The tight barrier between brain and blood vessels guards the brain from harm, but it also frustrates scientists trying to deliver therapeutic molecules to the central nervous system (CNS). In the May 25 Science Translational Medicine, researchers at Genentech, South San Francisco, California, part of the Roche Group of pharmaceutical companies, describe a new way to hijack the brain’s own transport system to ferry antibodies across this barricade. Led by Ryan Watts and Mark Dennis, the team made use of transferrin receptors, which import iron-containing molecules by way of endocytosis, and are expressed in the cells of the blood-brain barrier. Antibodies to this receptor can get across the barrier but usually travel no further, ending up stuck in the vasculature. The Genentech team found that by lowering the antibodies’ affinity for the transferrin receptor, the antibodies were more readily released into the brain.

The researchers wanted to test the potential of this method to deliver treatments for CNS disorders such as Alzheimer’s disease. In a companion paper, the scientists detail the development of an antibody against BACE1, the β-secretase that begins the process of cleaving APP into Aβ. They then engineered a “bispecific” antibody, which has one arm that recognizes the low-affinity transferrin receptor epitope, and the other binding the BACE1 epitope. In wild-type mice, the bispecific antibody penetrated the brain more easily than did mono-specific anti-BACE1, reaching up to 10-fold higher concentrations and dramatically lowering endogenous Aβ40 levels. The results suggest that this strategy can work to deliver active therapeutic antibodies across the blood-brain barrier, and might be broadly applicable to numerous CNS disorders. This is important because antibodies currently in trials for Alzheimer’s disease tend to enter the brain poorly, necessitating fairly high doses.

The method has generated excitement among those in the field. “It is a brilliant approach,” said Robert Vassar at Northwestern University, Evanston, Illinois. He noted that more work needs to be done before this technique would be ready for clinical trials. Michael Agadjanyan at the Institute for Molecular Medicine, Huntington Beach, California, agreed: “If the method proves to be safe, this will be a beautiful technique for getting antibodies into the brain.”

BACE1’s key role in AD pathology has made it an enticing target for pharmaceutical companies, but the secretase has also been elusive. Many companies are developing small-molecule BACE1 inhibitors, although after a decade of research, none has yet emerged as a viable therapy (see, e.g., ARF related news story). One problem with such inhibitors is that they can block the action of the related proteases BACE2 and cathepsin D, leading to side effects, Watts said. To get around this, Watts and colleagues chose instead to target BACE1 with an antibody, which can be more selective than small molecules and also lasts longer in the body. Their first paper describes the development of this mono-specific antibody. First author Jasvinder Atwal characterized a library of human synthetic antibodies for their ability to recognize and inhibit human BACE1. The screen turned up an anti-BACE1 that selectively blocked this enzyme but had no effect on BACE2 or cathepsin D. Crystallographic analysis revealed that this was because the antibody does not bind the enzyme’s active site. Instead, it latches on at a region of low homology, namely loops C, D, and F of BACE1, causing slight shape changes in the enzyme that seem to affect the behavior of the active site. Other scientists in academia and industry are on the same trail. In the March 11 Journal of Biological Chemistry, researchers led by Bart De Strooper at VIB, Leuven, Belgium, with coauthors at Lilly and Johnson and Johnson, describe an anti-BACE1 antibody, mAb 1A11, with a very similar mode of action. Their antibody also binds loops D and F, and blocks BACE1 activity both in vitro and in vivo. Lacking the anti-transferrin mechanism, De Strooper’s group stereotactically injected mAb 1A11 into the hippocampus of APP-transgenic mice, where it decreased Aβ production.

One concern about BACE1 antibodies as therapeutics is whether they can get into endosomes, where the enzyme normally processes APP. Atwal and colleagues showed in cell culture that their mono-specific anti-BACE1 gets internalized into these intracellular compartments, perhaps through binding to BACE1 on the cell surface. In neuronal cultures, the antibody lowered levels of soluble Aβ40 by up to 70 percent. Moving to in vivo tests, Atwal and colleagues first examined BACE1 knockout mice. Complete absence of the enzyme lowers plasma Aβ40 levels by about half, and brain Aβ40 by around 80 percent. Having established these maximal reductions, the authors then gave varying doses of intravenous mono-specific anti-BACE1 to wild-type mice. At 3 mg/kg, the antibody lowered plasma Aβ40 to near-maximal levels, but had no effect on brain Aβ. The authors attributed this to the fact that only about 0.1 percent of antibodies are estimated to cross the blood-brain barrier. When Atwal and colleagues administered multiple injections of 100 mg/kg anti-BACE1, they got brain antibody concentrations of around 20 nM, and saw about 40 percent reduction in brain Aβ40. Such high doses would be impractical for human therapy. Aβ42 levels were not detectible in wild-type mice with current assays, Watts told ARF. Injections in wild-type monkeys produced similar results to those in mice.

The researchers wanted to examine the antibody’s effect in an AD model mouse in order to see if treatment can reduce oligomers and plaque deposition and improve cognition. When they gave mono-specific anti-BACE1 to Tg2576 mice, which carry human APP with the Swedish mutation, however, they saw minimal effects on brain Aβ even at high doses. Swedish mutation APP is processed primarily in the secretory pathway, not in endosomes (see Haass et al., 1995; Yamakawa et al., 2010), placing the reaction out of reach for antibodies. The Genentech team is searching for a more suitable AD mouse model, Watts said. He added that, because most people with AD have wild-type APP, this antibody limitation should not affect therapy of the large majority of AD cases.

In the second paper, the researchers set out to solve the problem of poor brain penetrance by antibodies. They knew that antibodies against the transferrin receptor get taken up by brain, but remain stuck in the vasculature. Dennis, an antibody expert, hypothesized that lower-affinity antibodies would be better able to let go of the receptor and get released into the brain. Co-first author Yin Zhang mutated the binding site to develop several anti-transferrin receptor antibodies with varying affinities. When they administered these molecules to wild-type mice at the trace dosing levels normally used in such experiments, the low-affinity antibodies had trouble binding the receptor, and entered the brain more poorly than high-affinity antibodies. Clinicians routinely administer much higher antibody doses, however. At these therapeutic dosing levels, the transferrin receptor becomes saturated. Under these conditions, even low-affinity antibodies easily dock at the receptor and get ferried across the blood-brain barrier, where they disembark more easily than high-affinity antibodies. Indeed, under therapeutic dosing conditions, lead first author Y. Joy Yu found that the lowest-affinity antibodies entered the brain the best, reaching fivefold higher concentrations than their high-affinity cousins. Staining revealed that low-affinity antibodies spread out through brain tissue instead of remaining tethered to blood vessels.

To test whether this strategy could be used to deliver a therapeutic antibody, the authors built an antibody with two different arms, one containing the anti-BACE1 epitope and the other the low-affinity anti-transferrin receptor epitope. This bispecific antibody penetrated the brain at five- to 10-fold the concentration of mono-specific anti-BACE1. A single dose of 25 mg/kg resulted in brain concentrations around 25 nM, which Watts noted is in the therapeutic range. The bispecific antibody also produced correspondingly better reductions in Aβ40 than anti-BACE1 did. At 50 mg/kg, the dual antibody lowered Aβ40 levels by about half. This may be close to the maximum reduction achievable with antibody treatment, since some Aβ is produced in the secretory pathway and is not accessible to antibodies, Watts told ARF. In comparison, a new antibody trial just gearing up is examining doses from 0.1 to 15 mg/kg (see ARF related conference story).

Many steps remain before this anti-BACE1 approach could enter clinical trials. Vassar noted that the authors will need to look at the effect of chronic antibody dosing, as well as perform various safety studies. BACE1 knockout mice have several subtle defects, suggesting that inhibiting this enzyme too strongly could have side effects. For example, BACE1 cuts neuregulin, a transmembrane protein believed to play a role at synapses and as a myelin regulator (see ARF conference story). In the May 16 Journal of Biological Chemistry, researchers led by Riqiang Yan at the Lerner Research Institute, Cleveland Clinic Foundation, Ohio, report that inhibiting BACE1 in vitro disrupts myelination. It is also possible that targeting the transferrin receptor could have adverse effects on iron homeostasis, although the authors point out that their low-affinity antibodies do not compete with the binding of transferrin to its receptor. These safety studies are on his agenda, Watts said.

Another question is whether inhibiting BACE1 will actually slow the progress of AD. Agadjanyan noted that, because the authors saw no results in Tg2576 AD mice, the paper was not able to prove that BACE1 inhibition is an effective AD treatment. Dave Morgan at the University of South Florida, Tampa, wrote to ARF that simply blocking Aβ production will not reduce existing deposits, and noted that plaques may release oligomers back into their surroundings under some conditions (e.g., Martins et al., 2008; see full comment below).

Even if this new approach to inhibiting BACE1 translates beautifully to humans, small-molecule inhibitors will remain important, Vassar said. Small molecules can be taken orally and are much cheaper than antibody therapy. Also, some patients might respond better to small molecules than to antibodies, Vassar suggested. He also sees potential in anti-Aβ immunotherapy. “Eventually, we will have a toolbox of different anti-amyloid types of drugs that we will be able to pick and choose from and maybe even use in combination to treat or prevent the disease,” Vassar predicted.

What most excited scientists interviewed for this article is the potential of this low-affinity transcytosis technique to deliver antibodies into the brain. Several researchers suggested that making a bispecific antibody with anti-Aβ would be a logical step for AD research. Watts pointed out that other transcytosis pathways might work better than the transferrin receptor. “We believe this is a broad principle. Reducing affinity for a receptor-mediated transcytosis target will enhance antibody uptake,” Watts said. “We are very keen on using this technology to go after CNS diseases.”—Madolyn Bowman Rogers


  1. Reading the two excellent papers from Genentech, I was quite impressed by their elegant approach to raise anti-BACE1 antibodies that reduce Aβ production in vivo in brain. They further engineered the anti-BACE1 antibody by coupling it with a low-affinity anti-transferrin receptor antibody into a bispecific antibody that efficiently passes the blood-brain barrier. The results are striking, achieving about five to 30 times more penetrance of the antibodies into the brain and about 50 percent Aβ reduction in mouse brains.

    I speculate that the Genentech researchers would already have set out to design humanized TfR/BACE1 antibodies for human clinical trials in the near future. A few additional preclinical results are of interest—for example, chronic effect of anti-TfR/BACE1 on Aβ burden, behavioral effects through reduction in Aβ oligomers or increases in sAPPα levels, and any side effects by saturating TfR-mediated BBB transcytosis, although these are minor issues.

    This approach would readily be applicable to other therapeutic antibodies against AD, especially on passive Aβ immunotherapy. Our colleagues at Tokyo University (Drs. Tomita, Hayashi, and Takatori) have raised a neutralizing antibody against the extracellular domain of nicastrin, a putative substrate receptor of the γ-secretase, initially aiming at reducing Aβ in the brain. The obstacle was the antibody's low penetration into the brain parenchyma, so they first tackled Notch-dependent tumors with the reagents (SfN meeting, 2010). We would love to adopt such an elegant technology in our antibodies.

    With such a powerful therapeutic in our hands, the ultimate questions are when to start Aβ lowering therapy (AD dementia, MCI due to AD, or at the preclinical AD stage?), how long to treat, and what would be the best endpoint(s).

  2. Here is my quick reaction to these intriguing papers, without the benefit of having had time to fully consider their impact.

    BACE1 is an intracellular target. This raises a level of challenge for antibody action beyond the blood-brain barrier, though it is not impossible to overcome.

    Blocking BACE1 will reduce new amyloid deposition, but accumulating data in the field are increasingly convincing that simply blocking production does not reduce pre-existing deposits, which in humans begin to form many years before symptoms. An argument might be made that BACE1 inhibition may nonetheless reduce oligomers. But this assumes the plaques are not breathing oligomers on their periphery, and that oligomers are the toxic agent in AD, which is still a big assumption.

    The idea of bispecific antibodies is an intriguing one. If this inhibits BACE activity, it might also inhibit transferrin receptor activity. The consequences of that remain to be seen.

    Clearly, inhibiting BACE is a meaningful target for AD. I suspect that an antibody may be as good or better at hitting the target with fewer side effects than a small-molecule drug. Still, a small molecule wins on the basis of cost. This is a creative approach that deserves further evaluation.

  3. This pair of papers describes the efficacy of an antibody approach to inhibiting BACE1 activity, as well as an ingenious method for getting higher concentrations of therapeutic antibody into the brain. Passive immunization with a highly specific antibody to BACE1 resulted in a substantial decrease in plasma Aβ40 levels and a lesser decrease in brain Aβ40 levels. By increasing central exposure using a bispecific antibody with a low affinity for the transferrin receptor and a high affinity for BACE1, the investigators were able to demonstrate improved central exposure and greater reductions in brain Aβ levels. The work provides important insights into the biology of BACE1, including the finding that plasma and brain Aβ40 levels were independently modulated by BACE1 inhibition, and significant reductions in plasma Aβ40 had no effect on brain Aβ40 levels. The efficacy of anti-BACE1 treatment was reduced in mouse models carrying the APPswe mutation, supporting the previously proposed hypothesis that this mutation alters the substrate-enzyme interaction. Finally, the work suggests that the BACE1 antibody is internalized into a cell and is capable of altering the activity of the enzyme in a subcellular compartment. This finding suggests that other antibody approaches to intracellular targets may be equally effective.

    However, this approach may have some downsides as a treatment for Alzheimer’s disease. The clearance of the mono-specific BACE1 antibody was dependent on the level of BACE1 expression; BACE1 expression has been reported to increase in Alzheimer’s disease, making it more challenging to predict the therapeutic level that would be needed in a patient to achieve efficacy. Even with the improved central exposure achieved with the bispecific antibody, the modest, but effective, brain antibody concentrations required a high plasma antibody level, which may be difficult to safely maintain. In non-human primates, the changes in CSF Aβ levels were also variable, which could be reflecting an inability to precisely control the degree of Aβ reduction that results from treatment. Theoretically, this approach will reduce both Aβ40 and Aβ42, as well as other Aβ species, although only Aβ40 was measured in this study. If the ratio of Aβ42/Aβ40 is an important aspect of the pathophysiology of AD, any approach that simultaneously lowers both peptides may not be as effective as a targeted approach to lower Aβ42 levels.

    Overall, the combined studies are a significant leap forward in understanding how passive immunization approaches could be harnessed to treat central nervous system disorders, and how efficacious concentrations of antibody can be achieved in brain. It will be exciting to watch the further evolution of this approach toward safe and effective treatments for CNS diseases.

  4. Delivery of antibodies into the brain using bispecific antibodies is a very interesting strategy. I would be interested to hear more from the authors about the so-called full-length IgGs generated by cloning VL and VH regions into LPG3 and LPG4 vectors. Have they truly generated full-length IgGs using this technique, or it is only F(ab)2 fragments?

Make a Comment

To make a comment you must login or register.


News Citations

  1. Getting to First BACE: BACE1 Inhibition Takes a Step Forward
  2. Barcelona: Antibody to Sweep Up Aβ Protofibrils in Human Brain
  3. Barcelona: Out of Left Field—Hit to The Eye Kills BACE Inhibitor

Paper Citations

  1. . The Swedish mutation causes early-onset Alzheimer's disease by beta-secretase cleavage within the secretory pathway. Nat Med. 1995 Dec;1(12):1291-6. PubMed.
  2. . beta-Secretase inhibitor potency is decreased by aberrant beta-cleavage location of the "Swedish mutant" amyloid precursor protein. J Biol Chem. 2010 Jan 15;285(3):1634-42. PubMed.
  3. . Lipids revert inert Abeta amyloid fibrils to neurotoxic protofibrils that affect learning in mice. EMBO J. 2008 Jan 9;27(1):224-33. PubMed.

Other Citations

  1. Tg2576 mice

Further Reading


  1. . Linkage of plasma Abeta42 to a quantitative locus on chromosome 10 in late-onset Alzheimer's disease pedigrees. Science. 2000 Dec 22;290(5500):2303-4. PubMed.

Primary Papers

  1. . A therapeutic antibody targeting BACE1 inhibits amyloid-β production in vivo. Sci Transl Med. 2011 May 25;3(84):84ra43. PubMed.
  2. . Boosting brain uptake of a therapeutic antibody by reducing its affinity for a transcytosis target. Sci Transl Med. 2011 May 25;3(84):84ra44. PubMed.
  3. . Inhibition of beta-secretase in vivo via antibody binding to unique loops (D and F) of BACE1. J Biol Chem. 2011 Mar 11;286(10):8677-87. PubMed.
  4. . Cleavage of neuregulin-1 by BACE1 or ADAM10 protein produces differential effects on myelination. J Biol Chem. 2011 Jul 8;286(27):23967-74. PubMed.