© 2005 Nature Publishing Group

masses, we report two independent tests ofE�mc2. The two comparisons find a mea-sured fractional difference between E and�mc2 of 2.1(5.2)�10�7 and –9.7(8.0)�10�7

with sulphur and silicon isotopes, respectively,and a combined value of –1.4(4.4)�10�7. Theerror on this comparison is currently domi-nated by the uncertainty on the �-ray mea-surements. This result is 55 times moreaccurate than the previous best direct test ofE�mc2, which was performed by comparingthe electron and positron masses to the energyreleased in their annihilation10. Simon Rainville*†, James K. Thompson*,Edmund G. Myers‡, John M. Brown§,Maynard S. Dewey||, Ernest G. Kessler Jr||,Richard D. Deslattes||, Hans G. Börner¶,Michael Jentschel¶, Paolo Mutti¶,David E. Pritchard**Research Laboratory of Electronics,MIT–Harvard Center for Ultracold Atoms, andDepartment of Physics, Massachusetts Instituteof Technology, Cambridge, Massachusetts 02139,USAe-mail: rainville@alum.mit.edu†Present address: Département de Physique,Université Laval, Québec G1K 7P4, Canada ‡Department of Physics, Florida State University,Tallahassee, Florida 32306-4350, USA§The Physical and Theoretical ChemistryLaboratory, Department of Chemistry, South

Parks Road, Oxford OX1 3QZ, UK||National Institute of Standards and Technology,Gaithersburg, Maryland 20899, USA¶Institut Laue-Langevin, 38042 Grenoble Cedex,France

1. Mohr, P. J. & Taylor, B. N. Rev. Mod. Phys. 77, 1–107 (2005). 2. Kessler, E. G. et al. Phys. Lett. A 255, 221–229 (1999).3. Kessler, E. G. et al. Nucl. Instrum. Meth. Phys. Res. A 457,

187–202 (2001).4. Dewey, M. S. et al. Preprint nucl-ex/0507011 at

http://arxiv.org (2005). 5. Rainville, S., Thompson, J. K. & Pritchard, D. E. Science 303,

334–338 (2004).6. Audi, G., Wapstra, A. H. & Thibault, C. Nucl. Phys. A 729,

337–676 (2003).7. Thompson, J. K., Rainville, S. & Pritchard, D. E. Nature 430,

58–61 (2004).8. NIST Chemistry WebBook, NIST Standard Reference Database

Number 69 (eds Linstrom, P. J. & Mallard, W. G.) (NationalInstitute of Standards and Technology, Gaithersburg,Maryland, 2003).

9. Chase, M. W. J. Phys. Chem. Ref. Data Monogr. 9, 1–1951 (1998).10. Greene, G. L., Dewey, M. S., Kessler, E. G. & Fischbach, E.

Phys. Rev. D 44, R2216–R2219 (1991).

Supplementary information accompanies thiscommunication on Nature’s website.Competing financial interests: declared none.doi:10.1038/4381096a

NATURE|Vol 438|22/29 December 2005 BRIEF COMMUNICATIONS

1097

10�7. A slightly revised value of fD from thatgiven in ref. 2 is also reported here (c is exact):fSi�c/(0.146318275(86)�10�12 m)fS�c/(0.143472991(54)�10�12 m)fD�c/(0.557341007(98)�10�12 m)These values of fSi and fS are, respectively, 25and 50 times more precise than earlier values4.

The mass difference was determined bydirect comparison of the cyclotron frequenciesof two different ions simultaneously confinedin a Penning trap5. This two-ion techniqueachieves mass comparisons with fractionalaccuracies below 10�11 by virtually eliminatingthe effect of many sources of noise such asmagnetic field fluctuations. During measure-ments, the two ions are placed on a commoncircular orbit (magnetron mode), on oppositesides of the centre of the trap and separated by a distance of about 1 mm. Figure 1b showsthat the measured cyclotron frequency ratio is almost completely independent of ion–ion separation, tightly constraining the size of the largest possible systematic errors5. Theion mass ratios reported here, namelyM[33S]/M[32SH]�0.9997441643450(89) andM[29Si]/M[28SiH]�0.9997151241812(65),are respectively more than 700 and 100 timesmore precise than previously known6.

The total uncertainties include the uncorre-lated uncertainties of 4�10�12 each fromion–ion interactions, trap field imperfections,and statistics. The ratios also include correc-tions of 45(5)�10�12 and 7.3(2)�10�12 toaccount for polarization-induced shifts of thecyclotron frequencies7 (see methods in sup-plementary information). After accounting forthe mass of the missing electron and the chem-ical binding energies8,9, the ion mass ratios givethe neutral mass differences: M[32S]M[H]�M[33S]�0.00843729682(30) AMU, andM[28Si]M[H]�M[29Si]�0.00825690198(24)AMU. And by the addition of M[D]�2M[H]��0.00154828629(40) AMU (refs 1,6) to eachone, we obtain the desired mass differences ofequation (1), with a fractional accuracy ofabout 7�10�8 for both.

Comparing the measured energies and

when the ratio alters. This precise control ofcommunication by molecular chirality offersinsight into societal interactions in elephants,and may be useful in implementing new con-servation protocols4.

Frontalin (1,5-dimethyl-6,8-dioxabicyclo[3.2.1]octane)1,2,5,6 is discharged during musthin male Asian elephants (Elephas maximus)from the temporal gland on the face. Weanalysed more than 100 secretion samplesfrom six males and found that this pheromonewas first detectable in the late teens, the quan-

tity secreted rising about 15-fold over a 25-yearage span (Fig. 1a). Both enantiomeric forms(designated plus and minus; Fig. 1b, inset) werepresent and each was quantified (for methods,see supplementary information).

In young males, significantly more ()than (�) frontalin was secreted, but concen-trations of these became almost equal (1:1, aracemic mixture) as the elephant matured(Fig. 1b). Serial samples collected throughoutthe musth episodes of two young and two oldmales confirmed that this change in ratio wassignificantly associated with the elephants’ agegroups (results not shown).

Musth periods get longer as males age. Theenantiomeric composition of the frontalinsecreted becomes significantly more varied in

Interferometer angle (arcsec)

0

200

400

600

Cou

nts

in 8

8 s

0.14�

5,421 keV(1,–2) reflection

5,421 keV(1,2) reflection

a

–245.20 –245.15 246.15 246.20 246.25

600

7.4 × 10–12

–0.9

99 7

44

167

(10

–12 )

M[33

S+]

M[32

SH+]

b

–2,750

–2,730

–2,710

–2,690

–2,670

–2,650

1,2001,000800400Ion–ion separation (mm)

Figure 1 | Typical data for testing E��mc2. a, 33Swavelength data and fitted curves. The energy ofthe emitted 5,421-keV �-rays is measured fromthe angular separation of the second-order Braggpeaks resulting from diffraction from a siliconcrystal. b, Measured mass ratio M[33S]/M[32SH](which largely determines the mass change, �m)as a function of the distance between the ions, �s,in the Penning trap. The lack of variation stronglysuggests that the predicted uncertainties fromion–ion interactions and field inhomogeneities(turquoise and blue-hatched bands, respectively)are too conservative. The final mass ratio (solidline) is determined with fractional accuracy of 7 parts in 1012 (dashed lines). All error bars areone standard deviation.

CHEMICAL COMMUNICATION

Chirality in elephant pheromones Musth in male elephants is an annual period ofheightened sexual activity and aggression1–3

that is linked to physical, sexual and socialmaturation. It is mediated by the release ofchemical signals such as the pheromonefrontalin, which exists in two chiral forms(molecular mirror images, or enantiomers).Here we show that enantiomers of frontalinare released by Asian elephants in a specificratio that depends on the animal’s age andstage of musth, and that different responses are elicited in male and female conspecifics

22.12 brief comms MH 15/12/05 5:37 PM Page 1097

Nature Publishing Group© 2005

© 2005 Nature Publishing Group

BRIEF COMMUNICATIONS NATURE|Vol 438|22/29 December 2005

1098

the longer musths of older males but, critically,the ratios are nearly racemic during mid-musth — the period of prime signalling(Fig. 1c, right). Wide fluctuations in the ratio of frontalin enantiomers from young males(Fig. 1c, left) are akin to the variations in serumandrogens that occur during the progression oftheir shorter ‘moda’, or honey musth1.

To test whether the enantiomeric composi-tion of frontalin affects its activity as a multi-purpose pheromone (attractant, repellent,deterrent and dominance enhancer), we inves-tigated the effects on five categories of con-specific (females: follicular, luteal-phase orpregnant; males: young or old). Total frontalinconcentrations and frontalin enantiomerratios were determined for samples of mid-musth, temporal-gland secretions from maleswith a range of ages; the responses of otherconspecifics to these samples were thenrecorded — behaviour such as sniffing andchecking indicated attraction, whereasresponses such as running away and circlingthe sample indicated repulsion.

Other elephants were generally indifferentto secretions containing no frontalin, apartfrom young males (Fig. 1d, left). Secretionswith low concentrations of frontalin, predom-inantly as the () enantiomer, aroused theinterest of young males and were of mild inter-est to old males (Fig. 1d, middle). Secretionscontaining high concentrations of frontalin atracemic ratios actively repulsed males of allages, as well as luteal-phase and pregnantfemales, whereas they attracted follicular-phase females (Fig. 1d, right).

Our results indicate that the ratio offrontalin enantiomers enables other elephantsto distinguish both the maturity of male ele-phants in musth and the phase of musth. Theycorroborate earlier findings that syntheticracemic frontalin can elicit different behav-ioural responses from conspecifics, dependingon their sex, age and reproductive status2.

The length of musth increases as malesmature, with the fittest sustaining a long mid-musth phase and releasing an optimal ratio of frontalin enantiomers, thereby improvingtheir status and enjoying easier access tofemales3; these mid-musth emissions arepicked up almost exclusively by ovulatingfemales. A similar quantitative blending offrontalin enantiomers is also necessary to elicit

a full biological response in Dendroctonus beetles7–12. Our discovery of a stereospecifi-cally tailored message in mammals shouldhelp in resolving the interactions of phero-mones with their respective receptor proteins,which are also chiral.David R. Greenwood*, Dan Comeskey†, Martin B. Hunt†, L. Elizabeth L. Rasmussen‡ *HortResearch, Mount Albert Research Centre,Private Bag 92-169, Auckland, New Zealand†Palmerston North Research Centre, HortResearch, Private Bag 11-030, Palmerston North,New Zealand

‡Department of Environmental and BiomolecularSystems, OGI School of Science and Engineering,Oregon Health and Sciences University,Beaverton, Oregon 97006, USAe-mail: betsr@ebs.ogi.edu.

1. Rasmussen, L. E. L., Riddle, H. S. & Krishnamurthy, V. Nature415, 975–976 (2002).

2. Rasmussen, L. E. L. & Greenwood, D. R. Chem. Senses 28,433–446 (2003).

3. Rasmussen, L. E. L., Krishnamurthy, V. & Sukumar, R.Behaviour 142, 351–396 (2005).

4. Rasmussen, L. E. L. & Riddle, S. W. J. Elephant Man. Assoc.15, 30–38 (2004).

5. Rasmussen, L. E. L. & Perrin, T. E. Physiol. Behav. 67,539–549 (1999).

6. Perrin, T. E., Rasmussen, L. E. L., Gunawardena, R. &Rasmussen, R. A. J. Chem. Ecol. 21, 207–221 (1996).

7. Silverstein, R. M. J. Chem. Ecol. 14, 1981–2003 (1988).

8. Mori, K. Chirality 10, 578–585 (1998).9. Paine, T. D., Millar, J. G., Hanlon, C. C. & Hwang, J. S.

J. Chem. Ecol. 25, 433–453 (1999).10. Kinzer, G. W. et al. Nature 221, 477–478 (1969).11. Francke, W. et al. J. Chem. Ecol. 21, 1043–1063 (1995).12. Stewart, T. E., Plummer, E. L., McCandless, L. L., West, J. R.

& Silverstein, R. M. J. Chem. Ecol. 3, 27–43 (1977).

Supplementary information accompanies thiscommunication on Nature’s website. Competing financial interests: declared none.doi:10.1038/4381097a

13–20

40a

30

20

10

021–30

Elephant age (yr)

Frontalin

Fron

talin

con

cent

ratio

n(n

g m

l–1 )

31–43

None

4

5 d

3

2

1

0Low

YMYM

LF*PF*

FF

OM

Amount of frontalin secreted

Mea

n re

spon

se p

er h

our

High13–20

80b

60

0

0

(+) (–)

40

20

021–30

Elephant age (yr)

Med

ian

prop

ortio

n of

(+)

enan

tiom

er (

%)

31–43

80

100 c

60

40

20

0

Med

ian

prop

ortio

n of

(+)

enan

tiom

er (

%)

Early

3

Middle

4,5

1,2

3,4

13 4

YM*

OM*

2

Late

Stage of musth

4,6

Ending

1,2,3, 5,6

Start End

1

0

0

2

Figure 1 | Secretion of the pheromone frontalin by male Asian elephants in musth. a, Medianconcentrations of frontalin from elephants in three age groups (number of animals in each was 37, 41and 14, respectively; statistically significant differences (s.s.d.) between groups, P 0.05). b, Changingmedian proportions of () and (�) enantiomers of frontalin (inset) with increasing male age(numbers in each age group were 31, 48 and 16, respectively); s.s.d., P 0.05, as indicated by pairednumbers. c, Changes in median proportion of () enantiomer during musth progression for young(blue) and old males (red); s.s.d., P 0.05, between paired numbers. d, Mean responses by conspecificsto different frontalin concentrations; s.s.d. (t-test), P�0.001, between paired numbers. YM, youngmales; OM, older males; FF, follicular females; LF, luteal females; PF, pregnant females; asterisks,avoidance of musth male; no asterisk, attraction to musth male. Secretions with no frontalin inducedno response in older males and a very low response in females; females were not tested against lowfrontalin concentrations. For details of methods, see supplementary information.

A musth have: mature Asianelephants (Elephas maximus)secrete a chiral chemical in a ratiothat is especially attractive tofemales.

22.12 brief comms MH 15/12/05 5:37 PM Page 1098

Nature Publishing Group© 2005