We initially set out to determine the chromosomal localization of BACE by identifying mapped ESTs (expressed sequence tags) encoding identical amino acid sequences. For this purpose, the complete amino acid sequence of BACE was used to screen various GenBank databases in the National Center for Biotechnology Information (NCBI) website (www.ncbi.nlm.nih.gov). Specifically, BACE was first screened against the "human ests" database using the program "tblastn". ESTs with identical coding sequences included AI290317, R55398, AF150387, H60273, AA293614, and AA298516, among others. The largest of these was AF150387 (1114 nucleotides). The previous chromosomal mapping of several of these ESTs permitted the chromosomal localization of BACE to the long arm of chromosome 11. Among these, R55298 had been previously mapped to 11q23.2-11q23.3 (see accompanying Technical Response by Fan et al.). A similar screen of the "dbest" database identified the exact match EST, H60581. This EST had been previously localized to centimorgan position 121.037 on the summary map for chromosome 11, which can be accessed in "The Genetic Location Database" at web address (cedar.genetics.soton.ac.uk/pub/chrom11/map.htm).

Additional screening with the BACE amino acid sequence led to the retrieval of two full-length cDNAs (AF050171, AF117892) from the "nr" database. Both of these cDNAs encode a novel transmembrane aspartyl protease (BACE2) that is highly homologous to BACE. One of the BACE2 cDNAs, AF050171, was previously localized to the long arm of chromosome 21 at 21q22.3. Further screening of the "htgs" database led to the retrieval of a phase 2 ordered cosmid contig (AJ011929; HSMX1). By comparing the nucleotide sequence of AF050171 with that of the cosmid contig, we determined that the contig contains a portion of the genomic sequence of BACE2 starting at nucleotide 418 and ending at nucleotide 1863. The cosmid contig containing the BACE2 gene is derived from chromosome 21, specifically between the loci D21S349 and MX1. The map position of these loci is consistent with the cytogenetic localization of BACE2 to chromosome 21q22.2-21q22.3. This portion of chromosome 21 contains the obligate Down syndrome (DS; trisomy 21) region, a minimal region of chromosome 21 that must be inherited in a trisomic state for expression of the constellation of features and symptoms characteristic of DS.

BACE and BACE2 exhibit 52% amino acid sequence identity and 68% similarity. BACE2 contains two aspartyl protease active sites at virtually identical positions to the two active sites in BACE (Fig. 1). The amino acid sequences of BACE and BACE2 are divergent only at the COOH-terminal 30 amino acids and the NH₂-terminal 80 amino acids. The divergent COOH-terminal regions in BACE and BACE2 both contain predicted single membrane-spanning domains (Fig. 1). The significant sequence similarity and predicted topology of BACE and BACE2 suggest that both proteins are transmembrane aspartyl proteases and that both are candidates for the -secretase involved in AD.

It is interesting that BACE2 maps to the obligatory DS region of chromosome 21. This implies that in addition to the contribution of a third copy of the APP gene (3), a third copy of BACE2 may also significantly contribute to enhanced amyloid- deposition inevitably observed in the brains of middle-aged DS patients. Based on these data, DS patients would be expected to not only express more substrate for amyloid- production (APP) but also more of a protease (BACE2) that would be predicted to cleave at the -secretase site of APP, thereby exacerbating the release of amyloid-.

The identification of BACE and BACE2 as strong candidates for -secretase provides novel molecular targets for mechanistic (protein-based) high-throughput drug screening for specific protease inhibitors. Identification of such inhibitors could afford a potentially powerful new generation of AD and DS therapeutics. Furthermore, the precise chromosomal mapping of BACE and BACE2 to the long arms of chromosomes 11 and 21, respectively, should greatly facilitate attempts to test these genes for genetic linkage to AD.

Aleister J. Saunders
Tae-Wan Kim
Rudolph E. Tanzi
Genetics and Aging Research Unit,
Department of Neurology,
Massachusetts General Hospital,
Harvard Medical School,
Charlestown, MA 02129, USA

REFERENCES

1. B. De Strooper, et al., Nature 391, 387 (1998) [CrossRef] [Medline] ; M. S. Wolfe, et al., Nature 398, 513 (1999) [CrossRef] [Medline] .
2. R. Vassar, et al., Science 286, 735 (1999) [Abstract/Free Full Text] .
3. R. E. Tanzi, et al., Science 235, 880 (1987) [Abstract/Free Full Text] .

Response: Recently, we described a novel transmembrane aspartic protease, BACE, which exhibits all the known characteristics of the -secretase thought to be implicated in Alzheimer's disease (AD) (1). BACE directly cleaves amyloid precursor protein (APP) substrates to initiate the formation of amyloid  peptide (A)--the primary component of the amyloid deposits found in AD brains. Therefore, BACE is a prime therapeutic target for the treatment of AD. It is of great interest to determine whether alleles of the BACE gene exist that confer either an increased or decreased risk of AD.

Shortly after cloning BACE, we identified a BACE homolog, referred to as BACE2 by Saunders et al. (see Technical Comment), by searching an Amgen-expressed sequence tags database. We are currently investigating the role of BACE2 in the processing of APP and its metabolites. Although it is interesting that the BACE2 gene maps to chromosome 21 in the obligatory Down syndrome region, the potential role of BACE2 as a -secretase candidate is as yet unclear. Our BACE antisense inhibition experiments suggest that BACE2 is not likely to be the major -secretase, at least in HEK 293 cells (1). BACE antisense oligonucleotides (AS oligos) inhibit both BACE mRNA and -secretase activity in parallel, with up to 80% efficiency. The efficiency of AS oligo delivery into HEK 293 cells is also approximately 80%, suggesting that BACE inhibition may be 100% in cells transfected with BACE AS oligos. The residual 20% of -secretase cleavage we observe is likely to derive from BACE activity in cells that have not been transfected with BACE AS oligos. Moreover, BACE AS oligos are not complementary to BACE2 mRNA and therefore cannot cross-hybridize with and inhibit BACE2 mRNA. We conclude that BACE is the major -secretase, at least in HEK 293 cells, and that BACE2 does not compensate for loss of BACE function in these cells.

We also mapped the BACE gene using radiation hybrid (RH) analyses to human chromosome 11q23.3. The analyses were performed using the Genebrige4 and the Stanford G3 RH panels from Research Genetics. The GenebridgeG4 panel mapping data indicated that the BACE gene mapped 3.36 cR₃₀₀₀ from WI-6355 on the placement map. On the comprehensive Whitehead RH map of chromosome 11, BACE resides between WI-13525 and WI-14206. Data obtained with the StanfordG3 RH panel (Fig. 1A) place the BACE gene within 0 cR₁₀₀₀₀ of SHGC-4187 (LOD 12.33) and confirmed the localization obtained with the G4 panel. The cytogenetic locations of these markers were approximated by submitting the markers to the Genome Database (www.gdb.org) and viewing the maps of the region, resulting in a localization to 11q23.3 (Fig. 1B). Furthermore, YACs containing markers D11S1327 and D11S1336, which flank SHGC-4187, physically map to 11q23.3 (2), thereby confirming the assignment of the BACE gene to 11q23.3.

A search of the OMIM database (www3.ncbi.nlm.nih.gov/omim/) revealed that Jacobsen Syndrome (MIM 147791) and Familial Nonchromaffin Paragangliomas (MIM 168000) result from deletions in the 11q23 region. The BACE gene, which maps to SHGC-4187, is located slightly outside of the critical regions described for these diseases and is therefore an unlikely candidate gene for these disorders (3, 4).

The chromosomal localization and genes involved in several forms of familial AD are known (MIM 104300). These include the APP gene of chromosome 21, presenilin 1 and 2 genes on chromosomes 14 and 1, respectively, and an association of the ApoE4 allele on chromosome 19 to AD susceptibility. In addition, two genome scans for late onset AD have been recently completed and neither have demonstrated linkage to chromosome 11 (5, 6). Although these data suggest that the BACE gene is not likely to be a major cause of familial AD, further studies are required to determine whether specific alleles of the BACE gene may contribute to the genetic risk of AD (see accompanying Technical Comment by Saunders et al.). Mutations in the BACE gene that increase the activity or expression of the BACE enzyme may increase the risk of AD. However, such mutations would necessarily be limited to a small number of specific sites within the BACE gene and thus are expected to occur infrequently within the population. In contrast, mutations leading to partial or complete inactivation of the BACE enzyme could reduce the risk of AD. Although it is unclear whether the BACE gene may be genetically associated with AD, as the rate limiting enzyme involved in amyloid- formation, the BACE aspartic protease is an important new target for the development of AD therapeutics.

Wei Fan
Brian D. Bennett
Safura Babu-Kahn
Yi Luo
Jean-Claude Louis
Michael McCaleb
Martin Citron
Robert Vassar
William G. Richards
Neuroscience Department,
Amgen,
Thousand Oaks, CA 91320
E-mail: brichard{at}amgen.com

REFERENCES

1. R. Vassar, et al., Science 286, 735 (1999) .
2. P. Bray-Ward, et al., Genomics 32, 1 (1996) [CrossRef] [Web of Science] [Medline] .
3. L. A. Penny, et al., Am. J. Hum. Genet. 56, 676 (1995) [Web of Science] [Medline] .
4. B. E. Baysal, et al., Hum. Genet. 104, 219 (1999) [CrossRef] [Web of Science] [Medline] .
5. P. Kehoe, et al., Hum. Mol. Genet. 8, 237 (1999) [Abstract/Free Full Text] .
6. M. A. Pericak-Vance, et al., Neurobiol. Aging 19, S39 (1998) [CrossRef] [Web of Science] [Medline] .