letter

nature genetics • volume 21 • february 1999 209

Taming of transposable elements by homology-dependent gene silencing

Silke Jensen, Marie-Pierre Gassama & Thierry Heidmann

Unité des Rétrovirus Endogènes et Eléments Rétroïdes des Eucaryotes Supérieurs, CNRS UMR 1573, Institut Gustave Roussy, 39 rue Camille Desmoulins,94805 Villejuif Cedex, France. Correspondence should be addressed to T.H. (e-mail: heidmann@igr.fr).

Transposable elements can invade virgin genomes within a fewgenerations, after which the elements are ‘tamed’ and retainonly limited transpositional activity. Introduction of the I ele-ment, a transposon similar to mammalian LINE elements, intoDrosophila melanogaster genomes devoid of such elements ini-tially results in high-frequency transposition of the incomingtransposon, high mutation rate, chromosomal nondisjunctionand female sterility, a syndrome referred to as hybrid dysgene-sis1 (for review, see refs 2−4); a related syndrome has also beendescribed in mammals5. High-frequency transposition is tran-sient, as the number of I elements reaches a finite value andtransposition ceases after approximately ten generations6,7. Ithas been proposed that the I elements encode a factor that neg-atively regulates their own transcription, but evidence for sucha mechanism is lacking8. Using the hybrid dysgenesis syndromein Drosophila1–4 as a model, we show here that transpositionalactivity of the I element can be repressed by prior introductionof transgenes expressing a small internal region of the I ele-ment. This autoregulation presents features characteristic ofhomology-dependent gene silencing, a process known as cosup-pression9–15. Repression does not require any translatablesequence, its severity correlates with transgene copy numberand it develops in a generation-dependent manner via germlinetransmission of a silencing effector in females only. These resultsdemonstrate that transposable elements are prone to and canbe tamed by homology-dependent gene silencing, a processthat may have emerged during the course of evolution as a spe-cific defense mechanism against these elements.To test whether the process of homology-dependent gene silenc-ing, known as cosuppression9–20, is responsible for I-elementsilencing in Drosophila, we introduced transgenes containingtransposition-incompetent I-element internal sequences into

Drosophila and assayed for their capacity to silence functional Ielements. The transgenes and rationale for this assay are shown(Fig. 1). The I-element sequence contained in the hsp[i2∆]pAtransgene corresponds with a 969-bp fragment from the 5´ part of

Fig. 1 Constructs and rationale of the assay for measurement of the level of I-element activity in transgenic strains. The structures of the full-length I elementwith its two ORFs, including the reverse transcriptase domain (in black), of the hsp[i2∆]pA and hsp[i2∆*]pA constructs containing an internal part of the I elementand of the control construct are shown on the left. Hsp[i2∆*]pA differs from hsp[i2∆]pA by a 20-bp sequence containing stop codons in all three reading frames(asterisk), inserted 34 bp downstream of the ATG codon. I-element sequences are under the control of the hsp70 promoter (hsp) and the Actin 5C polyadenyla-tion signal (pA). The control construct contains the same components as hsp[i2∆]pA and hsp[i2∆*]pA except the I-element sequence. The constructs were intro-duced into Drosophila of the reactive wK strain (devoid of functional I elements) by P-mediated transgenesis, and several transgenic strains were established foreach of the three constructs. I-element activity was assessed at different generations after transgenesis by introduction by crossing of full-length active I elementsand subsequent measurement of the percentage of dead embryos laid by the progeny of the cross.

Fig. 2 Regulation of I-element activity by transgenic strains is transgene copy-number dependent. I-element activity was quantitated by measuring the per-centage of dead embryos laid by the progeny resulting from the cross oftransgenic females with w1118 males containing full-length functional I ele-ments at generation 20 after transgenesis for the hsp[i2∆]pA and hsp[i2∆*]pAstrains on the left; strains containing stop-codons in the I sequence of the trans-gene are annotated by an asterisk. The results for the transgenic strains contain-ing the control construct devoid of I-element sequence (T strains), tested 25generations after transgenesis, are indicated on the right; the data shown arethe mean values (±s.d.) for 5 batches of 40 embryos. The transgene copy-num-bers as determined by Southern-blot analysis are indicated.

© 1999 Nature America Inc. • http://genetics.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://g

enet

ics.

nat

ure

.co

m

letter

210 nature genetics • volume 21 • february 1999

ORF2. As this part of the I element is translatable, we also con-structed a mutated transgene by the introduction of stop codonsin all three reading frames, just downstream of the ATG initiationcodon (transgene hsp[i2∆*]pA). These constructs were intro-duced into a Drosophila strain devoid of active I elements (thereactive wK strain) by P-mediated transgenesis, and several inde-pendent transgenic strains were established for both constructs.Transgene integrity and copy number were assessed by Southernblot. The ability of the transgenes to repress I-element activity wasthen determined at different generations after transgenesis byintroducing functional I elements via crosses. Activity was mea-sured by quantifying lethality in the progeny of crosses21. Repres-sion of I-element activity can be seen in all strains containingI-derived transgenes (Figs 2,3). Control experiments with trans-genes containing all sequences present in hsp[i2∆]pA andhsp[i2∆*]pA except the I sequence showed no effect on I-elementactivity (Fig. 2, ‘T’ strains).

The characteristics of these transgenic strains revealed severalfeatures of the regulation process. The extent of downregulation isdependent on transgene copy number (Fig. 2). Strains wereobtained with a transgene copy number ranging from one to four;there was a direct correlation between extent of repression andtransgene dosage. The degree of repression is also independent ofany translation products (Fig. 2) because both the stop codon-containing hsp[i2∆*]pA and the non-mutated hsp[i2∆]pA trans-genes produce similar repression levels for given transgenedosage. A third feature of I silencing is that reduction of I-element

activity is not instantaneous after transgenesis, but generation-dependent (Fig. 3). In different strains, repression of 50% of I-ele-ment activity was observed within 3–40 generations aftertransgenesis. The kinetics for the establishment of repression weredependent on transgene copy number, with a higher rate observedfor higher transgene dosage. For strains containing less than threecopies of the transgenes, a lag of a few generations was observedbefore downregulation of I-element activity occurred, suggestingthat a threshold level of a silencing factor must be attained. Thisgeneration-dependent effect was further analysed for its transmis-sion profile (Fig. 4). Maternal versus paternal transmission of thetransgenes via crossing of the transgenic Drosophila with fliesfrom the initial non-transgenic strain resulted in heterozygousindividuals, which were then tested as above. There is a differencein the level of repression of I-element activity for all tested trans-genic strains (Fig. 4) depending on whether the transgenes werematernally or paternally transmitted. In this assay, almost com-plete reversion is observed in one generation upon paternal trans-mission of the transgene.

To further characterize the repression process, three additionalseries of transgenic Drosophila were developed. In a first series, athe fragment of the I element (still regulated by the hsp70 pro-moter) corresponding with the ORF1 domain (nt 167–1,492;non-overlapping with that in Figs 2−4) was introduced in place ofthe ‘i2∆’ fragment. This transgene induced repression of I-ele-ment activity (Fig. 5), demonstrating that repression is not depen-dent on a specific I-element sequence. Repression was similarly

Fig. 3 Kinetics of the repressionprocess in different hsp[i2∆]pA andhsp[i2∆*]pA transgenic strains. I-element activity was assessed atdifferent generations after trans-genesis as described (Fig. 2). Theresults for four independenthsp[i2∆]pA strains and four inde-pendent hsp[i2∆*]pA strains (aster-isk) are presented, sorted for therapidity of the downregulation of I-element activity. The data shownare the mean values (±s.d.) for 5batches of 40 embryos. Brackets,transgene copy-numbers for eachtransgenic strain.

Fig. 4 The repressed phenotype is maternally transmitted. a, Mating schemesfor maternal and paternal transmission of the transgene and rationale of theassay. Female or male transgenic Drosophila (in black) were crossed, 37 genera-tions after transgenesis, with wK flies (in white); the resulting heterozygousfemales were crossed with w1118 males to introduce active I elements. I-ele-ment activity was quantitated by measuring the percentage of dead embryoslaid by the transgenic female progeny. b, Results (±s.d.) for maternal (M) andpaternal (P) transmission of the transgenes. Left, hsp[i2∆]pA and hsp[i2∆*]pAstrains; right, control (T) strains. The corresponding transgene copy numbersare indicated; standard deviation was calculated from 5 batches of 40 embryos.

a

b

© 1999 Nature America Inc. • http://genetics.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://g

enet

ics.

nat

ure

.co

m

letter

nature genetics • volume 21 • february 1999 211

observed in another series of transgenic Drosophila containingboth domains (‘i1-2∆’, nt 167−2,484); in this case, a strongersilencing effect was observed, evidenced both by the repressionlevel at generation 20 after transgenesis (Fig. 5) and the kineticsfor establishment of repression (one-half maximal repressionoccurred within less than 10 generations; data not shown). A thirdseries of transgenic Drosophila was constructed in which the lattertransgene was deleted for the hsp70 promoter, which was replacedby a polyadenylation signal to prevent transcription from crypticpromoters in the transgene or flanking DNA. No repression wasobserved under these conditions (Fig. 5), suggesting that repres-sion requires synthesis of an RNA transcript.

Our results show that Drosophila can repress the activity ofincoming I elements by a very simple process, requiring only thepresence of a transgene containing part of the I element. This‘immunization’ against invading I elements is not dependent onI-element proteins acting as negative regulators, but on transgenedosage, and can be reversed by paternal transmission. Theseresults are similar to the properties of cosuppression seen in sev-eral single-copy genes in plants (for review, see refs 12,13) and inAdh in Drosophila20. The kinetics for establishment of repressionand transmission of the repressed state via females only are simi-lar to those observed under ‘natural’ conditions of hybrid dysge-nesis for crosses between non-transgenic Drosophila: uponintroduction of functional I elements into Drosophila devoid ofsuch elements, the activity of the invading elements also takesseveral generations to be fully repressed, and the repressed phe-notype is only maternally transmitted (for review, see refs 3,4).These data suggest that Drosophila use cosuppression as a mecha-nism to protect themselves against the mutagenic effects—in thepresent case, embryo lethal—of transposable elements. Previousstudies in plants have shown that cosuppression may be involvedin virus resistance: plants expressing transgenes containingsequences derived from the genomes of RNA viruses becameresistant to that virus17,22,23. Cosuppression might thus be a gen-eral protection mechanism; at the evolutionary level, mobile ele-

Fig. 5 Regulation of I-element activity is not dependent on specific I sequenceswithin the transgenes, but requires their transcription. I-element activity wasquantitated as in Fig. 2, for transgenic Drosophila (at generation 20 after trans-genesis) with the indicated transgenes, including the I ORF1 domain (I frag-ment, nt 167–1,492, hsp[i1]pA), or the i1-2∆ domain (I fragment, nt 167–2,484;[i1-2∆]pA), in the presence or in the absence of the hsp70 promoter (±hsp);data shown are the mean values (±s.d.) for 5 batches of 40 embryos; the trans-gene copy-numbers were determined by Southern-blot analysis.

ments—as ‘naturally occurring’ endogenous multicopy genes—might constitute the raison d’être of cosuppression.

An important question in the field of cosuppression concernsthe nature of the regulating signal, as well as the molecular mech-anisms for both triggering and establishment of repres-sion12,13,18. In some cases, homology-dependent gene silencingdid not require transcription of the transgene, but occurredunder specific conditions where ‘silencing’ sequences were intro-duced as tandem repeats or concatamers16,24,25. This so-called‘repeat-induced gene silencing’ seems to be a consequence ofDNA-DNA interactions, and has been evoked as a possible mech-anism for I-element silencing26. Conversely, we show that tran-scription of the transgene is necessary to produce I-elementsilencing. This has also been shown for homology-dependentgene silencing in plants using dispersed copies of transgenes (forreview, see refs 12,14,15,18,27). In these instances, it has beenproposed that the cosuppression signal is an RNA molecule thatmust reach a threshold to produce cosuppression. This is consis-tent with the present evidence of a lag of several generationsbefore repression develops and our observation that the silencingfactor can persist for at least one generation in the absence ofDNA copies of the transgene; non-transgenic offspring fromsilenced heterozygous transgenic mothers still repress I-elementactivity (53±23% of remaining I-element activity for single-copyhsp[i2∆]pA strains; data not shown). Several studies in plants13

and one in Drosophila20 suggest that both an RNA signal andimprinting at the chromatin level are required for cosuppression.Furthermore, it has been shown that viral- or transgene-gener-ated RNA could direct de novo epigenetic modification of ahomologous sequence in the plant genome28. Consequently, onecould propose that the origin of the cosuppression signal is anRNA molecule, that this RNA has an effect on the chromatinstructure of homologous genes (possibly triggered directly in thecase of tandem repeats) and that modified chromatin structurehas a role in the production and nature of the RNA effector mol-ecule itself. In light of the observed regulation of I-element activ-ity, such a positive control of the cosuppression signal on its ownproduction, associated with transmission through the femalegerm line, presents an attractive explanation for its accumulationover generations and for the physiology of control of dispersedinvading elements.

MethodsDNA constructs. Hsp[i2∆]pA and hsp[i1-2∆]pA were obtained byinserting either the HindIII-XbaI fragment ‘i2∆’ from pI407 (nt 1,516−2,484; ref. 29) or the ‘i1-2∆’ fragment (nt 167−2,484 in ref. 29) obtainedby EcoRI-XbaI restriction from pAct[I] (ref. 30) into the pW8-hsp-pAvector restricted by HpaI; the latter was constructed by introduction ofthe hsp70 promoter as a XhoI-HindIII fragment from phspCAT (gift fromP. Herbomel) into the BamHI site of the pW8 transformation vector andfurther insertion of the 1.23-kb BamHI-SalI fragment from pAct5C-PPA(gift from M.S. Levine), containing the Actin 5C polyadenylation signal,into the SphI site of pW8. Hsp[i2∆*]pA was constructed as hsp[i2∆]pAafter introduction of a multiple stop-codon containing linker (5´−AGTACTAACTAGTTAGATCT−3´) into the unique EcoRV site of the Ielement. Hsp[i1]pA was constructed by eliminating the ‘i2∆’ fragment

© 1999 Nature America Inc. • http://genetics.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://g

enet

ics.

nat

ure

.co

m

letter

212 nature genetics • volume 21 • february 1999

1. Picard, G. & L’Héritier, P. A maternally inherited factor inducing sterility in D.melanogaster. Drosophila Information Service. 46, 54 (1971).

2. Bregliano, J.C. et al. Hybrid dysgenesis in Drosophila melanogaster. Science 207,606–611 (1980).

3. Finnegan, D.J. The I factor and I-R hybrid dysgenesis in Drosophila melanogaster.in Mobile DNA (eds Berg, D.E. & Howe, M.M.) 503–517 (American Society forMicrobiology, Washington D.C., 1989).

4. Bucheton, A. I transposable elements and I-R hybrid dysgenesis in Drosophila.Trends Genet. 6, 16–21 (1990).

5. Waugh O’Neill, R.J., O’Neill, M.J. & Marshall Graves, J.A. Undermethylationassociated with retroelement activation and chromosome remodelling in aninterspecific mammalian hybrid. Nature 393, 68–72 (1998).

6. Pélisson, A. & Bregliano, J.C. Evidence for rapid limitation of the I element copynumber in a genome submitted to several generations of I-R hybrid dysgenesis inDrosophila melanogaster. Mol. Gen. Genet. 207, 306–313 (1987).

7. Pritchard, M.A., Dura, J.M., Pélisson, A. & Finnegan, D.J. A cloned I-factor is fullyfunctional in Drosophila melanogaster. Mol. Gen. Genet. 214, 533–540 (1988).

8. Dawson, A., Hartswood, E., Paterson, T. & Finnegan, D.J. A LINE-like transposableelement in Drosophila, the I factor, encodes a protein with properties similar tothose of retroviral nucleocapsids. EMBO J. 16, 4448–4455 (1997).

9. Matzke, M.A. & Matzke, A.J.M. How and why do plants inactivate homologous(trans)genes? Plant Physiol. 107, 679–685 (1995).

10. Jorgensen, R.A. Cosuppression, flower color patterns, and metastable geneexpression states. Science 268, 686–691 (1995).

11. Vaucheret, H., Palauqui, J.-C., Elmayan, T. & Moffatt, B. Molecular and geneticanalysis of nitrite reductase co-suppression in transgenic tobacco plants. Mol.Gen. Genet. 248, 311–317 (1995).

12. Baulcombe, D.C. RNA as a target and an initiator of post-transcriptional genesilencing in transgenic plants. Plant Mol. Biol. 32, 79–88 (1996).

13. Meyer, P. & Saedler, H. Homology-dependant gene silencing in plants. Annu. Rev.Plant Physiol. Plant Mol. Biol. 47, 23–48 (1996).

14. Depicker, A. & Van Montagu, M. Post-transcriptional gene silencing in plants.Curr. Opin. Cell Biol. 9, 373–382 (1997).

15. Metzlaff, M., O’Dell, M., Cluster, P.D. & Flavell, R.B. RNA-mediated RNAdegradation and chalcone synthase A silencing in Petunia. Cell 88, 845–854(1997).

16. Dorer, D.R. & Henikoff, S. Expansions of transgene repeats cause heterochromatinformation and gene silencing in Drosophila. Cell 77, 993–1002 (1994).

17. Goodwin, J. et al. Genetic and biochemical dissection of transgenic RNA-mediated virus resistance. Plant Cell 8, 95–105 (1996).

18. Bingham, P.M. Cosuppression comes to the animals. Cell 90, 385–387 (1997).19. Rossignol, J.-L. & Faugeron, G. Gene inactivation triggered by recognition

between DNA repeats. Experientia 50, 307–317 (1994).20. Pal-Bhadra, M., Bhadra, U. & Birchler, J.A. Cosuppression in Drosophila: gene

silencing of Alcohol dehydrogenase by white-Adh transgenes is Polycombdependent. Cell 90, 479–490 (1997).

21. Jensen, S., Cavarec, L., Gassama, M.P. & Heidmann, T. Defective I elementsintroduced into Drosophila as transgenes can regulate reactivity and prevent I-Rhybrid dysgenesis. Mol. Gen. Genet. 248, 381–390 (1995).

22. English, J.J., Mueller, E. & Baulcombe, D.C. Suppression of virus accumulation intransgenic plants exhibiting silencing of nuclear genes. Plant Cell 8, 179–188(1996).

23. Ratcliff, F., Harrison, B.D. & Baulcombe, D.C. A similarity between viral defenseand gene silencing in plants. Science 276, 1558–1560 (1997).

24. Assaad, F.F., Tucker, K.L. & Signer, E.R. Epigenetic repeat-induced gene silencing(RIGS) in Arabidopsis. Plant Mol. Biol. 22, 1067–1085 (1993).

25. Garrick, D., Fiering, S., Martin, D.I.K. & Whitelaw, E. Repeat-induced genesilencing in mammals. Nature Genet. 18, 56–59 (1998).

26. Chaboissier, M.-C., Bucheton, A. & Finnegan, D.J. Copy number control of atransposable element, the I factor, a LINE-like element in Drosophila. Proc. NatlAcad. Sci. USA 95, 11781–11785 (1998).

27. Montgomery, M.K. & Fire, A. Double-stranded RNA as a mediator in sequence-specific genetic silencing and co-suppression. Trends Genet. 14, 255–258 (1998).

28. Wassenegger, M., Heimes, S., Riedel, L. & Sänger, H.L. RNA-directed de novomethylation of genomic sequences in plants. Cell 76, 567–576 (1994).

29. Fawcett, D.H., Lister, C.K., Kellet, E. & Finnegan, D.J. Transposable elementscontrolling I-R hybrid dysgenesis in D. melanogaster are similar to mammalianLINEs. Cell 47, 1007–1015 (1986).

30. Jensen, S., Cavarec, L., Dhellin, O. & Heidmann, T. Retrotransposition of a markedDrosophila LINE-like I element in cells in culture. Nucleic Acids Res. 22, 1484–1488(1994).

from hsp[i1-2∆]pA upon restriction with HpaI and BamHI. The pro-moterless pA[i1-2∆]pA was constructed as hsp[i1-2∆]pA, except that thehsp70 polyadenylation sequence as a BamHI-EcoRI fragment from HZ50(gift from W.J. Gehring) was inserted in the place of the hsp70 promoter.

Drosophila strains and P-mediated transformation. Flies were raised at22 °C±1 on standard medium, and strains were maintained using onlyyoung flies as described21. The w1118 and the reactive wK strains21 were giftsfrom D. Coen and C. McLean. P-mediated germline transformation wasperformed as described21. Transgene copy-number was assessed by South-ern blots of genomic DNA restricted with enzymes cutting both in and out-side the transgene and probed with a hsp70 promoter fragment, by count-ing the number of fragments containing flanking DNA. The results wereconfirmed by quantitating the intensity of an internal, transgene specificfragment using PhosphorImager technology (Storm 840 scanner). Trans-genic strains were controlled (at generations 14 and 63 after transgenesis)for the absence of contamination by functional I elements by crossingtransgenic males with reactive wK females21.

Measurements of the level of I-element activity. Groups of 15 females weremated with 20 w1118 males (containing functional I elements), when lessthan 4 d old. The first 20 females and 20 males born from each batch of testcrosses were collected and allowed to mate. When less than 4 d old, wetransferred these flies to an egg-collector. Sixteen hours later, 5 batches of40 eggs were deposited as 4×10 matrices, thus allowing unambiguouscounting (after 48 h) of hatched and non-hatched (dead) embryos. Wekept the temperature at 22±1 °C throughout, as the intensity of the hybriddysgenesis syndrome is influenced by temperature changes.

AcknowledgementsWe especially thank M. Bartozzi for invaluable technical assistance and M.Ashburner, C. Lavialle and L. Cavarec for critical reading of the manuscriptand helpful discussions.

Received 1 September; accepted 8 December 1998.

© 1999 Nature America Inc. • http://genetics.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://g

enet

ics.

nat

ure

.co

m