document

984 nature neuroscience • volume 2 no 11 • november 1999

articles

Familial British dementia, an autosomal dominant neurodegener-ative disorder, is characterized by progressive spastic tetraparesis,cerebellar ataxia and dementia2,3. The principal pathological hall-marks of FBD include the presence of non-neuritic plaques andamyloid angiopathy in the cerebellum, hippocampus, amygdalaand cerebral cortex, hippocampal neurofibrillary tangles andischemic white matter changes1–3. A peptide termed ABri was orig-inally identified as a component of highly insoluble amyloid fibrilsin leptomeninges and parenchymal deposits from a patient withFBD4. This peptide derives from a larger, 266-amino-acid precursorprotein, termed BRI. Affected individuals from the FBD pedigreehave a T to A transversion (TGA–AGA) at the termination codon ofthe BRI protein4. As a result, an arginine codon is created, and trans-lation of the open reading frame is extended for an additional tenamino acids, then terminates; the extended open reading frameencodes a mutant BRI molecule, termed BRI-L, of 277 amino acids.

ABri purified from the leptomeningeal amyloid fraction has anMr of 3,935, suggesting that the peptide is generated by endopro-teolytic processing of mutant BRI between arginine 243 and glu-tamic acid 244 (ref. 4). The amino-acid sequence of BRI that flanksthe predicted cleavage site, ...KGIQKR⇓ EA, is highly reminiscent ofa consensus sequence required for processing by the prohormoneconvertase, furin5,6. Here we demonstrate that BRI adopted thetopology of type-II integral membrane proteins. Moreover, bothBRI and BRI-L were constitutively processed between Arg243 andGlu244 in transfected mammalian cells, leading to the productionand secretion of carboxyl-terminal peptides; furin seems to be acritical, if not the only, factor that mediates endoproteolysis of theprecursor proteins. Secretion of peptides derived from the mutantBRI-L precursor was enhanced compared to peptides generatedfrom BRI, suggesting that the carboxyl-terminal eleven amino acidsin mutant BRI effect furin-mediated proteolysis in a dominant

fashion. Finally, electron microscopy (EM) studies revealed thatABri peptides assemble into irregular, short fibrils. Collectively,our results support the view that enhanced furin-mediated pro-cessing of mutant BRI generates fibrillogenic peptides that mayinitiate pathogenesis in FBD.

RESULTSTopology and processing of BRI in mammalian cellsWe chose to generate BRI cDNA that encodes BRI containing epi-tope tags at the amino and carboxyl termini to aid in the identi-fication of BRI-derived proteolytic derivatives. BRI cDNA wasgenerated by polymerase chain reaction (PCR) using reverse-tran-scribed human brain RNA as a template. The sense primer encod-ed the N-terminal six amino acids of BRI, and the antisenseprimer was complementary to sequences encoding the last eightamino acids of BRI fused to the C-terminal seven amino acids ofamyloid precursor-like protein 1 (APLP1). The PCR product wascloned in-frame with sequences encoding a 12-amino-acid seg-ment of the c-myc oncoprotein (Fig. 1a). Nucleotide sequencingof the resulting BRI cDNA confirmed its identity to the publishedhuman BRI cDNA sequence4 and a cDNA, ITM2B, identified in ascreen for transcripts upregulated during chondro-osteogenic dif-ferentiation7,8. In parallel, we generated BRI-L cDNA that encod-ed an epitope-tagged BRI with the predicted 11-amino-acidextension at the carboxyl terminus, a result of the T–A transver-sion at the normal BRI termination codon (Fig. 1a).

Secondary structure algorithms9,10, and the presence of ahydrophobic stretch between amino acids 52 and 76, led to theprediction that BRI is a type II integral membrane protein4. Toexamine the topology of BRI, we transiently transfected epitope-tagged BRI cDNA into mouse neuroblastoma (N2a) cells, thenhomogenized cells with a ball-bearing homogenizer to insure

Furin mediates enhanced productionof fibrillogenic ABri peptides infamilial British dementia

Seong-Hun Kim1, Rong Wang2, David J. Gordon3, Joseph Bass4, Donald F. Steiner4, David G.Lynn6, Gopal Thinakaran1, Stephen C. Meredith5 and Sangram S. Sisodia1

1 Department of Neurobiology, Pharmacology and Physiology, 3Medical Scientists Training Program and Department of Biochemistry and Molecular Biology, 4HowardHughes Medical Institute, 5Department of Pathology, 6Department of Chemistry, The University of Chicago, Abbott 316, 947 East 58th Street, Chicago, Illinois 60637, USA

2 Laboratory of Mass Spectrometry, Rockefeller University, New York, New York 10021, USA

Correspondence should be addressed to S.S.S. ([email protected])

The genetic lesion underlying familial British dementia (FBD), an autosomal dominant neurodegen-erative disorder, is a T–A transversion at the termination codon of the BRI gene. The mutant geneencodes BRI-L, the precursor of ABri peptides that accumulate in amyloid deposits in FBD brain. Wenow report that both BRI-L and its wild-type counterpart, BRI, were constitutively processed by theproprotein convertase, furin, resulting in the secretion of carboxyl-terminal peptides thatencompass all or part of ABri. Elevated levels of peptides were generated from the mutant BRIprecursor. Electron microscopic studies revealed that synthetic ABri peptides assembled intoirregular, short fibrils. Collectively, our results support the view that enhanced furin-mediatedprocessing of mutant BRI generates fibrillogenic peptides that initiate the pathogenesis of FBD.

© 1999 Nature America Inc. • http://neurosci.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://n

euro

sci.n

atu

re.c

om

nature neuroscience • volume 2 no 11 • november 1999 985

minimal disruption of organellar integrity11. A post-nuclear, mem-brane fraction was prepared by centrifugation at 100,000 × g, andaliquots of the membrane fraction were digested with increasingamounts of proteinase K. Each of the resulting reactions weredivided, fractionated by SDS-polyacrylamide gel electrophoresisand analyzed by western blotting with 9E10 (ref. 12) or CT11 (ref. 13) antibodies, specific for the myc- and APLP1-epitope tags,respectively (Fig. 1b). The ∼ 40 kDa, full-length BRI was detectedwith either antibody, and this molecule was resistant to digestionat 0.25 µg per ml proteinase K (Fig. 1b, lanes 1 and 2, 1′ and 2′).However, N-terminal, 9E10-immunoreactivity was markedlyreduced at 2.5 µg per ml proteinase K (Fig. 1b, top, lane 3) andalmost completely abolished at 25 µg per ml proteinase K (Fig. 1b, top, lane 5). Coincident with loss of N-terminal, 9E10-immunoreactivity, an ∼ 34 kDa polypeptide, reactive with the car-boxyl-terminal, CT11 antibody, became apparent (Fig. 1b, bottom). These data suggest that the hydrophobicsequence encompassing residues 52 and 76 serves as a membrane-spanning domain, and thus, affords protection of the lumenaldomain from proteinase K digestion. Hence, we conclude that thechimeric BRI molecule used in our studies adopts the topologyof a type II integral membrane protein.

To examine the metabolism of BRI and BRI-L, we transient-ly transfected cDNA encoding these molecules into African greenmonkey kidney (COS-1) or N2a cells, and analyzed detergentlysates of transfected cells by western blotting with CT11 or 9E10antibodies. As expected, the CT11 antibody detected ∼ 40 kDa

BRI and ∼ 42 kDa BRI-L in lysates of COS-1 or N2a cells (Fig. 1c,lanes 2, 5 and 3, 6). On the other hand, the 9E10 antibody detect-ed a doublet of ∼ 40 kDa and ∼ 37 kDa in lysates of cells express-ing BRI (Fig. 1c, lanes 2′, 5′) and ∼ 42 kDa and ∼ 37 kDa species inlysates of cells expressing BRI-L (Fig. 1c, lanes 3′, 6′). In paral-lel, we detected CT11-reactive peptides of ∼ 3 kDa and ∼ 4 kDa inthe medium of COS-1 and N2a cells expressing BRI and BRI-L,respectively (Fig. 1d, lanes 2, 5 and 3, 6). To establish the iden-tity of secreted peptides, we immunoprecipitated the peptidesthe from conditioned medium of N2a cells with CT11 antibody,and analyzed the recovered material by matrix-assisted, laser des-orption, time-of-flight, mass spectrometry (Fig. 1e). These analy-ses showed that N2a cells expressing BRI or BRI-L secreteddiscrete peptides of 3557.7 and 4881.5 Mr, respectively (Fig. 1e),which includes the mass of the seven-amino-acid APLP1 epitopetag. These studies indicate that endoproteolytic cleavage of theprecursor proteins occurs between Arg243 and Glu244.

BRI processing is mediated by furinThe sequence immediately N-terminal to the BRI cleavage site,...Arg-Gly-Ile-Gln-Lys-Arg, is highly reminiscent of the consen-sus recognition site for the subtilisin-like, proprotein convertase,furin5,6,14–17. To examine the role of furin in BRI endoproteolysis,we transiently transfected cDNA encoding BRI or BRI-L into Chi-nese hamster ovary (CHO-K1) cells, or a CHO-K1 derivative,RPE.40, which is furin-deficient18–20 (Fig. 2). Transfected cellswere biosynthetically labeled with [35S]cysteine for three hours,

articles

Fig. 1. Topology and pro-cessing of BRI in mam-malian cells. (a) Schematicrepresentation of myc- andAPLP1-epitope-tagged, wild-type human BRI and FBD-associated mutant BRI-Lmolecules analyzed in thisstudy. (b) Analysis of BRItopology in microsomalmembranes using pro-teinase K treatment.Immunoblots were probedwith 9E10 (top), or CT11(bottom) antibodies.Arrow, a protected ∼ 34kDa CT11-reactivespecies. (c, d) BRI and BRI-L expression in transfectedmammalian cells. COS-1(lanes 1–3) or N2a (lanes4–6) cells were transientlytransfected with emptypRK5 vector (lanes 1, 4) orcDNAs encoding epitope-tagged BRI (lanes 2, 5) orBRI-L (lanes 3, 6), anddetergent lysates of cellsand conditioned mediumwere analyzed by westernblot. (c) Cell lysates werefractionated by 16.5%Tris/Tricine (top) or 10% Tris/glycine (bottom) SDS-polyacrylamide gel electrophoresis and blotted with CT11 (top) or 9E10 (bottom) antibodies. (d) Conditioned medium of transfected cells was immunoprecipitated with CT11 antibody, separated by 16.5% Tris/Tricine SDS-polyacrylamide gelelectrophoresis, and immunoblotted with CT11. (e) Mass spectrometry of secreted peptides. Conditioned medium of N2a cells transfected withempty pRK5 vector (top), pBRI (middle) or pBRI-L (bottom) was immunoprecipitated with CT11 and analyzed by matrix-assisted laser desorptiontime-of-flight mass spectrometry. In (b–d), molecular weight markers are in kDa.

a b

ce

d


199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://n

euro

sci.n

atu

re.c

om


articles

and detergent lysates of cells, and the conditioned media wereimmunoprecipitated with CT11 antibody. As expected, CT11 anti-body detected ∼ 40 kDa BRI and ∼ 42 kDa BRI-L in lysates of trans-fected CHO-K1 cells (Fig. 2, lanes 2, 4). Parallel analysis of theconditioned medium of transfected CHO-K1 cells revealed thepresence of ∼ 3 kDa BRI- and ∼ 4 kDa BRI-L C-terminal peptides,respectively (Fig. 2, lanes 2′ and 4′). In con-trast, we failed to detect a CT11-precip-itable peptide in medium of RPE.40 cellsexpressing either BRI or BRI-L (Fig. 2, lanes

7′ and 9′), suggesting that endoproteolysis of the precursor pro-teins required furin. To verify that expression of furin was bothnecessary and sufficient for processing BRI and BRI-L, we cotrans-fected BRI or BRI-L cDNA with cDNA encoding human furininto RPE.40 cells. Cotransfection of furin led to the reappearanceof ∼ 3 kDa and ∼ 4 kDa peptides in the medium (Fig. 2, lanes 8′and10′). These results confirm the view that furin mediates endo-proteolytic processing of the precursors, resulting in the produc-tion of secreted C-terminal peptides.

Enhanced processing of BRI-L by furinGiven a role for furin in mediating BRI endoproteolysis, it wascurious that the levels of secreted peptides in medium of CHO-K1 or RPE.40 cells that were derived from the BRI-L precursorwere consistently elevated over the peptides generated from theBRI precursor (Fig. 2). We quantified the levels of BRI- and BRI-L-derived peptides in the medium of N2a cells that were transientlytransfected with BRI or BRI-L cDNA, individually, or in combi-nation with furin cDNA. Transfected cells were biosyntheticallylabeled with [35S]cysteine for 3 hours, and then cellular BRI andsecreted BRI derivatives were immunoprecipitated with CT11 anti-body (Fig. 3a and b). Levels of immunoprecipitated peptides inthe conditioned medium were quantified by phosphorimagingand statististically analyzed using ANOVA followed by Scheffe’spost-hoc test (Fig. 3c). Full-length BRI and BRI-L precursors and C-terminal peptides were found in cell lysates. The latter observa-tion indicated that a fraction of precursor proteins are processed byendoproteolysis in intracellular compartments, consistent withmany reports that the preponderant steady-state distribution andactivity of furin is within the trans-Golgi network (for review, seerefs. 5, 6). More importantly, we observed an increase in the levelsof secreted peptides generated from the BRI-L precursor, com-pared to cells expressing BRI (Fig. 3b, compare lanes 3 and 1, quan-tified in Fig. 3c), similar to our results in CHO-K1 and RPE.40cells (Fig. 2). These results suggested that the carboxyl-terminalextension present in the BRI-L molecule had a dominant effect onproteolysis at a scissile bond that most likely is mediated by furin.Although coexpression of furin only modestly enhanced produc-

Fig. 2. Furin is required for BRI processing. CHO-K1 (lanes 1–5) orfurin-defective, RPE.40 cells (lanes 6–10) were transiently transfectedwith empty pRK5 vector (lanes 1, 6), pBRI (lanes 2-3, 7-8) or pBRI-L(lanes 4, 5, 9, 10) without (lanes 1, 2, 4, 6, 7, 9) or with (3, 5, 8, 10)human furin cDNA, then labeled with [35S]cysteine for 3 h. Cell lysates(top) or conditioned media (bottom) were immunoprecipitated withCT11, and recovered proteins were fractionated by 16.5% Tris/TricineSDS-polyacrylamide gel electrophoresis. The bands were visualized byphorphorimaging. Molecular weight markers are in kDa.

Fig. 3. Enhanced processing of BRI-L by furin.N2a cells transiently expressing BRI (lanes 1, 2)or BRI-L (lanes 3, 4), without (lanes 1, 3) orwith (lanes 2, 4) furin were labeled with[35S]cysteine for 3 h. Cell lysates (a) and condi-tioned medium (b) were immunoprecipitatedwith CT11 and analyzed by 16.5% Tris/TricineSDS-polyacrylamide gel electrophoresis. Theexperiment was done in triplicate, and a repre-sentative set is shown. (c) The intensity of eachband was quantified by phosphorimaging, andrelative levels of secreted peptides normalizedto those of full-length molecules were calcu-lated and expressed as the mean (± s.e.) of trip-licate experiments. *p < 0.01, **p < 0.001 byScheffe’s post-hoc test, significant differencebetween secreted peptides generated fromcells expressing BRI (white bars) and BRI-L(gray bars). In (a) and (b), molecular weightmarkers are in kDa.

a b

c

Furin

*

Furin

**


199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://n

euro

sci.n

atu

re.c

om

nature neuroscience • volume 2 no 11 • november 1999 987

articles

tion of secreted peptides from BRI or BRI-L (Fig. 3b and c), thelevels of secreted peptides generated from cells expressing the BRI-L precursor were considerably elevated compared to cells express-ing the BRI precursor protein (19% versus 6%, respectively;quantified in Fig. 3c). These results reinforce our view that theextended C-terminal peptide in the BRI-L precursor has a domi-nant effect on endoproteolysis at the furin site.

Electron microscopyAlthough histological studies indicate that the amyloid depositsin FBD patients exhibit yellow-green birefringence under polar-ized light after staining with Congo Red4, neither ultrastructuralanalyses of these deposits, nor information about the structureof purified ABri fibrils from leptomeninges and parenchyma ofFBD patients, have emerged. To address this issue, we synthe-sized the 34-amino-acid mutant ABri peptide and incubated itat room temperature for one week. At the end of this period, thesamples were centrifuged, and the precipitates were analyzed byEM. The ABri peptides formed fibrils with an average diameter of∼ 50 Å, though they were irregular and varied from ∼ 40 to ∼ 75 Å(Fig. 4a), quite distinct from the linear, unbranched fibrils, 90 Åin diameter, generated by Aβ1–40 (Fig. 4b) . The ABri fibrils weretortuous and showed occasional branch points or crossing, and insome areas, the fibrils were highly aggregated and matted.Notably, and in contrast to the highly fibrillogenic ABri peptidegenerated from mutant BRI-L, the normal C-terminal 23-amino-acid peptide derived from the wild-type BRI precursor is highlysoluble in neutral aqueous buffers, and we have observed neitherfibril formation nor amorphous precipitations of this peptide(data not shown).

DISCUSSIONA mutation at the termination codon of the BRI gene is the under-lying genetic defect in FBD, an autosomal dominant neurode-generative disorder4. The mutant BRI gene encodes BRI-L, theprecursor of ∼ 4 kDa ABri peptides that accumulate in lep-tomeningeal and parenchymal deposits in brains of FBD patients4.The present report offers several insights on the structure andmetabolism of BRI-L and the nature of ABri fibril assemblies.Using protease-protection strategies, we confirmed that BRIadopts the topology of type II integral membrane proteins. Fur-thermore, both BRI and BRI-L undergo endoproteolytic pro-cessing when ectopically expressed in cultured mammalian cells.Endoproteolysis of these proteins occurs between Arg243 andGlu244, resulting in the generation of ∼ 3 kDa and ∼ 4 kDa pep-

tides, respectively, that are released into the conditioned medium.Because the sequence N-terminal to the scissile bond resembleda recognition site for furin5,6, we expressed BRI or BRI-L in CHO-K1 cells, or the CHO-derived cell line, RPE.4018–20, which lacksactive furin. We observed secreted peptides from transfectedCHO-K1 cells expressing BRI or BRI-L, but failed to observe thesepeptides in the medium of RPE.40 cells. Coexpression of BRI orBRI-L and furin in RPE.40 cells complemented the furin defi-ciency, leading to secretion of ∼ 3 and ∼ 4 kDa peptides, respec-tively. Thus, furin seems to be both necessary and sufficient forendoproteolytic processing of BRI and BRI-L and the productionof secreted, carboxyl-terminal peptides.

Although BRI endoproteolysis was mediated by furin, the levelsof secreted peptides in medium of CHO-K1, RPE.40 or N2a cellsthat were derived from the BRI-L precursor were markedly elevatedover the peptides generated from the BRI precursor. These resultssuggested that the carboxyl-terminal extension present in the BRI-Lmolecule had a dominant effect on proteolysis at the furin site. Howmight this occur? Although several mechanisms might be enter-tained, we find attractive the notion that the extended peptide intro-duces subtle conformational transitions within the carboxyl-terminaldomain of BRI-L, providing a substrate with enhanced susceptibil-ity to furin proteolysis. However, structural contributions of distaldomains to furin proteolysis is without precedent, and hence it willbe important to validate this model in vitro by reconstitution assayswith purified furin and BRI/BRI-L substrates.

In view of the absence of any ultrastructural informationregarding the nature of the ABri fibril assemblies, we used EMto show that synthetic ABri peptides form fibrils in vitro. In con-trast to the Alzheimer’s-disease-associated Aβ1–40 peptide, how-ever, which forms infinitely long, unbranched, even fibrils withuniform diameters, ABri forms tortuous, irregular fibrils of small-er average diameters with apparent branch or cross-over points.In any event, these findings support our view that enhancedfurin-mediated processing of mutant BRI generates highly fib-rillogenic peptides that may initiate the pathogenesis of FBD.

Despite the strengths of these conclusions, the role of ABripeptides in initiating the clinical syndromes and pathophysio-logical cascades in patients with FBD are not known. Anapproach to examining these issues would be to generate trans-genic mice that overexpress human BRI-L in the nervous system,with the expectation that these mice would recapitulate at leasta subset of the histological lesions and behavioral alterations thattypify FBD and aid in clarification of the processes that lead toneuronal dysfunction and death in this disorder.

Fig. 4. Electron micrographs of ABri and Aβ1–40 peptides. (a) Electron micrograph of ABri fibrils after one week in PBS. Magnification, ×100,000. (b)Electron micrograph of Aβ1–40 fibrils formed after one week incubation at room temperature in PBS. Magnification, ×100,000. (c) Photographic mag-nification of sample shown in (a). Magnification, ×300,000.

a cb


199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://n

euro

sci.n

atu

re.c

om


RECEIVED 16 AUGUST; ACCEPTED 17 SEPTEMBER 1999

1. Plant, G. T., Revesz, T., Barnard, R. O., Harding, A. E. & Gautier-Smith, P. C.Familial cerebral amyloid angiopathy with nonneuritic amyloid plaqueformation. Brain 113, 721–747 (1990).

2. Worster-Drought, C., Greenfield, J. G. & McMenemey, W. H. A form of familialpresenile dementia with spastic paralysis (including the pathologicalexamination of a case). Brain 63, 237–254 (1940).

3. Griffiths, R.A., Mortimer, T. F., Oppenheimer, D. R. & Spalding, J. M.Congophilic angiopathy of the brain: a clinical and pathological report on twosiblings. J. Neurol. Neurosurg. Psychiatry 45, 396–408 (1982).

4. Vidal, R. et al. A stop-codon mutation in the BRI gene associated with familialBritish dementia. Nature 399, 776–781 (1999).

5. Nakayama, K. Furin: a mammalian subtilisin/Kex2p-like endoproteaseinvolved in processing of a wide variety of precursor proteins. Biochem. J. 327,625–635 (1997).

6. Molloy, S. S., Anderson, E. D., Jean, F. & Thomas, G. Bi-cycling the furinpathway: from TGN localization to pathogen activation and embryogenesis.Trends Cell Biol. 9, 28–35 (1999).

7. Pittois, K., Deleersnijder, W. & Merregaert, J. cDNA sequence analysis,chromosomal assignment and expression pattern of the gene coding forintegral membrane protein 2B. Gene 217, 141–149 (1998).

8. Deleersnijder, W. et al. Isolation of markers for chondro-osteogenicdifferentiation using cDNA library subtraction. Molecular cloning andcharacterization of a gene belonging to a novel multigene family of integralmembrane proteins. J. Biol. Chem. 271, 19475–19482 (1996).

9. Kyte, J. & Doolittle, R. F. A simple method for displaying the hydropathiccharacter of a protein. J. Mol. Biol. 157, 105–132 (1982).

10. Claros, M. G. & von Heijne, G. TopPred II: an improved software for membraneprotein structure predictions. Comput. Appl. Biosci. 10, 685–686 (1994).

11. Balch, W. E. & Rothman, J. E. Characterization of protein transport betweensuccessive compartments of the Golgi apparatus: asymmetric properties ofdonor and acceptor activities in a cell free system. Arch. Biochem. Biophys.240, 413–425 (1985).

12. Evan, G. I., Lewis, G. K., Ramsay, G. & Bishop, J. M. Isolation of monoclonalantibodies specific for human c-myc proto-oncogene product. Mol. Cell Biol.5, 3610–3616 (1985).

13. von Koch, C. S. et al. Generation of APLP2 KO mice and early postnatallethality in APLP2/APP double KO mice. Neurobiol. Aging 18, 661–669(1997).

14. Watanabe, T. et al. Sequence requirements for precursor cleavage within theconstitutive secretory pathway. J. Biol. Chem. 267, 8270–8274 (1992).

15. Molloy, S. S., Bresnahan, P. A., Leppla, S. H., Klimpel, K. R. & Thomas, G.Human furin is a calcium-dependent serine endoprotease that recognizes thesequence Arg-X-X-Arg and efficiently cleaves anthrax toxin protectiveantigen. J. Biol. Chem. 267, 16396–16402 (1992).

16. Watanabe, T., Murakami, K. & Nakayama, K. Positional and additive effects ofbasic amino acids on processing of precursor proteins within the constitutivesecretory pathway. FEBS Lett. 320, 215–218 (1993).

17. Takahashi, S., Hatsuzawa, K., Watanabe, T., Murakami, K. & Nakayama, K.Sequence requirements for endoproteolytic processing of precursor proteinsby furin: transfection and in vitro experiments. J. Biochem. (Tokyo) 116,47–52 (1994).

18. Moehring, J. M. & Moehring, T. J. Strains of CHO-K1 cells resistant toPseudomonas exotoxin A and cross-resistant to diphtheria toxin and viruses.Infect. Immun. 41, 998–1009 (1983).

19. Moehring, J. M., Inocencio, N. M., Robertson, B. J. & Moehring, T. J.Expression of mouse furin in a Chinese hamster cell resistant toPseudomonas exotoxin A and viruses complements the genetic lesion. J. Biol.Chem. 268, 2590–2594 (1993).

20. Spence, M. J., Sucic, J. F., Foley, B. T. & Moehring, T. J. Analysis of mutationsin alleles of the fur gene from an endoprotease-deficient Chinese hamsterovary cell strain. Somat. Cell Mol. Genet. 21, 1–18 (1995).

21. Duguay, S. J., Milewski, W. M., Young, B. D., Nakayama, K. & Steiner, D. F.Processing of wild-type and mutant proinsulin-like growth factor-IA bysubtilisin-related proprotein convertases. J. Biol. Chem. 272, 6663–6670(1997).

22. Sisodia, S. S., Koo, E. H., Beyreuther, K., Unterbeck, A. & Price, D. L. Evidencethat beta-amyloid protein in Alzheimer’s disease is not derived by normalprocessing. Science 248, 492–495 (1990).

22. Yang, M., Ellenberg, J., Bonicacino, J. S. & Weissman, A. M. The transmembranedomain of a carboxyl-terminal anchored protein determines localization to theendoplasmic reticulum. J. Biol. Chem. 272, 1970–1975 (1997).

24. Pan, C. J., Lin, B. & Chou, J. Y. Transmembrane topology of glucose-6-phosphatase. J. Biol. Chem. 273, 6144–6148 (1998).

25. Wang, R., Sweeney, D., Gandy, S. E. & Sisodia, S. S. The profile of solubleamyloid beta protein in cultured cell media. Detection and quantification ofamyloid beta protein and variants by immunoprecipitation-massspectrometry. J. Biol. Chem. 271, 31894–31902 (1996).

articles

METHODSGeneration of BRI cDNA expression vector. To isolate cDNA encodingBRI, we reverse transcribed adult human cerebral cortex RNA as a tem-plate for PCR with a primer complementary to the carboxyl-terminal sevenamino acids of APLP1. The insert was subcloned into the cytomegalovirus-based expression vector pRK5 (A. Lanahan and P. Worley, Johns HopkinsSchool of Medicine, Baltimore, Maryland, personal communication)encoding an N-terminal 12-amino-acid myc epitope to generate plasmidpBRI. Using PCR-based mutagenesis, we also generated pBRI-L encodingmutant BRI-L. The pCMV-furin construct21 was provided by Steve Duguay(Transkaryotic Therapies, Cambridge, Massachusetts).

Transfections, western blot and immunoprecipitation. Mouse neurob-lastoma N2a cells, CHO-K1 and furin-defective RPE.40 cells (a gift fromT. Moehring, University of Vermont, Burlington, Vermont) or Africangreen monkey kidney (COS-1) cells were transiently transfected withsupercoiled plasmid DNA using LipofectAMINE PLUS (Life Technolo-gies, Gaithersburg, Maryland). Forty-eight hours after transfection, cellswere lysed in immunoprecipitation buffer22, and conditioned mediumwas collected. Immunoreactive bands on western blots were visualizedusing an enhanced chemiluminescence (ECL) detection system (NewEngland Nuclear, Boston, Massachusetts). For metabolic labeling andimmunoprecipitation analysis, cells were labeled with [35S]cysteine (NewEngland Nuclear), and conditioned medium and detergent lysates of cellswere prepared, as described above. Immunoprecipitations were per-formed as described22. Immunoprecipitates were fractionated by SDS-polyacrylamide gel electrophoresis, visualized and quantified using aPhosphorimager (Molecular Dynamics, Sunnyvale, California).

Proteinase protection assay. N2a cells expressing human BRI werescraped in buffer containing 212.5 mM sucrose23 and protease inhibitorsand homogenized with a ball-bearing device11 with a clearance of 0.0012inches. The 100,000 × g microsomal fraction was used for proteinase pro-tection assays, as described24.

Mass spectrometric analysis. Conditioned media from transiently trans-fected N2a cells were immunoprecipitated using CT11 antibody.Immunoprecipitated peptides were analyzed by matrix-assisted laser des-orption time-of-flight mass spectrometry (Voyager-DE STR BioSpec-trometry Workstation, PerSeptive Biosystem) as described25. Each massspectrum was averaged from 256 measurements, and bovine insulin wasincluded as an internal mass calibrant.

Peptide synthesis. Peptides were synthesized with an ABI Model 431Apeptide synthesizer. The ABri peptide and the normal, BRI-derived pep-tide were purified using a C18 reverse-phase HPLC column (Dynax) at45°C, and a 0.1% TFA in water to 0.1% TFA in water/acetonitrile (40/60,v per v) linear gradient, with all solvents containing 0.5 M urea. Peptideidentity was verified by mass spectrometry and amino acid analysis; pep-tide purity was assessed by analytical HPLC using a Vydac C18 column.

Electron microscopy. For EM, fibrils were generated by allowing ABri orAβ1–40 peptides at 1 mg per ml to stand at 21°C in PBS containing 0.1%(w per v) NaN3 for one week. The samples were then centrifuged for 15min at 14,000 × g, and the precipitates were applied to a glow-discharge,400-mesh, carbon-coated support film and then stained with 1% uranylacetate. Micrographs were recorded using Philips EM 300 at magnifica-tions of × 100,000 or × 300,000.

ACKNOWLEDGEMENTSThe authors thank Anthony A. Lanahan and Paul F. Worley for pRK5 vector, Steve

Duguay for the pCMV-furin construct and Thomas Moehring for RPE.40 cells.

The authors also thank Takeshi Iwatsubo (University of Tokyo) for discussions.

This study was supported by National Institute of Health Grants AG14248 (S.S.S.)

and 5 T32 GM07281 (D.J.G.), The Alzheimer’s Association (S.M., IIRG # 98 134

and R.W.) and The Stasia Borsuk Memorial Fund (R.W.; RG1-96-070).


199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://n

euro

sci.n

atu

re.c

om

document

Documents