Integration of a vector containing a repetitive LINE-1 element in the human genome.

Mammalian cells contain numerous nonallelic repeated sequences, such as multicopy genes, gene families, and repeated elements. One common feature of nonallelic repeated sequences is that they are homeologous (not perfectly identical). Our laboratory has been studying recombination between homeologous sequences by using LINE-1 (L1) elements as substrates. We showed previously that an exogenous L1 element could readily acquire endogenous L1 sequences by nonreciprocal homologous recombination. In the study presented here, we have investigated the propensity of exogenous L1 elements to be involved in a reciprocal process, namely, crossing-overs. This would result in the integration of the exogenous L1 element into an endogenous L1 element. Of over 400 distinct integration events analyzed, only 2% involved homologous recombination between exogenous and endogenous L1 elements. These homologous recombination events were imprecise, with the integrated vector being flanked by one homologous and one illegitimate junction. This type of structure is not consistent with classical crossing-overs that would result in two homologous junctions but rather is consistent with one-sided homologous recombination followed by illegitimate integration. Contrary to what has been found for reciprocal homologous integration, the degree of homology between the exogenous and endogenous L1 elements did not seem to play an important role in the choice of recombination partners. These results suggest that although exogenous and endogenous L1 elements are capable of homologous recombination, this seldom leads to crossing-overs. This observation could have implications for the stability of mammalian genomes.

One common feature of repeated sequences is that they are homeologous (not perfectly identical). Our laboratory has been studying recombination between homeologous sequences by using the LINE-1 (long interspersed nuclear elements or Li) repeated elements as substrates. The Li family of interspersed repetitive elements is ubiquitous in mammals (17). Their number is in the order of 105 copies dispersed throughout the genome, such that the density of Li elements is of about one copy per 30 to 60 kb of genomic DNA (14,18). Li elements are estimated to account for 5 to 10% of mammalian genomes. A full-length Li element is 6 to 7 kb long, but over 90% of Li elements are truncated at the 5' end. In a given species, there is a very large spectrum of homology between individual copies that can range from 99% to less than 75%, with most copies in the range of 98 to 90% (17,27,28). Thus, Li elements offer the possibility of examining recombination between sequences that have a very wide range of homeology.
Using an assay monitoring the acquisition of chromosomal sequences by an extrachromosomal vector, we previously showed that an exogenous molecule containing Li sequences could acquire endogenous Li sequences by nonreciprocal homologous recombination (7). Numerous distinct endogenous Li elements had participated in this process. Interestingly, in about half of the cases, homologous recombination involved only one end, such that the acquired Li sequences were delimited on one side by a homologous and on the other side by an illegitimate junction. In the present study, we have investigated the capacity of exogenous Li elements to be involved in a reciprocal process, that is, integration of an exogenous Li element into endogenous Li elements by crossing-overs. We were interested in determining if the presence of such a high density of homeologous targets would affect the propensity of the extrachromosomal vector to integrate by homologous recombination and if the degree of homology would play a role in the choice of partners. The results were that despite the very high density of Li elements in the genome, out of over 400 distinct integration events analyzed, only 2% involved homologous recombination. This homologous recombination did not result in classical crossing-overs, which would lead to homologous junctions flanking each side of the insertion. Instead, we found one homologous junction and one illegitimate junction. This structure is more consistent with one-sided homologous recombination (5) followed by illegitimate integration. For these types of recombination events, the degree of homology did not seem to play an important role in the choice of partners.  (32) was replaced by the same fragment from pSV2neo (29) to augment the G418' activity (35). Then the 2.5-kb EcoRI fragment from the human Li cDNA cDllB (28) was cloned into the BamHI site of the modified pMClneo vector. The herpes simplex virus type 1 tk gene, obtained as a 2.0-kb restriction fragment from pAGO (10), was added at the HincIl site. The resulting vector was called pLlHSIV1.2 (Fig. 1). This vector was further modified by deleting a 23-bp NheI-BamHI fragment in the middle of the Li sequences, resulting in pLlHS1.2del.

MATERIALS AND METHODS
Transfection. Prior to transfection, pLlHSIV1.2 and pLlHS 1.2del were linearized with BamHI. Then 1 jig of vector DNA was electroporated in 5 x 106 GM00637E cells with a gene Zapper 450/2500 apparatus (IBI, New Haven, Conn.) as described by Chu et al. (9) except that incubations were at 4°C. At 72 h after electroporation, 200 ,ug of G418 per ml was added. After 2 weeks, individual colonies were picked or mixed in pools of 10 to 20 colonies.
Selection was performed with ampicillin (50 ,ug/ml). DNA sequencing. DNA sequencing was performed by the dideoxy method (26), using the T7 sequencing kit (Pharmacia LKB Biotechnology AB, Uppsala, Sweden). Sequences were analyzed with the DNasis program (Pharmacia LKB Biotechnology AB).
Comparison with human Li elements. Li elements used for sequence comparison with the vector (see Fig. 6C) were identified by Jurka (19) and obtained from GenBank.

RESULTS
Experimental strategy. To investigate the propensity of cellular Li elements to undergo crossing-overs, we studied the integration of exogenous Li sequences into the human genome. Two types of events were expected: homologous integrations resulting from crossing-overs between the exogenous and endogenous Li elements and illegitimate integrations.
The exogenous human Li sequences were taken from cDii (28) as a 2.5-kb fragment (X-Y in Fig. 1A) and cloned in a vector containing a selectable neo gene. Two plasmids, pLl HSIV1.2 and pLlHS1.2del, that differed one from the other by a 23-bp deletion at the site of linearization in the latter were produced (Fig. 1B). BamHI restriction digestion of these two plasmids produced linear insertion vectors with 1.2 kb of human Li sequences at each end (X and Y). The two vectors were electroporated in human fibroblasts. Cells in which integration had occurred were selected with G418. The ensuing colonies were either picked individually or pooled (10 to 20 colonies per pool). Five separate electroporations were performed. The first was done with pLlHSIV1.2, and the four others were done with pLlHS1.2del. The frequency of G418r colonies obtained varied between 10-' and 10-5 colonies per cell per ,ug of vector DNA.
To determine the number of integration events per clone, 38 independent G418r clones were analyzed by BglII digestion, which cuts once in the vector (Fig. 1C), followed by Southern analysis using the neo gene as a probe. Over 90% of the clones analyzed contained only one integrated vector, while the remainder contained two or three (Fig. 2).
Homologous recombination between Li elements does not lead to a selectable phenotype. Thus, to determine the nature of the integration events, homologous or illegitimate, we cloned the junctions. Genomic BgllI fragments containing plasmid DNA of the integrated vector plus the flanking cellular sequences present on the right side ( Fig. 1C) were rescued by ligation followed by transformation in bacteria. Analysis of the integration events rescued from individual clones indicated that the length of each fragment rescued in bacteria corresponded to the length expected from Southern analysis of the genomic DNA. Thus, the plasmid rescue procedure did not produce rearrangements. RECOMBINATION  Characterization of the integration events. Rescued integration events were analyzed by restriction enzyme mapping, hybridization, and sequencing. The restriction enzymes used were BglII and KpnI, while hybridization was done with probes Li, 3', and P (Fig. 1). The scheme for analysis of the rescued junctions is illustrated in Fig. 3.   (Fig. 3D). These would be discriminated from bona fide integration junctions by the fact that the sequences that should have corresponded to cellular sequences would hybridize with -± +/_ the P probe. Examples of the different types of rescued integration events obtained are shown in Fig. 4A   Hybridization of the rescued integration events with probes P, Li, and 3' is shown in Fig. 4B to D. With probe P (Fig. 4B), the 4.1-kb plasmidic band can be seen as expected, as well as other hybridizing bands in lanes 1 and 2. The additional band in lane 1 corresponds to the remaining fragment of pLlHS1.2del that contains plasmidic sequences. Hybridization of the additional fragment with probe P in lane 2 indicates that this rescued junction is of the type vector-vector junction illustrated in Fig. 3D. About 4% of the rescued junctions were of this type. They were not included in the tabulation of the distinct rescued junctions analyzed (see below) since they represented a vector-vector rather than a vector-cell recombination event. With probe Li (Fig. 4C), the KpnI 1.2-kb fragments can be seen as expected except in lane 5. As explained above, in this case, the KpnI 1.2-kb fragment has been replaced by a plasmid-cell fusion fragment that hybridizes with probe Li because of the remaining X sequences. There are also additional hybridizing bands in lanes 1, 2, and 4. In the case of lanes 1 and 2, this is due to the presence of vector Li sequences in these bands as explained above. The additional band in lane 4 hybridizes with probe Li, which indicates the presence of endogenous Li sequences in the rescued cellular sequences. Finally with probe 3' (Fig. 4D), there are bands hybridizing in lanes 1, 2, and 4. Again, the reason for the bands in lanes 1 and 2 is as explained above. The fact that the cellular sequences in lane 4 hybridize with probe 3' raises the possibility that the rescued integration junction is of the homologous type (Fig. 3A) or alternatively that integration occurred near an endogenous Li element (Fig. 3C). Restriction enzyme analysis and sequencing were used to discriminate between these two possibilities. Note that the absence of bands hybridizing with probe 3' in lanes 3 and 5 indicates that the rescued integration events in these lanes are of the illegitimate type (Fig. 3B).
More than 400 distinct rescued integration events were analyzed as described above (Table 1). The presence of endogenous cellular Li sequences in the rescued junctions was detected in 4% of them, and half of these represented homologous junctions. The variable degree of homology of the Li elements rescued compared with the probes used for their detection (see below) indicated that the hybridization conditions used for analysis enabled us to detect the vast majority of Li elements that could have been present in the rescued plasmids.
Analysis of homologous junctions. The seven homologous integration junctions that were rescued are presented in Fig. 5. All seven junctions had distinct lengths, and restriction maps indicating that they represented distinct integration events. In six of seven cases, the BamHI site used for linearization had been regenerated at the expected position. This BamHI site had most likely been reconstituted by the endogenous Li element, which indicated that the endogenous Li elements were located at the expected position for a homologous junction. To ensure further that these were truly homologous junctions, we sequenced the Li sequences present in the rescued plasmid (data not shown). In all cases, we found that the 23 bp that had been deleted from the vector were present at the expected location. Furthermore, we found in all cases a switch from vector Li sequences to cellular Li sequences, the latter being identified by the variation in sequences compared with the vector Li sequences. These switches occurred at different positions for the different homologous junctions. All seven endogenous Li sequences were unique compared one with the other and with the vector Li sequences.
If these homologous junctions were the result of classical crossing-overs, the left junction should also have been homologous. We rescued and analyzed the left junction of three of VOL 14, 1994~~~RECOMBINATION  these homologous integration events. In all cases, the left junction was found to be illegitimate and did not involve endogenous Li sequences (data not shown). Analysis of the left junctions of other integration events indicated that the left side of the vector could undergo homologous recombination with endogenous Li elements. However, this also resulted in one homologous and one illegitimate junction.
Homology of the endogenous Ll elements. The overall degree of homology between the endogenous Li sequences present at the homologous junction downstream from the vector X sequences and the Li sequences originally present in the vector at that position (Y sequences in Fig. 1) ranged from 98 to 94% (Fig. 6A). This distribution is similar to the one found when the Li sequences present in the vector are compared with the endogenous Li elements which were found near illegitimate recombination junctions (Fig. 6B) and also when the Li sequences of the vector are compared with human Li elements obtained from GenBank (Fig. 6C) (19). Thus, there was no indication that the homologous recombination events involved a marked selection for endogenous Li elements with a higher degree of homology to the Li sequences present in the vector.
Density of LI elements in rescued cellular sequences. To verify that the density of Li elements present in the rescued cellular sequences was representative of the density found in the genome, we calculated the ratio of Li elements per kilobase of cellular DNA rescued in experiments 1 and 2 ( Table 1). There were five distinct elements in a total of 251 kb of cellular DNA. This corresponds to a density of one endog-enous Li element per 50 kb of rescued cellular DNA, which is comparable to the density found in random genomic DNA (one per 30 to 60 kb [14,18]). Thus, it seems that the exogenous Li sequences had access to representative genomic DNA in terms of endogenous Li element density.

DISCUSSION
In this study, we examined the integration of an exogenous Li element into the genome of human fibroblasts. We analyzed over 400 independent integration events and found that only 2% had involved homologous recombination with endogenous Li elements. Although this frequency is in the upper limit of what has been observed for single-copy genes, one might have expected it to be even higher considering the very high copy number of Li elements in the genome. It was shown previously that in mammalian cells, increasing the number of homologous targets did not result in an increase in the ratio of homologous to illegitimate integration events (33,36). In these studies, the targets were localized at very few distinct loci and numbered less than 103. The results presented here indicate that even with i05 potential homologous targets and a density of one homologous target per 30 to 60 kb, the vast majority of integration events were still not homologous.
Since the endogenous Li sequences were homeologous to the exogenous Li sequences, their sequence divergence could have affected the frequency of homologous integration. Previous studies showed that the frequency of homologous integration for single-copy genes was extremely sensitive to any mismatch between the exogenous and endogenous sequences. Divergence of less than 1% resulted in a more than 10-fold reduction in the frequency of homologous integration (ii1, 3 1). Furthermore, when competing homologous targets were present, divergence of less than 5% resulted in at least a 15-fold preference for the homologous versus homeologous target (23). In the study reported here, the exogenous Li sequences had access to targets with a very wide range of homology, yet homologous recombination did not seem to favor endogenous Li sequences with a higher degree of homology to the exogenous sequences. The various degrees of homology of Li sequences involved in homologous recombination were what would be expected if the endogenous Li sequences had been selected randomly. This was also found to be the case when endogenous Li sequences were rescued by exogenous Li vectors with various degrees of homology (6). One possibility for the discrepancy between our results and those cited above (ii, 23,31) is that the latter case involved assays that looked specifically at classical crossing-overs with two homologous junctions, one on each side of the integrated vector. Classical crossing-overs were not observed in the present study. Instead, we found that only one integration junction was homologous, the other being illegitimate.
Integration events with one homologous and one illegitimate junction have been observed previously in gene targeting experiments involving perfectly homologous sequences (1,2,8,13). When both ends of the vector were homologous to the cellular target, as was the case here, integration events with one homologous and one illegitimate junction were associated with a two-step process involving nonreciprocal homologous recombination with the cellular target followed by illegitimate integration elsewhere in the genome. This process left the cellular target unmodified. When only one end of the vector was homologous, integrations with one homologous and one illegitimate junction directly in the cellular target (8) as well as integration by the two-step process described above (2) were reported. Integration events with one homologous and one VOL. 14,1994 on April 26, 2019 by guest http://mcb.asm.org/ Downloaded from 6694 illegitimate junction are not consistent with the double-strand break repair model, which predicts products with two homologous junctions (30), but could be explained by the one-sided invasion model (5,7). In the present study, we cannot differentiate between integration into an endogenous Li element and the two-step process. Nevertheless, despite the presence of homologous ends on both sides of the vector, classical crossingovers were not observed, contrary to what is seen with perfectly homologous sequences.
The considerations presented above raise the possibility that in mammalian cells, homeology interferes with the outcome of the homologous recombination process such as to favor nonreciprocal homologous recombination without crossing-over. This is the case in yeast cells; in this case, it has been observed that homeologous sequences undergo gene conversion but are rarely associated with crossing-overs, contrary to homologous sequences (24,34). This observation is also consistent with data from our laboratory indicating that the reduced homology between two Li sequences does not interfere with the frequency of homologous recombination but favors the production of recombinants with one homologous and one illegitimate junction rather than two homologous junctions (6). We do not know to what extent exogenous-endogenous interactions seen in transfection experiments reflect the normal behavior of genomic DNA. However, if genomic nonallelic repeated sequences interact among themselves as they do with exogenous sequences, homologous recombination could contribute to their concerted evolution with minimal risk of genome instability resulting from crossing-overs.