We presume that hornets collect thesecrystals from the local environment, but itis possible that they secrete them them-selves, as titanium and iron are present inthe hornet body5.Ietse Stokroos*, Luba Litinetsky†, Johannes J. L. van der Want*, Jacob S. Ishay†*Department of Cell Biology, Faculty of MedicalSciences, Graduate School of Brain Cognition andNeurosciences, University of Groningen,Oostersingel 69-2, 9713 EZ Groningen, The Netherlands†Department of Physiology and Pharmacology,Sackler Faculty of Medicine, Tel Aviv University,Ramat-Aviv 69978, Israele-mail: physio7@post.tau.ac.il1. Welham, N. J. J. Mater. Sci. 33, 1795–1799 (1998).

2. Wilson, G. S. & Roberts, A. P. in Paleomagnetism and Diagenesis

in Sediments (eds Tarling, D. H. & Turner, P.) 151, 95–108

(Geol. Soc., London, 1999).

3. Ishay, J. S., Rosenzweig, E. & Paniry, V. A. Insectes Sociaux 29,

34–43 (1982).

4. Spradbery, J. P. Wasps (Sidgwick & Jackson, London, 1973).

5. Ishay, J. S., Landsberg, A. & Pelah, S. Scan. Microsc. 10,

187–208 (1996).

6. Jongebloed, W. L., Rosenzweig, E. Kalicharan, D.,

van der Want, J. J. L. & Ishay, J. S. Electron Microsc. 48,

63–75 (1999).

Alzheimer’s disease

Molecular consequencesof presenilin-1 mutation

Alzheimer’s disease is characterized byaccumulation in the brain of a familyof insoluble amyloid peptides (Ab

peptides)1, which are produced as a result ofthe normal processing of b-amyloid precur-sor protein (b-APP). Russo et al.2 claim thata truncated Ab peptide that lacks the firstten amino acids accumulates in the brainsof patients carrying a mutant form of pre-senilin 1 (PS1), a protein that is involved incleavage of b-APP. However, we have foundthat this same species is also overrepresent-ed in Alzheimer’s patients with mutationsin b-APP itself 3. Our findings do not sup-port the conclusion of Russo et al. thatpathogenic PS1 mutations may controlcleavage of b-APP by b-secretase4,5.

Harmful mutations can occur at residue717 in b-APP as well as in PS1, favouringgeneration of the Ab species Ab1–42 (refs 6,7). Tandem missense mutations at residues670 and 671 of b-APP promote cleavage atthe Ab amino terminus8, the rate-limitingstep in producing Ab.

Russo et al.2 characterized Ab species inthe brains of patients suffering from spo-radic or familial Alzheimer’s disease due tomutations in PS1 or b-APP. Because N-ter-minally truncated Ab species (particularlythose starting with glutamate at residue 11)were overrepresented in PS1-mutant brainsof familial Alzheimer’s sufferers, theyinferred that pathogenic PS1 mutationsaffect b- as well as g-secretase reactions.

654 NATURE | VOL 411 | 7 JUNE 2001 | www.nature.com

brief communications

Magnetic minerals

Keystone-like crystals incells of hornet combs

The hexagonal brood-rearing cells insidethe nest combs of the hornet Vespa ori-entalis are uniform in both their archi-

tecture and orientation. We have discoveredthat each cell contains a minute crystal thatprojects down from the centre of its domedroof and has a composition typical of themagnetic mineral ilmenite1,2. These tinycrystals form a network that may act like asurveyor’s spirit-level, helping the hornetsto assess the symmetry and balance of thecells and the direction of gravity while theyare building the comb.

The keystone-like crystal we describehere is glued with hornet saliva to the centreof the roof on the inside of each cell (Fig. 1);it is positioned away from the egg, which isfastened onto a side wall3,4, the site at whichthe hatching larva also develops. The crystalis opaque, round, about 100 mm in dia-meter, and consists of polydomains. Dif-fraction analysis of the element composi-tion (data not shown) reveals that thesecrystals contain large quantities of titanium,

iron and oxygen, with some carbon andtrace amounts of manganese (58 out of 60comb cells we inspected contained a crystalwith the following composition in terms ofnumber of atoms: 25–35% Ti, 15–20% Fe,34–38% O, 12–18% C). This composition issimilar to that of ilmenite crystals (FeTiO3).The carbon component may originate fromthe hornet’s saliva.

Comb building in social wasps occurs instages. First, the hornet (whether a queen inthe spring or workers thereafter) buildsthree to five concentric bands, which ulti-mately form the roof of the comb cell. Oncethe roof is completed, the hornet polishes itfrom the inside and deposits another con-struction layer of small particles or organicfibres. At the centre of the dome, thebuilder leaves a hollow where it attaches thecrystal and which it fills with saliva thatrapidly hardens into a polymer.

As the roof of the cell moves, the hang-ing crystals move in response to gravity.The downward-sloping cell walls that sepa-rate each domed roof from its six neigh-bours prevent the crystal from being shakenloose by workers or larvae communicatingwith other members of the colony, eitheracoustically or by tapping.

Figure 1 Macroscopic images and electron micrographs showing the locations of ilmenite crystals in cells of the comb of the oriental

hornet. a, Inverted comb (with respect to the normal gravity-orientated position), showing the interior of comb cells and the inner side of

the domed roof in each cell. b, Inverted hexagonal cells, some of which contain eggs. c, Small section of the comb at higher magnifica-

tion, showing hexagonal roofs from the outside. d, e, At higher magnifications, an ilmenite crystal is seen (arrows in d) in the centre of the

roof of different cells. For electron microscopy, the tops of cell walls were removed so that the roof could be seen; samples were prepared

as described6. Scale bars: a, 5 cm; b, c, 1 cm; d, 1 mm; e, 100 mm.

a b c

d

e

© 2001 Macmillan Magazines Ltd

We used size-exclusion chromatographyand electrospray ionization mass spectrom-etry to study extractable amyloid in brainsfrom elderly, non-demented individualsand from cases of either sporadic or familialAlzheimer’s disease resulting fromK670N/M671L or V717I mutations3 (whereK, M and V represent the amino acid at thatposition in normal b-APP, and N, L and Ithe respective substituent in the mutant).We investigated their total Ab content,whereas Russo et al.2 looked primarily atsoluble Ab species.

Of the Ab species we found, Ab1–42 wasthe most abundant in four of five non-demented elderly brains, and Ab1–40 was themost abundant in eight of ten brains withsporadic Alzheimer’s. N-terminally trun-cated Ab species were frequently identifiedin all patients: Ab11–42 was the most abun-dant in one of ten sporadic patients and inthe single b-APP(V717I) familial case, andthe second most abundant in two of fivenon-demented individuals.

However, Ab11–42 was more abundantthan Ab1–42 in four of ten sporadic patientsand in those with V717I andK670N/M671L b-APP mutations. TheK670N/M671L mutations at the Ab aminoterminus help b-secretase to cleave b-APP,yielding Ab with aspartate as its firstresidue8, but in post-mortem tissue from apatient bearing these mutations, Ab11–42 wasthe predominant species recovered from thedeposited amyloid plaque, which is incon-sistent with the conclusions of Russo et al.2.

The Glu-11 Ab species is the main pep-tide generated from endogenous b-APP inrodent neurons9 and in PS1-deficientmouse neurons10. The peptide bond atAb10–11 is also the preferred alternative cleav-age site11 for BACE, the protease that isresponsible for most b-secretase reactions.These results argue against the model ofRusso et al.2 by showing that cleavage at thissite occurs even in the absence of PS1.

N-terminally truncated Ab peptides inamyloid plaques could be generated bypost-production modification of the N terminus, rather than during cleavage by b-secretase. The pyrrolic modification ofglutamate at residue 3 or 11 (ref. 12) of Aboccurs after b-secretase cleavage. Ab pep-tides that end at residue 42 may be relativelylong-lived and so more susceptible toenzymes that cut at the N terminus,enhancing their heterogeneity by increasingtheir exposure to chemical modificationssuch as pyrrolation or oxidation.

Despite the technical difficulties inobtaining a full and accurate profile of allthe Ab species present in the amyloidplaques associated with various forms ofAlzheimer’s disease, we consider thatamino-terminal truncation of Ab is notunique to PS1-mutant familial Alzheimer’s,and therefore question any proposed effect

of PS1 mutations on b-secretase activity.Sam Gandy*, Jan Naslund†, Christer Nordstedt‡*Sir James McCusker Alzheimer Research Unit,University of Western Australia, Perth, Australiaand Department of Psychiatry, New York UniversitySchool of Medicine, New York 10962, USAe-mail: samgandy@earthlink.net†Karolinska Institutet, 141 86 Huddinge, Sweden‡AstraZeneca, 151 85 Södertälje, Sweden1. Masters, C. L. et al. Proc. Natl Acad. Sci. USA 82,

4245–4249 (1985).

2. Russo, C. et al. Nature 405, 531–532 (2000).

3. Naslund, J. et al. Proc. Natl Acad. Sci. USA 91,

8378–8382 (1994).

4. Li, Y. M. et al. Nature 405, 689–694 (2000).

5. Esler, W. P. et al. Nature Cell Biol. 2, 428–434 (2000).

6. Sherrington, R. et al. Nature 375, 754–760 (1995).

7. Cai, X. D., Golde, T. E. & Younkin, S. G. Science 259,

514–516 (1993).

8. Citron, M. et al. Nature 360, 672–674 (1992).

9. Gouras, G. K. et al. J. Neurochem. 71, 1920–1925 (1998).

10.Chen, F. et al. J. Biol. Chem. 275, 36794–36802 (2000).

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

12.Miller, D. L. et al. Arch. Biochem. Biophys. 301, 41–52 (1993).

Russo et al. reply — Gandy et al. compareour results1 with their 1994 findings2 thatthe amino-terminally truncated amyloidAb11–42 was relatively abundant in two casesof familial Alzheimer’s disease involvingtwo distinct mutations in b-APP. However,four important differences should be bornein mind: the authors compare Ab11–42 withAb1–42 and ignore Ab1–40, although bothAb1–40 and Ab1–42 are generated by b-secre-tase/BACE cleavage at residue Asp 1 (ref. 3);their data are not correlated with featuresrelated to disease severity, such as age atonset and duration; they did not examinebrains with PS1 mutations (these were notknown at that time); and their characteriza-tion was based on the use of size-exclusionchromatography and electrospray massspectrometry to quantify formic-acid-extracted Ab, whereas we used quantitativeanalysis of immunoprecipitated water-soluble Ab on western blots, mass spec-trometry to identify Ab variants, andimmunohistochemistry to reveal amino-truncated Ab peptides in plaques.

Table 1 gives ratios of the amounts offull-length Ab1–40 and Ab1–42 to that ofAb11–40/42 species in brains of Alzheimer’spatients. The results of Naslund et al.2 givevalues that are consistently greater thanunity even when all other N-terminallytruncated Ab species they detected in spo-radic Alzheimer brains are added to Ab11–42.The main difference with our data is in

brief communications

NATURE | VOL 411 | 7 JUNE 2001 | www.nature.com 655

sporadic rather than familial cases involvingmutations in b-APP. The ratios betweenAb1–40/42 and Ab11–40/42 that we found in PS1-mutant cases with early disease onset aremuch lower (less than unity) than thosethat Naslund et al. (and we) observed infamilial Alzheimer’s disease with b-APPmutations.

We proposed that increased formationof N-terminally truncated Ab correlateswith the early-onset phenotype of familialAlzeimer’s disease linked to mutated PS1,and that a similar Ab pattern may be sharedby other malignant familial forms involvingb-APP mutations4,5. We explained the over-representation of these truncated Ab formsin PS1-mutant brains by postulating thatsome PS1 mutations directly or indirectlyaffect b-secretase cleavage. A variety ofmechanisms that do not require a directeffect of PS1 on b-secretase might beinvolved.

Gandy et al. suggest that late progres-sive proteolysis of the Ab amino terminusoccurs after b-secretase cleavage, afterwhich glutamate is cyclized at the N-terminal residues 3 and 11. This sequenceof events seems unlikely, as N-terminallytruncated and pyroglutamylated carboxy-terminal fragments of b-APP (createdbefore g-secretase cleavage), which aredirect precursors of similarly modified Abpeptides, are present in the brains ofAlzheimer’s patients6. The finding that N-terminally truncated C-terminal fragmentsof b-APP are more abundant in PS1-deficient neurons7, quoted by Gandy et al.as evidence against our hypothesis, couldalso mean that PS1, either through itsinteraction with b-APP or its lack thereof,influences b-secretase cleavage. Some PS1mutations may have the same effect on b-secretase cleavage as that produced byPS1 deficiency in a mouse knockoutmodel.

We maintain that overrepresentation ofN-terminally truncated Ab forms may becharacteristic of Alzheimer’s disease that isassociated with PS1 or other mutations andcharacterized by early onset and short dura-tion, or other malignant phenotypic fea-tures. Whether PS1 mutations have aneffect, be it direct or indirect, on b-secretasecleavage remains to be determined.C. Russo, G. Schettini, T. C. Saido, C. Hulette, C. Lippa, L. Lannfelt, B. Ghetti,P. Gambetti*, M. Tabaton, J. K. TellerDivision of Neuropathology, Institute of Pathology,

Table 1 Ratios of Ab1–40/42 to Ab11–42 in brains from Alzheimer’s disease patients

Disease type Naslund et al.2 Russo et al.1

Sporadic 3.4 (n410) 1.4 (n417; 66–84 years)

APPK670M/N671L 1.7 (n41) NA

APPV717I 1.2 (n41) 1.3 (n42; 59–63 years)

PS1 mutations (all) NA 0.75 (n411; 40–81 years)

PS1 M139I and H163A NA 0.3 (n44; 40–48 years)

Number of cases (n ) and age at death are shown in parentheses. NA, not applicable.

© 2001 Macmillan Magazines Ltd