To identify IAP elements encoding the proteins necessary for retrotransposition, we first devised an ex vivo assay in heterologous cells (Fig. 1), which combined a reporter IAP element marked with an indicator gene for retrotransposition and expression vectors for full-length genomic IAP elements. The marked reporter is derived from an internally deleted IAP (a naturally occurring type IΔ1 element1 that recently transposed within the gene Il3 and is still competent but not autonomous for transposition2), which we marked with the neoTNF indicator gene for retrotransposition3 (described in Fig. 1b,c). The full-length IAP elements were sorted by an in silico search on the available sequence of the mouse genome, by imposing that the three retroviral genes gag, prt and pol be fully coding. A search on the nr database in GenBank (2% of the mouse genome, May 2001 release) identified 18 full-length elements, of which 9 had three canonical open reading frames (ORFs) separated from one another by a −1 frameshift4 (Fig. 1a). In addition, their two LTRs shared a high degree of identity (0 or 1 nucleotides (nt) different), consistent with recent integration of these elements. The search also identified copies with three intact ORFs but with a 1-nt insertion into the prt-pol frameshift signal resulting in fusion of these two genes. We retained a series of six full-length IAP elements for which BAC clones could be obtained, including four canonical elements with the −1 frameshift and two 'mutant' elements with prt-pol fusion (Fig. 2). For the assay in trans, we subcloned these elements into a unique expression vector, in which we inserted the 5′ U5-gag-prt-pol-3′ U5 IAP proviral fragment in place of the corresponding region of the internally deleted Il3 IAP (whose LTR has strong promoter activity, possibly owing to the presence of flanking Il3 enhancer sequences2).

Figure 1: Structure of IAP elements and description of the assay for retrotransposition.
figure 1

(a) Genomic organization of full-length IAP elements. The LTRs (U3-R-U5) border three ORFs homologous to the retroviral genes gag, prt (protease) and pol and are separated from one another by a −1-nt frameshift at the indicated heptameric slippery sites4. The transcription start site is marked with an arrow, and the signals necessary for retroviral reverse-transcription are indicated. PBS, primer binding site; PPT, polypurine tract. (b) Schematic representation of an IAP marked with the reporter gene neoTNF (light gray), in which neo placed in reverse orientation with its own promoter (Pr) is rendered inactive by the presence of a forward intron (top), which should be spliced out in the transposition RNA intermediate (middle), resulting in an active neo gene in the de novo IAP insertion (bottom). (c) Experimental procedure for detection of IAP retrotransposition. Cells were transfected either with a neoTNF-marked defective IAP and an expression vector for the selected IAPs or with a full-length neoTNF-marked IAP alone. Cells were amplified for 1 wk and retrotransposition events detected after G418 selection.

Figure 2: Mobilization of a marked IAP reporter using expression vectors for cloned genomic IAPs.
figure 2

(a) The neoTNF-marked IAP reporter was assayed for retrotransposition in heterologous feline cells after cotransfection with a control plasmid (pCMVβ) or an expression vector for the cloned IAP elements (BAC clone number indicated, structural organization schematized). Results of the G418 selections are illustrated (in the plates shown, the G418R foci were fixed and stained). The number of foci is given for each construct as the mean ± s.e. of 2–4 independent experiments, carried out with at least 4 × 106 cells. The resulting retrotransposition frequencies are also shown. (b) G418R clones are associated with retrotransposition. Genomic DNA of individual clones from a was assayed by PCR for the presence of retrotransposed copies with the intron of the neoTNF indicator spliced out (using primers bracketing the intronic domain as indicated), and the PCR products were separated by electrophoresis in agarose gels stained with ethidium bromide. All clones tested had a band corresponding to the size of an intronless copy of the neo gene (intron-containing and intron-less PCR size markers indicated for each gel), and no band was detected with mock-transfected cells (control).

We cotransfected heterologous cells with the IAP reporter and the expression vector for each of the cloned full-length IAP proviruses and then selected G418-resistant clones (Fig. 2a). These assays resulted in a high number of clones (frequency in the range of 10−4) for the four canonical IAP elements but very few or no clones for a control vector or for the two mutant IAP copies with the prt-pol fusion. To confirm that the G418R clones resulted from retrotransposition of the marked element, we verified that the intron present on the indicator gene in the initial marked copy was spliced out in the transposed IAPs by PCR analysis of genomic DNA for a series of randomly selected G418R clones (Fig. 2b).

We then selectively inactivated each of the three IAP genes, by deletions in either gag or prt (in-phase so as to preserve translation of the downstream genes) and by introducing an early termination codon in pol. None of these modifications altered the RNA level of the transfected IAPs (Fig. 3a). All three mutants had a much lower rate of transposition (<10−7 versus 10−4 for the wild-type element), indicating that the three genes are necessary for transposition. It has been reported that the Gag precursor remains uncleaved1,5, which seems to contradict our observation that the protease gene is required for transposition. We therefore carried out western-blot experiments with cells transfected with the wild-type element and the prt mutant. Comparison of the two profiles (Fig. 3b) indicates that partial processing of the Gag precursor specifically occurred in the presence of a full-length prt ORF (with evidence for cleavage products of 35 kDa and 24 kDa), consistent with the functional data for retrotransposition.

Figure 3: IAP gene products required for retrotransposition.
figure 3

(a) Assay for IAP transposition was done as in Figure 2a, with expression vectors (schematized on the left) encoding either the wild-type IAP 440N1 (WT) or the same element rendered defective for gag, prt (in-phase deletion so as not to alter translation of the downstream ORFs) or pol (insertion of a premature termination codon). Transcript levels for each IAP construct (determined by northern blots) are shown. The retrotransposition frequencies obtained for each construct are given as the mean ± s.e. of 2–4 independent experiments, carried out with at least 4 × 106 cells. In the plates shown, the foci were fixed and stained. (b) Western-blot analysis for cleavage of the IAP Gag polyprotein. Lysates of cells transfected with a control plasmid (mock) or an expression vector for either the wild-type IAP 440N1 (WT) or the same element rendered defective for the prt gene (prt) were separated by electrophoresis, blotted and hybridized with a rabbit antiserum directed against the capsid domain of the Gag protein. Both IAP expression vectors drive the synthesis of the full-length Gag protein (black arrowhead, 73 kDa) with a few degradation products in common (bracket), whereas only the wild-type prt ORF results in the occurrence of two additional specific products (empty arrowheads) of 35 and 24 kDa (prolonged exposure shown below) corresponding to a partial processing of the Gag protein.

We then tried to rescue the two IAP elements with a prt-pol fusion: we deleted the 1-nt insertion that places prt and pol in-frame in these elements and assayed the 'corrected' copies for retrotransposition (data not shown). For the 262J21 element, this single mutation increased retrotransposition efficiency by a factor of 10, suggesting that, as for Ty1 elements6,7,8, the ratios of the various 'viral' polyproteins, which depend on the frameshift signals, are essential for retrotransposition. The same correction made on the 31B18 IAP element did not restore any activity, indicating that in this case, other mutations impair the function of the encoded proteins. The first and most extensively studied full-length IAP element (IAP14; refs. 4,5,9) was defective in this assay for unknown reasons (data not shown).

We next determined whether the four canonical IAP elements described above could promote their own mobilization, that is, whether they also contained the signals required in cis for their efficient autonomous retrotransposition. We recloned each IAP element as a complete LTR-gag-prt-pol-LTR provirus, marked them with the indicator gene downstream of the coding sequences (Fig. 4a) and assayed their autonomous retrotransposition in heterologous cells. Three of the elements had high retrotransposition activity (Fig. 4a). We found that these three active, autonomous copies belong to the previously described PC subfamily10, whereas the nearly inactive one is a member of the LS subfamily11. Because this classification was based on a few diagnostic positions in the LTR (notably in the TATA box), we tested the promoter activity of each 5′ LTR using LTR-luciferase constructs. These assays confirmed (Fig. 4a) that the LTR promoter efficiency of the LS element was 3–30 times lower than that of the PC elements; this probably accounts for its lower rate of autonomous retrotransposition. For the three PC elements, transposition rates measured in the cis assay do not correlate perfectly with their relative LTR promoter strength, probably because of subtle differences in their enzymatic transposition machinery. This is supported by their different transposition rates in the assay in trans where they are placed under the control of the same promoter (Fig. 2a) but could not be directly inferred from a comparison of their amino acids sequences.

Figure 4: Assay for autonomous IAP retrotransposition and characterization of the de novo insertions.
figure 4

(a) Autonomous IAP retrotransposition. Top and middle sections: each of the four IAP elements encoding functional proteins (copy name indicated) was cloned as a complete provirus with the neoTNF indicator inserted downstream of the IAP ORFs as schematized and tested for its autonomous retrotransposition in heterologous feline (G355.5) and human (HeLa) cells as described in Figure 1c; retrotransposition frequencies are the mean ± s.e. of 2–3 independent experiments. The marked L1-2B human LINE element used as a control (data not shown) yielded transposition frequencies of 1.5 × 10−3 and 1.1 × 10−2 in G355.5 and HeLa cells, respectively. Bottom section: LTR-luciferase activities of the corresponding IAPs in both cell types. Values expressed in arbitrary units are the mean ± s.e. of 2–3 independent experiments. (b) Structure and chromosomal localization of nine transposed IAPs. The complete characterization of transposed IAP copies and insertion sites was done using the genomic IAP clone 92L23 marked as indicated in a and assayed for transposition in HeLa cells. A de novo IAP insertion and the corresponding empty site is schematized. The sequences of nine insertions are given below, with the flanking DNA in uppercase and the proviral sequences in lowercase; target-site duplications (TSD, shaded) are found in all nine cases, associated with complete LTRs. The GenBank accession number of each insertion site is given, together with its chromosomal localization (indicated with an arrowhead; R bands are shown in white and G bands in dark gray); seven of nine insertions are in R bands.

We confirmed canonical retrotransposition for the most active IAP copy by characterizing (by inverted PCR) several transposed copies and their insertion sites (Fig. 4b). These newly inserted elements had perfect 6-bp target-site duplications as well as the canonical structure of endogenous and de novo transposed IAPs1,2. In six of nine cases, insertion occurred in actively transcribed domains (as inferred from an expressed-sequence tag–based analysis using the University of California Santa Cruz Genome Bioinformatic site; data not shown), consistent with the integration preference of retroviral elements12.

We next investigated whether IAP retrotransposons had a cis preference effect similar to that reported for non-LTR LINE retrotransposons, in which a coding mRNA is a far better substrate for retrotransposition than a noncoding one complemented in trans13,14. We compared the retrotransposition rate of a full-length autonomous marked IAP with that of the same marked IAP rendered defective for protein synthesis (premature stop codon in the gag gene) and complemented in trans by an expression vector for the IAP ORFs. All three constructs are derived from the same provirus (440N1), placed (as in Fig. 2a) under the control of the same strong LTR promoter (from Il3 IAP). G418R clones were 30 times more numerous when the IAP proteins were encoded by the marked mRNA itself than when they were provided in trans (Fig. 5).

Figure 5: Cis preference for IAP retrotransposition.
figure 5

Assays for retrotransposition were carried out in feline G355.5 cells transfected with (top) an autonomous marked IAP element (0.75 μg of DJ33/440N1neoTNF and 0.75 μg of control pCMVβ) or (bottom) the unmarked IAP element acting in trans on the IAP marked with neoTNF and rendered defective for the IAP proteins (premature stop codon in gag indicated by the asterisk), using 0.75 μg of each construct. In the plates shown, the foci were fixed and stained. The corresponding retrotransposition frequencies (mean ± s.e. of 2–4 independent experiments) are also shown. The higher transposition rates compared with those in Figure 4a are caused by the use of a strong LTR promoter (also used in Fig. 2a).

The identification of functional retrotransposition elements provides a unique opportunity to visualize and characterize the IAP gene products. Transfection of cells with a functional IAP yielded numerous viral-like particles in the cisternae of the endoplasmic reticulum (Fig. 6a), whereas cells transfected with a control vector did not. The particles were 90 nm in diameter with an A-type, immature-like morphology (Fig. 6b) and were not substantially different from particles spontaneously expressed in mouse tumor cells1. Some of them were associated in doublets (Fig. 6c), probably resulting from fusion of individual particles (due to overexpression in the transient transfection). Budding could be detected at the endoplasmic reticulum membrane (Fig. 6d), and no particles were found outside the cells. Close examination of a large (>100) set of particles did not show any obvious heterogeneity among them. Transfection of cells with the IAP protease mutant vector (Fig. 6e) resulted in a similarly large number of intracisternal particles undistinguishable from those produced by the wild-type IAP. These observations are consistent with the low level of protein cleavage shown by western-blot analysis (Fig. 3b). Although we cannot exclude the possibility that the functional particles (mediating retrotransposition) underwent a complete cleavage and had a mature morphology too rare to be observed, it is more likely that all the particles were subjected to a limited, protease-dependent processing that did not change their morphology but yielded a sufficient amount of pol gene enzymatic products to drive reverse-transcription and integration.

Figure 6: Electron microscopy of the viral-like particles generated by a retrotransposition-competent IAP.
figure 6

G355.5 feline cells were transfected with the IAP element in Figure 5 (wild-type in ad, prt mutant in e) and observed 48 h after transfection. (a) Representative low-magnification image of a transfected cell, with numerous particles grouped in the cisternae of the endoplasmic reticulum. The nucleus (N) is indicated. (b) High-magnification image of a representative particle, with two concentric shells surrounding an electron-lucent core. (c) Example of merged particles (a doublet in the present view), as frequently observed in the assay. (d) Example of particles budding into the cisternae of the rough endoplamic reticulum. (e) Particles obtained with the expression vector for the protease mutant; morphology is similar to that of wild-type particles, and merged particles (as in c) are also observed. Scale bars: a, 500 nm; be, 100 nm.

We identified, for the first time in mammals, LTR retrotransposons that encode all the proteins required for their mobility and that transpose autonomously in heterologous cells with a high efficiency. Transposition rates (up to 10−3) were similar to those of L1 LINEs3,15,16,17, which are non-LTR retrotransposons responsible for most of the insertional mutations observed in humans18. We found that all three genes gag, prt and pol are required for transposition and that transposition of coding IAP elements is more efficient than that of defective IAPs complemented in trans by an active element. This result is somewhat unexpected for a retroviral-like element but reminiscent of the strong cis effect observed for the non-LTR LINE retrotransposons13,14. This result could be interpreted, as for LINEs, by assuming that the coding transcript during translation is a preferred target for the retrotransposon nascent proteins. This would ensure that active copies are preferentially transposed, preventing transposon extinction. Close examination of the mouse genome shows that most IAPs (700 of the 1,000 genomic copies) are full-length elements, and computations from our in silico analysis indicate that more than half of them contain three canonical ORFs. Our functional analysis of six such randomly selected elements suggests that a large fraction of these full-length coding IAPs (possibly 300) could be transposition-competent, autonomous elements.

Finally, the identified autonomous IAP copies could be used as safe (noninfectious) insertional mutagens in a large series of cells owing to their high transposition rate, being possibly useful complements to the Mariner DNA-transposon19 as well as to LINEs. Because they are LTR retrotransposons, IAPs might insert with distinct target site specificities12. In addition, IAPs are only present in some rodents, and their mobility therefore might not be restricted by specific 'epigenetic' taming processes (e.g., homology-dependent gene silencing) as observed for transposons in several organisms20,21,22,23, making them appropriate mutagenic agents.



We derived full-length IAP-containing plasmids from commercial BACs (BACPAC Resources). GenBank accession numbers, positions of the first and last nucleotides of the IAP proviruses and strand orientation were as follows: RP23-231P12: AL590616, 87,256–94,327, +; RP23-262J21: AL450406, 135,142–142,227, +; RP23-31B18: AC074106, 150,810–143,692, −; RP23-324C9: AC090431, 15,424–22,549, +; RP23-440N1: AC087891, 163,007–155,853, −; RP23-92L23: AC012382, 161,601–168,684, +.

Constructs for the assay in trans.

We constructed the marked defective IAP element (DJ33 neoTNF reporter plasmid) by removing the SalI fragment corresponding to neoRT from DJ33 (ref. 2) and replacing it by a blunt-ended HindIII-SalI fragment from pSVneoTNF (ref. 3) containing the indicator gene neoTNF. We obtained the expression vectors for the six IAP elements by introducing the corresponding MluI-MluI IAP internal fragment (6.9 kb extending from the end of 5′ LTR to the end of 3′ LTR) excised from the BACs into the IAP-DJ33 reporter digested by MluI. We constructed the gag, prt and pol IAP mutants using the IAP 440N1 expression vector by either in-frame deletions (gag: 813-nt deletion, 891–1,703; prt: 411-nt deletion, 2,410–2,820) or insertion of a linker containing multiple stop codons (for the pol mutant) at the unique PmlI site (nucleotide position 3,417).

Construction of the autonomous marked IAPs.

We amplified the 5′ LTRs of the four functional IAPs on the corresponding BAC by PCR and cloned them into the polylinker of the pGL3 Basic Vector (Promega; pGL3/LTR IAP plasmids). We excised the MluI IAP internal fragments from the BACs, inserted them into the unique MluI site of the corresponding pGL3/LTR IAP plasmids and inserted the neoTNF reporter gene at the SalI site (IAP nucleotide position 5,744) of the resulting plasmids as a PCR fragment (SalI is located 36 nt upstream of the pol stop codon, and pol was restored by introducing the lacking sequence in the primer).

Constructs for the cis-trans assay.

We constructed the expression vector for the functional marked IAP (DJ33/440N1neoTNF) used in the test in cis by inserting the neoTNF indicator gene at the NdeI site (downstream of pol, nucleotide position 6,566) of the IAP-440N1 expression vector. This marked IAP was additionally rendered defective by introducing a 2-nt frameshift in gag (by religating the ClaI opened and blunt-ended vector), thus resulting in a premature stop codon.

Cells, transfection and transposition assays.

We grew G355.5 feline and human HeLa cells in Dulbecco's modified Eagle medium supplemented with 10% fetal calf serum (Gibco BRL), 100 μg ml−1 streptomycin and 100 U ml−1 penicillin. The day before transfection, we seeded 5 × 105 cells per 60-mm dish. For transfection, we used 12 μl of lipofectamine, 8 μl of Plus Reagent (Gibco BRL) and 1 μg (single transfection) or 1.5 μg (cotransfection) of DNA for the G355.5 cells or 4 μg of DNA for the HeLa cells, following the manufacturer's instructions. To assay for retrotransposition, we expanded transfected cells for 6 d, seeded them at 5 × 105 cells per 100-mm dish and allowed them to settle for 24 h before adding G418 (560 μg ml−1, Gibco-BRL). After selection for 10–15 d, we fixed, stained and counted or individually pipetted and plated G418R foci. We extracted genomic DNA of some clones and subjected it to a PCR analysis for the presence of a spliced copy of the neo reporter gene as described17, using primers neo12 and sv3 (sequences available on request). We carried out northern-blot analysis as described17 using the PmlI-SalI IAP fragment as a probe.

Characterization of retrotransposed copies and flanking sequences.

We determined integration sites by inverted PCR, essentially as described17. We digested 5 μg of each clone with DraI, SspI or a mix of BamHI and BclI. After removing the enzymes, we diluted and self-ligated the DNA overnight. We precipitated it with ethanol, resuspended it in 20 μl of 10 mM Tris and subjected it to one round (or in some cases two rounds) of PCR with sets of divergent primers (selected for the amplification of either the 5′ or 3′ proviral ends). We gel-purified, cloned and sequenced the amplified DNAs, yielding insertion sites. We further analyzed the corresponding inserted IAP copies by direct PCR amplification (and sequencing), using primers derived from the flanking sequences and the neo reporter gene. All primer sequences are available on request.

Gag antiserum and western-blot analysis.

We cloned a HpaI-NotI blunt-ended fragment from the MIA14 IAP genome (nucleotide positions 1,547–1,914) corresponding to the 3′ half of the putative Gag capsid domain into the pET-19b plasmid (Novagen) opened by XhoI and blunt-ended, to allow prokaryotic expression as a histidine-tagged protein. We then purified the recombinant protein on a HiTrap column (Amersham) loaded with NiSO4 and used it to immunize rabbits. We affinity-purified the antiserum (Agrobio, France) before use. For analysis of Gag cleavage, we lysed cells transfected with the expression plasmids for the wild-type IAP, the prt mutant or a control in Laemmli buffer 48 h after transfection. We separated whole-cell lysates by SDS-PAGE using gels containing 12% polyacrylamide (37.5:1 ratio of acrylamide to N,N-methylenebisacrylamide). We transferred proteins to nitrocellulose membranes (Schleicher & Schuell) and incubated them with the affinity-purified rabbit antiserum (1:500 dilution). We used goat antibody to rabbit IgG conjugated to horseradish peroxidase (Amersham) as the secondary antiserum.

LTR promoter activity.

To assay for IAP LTR activity, we transfected feline G355.5 cells and human HeLa cells with an equal mass of pGL3-LTR IAP plasmid and pRL-TK vector (Promega) to normalize transfection efficiencies. We recovered cell lysates 48 h after transfection and tested them for activities of both luciferases with the Dual Luciferase Reporter Assay System (Promega) following the manufacturer's instructions.

Electron microscopy.

For morphological studies, we grew G355.5 cells in monolayer and fixed them in situ for 1 h at 4 °C with 1.6% glutaraldehyde (Taab Lab Equip Ttd) in 0.1 M Sörensen phosphate buffer, pH 7.3. We scraped the cells from the plates during fixation and centrifuged them at 4 °C. We rinsed the fixed pellets for 2 h in ice-cold phosphate buffer, post-fixed them with 2% aqueous osmium tetroxide (Pelanne Instruments) and dehydrated them in increasing concentrations of ethanol before embedding them in Epon. We collected ultrathin sections on 200-mesh grids coated with Formvar and carbon, stained them with uranyl acetate and lead citrate and observed them with a Zeiss EM902 transmission electron microscope.


We determined the chromosomal localization of the transposed IAP insertions using Ensembl (available at