The inhibitors of the β-site APP-cleaving enzyme 1 (BACE1) that are in clinical trials to prevent Alzheimer’s dementia block the activity of BACE2, as well. Alas, the function of this less-studied enzyme in the brain is entirely unknown. To shed some light, Bart De Strooper and colleagues at KU Leuven in Belgium analyzed the distribution of BACE2 in the brain, pinpointing spots of expression in neurons, glia, and oligodendrocytes. The investigators also identified BACE2 substrates in glia. Importantly, while BACE2 seemed quiet in normal mice, inflammatory cytokines boosted BACE2-mediated processing of at least one substrate, the vascular cell adhesion molecule 1 (VCAM1), and its concentration in CSF rose. The results appeared online February 15 in Life Science Alliance, a new open-access journal launched by partners EMBO Press, Rockefeller Press, and Cold Spring Harbor Laboratory Press.

  • BACE1 inhibitors also shut down BACE2, but little is known about BACE2 function.
  • In brain, BACE2 mostly appears in oligodendrocytes, but also in some neurons and astrocytes.
  • BACE2 lays low in normal brain, but ramps up with inflammation.

“This appears to be an important finding,” said Robert Vassar of Northwestern University, Chicago. “Not much is known about BACE2, but the possibility of BACE2 side effect liabilities is a real concern,” he wrote in an email to Alzforum.

The work was done in mice and will require confirmation in people. De Strooper says it is a step toward better understanding the risks involved in BACE2 inhibition. “While work up until now suggests that BACE2 inhibition is not much of a concern, we should be very careful,” he told Alzforum. “We’ve shown that BACE2 is present in the brain in different cell types, and there are substrates that will be affected by BACE1/2 inhibitors. The role of inflammation is important as AD patients display early inflammatory changes,” he said.

Second BACE.

In situ hybridization detects BACE2 (left) and BACE1 (right) mRNA in myelin basic protein-expressing striatal oligodendrocytes (pink). [Image from Voytyuk et al., LSA, 2018.]

BACE2 has always taken a back seat to BACE1 in the Alzheimer’s field. The enzyme is considered irrelevant to amyloid pathogenesis, because it is expressed at far lower levels than BACE1 in the brain, and because it does not contribute to APP cleavage in neurons (Dominguez et al., 2005). However, could it be relevant for producing side effects of BACE inhibitor treatment in other ways?

To survey BACE2 expression, first author Iryna Voytyuk coupled ultrasensitive in situ hybridization for BACE1 and BACE2 with cell-type-specific markers in brain slices from four-month-old mice. As expected, BACE2 expression was much lower, and more limited in distribution, than BACE1. While BACE1 is expressed in most neurons, Voytyuk detected BACE2 only in a few, with highest levels in the ventral hippocampus. BACE2 was present in astrocytes, but only in those lining the lateral ventricle. Oligodendrocytes throughout the brain expressed BACE2, which appeared along many fiber tracts.

The expression results agree with earlier RNA-sequencing studies that found BACE2 mainly in oligodendrocytes in the mouse CNS. Sten Linnarsson’s group at the Karolinska Institute in Stockholm has extensively characterized single cells from mouse central nervous system (Zeisel et al., 2015). Their latest unpublished data, which includes 265 cell types from the brain and peripheral nervous system, reveals definite BACE2 expression in newly formed and mature oligodendrocytes and in ependymal cells, a type of glia that line the ventricles and participate in the production of CSF. Those cells, rather than astrocytes, may be the BACE2-positive cells Voytyuk and De Strooper detect around the lateral ventricles, Linnarsson wrote in an email to Alzforum. The astrocytes are also there, but are negative for BACE2 in his data, Linnarsson wrote. Another RNAseq study (Zhang et al., 2014) reported highest levels of BACE2 in oligodendrocytes, with far lower expression in neurons, astrocytes, or microglia.

Next, the investigators wanted to identify BACE2 substrates in the brain. They first tried mass spec-based proteomic analysis of CSF, looking for changes in protein composition in BACE1/2 double-knockout mice compared with BACE1 knockouts. Voytyuk reasoned that loss of BACE2 would decrease shedding of its substrates in the brain, which would show up in CSF. Unfortunately, the approach was not very informative: Of 579 CSF proteins quantified, 60 showed a significant decrease or increase in mice lacking BACE2, but most were cytosolic proteins, and not direct BACE substrates. They did find reduction in several previously identified BACE1 substrates, suggesting they are also cleaved by BACE2. They did not detect any new BACE2 substrates. 

To take a more targeted approach, Voytyuk turned to cultured primary glia, where BACE2 sheds its substrates directly into the medium. She isolated mixed glia from BACE1 knockout mice, treated them with the BACE1/2 inhibitor CpJ, and then analyzed the extracellular medium by mass spec. Of 246 proteins quantitated, four were decreased more than 30 percent by the inhibitor. All four—VCAM1, delta and notch-like epidermal growth factor-related receptor (DNER), fibroblast growth factor receptor 1 (FGFR1), and plexin domain containing 2 (PLXDC2)—are single-transmembrane proteins with large, glycosylated extracellular domains released by BACE2. In additional validation experiments in cultured cells, Voytyuk determined that BACE2 was solely responsible for VCAM1 and DNER cleavage, while both BACE1 and BACE2 can process FGFR1 and PLXDC2.

Are the substrates also cleaved in vivo? The scientists looked for evidence of BACE2 processing in mice by western blotting of CSF proteins. They detected the shedded portion of known BACE1 substrate SEZ6 in CSF of wild-type or BACE2 knockout mice, but not in BACE1 or double-knockout, as expected. However, VCAM1 and DNER were present in equal amounts in wild-type, BACE1, BACE2, or double-knockout CSF, suggesting they were not cleaved at all by BACE2 in vivo. FGFR1 and PLXDC2 were not tested, because the scientists have no good antibodies to detect the endogenous proteins.

What was the difference between cell culture and in vivo? Voytyuk knew inflammatory stimuli increase VCAM1 expression, so she tried adding the inflammatory cytokines TNF or IL1β to the glial cultures. The result was increased overall VCAM1 expression and a significant uptick in BACE2-dependent shedding, which was completely blocked by CpJ. When the investigators evoked brain inflammation in vivo by injecting TNF, they saw a 2.2-fold uptick in CSF VCAM1 in wild-type mice, but no change in the BACE2 knockout.

This indicates that BACE2 may play a role in inflammation, said De Strooper. “We’ve studied the BACE2 knockouts for a while, and they don't show much of a phenotype. We probably need to have some additional stress on the mice, like inflammation, and then phenotypes may emerge which could inform us about the function of BACE2,” he said. “In addition, now that we’ve identified substrates in glia, we can start to elucidate the role of BACE2 in those cells,” De Strooper said.

The work suggests VCAM1’s potential as a CSF marker of brain BACE2 activity or inflammation. Previously, Doug Galasko at University of California, San Diego, evaluated CSF VCAM1 as a potential indicator of vascular injury, and found it increased with age, but not AD status (Li et al., 2017; Li et al., 2015). Galasko told Alzforum he is unaware of data on VCAM1 in CSF from people receiving BACE inhibitors. He doubts that cleavage by BACE will alter bulk levels of VCAM1 in the CSF, because while most BACE2 seemed to be in neurons, astrocytes, and oligodendrocytes, most VCAM1 likely originates from pathways in vascular endothelial cells (see complete comment below). De Strooper would very much like to do this experiment, but so far has been unable to get access to CSF samples from BACE trials, he told Alzforum.

Philip Wong, Johns Hopkins University in Baltimore, suggested that brain and peripheral tissues of patients treated with the discontinued Merck BACE inhibitor verubecestat be examined for evidence of alterations in BACE2 substrates, and for possible abnormalities related to these targets (see complete comment below).—Pat McCaffrey


  1. I read this paper with interest. There are many substrates besides VCAM1 that might be altered as a result of BACE-2 inhibition, with potentially unknown effects on the brain. The magnitude of alteration was clearly higher in the presence of TNF-α, which may be relevant to AD as this may be produced by microglia and astrocytes.

    I am not clear that BACE2 cleavage would alter overall bulk levels of VCAM1 in CSF, because the BACE2 mRNA was predominantly expressed in neurons, oligodendrocytes, and cells lining the lateral ventricle, and VCAM1 release may be the predominant result of other pathways taking place in vascular cells. 

    We had previously looked at CSF VCAM1 with the idea that it might be a vascular-related biomarker, because it is expressed at high levels in CNS endothelial cells. I am unaware of VCAM1 or other proteins being reported from CSF in humans receiving BACE inhibitors. The neuropsychiatric symptoms reported in the verubacestat clinical trial have received some discussion at CTAD 2017 (Dec 2017 conference news), and I wonder if a subtle dysregulation of secreted molecules cleaved by BACE2 might have contributed or serve as markers of those symptoms?

  2. This is a good study that raises the potential side effects of current BACE inhibitors (which do not discriminate BACE1 versus BACE2) due to impact on substrates of BACE2, particularly under pathological conditions.

    As it is not clear whether this information can be translated directly to humans, or whether such off-target effects would be substantial in patients with AD, the main lesson here is that it will be important to monitor for potential side effects related to pathways impacted by BACE2 inhibition during BACE1 inhibitor clinical trials.

    It would be useful to examine the brains and periphery of patients treated with the Merck BACE inhibitor verubecestat for evidence of whether these substrates of BACE2 are altered, and secondly whether abnormalities related to these targets can be shown.​

Make a Comment

To make a comment you must login or register.


Therapeutics Citations

  1. Verubecestat

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. . Brain structure. Cell types in the mouse cortex and hippocampus revealed by single-cell RNA-seq. Science. 2015 Mar 6;347(6226):1138-42. Epub 2015 Feb 19 PubMed.
  3. . An RNA-sequencing transcriptome and splicing database of glia, neurons, and vascular cells of the cerebral cortex. J Neurosci. 2014 Sep 3;34(36):11929-47. PubMed.
  4. . Cerebrospinal fluid biomarkers for Alzheimer's and vascular disease vary by age, gender, and APOE genotype in cognitively normal adults. Alzheimers Res Ther. 2017 Jul 3;9(1):48. PubMed.
  5. . Increased CSF E-Selectin in Clinical Alzheimer's Disease without Altered CSF Aβ42 and Tau. J Alzheimers Dis. 2015;47(4):883-7. PubMed.

Further Reading

No Available Further Reading

Primary Papers

  1. . BACE2 distribution in major brain cell types and identification of novel substrates. Life Science Alliance Feb 2018