Brain Delivery of Biologics with Trojan Horse Shuttles: Remembering Paracelsus

Paracelsus in the 16th century established the principle of toxicology that the dose of a medicinal defines the difference between a remedy and a toxic effect. This difference is the therapeutic index, which is the ratio of the mean toxic dose (TD50) relative to the mean effective dose (ED50). A drug with a narrow therapeutic index (NTI) has a TD50/ED50 ≤ 2-3. Examples of small molecule drugs with a NTI are warfarin, phenytoin, chemotherapeutic agents, and levothyroxine. Levothyroxine, an endogenous hormone, shows that even a natural product can give rise to toxicity at a given dose. The same principles of drug action and drug toxicity apply to biologics, such as monoclonal antibodies (MAb), which are emerging as the fastest growing sector of the pharmaceutical industry.

A new application of a receptor-specific MAb is now taking place in the drug development of biologics for brain disorders. Biologic agents are genetically engineered recombinant proteins, and include therapeutic antibodies, decoy receptors, neurotrophins, or lysosomal enzymes. Until recently, biologics have not been developed as new therapeutics for disorders of the brain or spinal cord, because these large molecule pharmaceuticals do not cross the blood-brain barrier (BBB).  However, this situation is changing as biologics are now being re-engineered for BBB delivery as IgG fusion proteins. The IgG domain is a MAb that binds an exofacial epitope on an endogenous peptide receptor on the BBB, such as the insulin receptor (IR), transferrin receptor (TfR), or insulin-like growth factor receptor (IGFR). Owing to the presence of these receptors at the BBB, the respective ligands (insulin, transferrin, insulin-like growth factors) undergo receptor-mediated transport (RMT) across the BBB in vivo. Similarly, a MAb, or fragment thereof, that targets these BBB receptors may also undergo RMT across the BBB. This transport through the BBB involves sequential steps: (i) binding of the MAb to the receptor at the luminal membrane of the brain capillary endothelium, i.e., the blood side of the BBB, followed by receptor-mediated endocytosis of the MAb-receptor complex into the intracellular compartment of the endothelium; (ii) movement (transcytosis) of the MAb-receptor complex through the narrow endothelial cell, which is a distance of only 200 nm between luminal and abluminal endothelial membranes, followed by separation of the MAb and intracellular receptor; and (iii) recycling of the unoccupied receptor back to the luminal endothelial membrane, and exocytosis of the MAb across the abluminal membrane of the endothelium, i.e., the brain side of the BBB, which results in MAb distribution into the interstitial space of brain. If fused with a therapeutic protein, the MAb acts as a molecular Trojan horse to shuttle across the BBB any biologic agent that is genetically fused to the MAb.

The first steps in the re-engineering of a biologic for RMT across the BBB is the identification of the receptor-specific MAb that serves as the lead BBB molecular Trojan horse, and the design of the engineering strategy for genetic fusion of the biologic to the targeting MAb.  The initial strategy for engineering BBB-penetrating IgG fusion proteins developed over 15 years ago fused the biologic to the carboxyl terminus of each heavy chain of a MAb that bound either the IR or TfR via a bivalent, high affinity mechanism with a receptor binding dissociation constant, KD, of 0.5-5 nM, and this approach was reduced to practice for all 4 classes of recombinant proteins: neurotrophins, lysosomal enzymes, decoy receptors, or therapeutic antibodies.  Owing to the high affinity of the bivalent binding of the MAb to the BBB receptor, therapeutic effects of the IgG fusion proteins were observed at low injection doses (ID) of 1 mg/kg of fusion protein in mouse models of neural disease, including Parkinson’s disease, Alzheimer’s disease, stroke, and lysosomal storage disease of brain. This early work led to the issuance of intellectual property with broad claims on the BBB delivery of biologics with high affinity, bivalent antibodies targeting endogenous BBB receptors.

About 10 years ago, a low affinity MAb was developed as a preferred BBB Trojan horse, based on the hypothesis that a MAb with a high affinity for the BBB receptor was preferentially retained within the intra-endothelial compartment of brain. This hypothesis of intra-endothelial sequestration of a high affinity MAb ignored the conflicting morphologic data in rodents and primates, which demonstrated rapid transcytosis through the BBB of high affinity, bivalent antibodies targeting either the IR or TfR. The endothelial sequestration hypothesis also ignored the therapeutic effects of high affinity IgG fusion proteins in animal models of neural disease at a low ID of 1 mg/kg. Such therapeutic effects in vivo in animal models would not be possible if the high affinity MAb was sequestered within the brain endothelium. The brain uptake of the high affinity and low affinity MAb Trojan horses was compared in mice after the intravenous administration of an ID as high as 50 mg/kg. The brain uptake of a high affinity Trojan horse is saturated at an ID of 3 mg/kg, so further increases in brain uptake are not expected at an ID > 3 mg/kg. Conversely, the brain uptake of a low affinity Trojan horse increases when the ID is increased above 1-3 mg/kg. The brain uptake of a low affinity Trojan horse is comparable to the brain uptake of a high affinity Trojan horse, but at the expense of the requirement for a much higher ID, e.g., 10 mg/kg or greater. The high ID of a low affinity Trojan horse reduces the therapeutic index of the fusion protein, as the high ID increases the likelihood of toxicity produced either in preclinical studies in primates or later in human clinical trials.

The low affinity Trojan horse is typically engineered in a format that produces monovalent binding of the MAb domain of the fusion protein to the BBB receptor. Monovalent low affinity binding can be accomplished by alternate strategies. In one approach to engineering a mono-valent, low affinity anti-receptor antibody, a single chain anti-receptor antibody is genetically fused to the carboxyl terminus of a single heavy chain of a second antibody. The second antibody may be either a bivalent therapeutic antibody or a non-specific bivalent antibody that acts as a scaffold for fusion of the therapeutic domain to the amino terminus of one or both chains of the scaffold antibody. Since the anti-receptor single chain antibody is only fused to a single half antibody molecule, it is necessary to use additional technology, such as knob-in-hole genetic engineering, to pair the 2 distinct half antibodies within the host cell. The affinity of this type of monovalent antibody for the BBB receptor is typically low, with a KD of 50 nM to >100 nM, which is 1-2 log orders higher than the KD of a bivalent, high affinity anti-receptor antibody. In a second approach to engineering a mono-valent, low affinity anti-receptor antibody, several amino acids within the CH3 region of the Fc domain of an antibody are mutated until a binding site for the BBB receptor is created within the CH3 region of one half-antibody. This half-antibody is then paired, e.g., with knob-in-hole technology, with another half antibody that expresses no engineered anti-receptor binding site. The KD of this type of monovalent Fc-directed antibody for the BBB receptor is typically very low, e.g., 300 nM to 1400 nM, which is 2-3 log orders higher than the KD of a bivalent, high affinity anti-receptor antibody.

In most branches of pharmaceutics, it is recognized that a reduction in drug affinity for the target receptor requires the administration of a higher ID of the pharmaceutical. Moreover, as the ID of the drug is increased, the likelihood of untoward drug side effects is increased. Drug toxicity accounts for a significant attrition of novel drug candidates, often not recognized until drug testing in clinical trials or even post-marketing. The relationship between drug dosage and drug toxicity was observed by Paracelsus nearly 500 years ago.  In an effort to minimize dose-related toxicity, the developer of a monovalent, low affinity anti-receptor Trojan horse for BBB delivery of biologics may choose to lower the ID in clinical trials, e.g., from 10-15 mg/kg to 3-5 mg/kg. This reduced ID may minimize toxicity, but at the cost of a reduction in drug efficacy.

Adherents of the low affinity, monovalent approach to engineering biologics that penetrate the BBB via RMT for the treatment of neural disease are forced to make several compromises, including the administration of IgG fusion proteins at high IDs, dealing with dose-related toxicity, and adjustments to fusion protein downstream processing to remove homo-dimers associated with the use of knob-in-hole pairing of different half antibodies. An additional problem is the increased cost of goods to support the high injection doses for a disease such as Alzheimer’s disease than may affect 50 million subjects worldwide. These problems are largely obviated by the choice to engineer BBB-penetrating IgG-biologic fusion proteins that bind to the target receptor on the BBB via a classical bivalent, high affinity mechanism. Such bivalent, high affinity BBB Trojan horses enable the reduction in the dosing of the pharmaceutical, and as taught by Paracelsus, reduction in toxicity.