BACE1 knockout mice have more than their fair share of problems, but is it because they lack the enzyme during development or during adulthood? New conditional knockouts from the lab of Robert Vassar, Northwestern University, Chicago, suggest a bit of both. In the September 19 Science Translational Medicine, first author Ming-Hsuan Ou-Yang and colleagues describe how mice that lose BACE1 activity as adults avoid many troublesome characteristics of germline knockouts, such as seizures and weight loss. However, one pesky phenotype remains: The mice have shortened, somewhat disheveled mossy fiber bundles in their hippocampi. The results hint that strongly inhibiting BACE1 in adults could lead to hippocampal defects.

  • Conditional BACE1 knockouts escape many of the problems of germline knockouts.
  • cKOs still have short, disorganized hippocampal mossy fibers.
  • Scientists caution against strong BACE1 inhibition in people.

“The work is very careful, and the data convincing,” said Riqiang Yan, University of Connecticut Health, Farmington. Yan was not involved in the work but has generated his own line of conditional BACE1 knockouts.

“The findings by the Vassar lab are extremely important for the safety of clinical trials using BACE inhibitors,” wrote Christian Haass, Ludwig-Maximilians University Munich, to Alzforum. Clinical trials that aim to almost completely inhibit BACE1 function may have a deleterious outcome, he said. Other experts suggested that lowering the dose of BACE1 inhibitors could avoid this side effect and prove safe for long-term use in normal people at risk for Alzheimer’s disease.

BACE1 works alongside γ-secretase to snip the amyloid precursor protein to release Aβ, the peptide that forms the senile plaques that characterize Alzheimer’s disease. Researchers have considered inhibiting both proteins to treat AD. In the case of γ-secretase inhibitors, this proved detrimental. Notably, in a Phase 3 trial, patients on the γ-secretase inhibitor semagacestat tended to have worse cognitive outcomes than those on placebo, along with significant psychiatric and other side effects. Scientists have attributed these changes to decreased processing of another γ-secretase substrate, Notch (Aug 2010 news; Nov 2012 news). 

Cut Short. In the CA3 of the hippocampus (light green), the infrapyramidal bundle (bounded by white arrows) appears shorter in conditional knockouts (left), compared to control mice (right). [Courtesy of Ou-Yang et al., 2018; Science Translational Medicine/AAAS.]

The failure of γ-secretase inhibitors has made scientists extra cautious about inhibiting BACE1, which has many substrates beyond APP. In keeping with this, BACE1 knockouts have numerous problems. They lose myelin from their axons, develop spontaneous seizures, have trouble navigating and remembering objects, and they die younger than wild-type animals (Dominguez et al., 2005; Willem et al., 2006; Hu et al., 2010; Laird et al., 2005). However, scientists are unsure whether those traits arise from developmental deficits, or from lack of BACE1 in adulthood. In one conditional knockout created by Yan’s group, gradual reduction in BACE1 expression in the whole body starting about a month after birth avoided many of the phenotypes observed in complete knockouts, with the exception of reduced neural plasticity (Feb 2018 news). When crossed with an AD mouse model, it developed fewer plaques than transgenic controls.

Ou-Yang and colleagues took a similar approach, using the Cre/lox genetic recombination system to carve out exon 2 of BACE1. In one mouse they expressed Cre shortly after birth, limiting BACE1 loss to the forebrain on postnatal day four and beyond. The cerebellum and brain stem still had some expression. In another mouse, Ou-Yang induced Cre by injecting mice with tamoxifen. This allowed the researchers to raise the mice until full adulthood (three months of age) before they wiped out BACE1 throughout most of the body, including the cortex and hippocampus. The subcortical regions and olfactory bulb expressed some residual BACE1 in this model, too, possibly because tamoxifen did not penetrate those regions fully, the authors wrote. BACE1 expression was reduced in all mice for at least six months before behavioral testing or histopathology.

In the end, both mouse models, but especially the adult knockouts, were spared many of the problems seen in BACE1-/- knockouts. When BACE1 was knocked out just after birth, animals lived about as long as controls. They took slightly longer to learn the location of a platform in the Morris water maze, but otherwise their memories were on par with wild-type mice. These cKOs had fewer spontaneous seizures and their degree of hypomyelination was considerably less than that seen in BACE1-/- embryonic knockouts. The adult knockouts also lived as long as controls. They gained more weight than wild-type mice, but the researchers are unsure why. These mice had normal learning and memory and had no spontaneous seizures or reduced myelination in any brain region.

However, one trait of embryonic BACE1-/- mice was still apparent in both conditional knockouts. The infrapyramidal bundle (IPB)—a tract of mossy fiber axons that extends from the dentate gyrus—was uncharacteristically short. Normally, adult-born neurons in the dentate gyrus extend their axons along the ventral side of the CA3 layer of the hippocampus, forming the IPB, before they cross over the CA3 layer to its dorsal side. However, in BACE KOs, the IPB was 30 percent shorter, and crossed the CA3 layer prematurely (see image above).

Although behavioral testing picked up no learning or memory problems in the adult knockouts, the authors note that the length of the IPB in different mouse strains correlates with spatial learning (Crusio and Schwegler, 2005). Future studies should use more sensitive tests to detect any impairments, they wrote.

In a previous study, the authors attributed the IPB deficit in BACE-/- knockouts to reduced cleavage of a BACE1 substrate called close homolog of L1 (Hitt et al., 2012). CHL1 cleavage is important for guiding axons to their final destination. In the knockouts, Ou-Yang and colleagues found more full-length CHL1 in the hippocampus and less of the cleaved fragment than in control mice. The more full-length they found, the shorter the IPB. The findings indicate that BACE1 cleavage of CHL1 in adult newborn neurons is needed for axons to navigate their way through the hippocampus.

The results suggest both good news and bad news for BACE inhibitors, Vassar told Alzforum. “The good news is that most of the phenotypes we were concerned about in the germline BACE1 knockouts go away when you knock out BACE1 in adults,” he said. “The bad news is that the wiring problem remains, at least in the hippocampus.” To his mind, this does not mean BACE1 inhibitor programs are in jeopardy or that clinical trials should be stopped. “This is a note of caution; drug makers should be aware of this and consider dialing back doses,” Vassar said. The level of BACE1 knockout in these models—in the neighborhood of 90 percent—approximates that with the highest doses of BACE inhibitors given in clinical trials, he said.

Eric Siemers, formerly at Lilly and now of Siemers Integration LLC in Indianapolis, noted that the hippocampus shrank slightly in treated patients in the Phase 3 EPOCH trial of Merck’s BACE1 inhibitor, verubecestat (Dec 2017 conference news). “Whether that’s related to mossy fiber bundle shrinkage is unknown, but people need to take this paper seriously,” said Siemers. “It doesn’t mean we need to stop the [BACE1] studies, but it does mean we need to pay attention to hippocampal morphology.” He suggested that volumetric magnetic resonance measures of the hippocampus would be an important addition to trials of BACE inhibitors. In EPOCH, verubecestat provided no cognitive benefit and a subgroup analysis even suggested a slight worsening on the drug.

Philip Wong, Johns Hopkins University of Medicine, Baltimore, recommends that researchers aim to reduce BACE1 function by a maximum of 50 percent to preserve healthy BACE1 cleavage of other substrates. Companies currently exceed that in clinical trials, he said. “If you can get away with a lower dose and still be effective, you should do that.” Stefan Lichtenthaler from the German Center for Neurodegenerative Diseases, Munich, agreed. “A careful choice of the inhibitor dose may allow enough BACE1 activity to prevent such phenotypes in patients. In fact, BACE inhibitors may be more efficient in prevention than in treatment trials and a lower dose may be sufficient for prevention,” he wrote (see full comment below). Wong said it would be interesting to knock out BACE1 in 12-month-old mice, because the timing would be more comparable to shutting down BACE1 in middle-aged people with a drug.—Gwyneth Dickey Zakaib


  1. This excellent and carefully conducted study finally demonstrates that most of the phenotypes observed so far in BACE1-deficient mice are of developmental origin and may not be a cause of concern for the BACE inhibitors tested in the clinic. Yet, the authors also show that one phenotype persists in adult BACE1 KO mice, namely a structural disorganization within the mossy fibers. This phenotype adds to a, so far, short list of adult phenotypes that were detected in BACE inhibitor-treated mice, which comprise muscle spindle alterations as well as changes in LTP and dendritic spines.

    Is the new adult phenotype a concern for BACE inhibitors in clinical trials? The authors carefully discuss this point, and I agree that a careful choice of the inhibitor dose may allow enough BACE1 activity to prevent such phenotypes in patients. In fact, BACE inhibitors may be more efficient in prevention than in treatment trials and a lower dose may be sufficient for prevention than for treatment.

  2. Finding a safe therapeutic window in which the dose range of BACE1 inhibitors is adjusted between tolerable mechanism-based toxic effects and Aβ reduction will be crucial to achieve a safe AD therapy. Although Bace1null mice abolish amyloid generation, they present complex cognitive and neurochemical phenotypes suggesting that BACE1 has diverse physiological functions beyond APP processing. However, little was known of whether the Bace1null phenotypes were caused by absence of BACE1 during development or whether they would also be present when BACE1 activity was inhibited in adult brain. The group of Robert Vassar shows now that the conditional knockout mice for BACE1 clearly overcome most of the Bace1null defects. This strongly suggests that BACE1 plays very important roles during brain development.

    However, BACE1 conditional knockout mice still exhibit axonal disorganization in the mossy fiber pathway of the hippocampus. This phenotype mimics the CHL1null phenotype and has been previously linked to the abnormal processing of CHL1 by BACE1. L1 family cell adhesion molecules have been linked to diverse functions in the developing brain but little is known about their functions in adults. Ou-Yang’s findings once again emphasize that BACE inhibitors for Alzheimer's disease therapy should be monitored with caution and take into consideration potential mechanism-based side effects. Nonetheless, it is important to emphasize that the risk of mechanism-based toxic effects might depend on the level of BACE1 inhibition. What are the doses of inhibition that would allow BACE1 function in the mossy fiber pathway and effective reduction of amyloid levels? Finding ways to monitor the mossy fiber circuitry in ongoing clinical trials will be critical to understand the risks of BACE1 inhibition and its potential as a therapeutic target for the disease.

  3. This is a study the entire field has been waiting for, for a long time. After the initial finding that BACE1 is required for myelination via neuregulin 1 processing back in 2006 (!) (Hu et al., 2006; Willem et al., 2006), a number of other substrate-related phenotypes were observed in BACE1 knockout mice (Lichtenthaler et al., 2018). However, phenotypes like hypomyelination were most likely developmental phenotypes and may not be observed upon reduction of BACE1 in adulthood. The discrimination of developmental versus adult phenotypes is not just of academic interest, but is of greatest importance for the safety of current clinical trials using BACE1 inhibitors.

    Ou-Yang et al. now generated adult conditional BACE1 knockout mice, which allowed them to differentiate between developmental and adult phenotypes. Strikingly, they found that most phenotypes that were observed in germline-deleted BACE1 mice were not observed in the conditional knockouts. This sounds, at first blush, like good news, and may suggest that BACE inhibitors in clinical trials may show little if any substrate-related side effects.

    However, Ou-Yang and colleagues still found reduced length and massive disorganization of the hippocampal mossy fiber infrapyramidal bundle in adult BACE1 knockout mice. Moreover, this phenotype was apparently related to the reduction of BACE1-mediated cleavage of the neuronal cell adhesion protein close homolog of L1 (CHL1), which previously was shown to be a BACE1 substrate that is involved in axonal guidance (Dislich et al., 2015; Hitt et al., 2012; Zhou et al., 2012). If a similar phenotype were to be found after treating wild-type mice with BACE1 inhibitors, it would be a warning for clinical trials.

    But we are already aware that almost complete inhibition of BACE1 in humans could result in mechanism-based side effects. In a worst-case scenario, efficient BACE1 inhibition could even worsen the clinical outcome of Alzheimer patients.

    Furthermore, previous work from our lab also showed that BACE inhibition/knockout induces an alternative pathway for proteolytic processing of APP, which results in the enhanced production of a peptide we have called Aη-α (Willem et al., 2015). This peptide blocked neuronal activity in vivo, a finding that also raises concerns about BACE1 inhibitors and could ultimately worsen cognition in Alzheimer's disease patients. After the rather disastrous, and at least partially predictable, outcome of γ-secretase inhibition in humans, the field would be well advised to carefully address mechanism-based side effects of BACE1 inhibition—and that's exactly what Vassar and colleagues did in a careful and detailed manner.


    . Label-free Quantitative Proteomics of Mouse Cerebrospinal Fluid Detects β-Site APP Cleaving Enzyme (BACE1) Protease Substrates In Vivo. Mol Cell Proteomics. 2015 Oct;14(10):2550-63. Epub 2015 Jul 2 PubMed.

    . β-Site Amyloid Precursor Protein (APP)-cleaving Enzyme 1 (BACE1)-deficient Mice Exhibit a Close Homolog of L1 (CHL1) Loss-of-function Phenotype Involving Axon Guidance Defects. J Biol Chem. 2012 Nov 9;287(46):38408-25. PubMed.

    . Proteolytic ectodomain shedding of membrane proteins in mammals-hardware, concepts, and recent developments. EMBO J. 2018 Aug 1;37(15) Epub 2018 Jul 5 PubMed.

    . Control of peripheral nerve myelination by the beta-secretase BACE1. Science. 2006 Oct 27;314(5799):664-6. PubMed.

    . η-Secretase processing of APP inhibits neuronal activity in the hippocampus. Nature. 2015 Oct 15;526(7573):443-7. Epub 2015 Aug 31 PubMed.

    . The neural cell adhesion molecules L1 and CHL1 are cleaved by BACE1 protease in vivo. J Biol Chem. 2012 Jul 27;287(31):25927-40. PubMed.

Make a Comment

To make a comment you must login or register.


Therapeutics Citations

  1. Semagacestat
  2. Verubecestat

News Citations

  1. Lilly Halts IDENTITY Trials as Patients Worsen on Secretase Inhibitor
  2. Déjà Vu? AD Patients Again Look Worse on γ-Secretase Inhibitor
  3. BACE1 Conditional Knockouts Model Adult BACE Inhibition
  4. Verubecestat Negative Trial Data: What Does it Mean for BACE Inhibition?

Paper Citations

  1. . Phenotypic and biochemical analyses of BACE1- and BACE2-deficient mice. J Biol Chem. 2005 Sep 2;280(35):30797-806. Epub 2005 Jun 29 PubMed.
  2. . Control of peripheral nerve myelination by the beta-secretase BACE1. Science. 2006 Oct 27;314(5799):664-6. PubMed.
  3. . BACE1 deficiency causes altered neuronal activity and neurodegeneration. J Neurosci. 2010 Jun 30;30(26):8819-29. PubMed.
  4. . BACE1, a major determinant of selective vulnerability of the brain to amyloid-beta amyloidogenesis, is essential for cognitive, emotional, and synaptic functions. J Neurosci. 2005 Dec 14;25(50):11693-709. PubMed.
  5. . Learning spatial orientation tasks in the radial-maze and structural variation in the hippocampus in inbred mice. Behav Brain Funct. 2005 Apr 22;1(1):3. PubMed.
  6. . β-Site Amyloid Precursor Protein (APP)-cleaving Enzyme 1 (BACE1)-deficient Mice Exhibit a Close Homolog of L1 (CHL1) Loss-of-function Phenotype Involving Axon Guidance Defects. J Biol Chem. 2012 Nov 9;287(46):38408-25. PubMed.

Further Reading

Primary Papers

  1. . Axonal organization defects in the hippocampus of adult conditional BACE1 knockout mice. Sci Transl Med. 2018 Sep 19;10(459) PubMed.