the human TfR. However, antibodies binding human TfR1 at the same domain as 8D3 have been generated and shown to successfully shuttle biologics across the in vitro human BBB and in vivo in monkeys [59, 60]. Attempts to generate species-independent TfR1 antibodies, which would aid in translation, have been described, but with limited success. One exception is perhaps the generation of variable new antigen receptors (VNARs) that appear to bind both mTfR1 and hTfR1 . Published data on the capacity of the VNARs to shuttle antibody cargos across the BBB is limited, however, and it appears that the VNARS may be less efficient for BBB transport of full-sized antibodies compared with smaller fragments. This is not the case for 8D3. In summary, it appears that TfR1 can be used in both mice and men, but may require species-specific TfR1 binders. The expression of TfR1 at the murine and human BBB has been reported to be similar .
The development of antibody-based PET ligands for brain disorders has so far not been feasible, as techniques to facilitate large protein delivery to the brain have been lacking. Still, antibodies have many benefits, particularly their ability to bind specifically to their target, which is advantageous for obtaining PET images without background noise caused by nonspecific binding. Thus, the development of protein engineering strategies to increase antibody concentrations in the brain may enable a completely new class of radioligands with no or very low nonspecific binding.
The high affinity and specificity of antibodies are attractive, especially when specific aggregation forms of a protein are of interest. This is the case for many misfolded proteins, e.g. Aβ and α-syn, for which protein aggregates representing intermediate stages in the aggregation pathway is believed to be more toxic and dynamic than the insoluble end state deposits. In addition, numerous antibodies have been described for inflammation markers on astrocytes and microglia. Generation of bispecific antibodies able to pass the BBB based on these well-characterized and frequently used antibodies for immunohistochemical analyses could likely lead to novel in vivo imaging biomarkers.
Although already shown to be successful in the preclinical setting, the translation into clinical use will require the development of new radiochemistry for incorporation of more clinically suitable radionuclides such as 18F. A number of radiochemical strategies have been described, but large-scale synthesis and reproducibility has to be improved. However, labelling antibodies with 18F, which has a half-life of less than 2 h, will also require fast clearance of the ligand from blood and of unbound ligand from the brain. Further research on the pharmacokinetics of bispecific antibody ligands of different formats is needed to optimize these parameters for clinical development.
Another challenge is finding species-independent BBB shuttles, or validated shuttles in higher species. This has to some extent been accomplished for TfR1, although no species-independent binders have been proven to be as efficient as current mTfR1 binders. Several large-scale projects aiming to discover novel shuttles beyond TfR binders are ongoing.
In conclusion, we have already entered the era of biologics, both for the periphery and for the CNS, and it is likely that antibody- or protein-based radioligands will also become an important class of PET radioligands. A number of preclinical studies have shown the feasibility of antibody-based imaging of Aβ pathology in the brain, and although some hurdles remain, this novel class of tracers is likely to enter clinical development within the next few years.
We would like to thank Tobias Gustavsson for the generation of the autoradiography and immunohistochemistry images shown in Fig. 3. This work was supported by grants from the Swedish Research Council (2017-02413, 2018-02715), Swedish Innovation Agency, the Swedish Alzheimer's Foundation and the Swedish Brain Foundation.
Faculty of the Multimodal Imaging in Neurodegeneration Cologne (MINC) symposium
Bénédicte Ballanger, Lyon Neuroscience Research Center, Lyon, France, EU
Henryk Barthel, University Hospital Leipzig, University of Leipzig, Germany, EU
Gérard N Bischof, University Hospital Cologne, University of Cologne, Germany, EU
Delphine Boche, University of Southampton, Southampton, United Kingdom, EU
Hennig Boecker, German Center for Neurodegenerative Disease (DZNE), Bonn Germany, EU
Karl Peter Bohn, University Hospital Cologne, University of Cologne, Germany, EU
Per Borghammer, Aarhus University, Denmark, EU
Donna Cross, University of Utah, United States of America
Donato Di Monte, German Center for Neurodegenerative Disease (DZNE), Bonn, Germany
Alexander Drzezga, University Hospital Cologne, University of Cologne, Germany, EU, & Research Center Jülich, Germany, EU
Heike Endepols, University Hospital Cologne, University of Cologne, Germany, EU
Kathrin Giehl, University Hospital Cologne, University of Cologne, Germany, EU
Michel Goedert, Medical Research Council, Laboratory of Molecular Biology, Cambridge, United Kingdom, EU
Jochen Hammes, University Hospital Cologne, University of Cologne, Germany, EU
Oskar Hansson, Lund University, Sweden, EU
Karl Herholz, The University of Manchester, Manchester, United Kingdom, EU
Günter Höglinger, German Center for Neurodegenerative Disease (DZNE), Munich, Germany, EU
Merle Hönig, University Hospital Cologne, University of Cologne, Germany, EU
Frank Jessen, University Hospital Cologne, University of Cologne, Germany, EU
Thomas Klockgether, German Center for Neurodegenerative Disease (DZNE), Bonn Germany, EU
Pierre Lafaye, Institut Pasteur, Paris, France, EU
Adriaan Lammerstma, Amsterdam University Medical Center, Amsterdam, Netherlands, EU
Eckhard Mandelkow, German Center for Neurodegenerative Disease (DZNE), Bonn Germany, EU
Eva-Maria Mandelkow, German Center for Neurodegenerative Disease (DZNE), Bonn Germany, EU
Andreas Maurer, University Hospital Tübingen, Germany, EU
Brit Mollenhauer, University Medical Center Göttingen, Germany, EU
Bernd Neumaier, Research Center Jülich, Germany, EU
Agneta Nordberg, Karolinska Institute, Stockholm, Sweden, EU
Özgur Onur, University Hospital Cologne, University of Cologne, Germany, EU
Kathrin Reetz, University Hospital Aachen, Germany, EU
Elena Rodriguez-Vietez, Karolinska Institute, Stockholm, Sweden, EU
Axel Rominger, University of Bern, Switzerland
James Rowe, University of Cambridge, Cambridge, United Kingdom, EU & Medical Research Council Cognition and Brain Sciences Unit, Cambridge, United Kingdom, EU
Osama Sabri, University Hospital Leipzig, University of Leipzig, Germany, EU
Anja Schneider, German Center for Neurodegenerative Disease (DZNE), Bonn, Germany, EU
Antonio Strafella, University of Toronto & Toronto Western Hospital, UHN, Toronto, Canada
Stina Syvänen, Uppsala University, Sweden, EU
Thilo van Eimeren, University Hospital Cologne, University of Cologne, Germany, EU
Neil Vasdev, The Centre for Addiction and Mental Health (CAMH), Toronto, ON, Canada
Victor Villemagne, Austin Health, Heidelberg, Victoria, Australia
Dieter Willbold, Research Center Jülich, Germany, EU
Open access funding provided by Uppsala University.
Compliance with ethical standards
Conflict of interest
The authors declare they have no conflict of interest.
- 9.Giugliani R, et al. Neurocognitive and somatic stabilization in pediatric patients with severe mucopolysaccharidosis type I after 52 weeks of intravenous brain-penetrating insulin receptor antibody-iduronidase fusion protein (valanafusp alpha): an open label phase 1-2 trial. Orphanet J Rare Dis. 2018;13(1):110.CrossRefPubMedPubMedCentralGoogle Scholar
- 46.Eberling JL, Dave KD, Frasier MA. Alpha-synuclein imaging: a critical need for Parkinson's disease research. J Park Dis. 2013;3(4):565–7.Google Scholar
- 61.Stocki PWK, Szary J, Demydchuk M, Logand D, Walsh FS, Rutkowski L. Combined in vitro/in vivo methods for selecting VNARs that shuttle large therapeutic molecules across the blood brain barrier 2016. Available from: https://cdn.website-editor.net/a3c3cad520224a3f90107eddfe79c79a/files/uploaded/Poster-PEGS-2016-for-website.pdf.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.