Alzheimer's disease

A firm base for drug development

Article metrics

Alzheimer's disease is a progressive dementia in which massive deposits of aggregated protein-breakdown products — amyloid plaques and neurofibrillary tangles — accumulate in the brain. The amyloid plaques are thought to be responsible for the devastating mental decline in patients with Alzheimer's disease, and the main component of these plaques is a small peptide called amyloid β (Aβ). This peptide is made when its precursor, the amyloid precursor protein (APP), is cleaved, and two distinct proteases are thought to do this job. The β-secretase cleaves at the β-site (Fig. 1), whereas the γ-secretase cuts APP at the γ-site1.

Figure 1: Proposed mechanism by which the amyloid β (Aβ) peptide is generated through cleavage by the newly identified β-site APP-cleaving enzyme, BACE2,3,4,5.

The APP molecule is inserted into the lumen of the endoplasmic reticulum (ER) and then transported to the cell surface. Either in the Golgi, or in early endosomes after endocytosis, APP is cut by the transmembrane-spanning BACE. This cleavage results in secreted soluble APP (β-sAPP) and the membrane-associated carboxy-terminal fragment β (CTF-β). The CTF-β is a substrate for the γ-secretase (model adapted according to ref. 9), which cleaves its substrate in a symmetrical fashion in the transmembrane domain. This results in secreted Aβ and rapidly degraded CTF-β.

The identity of these secretases has long been sought, but half of the mystery, at least, is now over. Papers in Science2 and Molecular and Cellular Neuroscience3, as well as reports by Yan et al.4 and Sinha et al.5 (on pages 533 and 537 of this issue), describe the identification of the β-secretase. And these exciting results could, perhaps, accelerate the development of drugs to stop or even reverse the neurodegenerative process.

Development of any medication requires, among other things, validated animal models and well-defined molecular targets. For research into Alzheimer's disease we lacked both just a few years ago. Then disease-causing mutations were discovered in the APP and presenilin genes, allowing animal models to be developed6. But none of the disease genes were the type of classical drug target that drug hunters enjoy chasing7. Proteases such as the secretases, by contrast, would be excellent pharmacological targets — take HIV research, for example, where the identification of viral proteases in the mid-1980s led to the successful development of marketed protease inhibitors about ten years later8.

Four groups (not surprisingly, all from the pharmaceutical and biotechnology industries) have now identified a convincing candidate ‘β-site APP-cleaving enzyme’ (BACE), also known as Asp-2. Vassar and colleagues2 inserted, in a screening-like fashion, pools of unknown human genes (complementary DNAs) into a cell line that they knew expressed their target protease. Of 860,000 clones, Vassar et al. identified just one candidate protease (BACE) that increased the release of Aβ from their chosen cells. Naturally occurring APP mutations at the β-site cause early-onset Alzheimer's disease, and the authors found that mutant APP was a better substrate for BACE than normal APP. In an impressive series of cell-biological experiments, Vassar and colleagues went on to show that BACE fulfils all possible criteria for a candidate β-secretase.

Yan and colleagues4 took a very different approach. Based on pharmacological data, these authors knew that the β-secretase must have aspartyl-protease-like characteristics — acidic pH requirements and a signature sequence (aspartic acid–serine/threonine –glycine) found in most aspartyl enzymes. Using this information, Yan et al. first scanned the genome of the nematode worm Caenorhabditis elegans, which has been almost completely sequenced, for candidate proteases. They then went back to the vertebrate genome and identified homologues of the ten hits they obtained. Several candidates were further validated by extensive cellular (antisense) experiments. This approach (which is often known to produce false-negative results) allowed them to identify Asp2 as the β-secretase, using the same criteria as Vassar and colleagues.

It is very reassuring that BACE has also been identified by a classical protein-fractionation approach. Sinha and colleagues5 established that the β-secretase is a membrane-associated aspartyl protease with an optimal pH of 5.5. They then observed, like the other groups2,4, that this presumably aspartyl-like activity is unusual in that it is not inhibited by pepstatin, a typical aspartyl-protease inhibitor. But to purify BACE, Sinha and colleagues needed a molecular hook. They designed several variants of the APP sequence, spanning the β-site, including so-called transition-state analogues. Such analogues ‘freeze’ a bound protease in the act of cleaving the substrate. Sinha et al. then tested their analogues against crude preparations of human brain containing abundant BACE activity. They found that one amino-acid substitution, from aspartate to valine with a statine analogue, at position +1 (that is, on the carboxy-terminal side of the cleavage site) resulted in a potent inhibitor with a half-maximum inhibition at 30 nM. Using this molecular hook, Sinha and colleagues pulled out their candidate protease from human brain extracts, with a 300,000-fold enrichment.

Finally Hussain and colleagues3, like Yan et al., named their β-secretase Asp2 — suggesting they have more than one candidate protease. However, Hussain et al. did not reveal in their paper how they obtained their cDNA clone from their proprietary expressed sequence tag (EST) database. They did show, however, that point mutations in Asp2 (or BACE) at both of its two active sites (the aspartic acid–serine/threonine–glycine sequence) mean that it can no longer process APP to Aβ.

Ever since it became clear that proteases chop APP down to Aβ, these proteases have been prime targets for drug discovery. In the absence of molecular targets, cellular reporter systems have been used to develop compounds that reduce the amount of Aβ produced, and we may soon see the first clinical trials of these drugs. But the isolation of BACE means we can now screen for drugs that act directly on the target protease. Future structural information from X-ray diffraction studies of BACE with a bound inhibitor might give valuable insights into the design of new structural classes of inhibitors.

Several challenges remain, however. The BACE and its homologue BACE-2 belong to a new class of membrane-bound aspartyl proteases. Are there other BACE homologues? And, if so, will these have to be considered for selectivity screens in drug-optimization studies? We also do not know which other precursor molecules or cellular processes depend on proper BACE activity. Transgenic mice with these genes knocked out, either conditionally or totally, will be very useful for resolving such questions. Another problem is the subcellular location of BACE, in the lumen of the Golgi body and endosomes (Fig. 1). This means that inhibitors will have to cross at least two lipid bilayers — a formidable penetration hurdle for even small-molecular-weight compounds. Moreover, any BACE inhibitor has to pass the blood–brain barrier to find its target in neurons. New compounds will therefore need to have excellent pharmacokinetic properties.

Despite all of this, the identification of the β-secretase means that the path towards specific inhibitors is now set, and it is time not only to test the amyloid hypothesis (‘in vivo veritas’), but to find a way of halting this dreadful disease.


  1. 1

    Haass, C. & Selkoe, D. J. Cell 75, 1039–1042 (1993).

  2. 2

    Vassar, R. et al. Science 286, 735–741 (1999).

  3. 3

    Hussain, I. et al. Mol. Cell. Neurosci.

  4. 4

    Yan, R. et al. Nature 402, 533–537 (1999).

  5. 5

    Sinha, S. et al. Nature 402, 537–540 (1999).

  6. 6

    Price, D. L. et al. Annu. Rev. Genet. 32, 461–493 (1998).

  7. 7

    Drews, J. Nature Biotechnol. 14, 1516–1518 (1996).

  8. 8

    Ren, S. & Lien, E. J. Prog. Drug Res. 51, 3–31 (1998).

  9. 9

    Lichtenthaler, S. F. et al. Proc. Natl Acad. Sci. USA 96, 3053–3058 (1999).

Download references

Author information

Correspondence to Bart De Strooper or Gerhard König.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

De Strooper, B., König, G. A firm base for drug development. Nature 402, 471–472 (1999) doi:10.1038/44973

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.