*4.3. Antibody Engineering Approaches to Increase Trans-BBB Transport*

A serious limitation to the use of many antibodies in the design of improved biotherapeutics is their non-optimal behavior in the organism, including their poor PK parameters, non-optimal distribution, inhibition of binding with FcRn, and toxicity [69]. At the same time, one of the main problems is their frequent administration at a large dosage, which increases the risk of immunogenicity and side effects and reduces patient tolerance for the antibody. As such, one should note that antibody production is a continual design process that involves the generation and optimization of antibodies to enhance their clinical potential [70]. Moreover, much of the development and clinical experience that is gained from the generation and optimization of one antibody is applicable to other antibodies, thereby streamlining certain activities and decreasing some of the risks that are intrinsic to

drug development. For example, to optimize the properties of an antibody for a particular indication, it would be preferable to improve or even delete particular characteristics. As an example, one major goal in developing therapeutic antibodies in neurological diseases is to improve the clinical utility of these reagents with respect to antigen targeting and better brain uptake (i.e., better brain-to-blood ratio) to encourage effective disease therapy. Achieving this goal depends not only upon a thorough understanding of molecular properties underlying antibody behavior and function but also upon the development of techniques to manipulate these properties in such a way that enhances their therapeutic potential [43]. For instance, PD response is often directly proportional to brain exposure and, thus, plasma half-life. As such, a typical goal in biotherapeutic development is to identify a candidate molecule having desirable PK properties or, alternatively, to manipulate a molecule's properties to improve its PK while preserving antigen recognition. PKPD properties of antibodies are governed by both molecule-dependent and species-dependent parameters. Biological processes such as antigen binding and receptor binding are important determinants in antibody PK. Since the field of therapeutic monoclonal antibodies has become extremely competitive, especially against validated antigens, it is necessary to develop highly optimized antibodies, above and beyond humanization [70]. A highly optimized humanized antibody would have superior pharmacological properties important for clinical efficacy, such as high antigen-binding activity and long half-life, as well as biophysical properties important for commercial development of the therapeutic antibody, including stability and expression yield in host cells. In order to generate such highly optimized antibodies, it is necessary to consider these pharmacological and biophysical properties during the process of humanization and manufacture. This section describes the critical properties of therapeutic antibodies that should be sufficiently qualified, including Fc and antigen binding affinity, and physiochemical properties and PK.

Fortunately, antibody fragments, such as single-chain Fv, diabody, triabody, Fab, F(ab )2, and full length antibodies, ranging in size from 30 to 150 kDa and valence from one to three binding sites can be also derived via molecular engineering [71]. The single chain Fv (scFv, 30 kDa) is one of smallest forms of antibody that consists of variable light and heavy domains connected by a flexible peptide linker of approximately 15 amino acids generating one binding site. Diabodies (60 kDa) consist of two single chains joined by a very short linker, while triabodies (90 kDa) do not have a linker, thereby forcing trimerization. For example, the ability of these fragments to bind to tumor lies in a fine balance between their ability to penetrate tumor tissues due to their small size and their fast clearance from the body by the kidneys [72]. While retaining their antigen-binding capabilities, these fragments not only cleared faster but were also shown to have much higher tumor/organ ratios compared with their larger counterparts [72]. These antibody fragments are also used in neuroimaging agents in various diseases [35,73]. In this aspect, these studies showed that a diabody and a triabody penetrate the brain parenchyma more rapidly than the full length antibody, which in turn enables in vivo imaging of Aβ pathology at an earlier time point after administration [73].

#### 4.3.1. Role of Fc Receptors

Receptors on the blood–brain barrier bind ligands to facilitate their transport to the CNS. Therefore, it is hypothesized that by targeting these receptors, therapeutic macromolecules (e.g., nanoparticles, antibodies) can be delivered to the CNS [74]. In this regard, FcRn, LRP, TfR, and IR receptors play an important role in regulating the endocytosis and transcytosis of IgGs, peptides, and proteins across the BBB (Table 2) [12]. The function and mechanism of FcRn in regulating IgG recycling have been well characterized. Because of the protective effects of FcRn against lysosomal degradation of IgG, generating Fc fusion proteins and modulating the pH dependent affinity between Fc and FcRn has been utilized to improve the PK of therapeutic antibodies [74]. In vitro studies have shown that FcRn regulates the transport of IgGs in both directions across the endothelial barriers of blood vessels, including those in brain, intestine, and placenta [75]. Importantly, studies using

immortalized rat brain endothelial cells suggested that the human Fc fragment transports faster in the brain-to-blood direction than in the opposite direction [76]. These studies showed that while FcRn mediated the transport of IgGs across peripheral vascular cells in both directions, FcRn only mediated transport across BBB in the brain-to-blood direction. Modulating the interaction between Fc and FcRn through protein engineering has been applied to improve the PK of the therapeutic antibodies. Various studies have shown that the prolonged half-life and exposures of therapeutic antibodies can be achieved by increasing the pH-dependent binding affinity between Fc and FcRn [74].

There have been controversial studies on the role of FcRn in regulating the efflux of IgG from the brain and whether FcRn behaves as an efflux receptor that can transport antibodies across the blood–brain barrier back into the systemic circulation. One study in rats suggested that BBB FcRn mediates the efflux of IgG from the brain to the blood [77]. This study showed this efflux mechanism can be avoided when using antibody fragments devoid of Fc regions (Fab, F(ab')2 and scFv fragments). Subsequently, another study investigated the mechanism of Aβ immunotherapy in the clearance of Aβ amyloid peptide in APPsw mice, a model that develops Alzheimer's disease-like amyloid pathology [76]. The study showed that anti-Aβ IgG-assisted efflux of Aβ amyloid peptide from the brain to the blood in wild-type mice was inhibited when the FcRn gene was knocked out. Taken together, these data suggest that FcRn at the BBB may play a role in regulating IgG-assisted Aβ amyloid peptide removal from the aging brain.

Unfortunately, other studies have shown that brain distribution and disposition of IgG is not regulated by FcRn and FcγR [78,79]. In these studies, IgG was injected intravenously to FcRn knockout and control mice [78]. As anticipated, the plasma clearance of IgG was increased by about 10 times, and the plasma exposures decreased by 4–5 times in FcRn deficient mice when compared with the controls. The brain exposure of IgG, however, was also reduced to a similar extent, and as a result, the brain-to-plasma ratios of IgG were not significantly different between the FcRn deficient and the controls. In another study, the role of FcRn in regulating brain IgG disposition was further investigated in the FcRn knockout, FcγR knockout, and control mice [79]. Compared with controls, the plasma and brain exposures from FcγR knockout mice were not significantly different, and the plasma and brain exposures from FcRn knockout mice decreased by 3–4 times as anticipated. However, similar to what was observed in the previous study, the brainto-blood exposure ratio was not significantly different among the knockout and control mice. Together, these two studies indicate that FcRn and FcγR do not contribute to the "BBB" that limits IgG uptake into the brain. Similar to what was reported by Garg et al. and Abuqayyas and Balthasar, recent results from other groups showed that there was no IgG uptake difference between FcRn knockout and wild-type mice in the brain, which suggested that FcRn has little effect on the distribution of IgG in the brain [80]. In support of these findings, a more current study evaluating IgG uptake in tissues for FcRn wild type and FcRn- constructs indicates that FcRn does not contribute significantly to the brain for IgG in mice [81]. This study also demonstrates that FcRn does not play a protective role in the brain. These data are not consistent with previous studies that showed higher brain uptake for the engineered high-IgG FcRn binder relative to the wild type [82]. As postulated above, higher brain concentrations of the IgG variant with enhanced binding to FcRn could result from a role in efflux of IgG, as opposed to influx [76,77]. Further evidence toward this notion is the fact that the brain expression of FcRn is co-localized with the glucose transporter 1 (Glut1) in the capillary endothelium, suggesting that FcRn is expressed in the proper location to potentially mediate reverse transcytosis of IgG from the brain to the blood [83]. In summary, there is no consensus on the role of FcRn in influencing the blood-to-brain transcytosis of IgG across the brain endothelial cells (BECs) despite several notable studies. Indirect evidence of potential FcRn-mediated recycling was provided by recent studies demonstrating that IgG transcytosis across an in vitro BBB exhibits a non-saturable and nonspecific mechanism and supports the use of RMT

approaches or modifications of biophysical properties, such as pI, to achieve improved brain uptake of therapeutic IgGs [7,84].

#### 4.3.2. Role of Antigen Binding

Trans-BBB delivery methods that use targeting antibodies are often hampered by limited flux through the BBB. A solution to this problem lies in the rational engineering of BBB-targeting antibodies. Leveraging knowledge of intracellular trafficking, researchers have begun to tune selected binding properties of the antibody–antigen receptor interaction. Engineered binding affinity, avidity, and pH sensitivity have been shown to affect binding, intracellular sorting, and release, ultimately leading to increased brain uptake of the targeting antibody and its associated cargo [7]. The first successful attempts for chimeric proteins targeting cell receptors initially relied on cationized albumin, which lacked brain selectivity, and then later IgGs directed against IR or TfR receptors [85]. However, the success of these initial antibodies was limited by their high affinity, which hindered an efficient release and penetration into the brain parenchyma [58]. A variety of protein shuttles have been investigated; most of them are ligands of receptors on the brain endothelium that compete with their endogeneous proteins (e.g., apolipoproteins A and E, receptor-associated protein, transferrin, lactotransferrin, melanotransferrin, and leptin). Although a few nonendogenous proteins have been used (e.g., wheat germ agglutinin, non-toxic mutant of diphtheria toxin), they also have shown moderate efficacy and selectivity [13]. Recent efforts have leveraged antibody engineering strategies to increase trans-BBB transport and have highlighted the importance of the antigen-binding and trafficking issues. As shown in Tables 1 and 2, each targeted receptor may exhibit differential responses to engineered binding properties, illustrating the need to better understand antibody–antigen receptor interactions and trafficking dynamics.

Regarding the above, well-designed experiments have engineered the binding properties of anti-transferrin (anti-TfR) antibodies to study their trafficking and delivery in vitro and in vivo. First, antibody affinity and avidity for TfR were evaluated, and it was shown that higher brain uptake of anti-TfR antibodies can be accomplished by lowering antibody affinity [58]. Intravenous administration of antibodies having a range of affinity to TfR (Kd = 6.9–111 nM) indicated that at trace doses, mouse brain uptake directly correlated with affinity, suggesting that receptor engagement at the blood side of the BBB was the key parameter governing uptake (Figure 1). This figure shows the diagram taken from Goulatis and Shusta (2017) to illustrate that high-affinity monovalent and bivalent anti-TfR antibodies internalize readily into the early endosomes but then direct the antibody–receptor complex toward lysosomal degradation, possibly by crosslinking the TfR and altering its intracellular trafficking [7]. While high-affinity monovalent anti-TfR antibodies can transcytose the BBB, they remain bound to the receptor on the abluminal side, limiting the dose to the brain. In contrast, low-affinity anti-TfR antibodies decrease antibody-TfR sorting to the lysosome and can either be recycled back to the luminal side or get transcytosed to the abluminal side where they dissociate from TfR, leading to increased brain accumulation. Further, pH-sensitive TfR-binding antibodies that can dissociate from TfR in the acidic endosome led to increased transcytosis compared with pH-insensitive antibodies. In the case of the single-domain antibody FC5 (Table 3), increased affinity toward the receptor leads to an increase in the amount of transcytosed antibody, highlighting the fact that antibodies utilizing different trafficking machinery may require customized optimization (Figure 1). However, at therapeutic dosing (20 mg/kg), an inverse correlation was observed, where the lowered affinity antibody demonstrated greater brain accumulation (up to 0.6% ID/ g) (Figure 2). These data demonstrate that lowered affinity allows for antibody release from the TfR at the abluminal membrane, while higher-affinity variants remain bound to the TfR. Further studies provided the evidence that affinity-derived effects on brain uptake of anti-TfR antibodies were at least in part caused by altered intracellular trafficking and lysosomal degradation [37]. The high-affinity anti-TfR antibody was found to be more prominently trafficked to the lysosome and degraded, resulting in reduced cortical TfR

levels. Thus, productive trans-BBB anti-TfR antibody trafficking could be increased by lowering antibody affinity.

**Figure 1.** A schematic depiction of the various engineering optimization strategies for increased transcytosis of antibodies and nanoparticles (NPs) across the BBB. High-affinity monovalent and bivalent anti-TfR antibodies internalize readily into the early endosome (EE) but then direct the antibody–receptor complex toward lysosomal degradation, possibly by crosslinking the TfR and altering its intracellular trafficking. While high-affinity monovalent anti-TfR antibodies can transcytose the BBB, they remain bound to the receptor on the abluminal side, limiting the dose to the brain. In contrast, low-affinity anti-TfR antibodies decrease antibody-TfR sorting to the lysosome and can either be recycled back to the luminal side or are transcytosed to the abluminal side where they dissociate from TfR, leading to increased brain accumulation. Similarly, Tf-coated nanoparticles show a higher transcytosis capability when lowering the Tf coating content, resulting in reduced avidity. Further, pH-sensitive TfR-binding antibodies that can dissociate from TfR in the acidic EE lead to increased transcytosis compared with pH-insensitive antibodies. In the case of the single domain antibody FC5, increased affinity toward the receptor leads to an increase in the amount of transcytosed antibody, highlighting the fact that vectors utilizing different trafficking machinery may require customized optimization. This figure is reproduced from Goulatis and Shusta (2017) with permission of the copyright owner [7].

**Figure 2.** Lower-affinity anti-TfR<sup>D</sup> antibodies (A > D) at therapeutic doses show increased brain uptake. This figure is reproduced from Yu et al. (2013) with permission of the copyright owner [58].

In addition to studies of TfR antibody-binding affinity, the role of avidity has also been explored with similar conclusions. In order to investigate avidity effects on trans- BBB transport, a Fab fragment targeting the TfR was fused to the carboxy-terminus of an anti-BACE1 (β-amyloid cleaving enzyme-1) antibody in a bivalent (dFab) or monovalent (sFab) format [57]. Monovalent binding of an anti-TfR antibody allowed for preferential transcellular transport in brain endothelia, while bivalent binding led to diversion of trafficking toward the lysosome. The results of this study strongly suggest that differences in TfRbinding mode led to major differences in intracellular trafficking, which ultimately allows sFab-associated cargos to cross the BBB. Similarly, another in vitro study has investigated whether or not pH-insensitive TfR binding could be an additional engineering approach for regulating anti-TfR antibody trafficking and increasing trans-BBB transport [86]. These data demonstrated that attenuated binding of an anti-TfR antibody at endosomal pH can lead to differential intracellular trafficking, ultimately enhancing transcytosis across the BBB.

#### 4.3.3. Role of Biophysical Properties

The majority of small drugs that are used to treat CNS disease have a molecular weight between 150 and 500 Da [11]. This does not indicate that drugs with a molecular weight less than 150 or greater than 500 Da are unable to cross. Characteristics that reduce the ability of small molecules to cross the BBB include a polar surface area in excess of 80 A ◦ and a high potential for hydrogen bond formation. Additionally, increased number of positive charges and increased flexibility contribute to BBB crossing. Lipid solubility is a clear indicator of small drugs that can pass through the BBB [87]. Rules for proteins have some similarities and some apparent differences from those for small drugs [11]. Most proteins are poorly soluble in lipids and so would not be expected to penetrate the BBB very well by trans-endothelial diffusion. However, lipid solubility was a predictor of BBB penetration for one series of peptides and proteins that had molecular weights ranging from 486 to 6000 Da. The largest substance found to cross the BBB using transmembrane diffusion thus far is cytokine-induced neutrophilchemoattractant-1, which has a molecular weight of 7800 Da [88]. This is thought to represent a direct correlation between BBB penetration and the ability of a drug to deliver cargoes across the cell membrane. One of the primary factors in determining whether a protein will cross the BBB is its lipophilicity. A strategy for enhancing the ability of a peptide to cross the BBB is increasing its lipophilicity. There are a number of techniques able to do this, including alteration of the protein structure, methylation, halogenation, or acylation. Structural changes, for example covalently binding the drug to lipidic moieties, such as long chain fatty acids, will increase the lipophilicity of a peptide [89]. Peptides with a high number of hydroxyl groups tend to promote hydrogen bonding with water, which leads to a decrease in membrane permeability. Decreasing hydrogen bonding therefore increases membrane permeability. Ideally, there should be potential for the formation of fewer than eight hydrogen bonds when developing new drugs. Methylation is one method used to reduce hydrogen bonding. This illustrates an important point for protein modifications: that the location and type of modification play a significant role in improving BBB transport of your peptide of interest [11]. Halogenation of peptides and proteins can also lead to increased lipophilicity and BBB permeability. The increase in BBB transport of peptides was dependent on which halogen was utilized; chloro and bromo additions increased BBB transport, while fluoro and iodo additions had no effect [90]. An alternative approach is acylation of the N-terminal amino acid, which can increase the lipophilicity of peptides and proteins. For example, acylation of insulin improved its ability to cross the BBB while maintaining its pharmacological effects. Another approach involves glycosylation and hyperglycosylation of therapeutic proteins [91]. In this case, the in vivo results showed that glycosylation is required to maintain protein exposure in blood and proved to increase protein uptake into the CNS [92].

Along these lines, the large change in biophysical properties induced by therapeutic cargoes and the distinct location of the targets of these drugs inside the brain has limited the universal aspiration of most BBB shuttles [13]. Hence, in general, each protein shuttle is prominent in the delivery of a particular family of cargoes. There are many receptors and carriers that are overexpressed on the BBB (Table 2), which can mediate the transport of specific ligands and their cargoes. Additionally, the membrane of the BBB is negativelycharged and shows high affinity with positively charged compounds, which could also trigger the internalization by cells [9]. Thus, these kinds of ligands could mediate the penetration of macromolecules through the BBB. Efficient transport of macromolecules across the BBB through endocytic mechanisms involves both specific (receptor-mediated transcytosis) and/or nonspecific (adsorptive-mediated transcytosis) interactions with proteins and receptors expressed on the brain endothelial cell surfaces. In adsorptive-mediated transcytosis, endocytosis is generally promoted by the interaction of the often positively charged molecule with membrane phospholipids and the glycocalyx [13]. The most common approach relies on enhancing positive charge, in order to mediate interaction with the anionic glycocalyx. However, this approach leads to higher unspecific uptake in many other tissues, often resulting in off-target effects, which in addition requires a high degree of tailoring for certain small molecules that are rarely applicable to biotherapeutics. Another approach for drug delivery to the CNS, as shown in the previous section, focuses on BBB shuttles that allow the transport of a wide range of molecules, comprising small molecules, proteins, nanoparticles, and IgGs across the BBB. Substrates of natural carriers such as glucose and neutral amino acids have been applied to transport small molecules through their natural carriers on the BBB, while for biomolecules the focus has been set on receptor ligand proteins (Table 2) since endocytic pathways tolerate a high cargo load [13].

A common goal for therapeutic antibodies is to extend plasma half-life as a means to increase exposure, often expressed in terms of the area under the plasma concentration-time curve [74]. In contrast, cationization tends to shorten plasma half-life due to an enhancement in both the rate and the magnitude of tissue distribution. However, cationization might prove to be a useful strategy in specific applications in which prolonged antibody exposure may be sacrificed for the sake of rapid, enhanced tissue uptake (e.g., targeting antibodies to efficiently cross the BBB) [93]. Cationization of antibodies has also been explored as a means to encourage extravasation, antigen binding, and receptor-mediated endocytosis of antibodies into target cells [94]. In contrast to native Abs, which are generally excluded from cell membranes in the absence of receptor-mediated endocytosis, cationized Abs are better able to reach the intracellular space via absorptive-mediated transcytosis (AMT). Electrostatic interactions between positively charged proteins and negatively charged cell membranes could permit cell entry via nonspecific membrane flow and have been implicated in the mechanism by which cationized antibodies are rapidly endocytosed by cells in vitro. A similar phenomenon is suggested to induce absorptive-mediated transcytosis across microvascular endothelial barriers in vivo [94]. This interaction also occurs further with sialic acid moieties on the luminal surface and heparin sulfate group on the abluminal surface. AMT of cationized albumin is triggered by this electrostatic interaction and results in the transport of the moiety across the BBB [93]. The use of cationized albumin for the transport of β-endorphin, a very large molecule that cannot cross the BBB, has been reported in rats [95]. After conjugation with cationized albumin brain uptake of β-endorphin was increased. When the isoelectric point of antibodies is raised from neutral to highly alkaline, cationized antibodies are formed. These antibodies are used mainly as neuroimaging agents in various diseases, including brain tumors, AD, and stroke [96]. Mechanisms governing the passage and partitioning of small molecule drugs and antibodies across the BBB have also been the subject of several reviews and modeling studies [29,97,98]. While the distribution of small molecules is influenced by multiple factors, including drug liposolubility, free vs bound concentrations in blood and brain fluids, and their (bidirectional) transport via BBB carriers and efflux pumps, some of these processes may not be important in the case of some biologic molecules [61]. For example, antibodies (including VHHs) are not substrates of efflux pumps and their "bound" concentration in body fluids can be considered negligible in most cases. Their paracellular "filtration" across the BBB is essentially completely restricted by tight junctions of brain endothelium and choroid

plexus epithelium, respectively. In the absence of specific transport mechanisms such as the RMT, they therefore can access the brain only via a low-rate nonspecific adsorptive endocytosis. To discover new antigen−ligand systems for transvascular brain delivery, a recent study developed a method for functional selection of brain microvascular endothelial cell-specific internalizing and transmigrating antibodies [62] from a phage-display llama single-domain antibody (sdAb) library (Table 3). sdAbs are the VHH fragments of the heavy-chain IgGs, which occur naturally in camelid species and lack light chain, and are half the size (15 kDa) of a single-chain antibody (scFv) [99]. These sdAbs have been shown to internalize into the brain's endothelial cell and transmigrate in an in vitro BBB model via a saturable, energy-dependent and charge-independent process; pretreatment of cells with highly cationic protamine sulfate did not affect sdAb transcytosis [100]. Similarly, two positively charged control antibodies, showed minimal "passive" transcytosis in an in vitro BBB model. As such, it can be assumed that CSF levels of control sdAb, which does not bind any known receptor in mammals, are representative of nonspecific passive uptake processes (macropinocytosis; adsorptive endocytosis) of large, hydrophilic, and positively charged biologic molecules at the BBB.

Nevertheless, as discussed above, other antibodies and single-chain antibody fragments have also been developed as "Trojan horse" bispecific antibodies for delivery of therapeutics via RMT, including IR and TfR receptor antibodies; the extensive literature on these antibodies reports a range of their serum/brain partitions (from 0.1% to 4% ID/g vs 0.06% ID/g for IgG) [58]. Due to a lack of comparative studies using the same experimental and analytical methods as well as vast differences in size and pharmacokinetics, the direct comparison between known antibody "Trojan horses" with unmodified single-domain antibodies (sdAbs) remains difficult. These sdAbs will require further engineering for improvement of their plasma half-life and potentially their binding properties before their comparative assessment with similar antibody RMT technologies can be properly performed [61].

#### **5. Conclusions and Outlook**

This review focuses mainly on the historic trends and current practices of research and development activities involving the BBB as a complex interface between the blood and the CNS, essentially for the targeted delivery of antibodies to treat neurodegenerative diseases. It is well established that nearly 0.1% of circulating biotherapeutics, i.e., recombinant proteins or gene-based medicines, cross the BBB [42,45]. Hence, improving brain exposure for at least some of these molecules is the ultimate goal of the brain delivery systems. It has been proposed that CNS diseases can be initiated by several mechanisms, including decreased cerebral blood flow, perturbation of transporters, BBB disruption, deformations of capillaries, and secretion of neurotoxic substances by the BBB. Thus, the BBB may have a fundamental role in brain diseases [40,98]. Nonetheless, the BBB is intimately involved in crosstalk with the rest of the CNS and peripheral tissues and is crucial for normal brain pressure and functioning, and therefore, perturbation of its function might have physiologic consequences. As shown in Tables 2 and 3, different physiological approaches are used to deliver biotherapeutics in the brain parenchyma. Generally, the techniques used involve direct injection or infusion of therapeutic compounds into the brain or the cerebro-ventricles or the CSF. All these approaches, however, are severely limited by poor distribution into brain parenchyma [14]. In fact, the most promising new technology uses a physiological approach to take advantage of endogenous receptors highly expressed at the BBB (e.g., TfR and IR). Fundamentally, this latter approach has been employed by cells of the BBB to enhance the delivery of antibodies across the BBB by receptor-mediated transcytosis. While the physiological approach has the potential to achieve improved brain delivery and to play a significant role in the treatment of CNS disease, in vivo preclinical studies quantifying antibody levels systematically and determining antibody activity relationship in the brain is difficult [56,101]. Despite these challenges, however, current efforts are focused on developing newer generations of antibody therapeutics that

can cross or otherwise interact with the BBB for optimal in vivo benefit. For now, exploiting TfR receptors for delivering antibodies across the BBB may not be the answer to the brain targeting question [68]. However, the research performed with the available anti-TfR antibodies has provided invaluable insight into the mechanisms of action of receptors at the BBB and has also helped to highlight protein engineering issues that must be addressed in order for a successful BBB shuttle to be developed. Additionally, perhaps the most challenging aspect of moving anti-TfR antibodies to the clinic is the lack of species cross reactivity observed in the available antibodies. To solve this problem, antibodies are being engineered for use in each species under investigation, or transgenic mice are generated to express human antigens that tolerize them to humanized antibodies, something that will add significantly to development costs.

Moreover, in search of a solution to increase the penetration of antibodies in the brain, it is likely that all of the parameters described earlier, such as target interaction, FcRn binding, molecular size, and surface charge, are likely to play a part in the trafficking of anti-receptor antibodies across the BBB. As mentioned above, there is no consensus on the role of FcRn in influencing the blood-to-brain transcytosis of IgG across the brain endothelial cells (BECs), despite several notable studies that support the modifications of biophysical properties, such as pI, to achieve improved brain uptake of therapeutic IgGs [7,84]. Special attention is also being given to lipophilicity and overall surface hydrophobicity, since there is a tendency for those parameters to play a significant role in the BBB transport of proteins [90,92,102]. Only through the systematic evaluation of all of these parameters will it become clear which, if any, is the most important to have an effect on the PK of the brain interstitial space [103]. In the past decade, innumerable preclinical studies have been reported on the use of real-time imaging with targeted drug delivery, and this strategy has now matured with promises to assess the distribution and uptake of protein drugs [104]. For example, it has been shown that molecular imaging technologies like PET and SPECT have made important contributions to enable brain imaging of recombinant antibodies that are engineered for BBB transport, particularly in determining drug pharmacokinetics of directly labeled antibodies [35,36,105]. Additional evidence for the sensitivity of antibody PET imaging is provided by a study where a recombinant bispecific antibody was radiolabeled with I-124 and then administered in two transgenic AD and wild-type mice at different ages [73]. This study demonstrates that antibody-based PET is able to visualize and quantify early formed Aβ assemblies (Figure 3) and may become a valuable tool for disease staging of AD patients and for monitoring the effects of Aβ-directed treatment. Additionally valuable in the setting of advanced CNS imaging is the assessment of brain uptake and changes in BBB integrity, which will help to accelerate drug development by assisting in understanding and defining the challenges to translating molecularly targeted agents to the brain [10]. In this aspect, the current trend in drug discovery is to consider classical absorption, distribution, metabolism, and excretion (ADME) studies in parallel with imaging studies, enabling differentiation between biophysical and binding properties and their delivery to the brain. Such classical rules have the advantage of being very simple, as well as being easy to interpret. Their drawback, however, is that they do not take into consideration uncertainties in measurements and calculations as well as the pharmacological effect and toxicity requirements. Meanwhile, the release of an antibody drug in the brain should be accurately monitored and controlled in situ or in real-time [106,107]. It is also important to keep in mind that although no single technology currently provides all the answers, integrating different modalities into other in vivo methodologies (e.g., QWBA, microautoradiography, PET, and SPECT) can enhance our quantitative understanding of spatial brain distribution [32,36]. Furthermore, the brain is a highly vascularized organ with a relatively high proportion of endothelial cells. In determining the concentration of a therapeutic protein in the brain parenchyma, it is critical to ensure that the methods used for assessing distribution are capable of distinguishing endothelial uptake from parenchymal uptake [31]. Thus, this approach indicates that it would be more appropriate to establish a high-quality preclinical assessment of ADME,

PK, and safety/toxicity studies with an emphasis on activity in the CNS, which will permit the parallel optimization of pharmacological response and druggability properties.

**Figure 3.** Sagittal PET images obtained at 3 days after administration of the bispecific antibody radiolabeled with I-124 in two transgenic mouse models of AD (ArcSwe and Swe) and wild-type (WT) mice at different ages (12, 18, and 24 months). Quantification of the radiolabeled antibody in brain tissue showed an increasing signal intensity with age (i.e., with increasing Aβ pathology) in the two transgenic AD animal models, while brains of WT mice were devoid of signal regardless of age. This figure is reproduced from Sehlin and Syvänen (2019) with permission of the copyright owner [73].

In addition to classical ADME studies, generation of in silico physiologically based pharmacokinetic (PBPK) models by incorporating PKPD data and safety profiles as a tool for the treatment of CNS diseases has attracted great interest from pharmaceutical scientists and are likely to be crucial to the development of novel antibody-based therapeutics [28,29,108–111]. Further, the high-throughput and low-cost nature of these models permit a more streamlined drug development process in which the identification of antibody structural optimization can be guided based on a parallel investigation of CNS uptake and safety, along with activity [112]. Hence, the development of in vivo and especially in silico models can be an instrumental tool for predicting the association between BBB penetration and the profile of expected human response for a specific antibody drug against a specific target. This approach will greatly help to simplify the practical difficulties and circumvent potential ethical controversies [106]. In essence, by simultaneously optimizing the antibody molecule in the light of their biophysical and molecular properties, BBB penetration, and activity, it should prove possible to identify a highly qualified clinical candidate and consequently enable faster development of therapeutic antibodies [70]. Although delivery of antibodies to the CNS shows great promise, a greater understanding of CNS physiology and pathophysiology is still needed. Accordingly, it would be vital to characterize further the physiological and vascular attributes such as perfusion, blood volume, and permeability to protein drugs in various CNS compartments [30,43]. Nevertheless, it is believed that the various methods historically used to assess BBB permeability or dysfunction in mouse models of human disease have led to many disparate findings. For example, early studies showed that the BBB is disrupted in Alzheimer's disease models, potentially increasing drug permeability [113]. However, more recent data have shown

that the BBB remains intact in multiple preclinical models of Alzheimer's disease (AD) [38]. Thus, the current lack of in vivo validation of BBB permeability in neurodegenerative diseases and the lack of controlled studies hinder the understanding of antibody delivery through the BBB. Moreover, robust characterization of preclinical disease models is necessary for predicting drug delivery to target tissues and for interpreting correctly the pharmacodynamic responses to disease-modifying biotherapeutic candidates.

After a decade of intensive engineering of antibodies followed by preclinical testing, scientists are still seeking a variety of strategies for optimizing their use as powerful therapeutic agents, particularly for targeted delivery to the CNS for the treatment of neurological disorders. With recent advances in scaffold design, construction, and selection methodologies, there is now a rapid process for recombinant synthesis of specific antibodies differing in affinity and molecular properties against virtually any BBB target [49,114]. For example, in vivo studies showed an enhanced brain uptake of antibodies with novel BBB targets but doing so, while remaining in the range of manageable safety and efficacy, is still challenging [115]. In the case of antibody modifications, it remains to be seen whether modern genetic engineering may affect the pharmacokinetics, biodistribution, therapeutic index, or safety profiles due to immunogenicity [116]. As a consequence of all these minor modifications, immunogenicity concerns are still under investigation by drug regulatory agencies, since their potential for immunogenicity can alter ADME properties, thereby greatly confounding the interpretation of PK/PD assessments [116,117].

To date, great progress has been made with the brain shuttle approach, which has proved to be successful in improving the CNS exposure of some of the large molecules with poor brain permeability, such as bispecific TfR antibodies [4,108,109]. However, further developments are still needed for this approach to become a more robust technology. Until then, fine-tuning of the biophysical and binding properties for optimal brain exposure will remain a staple of CNS drug discovery and development [42]. More importantly, in order to increase further our knowledge regarding the effects of antibody modification on brain-targeted uptake and efficacy, additional clinical studies using relevant animal species and disease models need to be implemented [118,119]. Finally, MRI-guided FUS delivery to the CNS still has great promise and provides the opportunity to improve biotherapeutics' bioavailability locally and to improve their therapeutic profiles. In summary, targeted CNS biotherapeutics is an ever expanding and challenging but important field of study [120,121]. In fact, until now, there is only one human monoclonal antibody aducanumab (under the brand name AdulhelmTM), that has been FDA-approved for the treatment of people with AD. The approval was granted this year based on data from clinical trials demonstrating that aducanumab targets aggregated forms of β-amyloid, a biomarker that is reasonably likely to predict clinical benefits. Thus, since more investigators across academia and industry have joined the race to increase the uptake of antibodies across the BBB, there is good reason for optimism for additional FDA-approved CNS biotherapeutics in the near future.

**Author Contributions:** All other authors (A.K., T.K., V.P., P.H., A.L.E. and L.A.K.) contributed substantially to the discussions, writing, and referencing of the paper. In addition, all authors contributed to editing and reviewing the manuscript during the submission process. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Data Availability Statement:** The data presented in this review are available and can be found in the cited references.

**Acknowledgments:** We thank our colleagues Kevin Brady at UCB (Slough, United Kingdom) and Sherri Dudal at Roche (Basel, Switzerland) for scientific exchange and discussions. Leslie Khawli conceived the paper and wrote the outline of the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.
