CORRESPONDENCE

www.nature.com/naturebiotechnology • MARCH 2003 • VOLUME 21 • nature biotechnology

Why reinvent risk?

To the editor:I read with great interest the two commen-taries in the November issue by Goklany(Nat. Biotechnol. 20, 1075, 2002) andAuberson-Huang (Nat. Biotechnol. 20,1076–1078, 2002) regarding the risk assess-ment and the “precautionary principle” asapplied to genetically modified organisms(GMOs). In my view, neither author makesa critical distinction between the scientificand analytical process of risk assessmentand the policy stance of the precautionaryprinciple, neither acknowledges the impactof uncertainty on decisions, and neitheroffers a useful framework to address fullythe complex decisionmaking associatedwith GMOs.

In theory, the process of risk assessmentand precautionary policy can be integratedinto an analytical decision framework,which I believe is what both authors areattempting to define, but their definitionsare incomplete. Risk assessment is a scien-tific process that estimates the probabilityand severity of adverse events1. Thisprocess is independent of stakeholder orrisk manager viewpoints. Essentially, theprecautionary principle is a policy guidingrisk management that states that reductionor elimination of risk is an overriding deci-sion objective (over and above that oftrade-offs like cost, competing risks, etc.).If that policy is adopted, it does not replaceor inform risk assessment; it simply pro-vides a means to guide risk managementbased on the results of risk assessments.

Thus, Auberson-Huang appears to haveit backwards. As typically applied, the pre-cautionary principle hamstrings truestakeholder involvement and proper riskmanagement by constraining elicitation ofstakeholder values and the scientificprocess of risk assessment with predeter-mined conclusions and the refusal toacknowledge trade-offs. Goklany arguesthat we should assess one set of these trade-offs through risk–risk analysis so as to be“precautionary,” but this is an inferior sub-stitute for a true decision framework usinga decision criterion such as net benefit.Risk–risk analysis does not directly address

stakeholder values, nor the complete rangeof trade-offs associated with adoption of aparticular policy. Neither author explicitlyaddresses the central problem that therisks, benefits, and costs associated withmany GMOs are highly uncertain, and thatrigorous, quantitative assessment of theimpact that uncertainty has on decisions iscritical to informing stakeholders and deci-sion makers, as well as informing primaryresearch.

A proper decision process will integratethe science of risk assessment with the pol-icy of risk management, but to my knowl-edge this is not occurring with GMOissues. It is unfortunate that many of thesame mistakes that have been made histor-ically in the environmental toxicology fieldand other fields are now being made withregard to GMOs. The appropriate andacknowledged way to evaluate these highlyuncertain risks is to employ risk assessmentas a scientific process within a decisionframework that directly addresses stake-holder values and the impact of uncertain-ty on decisions. There is a wealth of litera-ture and applications regarding suchframeworks, such as multiattribute utilitytheory and multicriteria decisionmaking,that do exactly this2. It is time to acknowl-edge that policy approaches, such as theprecautionary principle, and halfway mea-sures such as risk–risk analysis, are insuffi-cient to address the complex decisionmak-ing that is associated with GMOs.

Robert C. LeeUniversity of Calgary,

Community Health Sciences,Health Technology

Implementation Unit,Foothills Medical Centre,

1403 29th St. NW,Calgary, AB T2N2T9,

Canada (rclee@ucalgary.ca)

1. Haimes, Y.Y. Risk Modeling, Assessment, andManagement (Wiley, NY, 1998).

2. Cox, L.A. Risk Analysis: Foundations, Models, andMethods (Kluwer Academic Publishers,Dordrecht, 2001).

229

Letters may be edited for space and clarity.They should be addressed to:CorrespondenceNature Biotechnology345 Park Avenue SouthNew York, NY 10010-1707, USAor sent by e-mail to biotech@natureny.comPlease include your telephone and fax numbers.

Telomere-driven replicative senescence is a stress response

To the editor:In the July issue, Rubin (Nat. Biotechnol. 20,675–681, 2002) and Wright and Shay (Nat.Biotechnol. 20, 682–688, 2002) presentedtwo contrasting views of the nature of invitro replicative senescence and its impor-tance for ageing in vivo. Rubin emphasizedthe marked stochastic variation in doublingpotential of individual cells and its appar-ent inconsistency with the idea that telom-

ere reduction acts as a mitotic clock (as aresult of the end-replication problem—i.e.,the inability of conventional polymerases toreplicate fully the very end of a linear DNAmolecule). Wright and Shay argued that thestochastic loss of growth potential resultsfrom a “stimulation and stress-inducedsenescent-like arrest,” or stasis, which is not“true” senescence. In this exchange, stasisand senescence were presented as a pair ofcomplementary, but essentially differentprocesses, stasis being telomere-indepen-dent and, most often, stress-induced, senes-cence being induced by telomere dysfunc-tion and stress-independent. We believethis dichotomy is misleading and overlooksthe growing evidence that telomere short-ening is stress-dependent.

Although stress can induce telomere-independent growth arrest, and telomerescan shorten as a result of the end-replica-tion problem in the absence of stress, stressis one of the major influences on the rate oftelomere loss. Chronic mild oxidative stressaccelerates telomere shortening and short-ens replicative lifespan, whereas free-radicalscavengers1 or overexpression of the antiox-idant enzyme superoxide dismutase2 do thereverse.

Under constant levels of extrinsic oxida-tive stress, there is a substantial inverse cor-relation between antioxidant capacity andtelomere-shortening rate in human fibrob-last strains3. Telomeres are more sensitive tooxidative damage4, and single-strand breaksin telomeres are less well repaired than else-where in the genome. This leads to an accu-mulation of telomeric damage, which isquantitatively transformed into fastertelomere shortening during DNA replica-tion1. Together, these data show that on topof the end-replication problem, oxidativestress is a major cause of telomere shorten-ing. Moreover, they indicate that the short-ening of replicative lifespan induced bymild oxidative stress is mediated by itseffect on telomeres.

If oxidative stress is indeed playing amajor role in telomere loss, the heterogene-ity in cell-doubling potential of cultures likeWI-38 and MRC-5 fibroblasts should belinked to a corresponding stress-dependentheterogeneity in telomere-shortening rates1.To test this suggestion, we employed fluores-cence-activated cell sorting (FACS) to selecta subpopulation of midpassage MRC-5(PDL 31) fibroblasts exhibiting the senes-cence phenotype, as characterized by largesize (forward scatter) and high lipofuscincontent (autofluorescence in FL1; ref. 5).

Re-analysis of the sorted cells indicatedan at least 3-fold enrichment of phenotypi-cally senescent cells. Few of the sorted cellshad undergone DNA synthesis (according

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy

nature biotechnology • VOLUME 21 • MARCH 2003 • www.nature.com/naturebiotechnology

CORRESPONDENCE

230

Choosing CCR5 or Rev siRNA in HIV-1

To the editor:In a report in Nature Biotechnology1 andanother in Nature2, it was demonstratedthat small inhibitory RNAs (siRNAs) effi-ciently target and degrade human immun-odeficiency virus (HIV)-1 RNA sequencesand inhibit viral replication. Lee et al.1 usedsiRNA against Rev generated by indepen-dent transcription of sense and antisensestrands, whereas Jacque et al.2 targeted sev-eral sequences of the HIV-1 RNA tran-scripts using both synthetic siRNA andplasmid-mediated delivery of siRNA. In amore recent report, Novina et al.3 haveused synthetic siRNA to inhibit HIV-1infection by impairing CD4 expression.Thus, siRNA technology can be used todegrade efficiently both endogenous andvirus-derived RNA.

Although both approaches seem worth-while, we believe that treatments limited tothe inhibition of viral products by siRNAwill not be clinically feasible. Thus, it seemshighly likely that inhibition of Rev willresult in the selection of escape mutantsencoding Rev, but with silent mutationsimpairing siRNA recognition, and recent

experimental evidence supports thisnotion4.

To avoid escape mutants, we suggest tar-geting RNA encoding cellular proteins.However, when selecting a cellular target, itis important to identify a protein that doesnot cause deleterious effects when down-regulated. As Novina et al.3 themselvespoint out, silencing of CD4 may becomelimited by the role of this receptor in nor-mal host immune function, and co-recep-tors, in particular CC-motif receptor-5(CCR5), may be the preferred target in thathomozygous mutations in CCR5 (ref. 5)effectively confer protection from HIV-1 inthe absence of any known adverse effectson the immune system. Finally, the recentdevelopment of plasmid-mediated expres-sion of siRNA in a single transcript6, allow-ing long-term expression, represents fur-ther improvement toward clinical applica-tion of siRNA.

We have used plasmid-mediated deliveryof siRNA and successfully impaired HIV-1infection as well as replication by usingsiRNAs directed against the 3′ ends of theCCR5 and HIV-1 Rev transcripts, respec-tively. U937 cells (a human promonocyticcell line expressing CCR5) transientlytransfected with the plasmid pRS-siCCR5,showed ∼ 40% reduction of CCR5 expres-sion 48 hours post-transfection, as detect-ed by FACS analysis (Fig. 1A). This treat-ment inhibited infection with a primaryCCR5-dependent, M-tropic, HIV-1 isolate.Quantification of the HIV-1 p24 antigen insupernatants of the cell cultures by ELISAassay, 3 days post-infection, demonstrated68% inhibition (Fig. 1B; left panel) com-pared with cells transfected with a plasmidexpressing a hairpin against green fluores-cent protein (GFP). A comparison of thedecrease in CCR5 expression and the mag-nitude of protection against infection sug-gests that even partial reduction of CCR5co-receptors on the surface of target cellscould result in considerable protectionagainst viral infection.

Following submission of this correspon-dence, two articles have appeared using asimilar approach, but with differentsequences of the siRNA for CCR5 (refs.7,8), demonstrating the versatility ofchemokine receptor downregulation bythis method. Furthermore, the vector pSR-siRev, which expresses a hairpin directedagainst the 3′ end of the HIV-1 Rev tran-scripts, blocked >90% of expression of areporter plasmid expressing a Rev–GFPfusion protein, in transiently co-transfect-ed 293 cells (Fig. 1C,D; right panel).Moreover, transient transfection of HeLa-CD4 with the siRev vector, followed byinfection with the CXC chemokine recep-

to 5-bromo-2-deoxyuridine incorporatedsubsequent to a 48-hour labeling period),and most showed a typical senescent mor-phology at one week after replating.Interestingly, sorted cells had, on average,much shorter telomeres, whereas the sort-ing process itself did not influence telom-ere length (Fig. 1). The wide spread intelomere length of the sorted populationmost probably reflects the imprecision ofthe FACS sorting, rather than intracellularvariation. However, the data clearly indi-cate that human fibroblasts that acquire asenescent phenotype at an early stage do sowith short telomeres. It should be men-tioned that lipofuscin accumulation infibroblast cultures is an indicator of oxida-tive stress5, again supporting the associa-tion between oxidative stress and telomereshortening.

It seems highly likely that there is a spec-trum of stress-induced growth arrests.Intense stress causes a variety of cell dam-age, which will activate cell cycle check-points. Most of the damage caused by mildstress will be repaired before it causesarrest, but telomere shortening will beaccelerated, resulting in a cumulative indexof cell damage3. We anticipate that stress-independent arrest driven solely by telom-ere reduction resulting from the end-repli-cation problem may be the exception,which will tend to be seen only in cells thathave uncommonly good defense mecha-nisms, such as some BJ foreskin fibrob-lasts6.

There is growing evidence, from a combi-nation of empirical and theoretical7,8 stud-ies, that cellular ageing is a network process.Thus, it is not a case of this or that alterna-tive but rather of synergy and interactionbetween different mechanisms actingtogether. Stress-induced telomere reduc-tion is a strong candidate to reconcile theviews expressed by Rubin and by Wrightand Shay.

Thomas von Zglinicki,Joanne Petrie,

and Thomas B.L. KirkwoodDepartment of Gerontology,

Institute for Ageing and Health,

Newcastle University,Newcastle upon Tyne

NE4 6BE, UK(t.vonzglinicki@ncl.ac.uk)

1. von Zglinicki, T. Trends Biochem. Sci. 27, 339–344(2002).

2. Serra, V. et al. J. Biol. Chem., in press (2003).3. von Zglinicki, T. et al. Lab. Invest. 80, 1739–1747

(2000).4. Henle, E.S. et al. J. Biol. Chem. 274, 962–971

(1999).5. Sitte, N., Merker, K., Grune, T. & von Zglinicki, T.

Exp. Gerontol. 36, 475–486 (2001).6. Lorenz, M., Saretzki, G., Sitte, N., Metzkow, S. &

von Zglinicki, T. Free Radic. Biol. Med. 31, 824–831(2001).

7. Proctor, C.J. & Kirkwood, T.B.L. Mech. Ageing Dev.123, 351–363 (2002).

8. Sozou, P.D. & Kirkwood, T.B.L. J. Theor. Biol. 213,573–586 (2001).

Figure 1. Telomere Southern blot of MRC-5fibroblast cells. Lanes 1 and 6, λ-Hind marker;lane 2, unsorted cells; lane 3, cells sorted forlarge size and high autofluorescence; lane 4,cells sorted for small size and lowautofluorescence; lane 5, cells collected in thewaste from the sort for large size and highautofluorescence.

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy