N E W S A N D V I E W S

NATURE GENETICS VOLUME 35 | NUMBER 3 | NOVEMBER 2003 205

disruption. After selection for integration ofthe fragment and growth of the transfor-mants, DNA is isolated from each pool andthe tags are subjected to PCR amplification.Each tag is flanked by a common sequence,such that the same pair of PCR primers canamplify the tags for all 4,700 strains at once.The amplified DNA is hybridized to microar-rayed oligonucleotides to quantify thegrowth of each strain. Synthetic lethal strainsare identified by comparing control and dele-tion pool hybridizations.

An important advantage of SLAM as com-pared with SGA is reduced labor, as the dele-tion strains can be treated in pools instead ofbeing monitored for growth individually. Thedouble-mutant pools can be stored for addi-tional assays under other selection conditions.One problem specific to SLAM, however, isthat up to 15% of the integrative transformantsmay be incorrectly targeted in the genome.Another limitation is that only ∼ 60% of thestrains yield high-quality data for both UPTAGand DOWNTAG hybridizations, as a result oftag mutations, slow-growing strains or strainsthat are defective in integrative transforma-tion. By comparison, the SGA method alsoshows problems with slow-growing strains andwith strains defective in mating or sporulation.Limitations common to both methods are thatthey deal only with nonessential genes andanalyze only complete deletions and not par-tial loss-of-function mutations.

Ooi et al.5 characterized the syntheticlethal network of two genes that had alsobeen analyzed by Tong et al.3, allowing adirect comparison of the two methods. Withthe SGS1 gene, each screen found a commonset of synthetic interactors as well as ones thatwere not present in the other screen, imply-ing that neither screen was saturating. SGA,which found 10 of 12 known synthetic lethalinteractors with SGS1 and 14 not previouslyidentified, seems to be more sensitive thanSLAM, which identified 7 of 12 known inter-actors and 5 new ones. False positives were aproblem with both methods, to a similardegree (∼ 50–60%), but these can largely beeliminated by further analysis.

Networks of synthetic genesBoth SGA and SLAM can potentially gener-ate a large amount of data. An importantchallenge will be to develop ways to repre-sent these data and to integrate them withresults from other work, such as proteininteraction or expression studies, to maxi-mize the inferences that can be drawn aboutnew genes or gene functions. One simpleapproach to visualizing large-scale syntheticlethal interactions has been the use of inter-action maps, where a line is drawn betweentwo colethal genes3. But these are geneticinteractions, not physical ones, and as suchrepresent many possibilities: genes withredundant functions, genes with additive

effects on the same pathway or genes withindirect effects. These indirect effects canoccur because a deletion phenotype repre-sents not just the absence of one particulargene, but also the response of the cell to theabsence of that gene, which may includeupregulating or downregulating diversepathways. If we can generalize from yeast3,5,however, indirect effects may be rare, asmost synthetic lethal interactions occurbetween genes involved in the same or simi-lar processes.

Genomic methodologies for syntheticlethal studies are beginning to take shape inother organisms, in which techniques such asRNAi allow combinations of gene ‘knock-downs’ to be analyzed in worms, flies andeven human cells. Thus, it becomes possibleto envision screening a cell line mutant for adisease-related gene, such as a tumor sup-pressor, with a genome-wide array of RNAiconstructs to search for synthetic effects.When the data from these types of screensemerge, it’s likely that yeast will provide aguide for their analysis.

1. Lucchesi, J.C. Genetics 59, 37–44 (1968).2. Bender, A. & Pringle, J.R. Mol. Cell. Biol. 11,

1295–1305 (1991).3. Tong, A.H.Y. et al. Science 294, 2364–2368 (2001).4. Winzeler, E.A. et al. Science 285, 901–906 (2001).5. Ooi, S.L. et al. Nat. Genet. 35, 277–286 (2003).6. Hartman, J.L. et al. Science 291, 1001–1004 (2001).7. Demant, P. Nat. Rev. Genet. 4, 721–734 (2003).8. Shoemaker, D.D. et al. Nat. Genet. 14, 450–456

(1996).

Vibrating in the backgroundAlysson R Muotri & Fred H Gage

Retroviruses make up a large proportion of the mammalian genome. A new study shows that an mRNA nuclear exportreceptor can act as a modifier of endogenous retrovirus insertion mutations by interacting with the mutated pre-mRNA.

Alysson R. Muotri and Fred H. Gage are in theLaboratory of Genetics, The Salk Institute forBiological Studies, 10010 N. Torrey Pines Road,La Jolla, California 92037, USA.e-mail: gage@salk.edu

Mammalian genomes are notorious forhosting an incredible number of geneticparasites, known as transposable elements,which can interact with the surroundinggenomic environment and increase theability of the organism to evolve. Some ofthese mobile elements, called retrotrans-posons, are able to reproduce through anintermediate RNA, using the reverse tran-scriptase enzyme to insert a DNA copy in

the genome in a new position. Humans andmice share ∼ 40% of their genomes that isthought to have been derived from retro-transposons1. In humans, the activity ofmost of these parasites is believed to havebeen silenced about 40 million years ago(although we still have some of them jump-ing around in our cells), but mice have closeto 3,000 active elements, responsible for10–20% of spontaneous mutations2. Theenormous contrast between the number ofactive elements in human and mouse sug-gests that the reason for the decline oftransposon activity in humans may berelated to some primary disparity betweenhominids and rodents1.

Mouse intracisternal A-particles (IAPs)are retrotransposons similar to modernretroviruses but incapable of leaving thehost cell owing to mutations in the envelopegene (env)3. These elements are severelyrepressed in most tissues of the mouse, pos-sibly as a biological requisite for genomicstability and to reduce the transcriptionalnoise from pointless expression of RNAs.These observations suggest that the hostgenome has evolved effective epigeneticnuclear defenses that shield it from activeretroelements, such as methylation andprobably repressive chromatin structures.Consistent with this theory, homozygousDNA methyltransferase-1 (Dmnt1) knock-

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

N E W S A N D V I E W S

206 VOLUME 35 | NUMBER 3 | NOVEMBER 2003 NATURE GENETICS

of some endogenous retroviruses (class Delements) that may stochastically escapethe initial epigenetic host suppressorforces, with deleterious consequences tothe organism. The strategy relies on anuclear host factor that modifies expres-sion of some genes carrying specific IAPinsertions, attenuating the mutant pheno-type (Fig. 1).

The authors used the vibrator mousemutation that is responsible for early-onsetprogressive action tremor, neural degenera-tion in brain and spinal cord, and juveniledeath. The mutation has been characterizedas a sense-oriented IAP insertion in theintronic region of the gene encoding phos-phatidylinositol transfer protein α (Pitpn)

that decreases its expression7. Whencrossed to the CAST/Ei background strain,mice carrying this same mutation havefewer tremors and survive to adulthood.The same suppressor phenomenon isobserved with the mouse model for thehuman branchiootorenal syndrome(another intronic sense-oriented IAP inser-tion that decreases gene expression), whichhas phenotypes that include circling anddeafness8. In both cases, the expression ofthe correctly processed RNA made from themutated gene was elevated in the CAST/Eibackground in a single intercross. Proteinamounts also correlate well with changes inRNA levels. This evidence strongly supportsthe view that the CAST/Ei background actsby altering RNA levels from the mutatedallele, probably involving the machinery ofRNA transport and processing.

The suppressor element was previouslycharacterized as Mvb1 (modifier of vibra-tor-1)7. Mvb1 has now been identified bypositional complementation strategy as anatural allele of the cellular gene encodingmRNA nuclear export factor 1 (Nxf1). Thisgene controls the expression of genes at thelevel of exporting messenger RNA from thenucleus to cytoplasm. Nxf1 can bind andmediate the export of constitutive transportelements in the unspliced genomes of someretroviruses, including rodent IAPs9. Thedifference in the C57BL/6J and CAST/Eialleles of Nxf1 is only two amino acids thatreflect natural polymorphisms present inthe wild population. The study also pointsout that this new allele is probably increas-ing its frequency in mouse populations bynatural selection. One of these amino acidalterations is located in a highly conserveddomain among vertebrates that is thoughtto mediate RNA export by physical interac-tion with the nuclear pore. This domain hasa significant chemical shift after binding toa FG-nucleoporin peptide10, suggesting analternative interaction between the Nxf1factor and the pore. The Nxf1CAST/Ei allelecan influence the balance of aberrant andcorrect mRNA in two different ways: it caninfluence the ribonucleoprotein complexassembly or isolate the transcript from pro-cessing compartments in the nucleus.

Welcome to orientationMoreover, although the authors did not testlarge numbers or types of mutations, itseems that this variant of Nxf1 acts as anorientation-specific selective pressure fac-tor, specifically suppressing the effects ofsense, but not antisense, IAP insertions inthe CAST/Ei population. Supporting this

out mouse embryos have much higher levelsof IAP transcripts than their wild-type lit-termates4. More recently, the methyl-CpGbinding protein-1 (Mbd1) knockout mouseshowed increased genomic instability andelevated levels of IAP expression5. As a con-sequence of this derepression, most of thenew copies are accumulated in regions thatdo not cause excessive damage to the host.So far, the epigenetic modifications were theonly known defense against the injuriescaused by mobile DNA. But are they enoughto keep these elements quiet?

Suppress the mutated transcriptIn this issue, Floyd et al.6 describe a mecha-nism of controlling newly inserted copies

a

b

c

Figure 1 Background influence during the control of retrotransposition events. (a) Epigenetic defenses,such as methylation (CH3 added groups) and chromatin changes (associated proteins that bunch theDNA), can keep most of the endogenous retroviruses silenced. (b) Some elements can escape the firstcellular control, producing intermediated RNA that will be reverse transcribed (RT) and inserted back intoa new site elsewhere in the genome. The target site could be an essential gene for the organism, forexample, Pitpn, encoding the phosphatidylinositol transfer protein α (PITPα). (c) The newly inserted copymay cause aberrant mRNA production, decreasing the expression of the correct mRNA that can beexported to the nucleus by the Nxf1 machinery. The vibrator phenotype seen in the C57BL/6J backgroundmouse arises from this situation. On the other hand, the Nxf1 protein found in the CAST/Ei backgroundcan select the correct mRNA (by affecting the ribonucleoprotein complex assembly or the nucleus exportmachinery of the mutated mRNA), raising its level in the cytoplasm and, as a consequence, attenuatingthe vibrator phenotype.

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

N E W S A N D V I E W S

NATURE GENETICS VOLUME 35 | NUMBER 3 | NOVEMBER 2003 207

view, the database search for IAP-derivedtargets in introns showed that mostoccurred in the antisense orientation. It isnot clear why the frequencies of sense andantisense IAP insertions are different in twostrains of mouse living in a similar geo-graphic area. Notably, Nxf1CAST/Ei cannotalter transcription from the LTR promoter;that is, it does not affect IAP mutationsrelated to transcriptional initiation. Thisobservation suggests that Nxf1 is probablynot the only retroviral suppressor acting inthe host cell.

Suppression of retrovirus insertionalmutations is certainly a new defensive strat-egy against mobile DNA in the genome.The finding will have an impact on our cur-rent knowledge of retrovirus control and, asthe authors point out, it will be extremelyvaluable for titrating mutations caused byretroviruses in vivo. One can also imaginethat, in the near future, new forms of theNxf1 factor could be helpful in a gene ther-apy approach using retrovirus vectors,where the risk of insertional mutagenesis isstill present.

1. International Human Genome Sequencing Consortium.Nature 409, 860–921 (2001).

2. Goodier, J.L., Ostertag, E.M., Du, K. & Kazazian, H.H. Jr.Genome Res. 11, 1677–1685 (2001).

3. Kuff, E.L. & Lueders, K.K. Adv. Cancer Res. 51,183–276 (1988).

4. Walsh, C.P., Chaillet, J.R. & Bestor, T.H. Nat. Genet.20, 116–117 (1998).

5. Zhao, X. et al. Proc. Natl. Acad. Sci. USA 100,6777–6782 (2003).

6. Floyd, J.A. et al. Nat. Genet. 35, 221–228 (2003).7. Hamilton, B.A. et al. Neuron 18, 711–722 (1997).8. Johnson, K.R. et al. Hum. Mol. Genet. 8, 645–653

(1999).9. Tabernero C. et al. J. Virol. 71, 95–101 (1997).10. Grant, R.P., Hurt, E., Neuhaus, D. & Stewart, M. Nat.

Struct. Biol. 9, 247–251 (2002).

A molecular signature of behavior

This worker is returning to the hive laden with pollen, which makesher a forager in the world of Apis mellifera, the honey bee. Whatshe is also carrying, unbeknownst to her, is a gene expressionsignature in her brain that differentiates her from her sibs left tolook after the hive. This knowledge comes from a recent study byCharles Whitfield and colleagues (Science 320, 296–299; 2003)in which they profiled gene expression in the brains of bees whosejob it is to bring home food and those who stay home to tend thehive.

During the first 2–3 weeks of adult life, honey bees are assignedvarious tasks in the hive, including nursing the young. The beesare then promoted to tasks outside the hive, such as foraging fornectar and pollen, which they carry out until the end of their livessome 5–7 weeks later. The timing of this transition is flexibleaccording to the needs of the colony. When it occurs, it isassociated with changes in brain structure and neurochemistry.The degree to which these changes are associated with geneexpression changes has so far not been known.

Using a microarray with probes for 5,500 different genes, aninitial comparison of the brains of nurse and forager bees showedexpression differences for 39% of genes. This analysis, however, isconfounded by the fact that nurses are younger than foragers and,therefore, expression differences may be associated with age ratherthan behavior. To dissociate age from behavior, the authors createdcolonies with only young bees. In these colonies, some of theyounger bees became foragers earlier than usual and someremained nurses for much longer. Microarray analysis was

performed on the brains of these young foragers and old nurses.Class prediction analysis was then able to determine behavioralphenotype using a set of 50 genes.

The set of 50 genes encode proteins that could conceivably havea role in mediating changes in the brain and behavior. Examplesinclude cell adhesion molecules, molecules involved in intracellularsignaling and carbonic anhydrase, which is known to have a role inspatial learning and memory. As with other genetic signatures,however, determining whether these expression differences are acause or an effect of behavioral change awaits further investigation.

David Gresham©

T. D

avid

son/

Pho

to R

esea

rche

rs, I

nc.

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics