7/31/2019 392761a0

1/3Nature Macmillan Publishers Ltd 1998

8

news and views

NATURE | VOL 392 | 23 APRIL 1998 761

is that ... if pollution of the gene pool were allthat was involved, genes that tended to causewomen to continue having babies despite therisk involved would certainly outreproducethose that influenced women to stop havingbabies9.

Alternatively, menopause may haveevolved as a counter-strategy to senescence.According to this good mother theory, if

the mother has more babies (with associateddangers) even as the ravages of age becomesevere, she is having children she may not beable to care for, and she is risking the futuresuccess of her existing children. If, instead,she stops having children and devotes herselfto helping those she already has, she may havemore total offspring who grow up to repro-duce themselves7. This hypothesis interpretsearly reproductive cessation as an adaptiveresponse to prolonged infant dependency. Insupport, there are some wild species in whichfemales often live well beyond their last preg-nancy, including Japanese macaques10, chim-

panzees

10

, elephants

11

, and pilot and killerwhales2,12. In all of these species, offspringrequire extended maternal care.

Good mother theorists have debatedwhich targets of nepotism favourmenopause. Is it the last-born offspring5,grown daughters and their children6,8,13, oran individuals entire genetic clan9? Muchattention has been paid to grown daughtersand their young, with quantitative models oftraditional societies both disputing14,15 andaffirming13 that the fitness benefits of help-ing to rear grandchildren can exceed thecosts of reproductive cessation. Now, Packeretal.1 introduce a comparative perspective to

the good mother debate.The Gombe baboons and Serengeti lions

have been studied continuously for over 30years, beginning with Jane Goodall andGeorge Schaller, respectively. Two to tentimes each month, a census is made of recog-nizable animals, and births and deaths arerecorded. The data reveal that femalebaboons live up to 27 years and female lionsup to 17 years. However, fertility dropsprecipitously when baboons are about 20

years old and the lionesses around 13.Because juvenile baboons require at least two

years of maternal care and lion cubs at leastone, in both species the mothers can live longenough after ceasing reproduction to reartheir last-born young. Thus, lion-cubsurvival does not decline with maternal age,and, in baboons, survival of infants from oldmothers (1924 years) is just slightly lessthan those from younger mothers.

Packer etal. found that in neither speciesdo post-reproductive females consistentlyimprove the reproductive performanceof grown daughters. Whether a femalebaboons mother was alive did not affect herage at puberty, interbirth interval, rate ofsuccessful pregnancy or survival of herinfants to their first birthday. In lions, a

females litter size and first-year survival ofher cubs was also the same regardless ofwhether her mother was dead, or aliveand post-reproductive. Cub survival wasimproved by the presence of a reproducinggrandmother, however, due to better nour-ishment (females nurse their grandcubs).

The authors interpret their data as sup-port for the senescence theory. They argue

that those few lions and baboons that live toold age make a negligible contribution to fit-ness. Therefore, selection cannot stop theirreproductive machinery from deteriorating.Packer et al. believe that senescence alsoaccounts for the timing of menopause byextrapolating from the observed relation-ship between mortality and maternity inbaboons and lions, and assuming a child-hood dependency of 10 years in humans,they calculate that the expected lifespan ofwomen who have reached 40 years (whenreproduction starts to decline) would be5865 years.

Packer et al. argue that reproductive ces-sation in female mammals is a non-adaptiveby-product of life-history patterns, andherein lies my only disagreement. Senes-cence is something that natural selectioncannot prevent. Although senescence isnon-adaptive, menopause is not. Senescenceis inevitable, so in species with prolongedinfant dependency, females have evolved aprecisely choreographed, adaptive counter-strategy. They cease ovulating abruptly,when they are still young enough to assist kinthat would not survive or successfully repro-duce without their mother57,9. Males areusually not essential to the survival of off-

spring, so their fertility deteriorates

gradually, solely due to senescence.Female lions and olive baboons survive

only a few years after ceasing to reproduce apparently, their main targets of maternalnepotism are last-born offspring. Butwomen routinely live to nearly twice theaverage age at which menopause occurs,implying that children are dependent for agreater proportion of their lives than juve-

niles in the other species (that is, longer than10 years). This also suggests that a broadarray of relatives benefit from the accumu-lated wisdom, status and resources of post-reproductive women.Paul W. Sherman is in the Section of Neurobiology

and Behavior, Cornell University, Ithaca, New York

14853, USA.

1. Packer, C., Tatar, M. & Collins, A. Nature392, 807811

(1998).

2. Austad, S. N. Why We Age(Wiley, New York, 1997).

3. vom Saal, F. S. & Finch, C. E. in The Physiology of Reproduction

Vol. 2 (eds Knobil, E. et al.) 23512413 (Raven, New York,

1988).

4. Symons, D. The Evolution of Human Sexuality(Oxford Univ.

Press, 1979).

5. Williams, G. C. Evolution11, 398411 (1957).

6. Dawkins, R. The Selfish Gene(Oxford Univ. Press, 1976).

7. Nesse, R. M. & Williams, G. C. Why We Get Sick(Random

House, New York, 1994).

8. Lancaster, J. B. & King, B. J. in In Her Prime2nd edn (eds

Kerns, V. & Brown, J. K.) 715 (Univ. Illinois Press, Urbana,

IL, 1992).

9. Alexander, R. D. Univ. Michigan Mus. Zool. Spec. Publ. 1, 138

(1990).

10. Takahata, Y., Koyama, N. & Suzuki, S. Primates36, 169180

(1995).

11.Laws, R. M., Parker, I. S. C. & Johnstone, R. C. B. Elephants and

Their Habitats(Oxford Univ. Press, 1975).

12.Marsh, H. & Kasuya, T. Rep. Int. Whaling Comm. Spec. Issue8,

5774 (1986).

13.Hawkes, K., OConnell, J. F., Blurton Jones, N. G., Alvarez, H. &

Charnov, E. L. Proc. Natl Acad. Sci. USA95, 13361339 (1998).

14.Hill, K. & Hurtado, A. M. Hum. Nature2, 313350 (1991).

15.Rogers, A. Evol. Ecol. 7, 406420 (1993).

How does a protein fold? What is thefolding code, and how do inter-actions between amino-acid residues

determine the native structure of a protein?Some theoreticians have attempted to tacklethese complex problems of molecular recog-nition and self-assembly by simulationsusing simple lattice models (Fig. 1, overleaf) proteins are modelled as chains config-ured on regular lattices, and, because theylack high-resolution atomic representations,these are models for generic proteins. Theytry to rationalize general trends in real pro-teins, but they do not address properties spe-cific to, say, lysozyme or cytochrome c. So,theoreticians and experimentalists in thefolding field are dealing with differentobjects generic versus specific proteins.How can the two be reconciled?

The experimentally observed foldingtimes for different proteins can differ by

more than nine orders of magnitude fromtens of microseconds to hours. Can these berationalized by models for generic proteins?Unfortunately, theoreticians have not yetreached a consensus on the main determin-ing factor that gives rise to this tremendousdiversity in folding rates. In this months

Journal of Molecular Biology, Plaxco, Simonsand Baker1 provide much-needed systematiccomparisons between theory and experi-ment. By statistical analysis of the correla-tions between measured folding rates andthe various determinants proposed bytheoreticians, they have identified a mainstructural factor that governs folding speedfor a set of non-homologous, single-domainproteins that fold with essentially two-statekinetics. Remarkably, they find this to be theaverage sequence separation between con-tacting residues in the native state, normal-ized by the length of the protein chain an

Protein folding

Matching speed and localityHue Sun Chan

7/31/2019 392761a0

2/3Nature Macmillan Publishers Ltd 1998

8

lattice models, concluded5 that proteins tendto fold faster if (local) interactions amongresidues that are close to one another alongthe sequence are stronger than (non-local)interactions between residues that are farapart. Based on another lattice model, how-ever, Sali etal. proposed6 that the native statebeing a pronounced global minimum isthe necessary and sufficient condition for

rapid folding. Moreover, they reported thatthe relative number of local versus non-localcontacts in the native state does not deter-mine whether a sequence is fast-folding orvery slow-folding (that is, practically non-folding). Because these predictions weremade about generic proteins, it is hard to tellwhether they can be verified or falsified byexperiments on specific proteins, yet thesehypotheses must somehow be matched upwith real data to be relevant.

Theoreticians have much to learn fromthe work of Plaxcoetal.1, because their statis-tical analysis comes close to meeting the

generic protein on theoreticians ground.On the face of it, the new study vindicatestheories which predict that local interac-tions increase the speed of folding5,7, andcasts doubt on theories predicting the oppo-site8,9 (Fig. 1). But on closer examination therelation between different theories, and thecorrespondence between theory and experi-ment, are more complex. For example,owing to different modelling assumptions,different conclusions have been reached bystudies based on the same lattice model6,7. Inanother instance, two theories5,8 agree thatnon-local interactions increase the thermo-dynamic stability of the native structure,

but come to opposite conclusions aboutwhether local interactions speed up or slowdown folding.

Many models predict that non-localinteractions increase stability. As a corollary,theories that emphasize a positive correla-tion between stability and folding speed8,9

predict that non-local interactions shouldincrease the speed of folding. But, for thedata set analysed by Plaxco et al., stabilityturns out to be only a secondary determi-

news and views

762 NATURE | VOL 392 | 23 APRIL 1998

extremely simple measure that they callrelative contact order.

Plaxcoetal. use their own folding kineticsdata, as well as results from other groups.Figure 2 shows the native structures of twoproteins in their data set, cytochrome candacyl phosphatase. Although they are aboutthe same size, the folding rates of these twoproteins under similar external conditionsdiffer by more than four orders of magni-tude. Plaxco etal. suggest that the key to this

difference in rate is chain topology therelatively simple topology of cytochrome c(with many local contacts) can be formedmuch more rapidly than the complex topol-ogy of acyl phosphatase, which involves themany non-local contacts that make up its -sheet. For a data set of 12 proteins (whichwere selected from a larger set to exclude alltraces of statistical correlation produced bysequence similarity), Plaxco et al. find asignificant correlation between relativecontact order and folding speed. Also inter-esting is that native stability and chainlength, which have been thought to beimportant for folding kinetics, fail to showsignificant correlations with folding rates.

This statistical study confirms a generalpicture that has been emerging from recentexperiments that stronger local inter-actions, especially those conducive to helixformation, tend to lead to faster folding2.Nanosecond laser temperature-jump exper-iments on short peptides show that foldingof a-hairpin at room temperature is about30 times slower than formation of an -helix3. Moreover, in a set of protein engineer-ing experiments4, folding rate decreasedwith increasing length of an inserted poly-glycine loop. Local guidance seems to be

Figure 2 Two proteins in the data set analysed by Plaxco et al.1. Experiments show that single-domain

proteins whose native structures have mostly local contacts fold faster. a, Equine cytochrome c

(length, 104) has mostly local contacts (relative contact order, 11.2%), and folds quickly at a rate 14 of

6,400 s1 at 23 C. b, Muscle acyl phosphatase (length, 98) has more non-local contacts (relative

contact order, 21.2%), and folds at the much slower rate of 0.23 s 1 at 28 C. (Data of F. Chiti, N. A. J.

van Nuland, N. Taddei, F. Magherini, M. Stefani, G. Ramponi & C. M. Dobson.)

critical in reducing the time that is neededfor a protein to find its native structure.

Many theoreticians believe in a step-by-step approach to solving the protein-folding problem. They begin by trying tounderstand the mechanisms by whichprotein thermodynamics and folding ratesare related to low-resolution properties suchas overall chain topology, percentage ofhydrophobic residues, strength of intrachaininteractions, amount of helices and sheets,

and so on. These studies are expected to gen-erate increasingly refined questions andexperiments. Indeed, valuable insights havebeen gained by these approaches, especiallywith regard to how protein-like behaviourscan arise from general properties of chainmolecules.

But there is no general agreement amongtheoreticians on the question of how foldingspeed is affected by chain topology. In 1978,Go and Taketomi, two pioneers of protein

Figure 1 Theoretical proteins, adapted from the cubic lattice models of Abkevich et al.8. Each model

sequence consists of 36 residues, given by the one-letter codes for the amino acids. The model native

structures have 37 (a) and 36 (b) nearest-neighbour contacts between residues not adjacent along the

sequence. Dotted lines in a represent one non-local (high-order) contact (sequence separation, S1,36= 35) and one local (low-order) contact (S5,8 = 3). The average contact order sequence separation

between contacting residues in the native state is 10.0 for a and 15.8 for b. Normalized by chain

length, the relative contact order is 27.9% for a and 44.0% for b. Abkevich et al. reported that under

conditions in which both model native structures are stable, the sequence with mostly non-localcontacts (b) folds two orders of magnitude faster than the one with mostly local contacts (a).

(Figure courtesy of Kaizhi Yue.)

L

a b

KG

EG

G

L

AT

K

Y

KY

P

M

TW

LE

AT

W

KI

I

M

W

W

G

T

GA

K T

GM

EW

YK

G

G

G KW

S M

IK

E

K

I

MW

WM

M

K

WS

M

W

E

L

S1,36

S5,8

K

Y

G

G

A

E

7/31/2019 392761a0

3/3Nature Macmillan Publishers Ltd 1998

8

nant of folding rate a far less significantfactor than contact order. Furthermore,the expected correlation between stabilityand relative contact order is lacking, leadingone to wonder whether any of the theoriesis directly comparable to the data set ofPlaxcoetal.

The problem seems to be that most of themodel sequences studied by theoreticians

have a higher tendency to get stuck kineti-cally trapped in non-native conformationsduring folding1012 than the real, single-domain proteins of Plaxco et al. For manymodel sequences, the key to rapid folding isto overcome kinetic traps. Higher native sta-bility would allow folding to occur at a highersimulation temperature, making it easier toescape from traps. This is the basis for thepredictions8,9 that non-local interactionscause fast folding, because non-local con-tacts tend to increase native stability in thesemodels and higher temperatures can be usedto simulate folding of more stable sequences

with more non-local contacts (Fig. 1)

8

. How-ever, the proteins studied by Plaxco et al. donot have much problem with kinetic traps12,so they should be described by models thatminimize the effects of traps. Future model-ling will need to capture this generic featureof single-domain, two-state proteins.

It goes without saying that the new statis-tical analysis does not account for all of thefactors that may contribute to folding speed.For example, mutated proteins sharing acommon topology (and, thus, the same rela-tive contact order) can fold at different rates:in one case13, the variations cover a range ofabout one order of magnitude. However, if

the general picture suggested by the analysisis correct, these other effects should also sup-port the idea that stronger local interactionsproduce faster folding. This should betestable by further experiments. In themeantime, the work of Plaxco et al. is one ofmany signs that, in the community ofprotein folders, an increasing constructivefeedback between theory and experiment isunder way.Hue Sun Chan is in the Department of

Pharmaceutical Chemistry, University of California,

San Francisco, California 94143-1204, USA.

e-mail: chan@maxwell.ucsf.edu

1. Plaxco, K. W., Simons, K. T. & Baker, D.J. Mol. Biol. 277,

985994 (1998).

2. Viguera, A.-R., Villegas, V., Aviles, F. X. & Serrano, L. Folding

and Design2, 2333 (1997).

3. Muoz, V., Thompson, P. A., Hofrichter, J. & Eaton, W. A.

Nature390, 196199 (1997).

4. Viguera, A.-R. & Serrano, L. Nature Struct. Biol. 4, 939946

(1997).

5. Go, N. & Taketomi, H. Proc. Natl Acad. Sci. USA75, 559563

(1978).

6. Sali, A., Shakhnovich, E. & Karplus, M. Nature369, 248251

(1994).

7. Unger, R. & Moult, J.J. Mol. Biol. 259, 988994 (1996).

8. Abkevich, V. I., Gutin, A. M. & Shakhnovich, E. I.J. Mol. Biol.

252, 460471 (1995).

9. Govindarajan, S. & Goldstein, R. A. Proteins Struct. Funct.

Genet. 22, 413418 (1995).

10.Bryngelson, J. D., Onuchic, J. N., Socci, N. D. & Wolynes, P. G.

Proteins Struct. Funct. Genet. 21, 167195 (1995).

11.Klimov, D. K. & Thirumalai, D. Phys. Rev. Lett. 76, 40704073

(1996).

12.Chan, H. S. & Dill, K. A. Proteins Struct. Funct. Genet. 30, 233

(1998).

13.Burton, R. E., Huang, G. S., Daugherty, M. A., Calderone, T. L.

& Oas, T. G. Nature Struct. Biol. 4, 305310 (1997).

14.Mines, G. A., Pascher, T., Lee, S. C., Winkler, J. R. & Gray, H. B.

Chem. Biol. 3, 491497 (1996).

news and views

NATURE | VOL 392 | 23 APRIL 1998 763

Spongiform encephalopathies

The prions perplexing persistence

Adriano Aguzzi and Charles Weissmann

Since prion diseases were identified, wehave come to consider such infectionsas an irrevocable death sentence

preceded by severe clinical disease. Take,for example, the cases of iatrogenicCreutzfeldtJakob disease elicited by minutetraces of the infectious agent, administeredthrough contaminated surgical instruments,hormone preparations or tissue trans-plants1. On page 770 of this issue, however,Race and Chesebro2 report that infectivitycan persist long-term in brain and spleen

tissue, without causing the development ofclinical symptoms.There were reasons to believe that expo-

sure to infectious prions may not necessarilylead to clinical disease. For example, micelacking Prnp(the gene that encodes the nor-mal prion protein, PrPC) are completelyimmune to clinical disease. They cannotreplicate prions, and the traces of infectivitythat were occasionally found in the brainseveral months after injection3,4 have beenattributed to persistence in the absence ofreplication4. Moreover, mice that containhalf the normal level of PrPC take a very longtime to develop symptoms after intra-

cerebral inoculation5. Finally, if immuno-deficient mice are inoculated peripherallywith the scrapie agent, they do not fall ill,despite the occasional presence of large prionloads in their brains6.

Another example in which exposure toinfectivity does not always lead to disease isthe epidemiology of bovine spongiformencephalopathy (BSE). Typically, only singleanimals from affected farms develop disease,even though the genetic make-up of herdsis probably similar, and each cow is likely to

have been exposed to similar amounts ofprion infectivity. Moreover, although theBSE agent is quite likely the origin of newvariant CreutzfeldtJakob disease7,8, andalthough many people in the United King-dom and Europe may have ingested theinfectious agent, only a small minority hasdeveloped overt disease to date. Why?

One reason could be that the infectiousagent fails to penetrate the organism per-haps because predisposing factors, such aslesions in the digestive tract, are required.Alternatively, prions may not transfer fromthe site of entry to the central nervous sys-tem. Or, they may not replicate to the extent

Figure 1 Due to the species barrier between mice and hamsters, prions originating from one species

produce spongiform encephalopathy rather quickly when transferred to a further animal of the same

species, yet only produce disease if any with much longer latency when passed to the other

species. This barrier can be abolished by expressing appropriate transgenic prion proteins in the

recipient animals12,13.

120 days

>500 days

No disease

No disease

75 days

Disease 1

No disease

No disease

No disease

Prnp0/0