|Home | About | Journals | Submit | Contact Us | Français|
The retrovirus-like mobile genetic element of Saccharomyces cerevisiae, Ty1, transposes to new genomic locations via the element-encoded integrase (IN). Here we report that purified recombinant IN catalyzed correct integration of a linear DNA into a supercoiled target plasmid. Ty1 virus-like particles (VLPs) integrated donor DNA more efficiently than IN. VLP and IN-mediated insertions occurred at random sites in the target. Mg2+ was preferred over Mn2+ for correct integration, and neither cation enhanced nonspecific nuclease activity of IN. Products consistent with correct integration events were also obtained by Southern analysis. Recombinant IN and VLPs utilized many, but not all, linear donor fragments containing non-Ty1 ends, including a U3 mutation which has been shown to be defective for transposition in vivo. Together, our results suggest that IN is sufficient for Ty1 integration in vitro and IN interacts with exogenous donors less stringently than with endogenous elements.
Ty1 is a retrotransposon of Saccharomyces cerevisiae which is structurally and functionally similar to retroviruses (for review, see references 3 and 18). Transcription of a genomic element results in the formation of intracellular virus-like particles (VLPs) which contain the element-encoded catalytic enzymes required for the process of transposition: protease, reverse transcriptase, and integrase (IN) as well as Ty1 RNA and a cellular tRNAMet required to prime reverse transcription. The resulting full-length cDNA copy is correctly inserted into a new genomic location by IN-mediated strand transfer. Correct Ty1 integration events are characterized by a 5-bp target site duplication (TSD) without rearrangement or loss of sequence adjacent to the insertion site and by the coupled joining of both termini of the element to a single insertion site (concerted integration).
Both Ty1 VLPs and recombinant IN are active in a physical assay which monitors the insertion of a radioactively labeled long terminal repeat (LTR)-based oligoduplex into an identical target molecule (29). Although this assay demonstrates strand exchange activity of both recombinant IN and VLP-associated IN, it bears limited similarity to transposition in vivo. A transposition assay has been developed which detects the integration of supF-marked Ty1 elements into bacteriophage lambda and demonstrates that purified VLPs are sufficient for correct integration events (13). Gel electrophoresis and electron microscopy have been used to show VLP-mediated insertion of an exogenous linear donor molecule with Ty1 LTR-like termini into both linear and circular target molecules (4). A subset of products analyzed from these insertions resemble correct integration events. Devine and Boeke (11) have exploited the integration activity of Ty1 VLPs to insert a selectable artificial transposon into random sites of target DNA of unknown sequence as an aid to DNA mapping and sequencing.
Although correct integration has been demonstrated and characterized for retroviral INs, including human immunodeficiency virus 1 (HIV-1) (8, 9, 20, 22), Rous sarcoma virus (RSV) (28), avian sarcoma virus (22, 24), and avian myeloblastosis virus (AMV) (17, 27, 36, 37), the ability of an LTR-retrotransposon IN to carry out correct integration outside the context of the VLP has not been established. Using a genetic assay (11), we demonstrate that IN requires no additional VLP-associated proteins for correct integration of a donor substrate bearing Ty1 LTR sequences at each end into a supercoiled circular plasmid target. We have characterized this activity in regard to efficiency compared to VLPs, target site preference, divalent cation preference, and utilization of donor molecules containing non-Ty1 LTR ends. Our survey of mutated ends has shown that while not all sequences are acceptable IN substrates, a variety of mutant substrates undergo correct integration. In addition, Southern hybridization of the reaction reveals products consistent with correct integration events.
Ty1 VLPs were isolated from strain GRY458 using established methods (13, 19). The yeast system for the ectopic expression of Ty1 IN has been described previously (29). This study required several IN and VLP preparations, which led to some variability in concentration and activity. However, experiments involving quantitative comparisons were carried out with the same IN and VLP preparations.
An 864-bp fragment containing the dihydrofolate reductase gene, which confers resistance to the antibiotic trimethoprim (TMP), was used as a donor in the integration assay (11). Half XmnI sites (GAANN/NNTTC) on each end allowed the introduction of the terminal 5′-AC-3′ nucleotides of the Ty1 U3 LTR in the NN position followed by the two XmnI-specific Ts. The resulting sequence comprises a U3-like terminus at each end of the donor fragment. This donor fragment, which was purchased from PE Applied Biosystems (Foster City, Calif.), was cloned into the XmnI site of the plasmid pUC19. This construct was subsequently used as a template for PCR amplification of donor fragments. Typically, PCRs were carried out for 30 cycles using 2.8 U of Expand high-fidelity DNA polymerase (Boehringer Mannheim, Indianapolis, Ind.) and 5 ng of template for each 100-μl reaction mixture. Primers and deoxynucleoside triphosphates were removed from the reactions by Wizard PCR Prep (Promega, Madison, Wis.). The product was restricted with XmnI, which generated phosphorylated 5′ ends. The digested fragment was loaded onto a 1% low-melting-point agarose (Life Technologies, Rockville, Md.) gel and resolved at 6.5 V/cm2 for 15 h in the presence of 5 μg of ethidium bromide/ml. Excised DNA fragments were purified by QiaexII (Qiagen, Valencia, Calif.), and the concentrations were determined spectrophotometrically. In certain experiments, amplification using phosphorylated primers allowed inclusion of additional LTR or mutated sequences and eliminated the requirement for XmnI digestion and gel purification (see Table Table22 for primer combinations). To avoid untemplated nucleotides at the termini of donor molecules made by the phosphorylated primer method, Vent DNA polymerase (New England Biolabs, Beverly, Mass.) was used in the PCR.
The target plasmid, pMC1871, consisted of a pBR322 molecule with the lacZ gene and a linker of 38 bp inserted into the β-lactamase gene in reverse orientation at a PstI restriction site (position 3607). This produced a 7,510-bp target plasmid in which the only functional antibiotic resistance gene conferred tetracycline resistance. This plasmid was purified by cesium trifluoroacetate (Amersham-Pharmacia, Piscataway, N.J.) gradient centrifugation.
The standard 20-μl reaction mixture contained 10 mM Tris-HCl (pH 7.5), 10 mM MgCl2, 1 mM dithiothreitol, 5% polyethylene glycol 8000, 200 ng of donor fragment (0.35 pM), 1 μg of target DNA (0.2 pM), and either Ty1 IN or Ty1 VLPs. Reaction mixtures were incubated at 30°C for 1 h, after which 5 μl of stop mix (0.25 M EDTA, 1% sodium dodecyl sulfate, 5 μg of proteinase K/ml [EM Science, Gibbstown, N.J.]) was added. The reaction mixtures were then incubated at 65°C for 30 min. Ammonium acetate was added to a final concentration of 0.33 M, and the DNA was precipitated by the addition of 2.5 volumes of cold ethanol. The DNA pellet was washed once in 70% ethanol, dried, and resuspended in 20 μl of H2O. An aliquot (3 μl for the experiment shown in Fig. Fig.1,1, 6 μl for all other experiments) was introduced into 40 μl of HB101 cells at an approximate density of 2 × 1010 cells/ml by electroporation (1.8 kV, 200 Ω resistance, 25 μF capacitance) with electroporation cuvettes having a gap size of 1 mm (BTX, San Diego, Calif.). After incubation at 37°C for 1 h, cells were diluted appropriately and plated on L agar plates containing either 100 μg of TMP/ml and 15 μg of tetracycline (TET) or TET only. Typically, three dilutions of the cell suspension were each plated in triplicate on TMP-plus-TET plates to select colonies containing integrant plasmids. Two dilutions were each plated in triplicate on TET-only plates to determine electroporation efficiency. After overnight incubation at 37°C, colonies were counted and integration efficiency was calculated as the number of TMP-and TET-resistant colonies divided by the number of TET-only-resistant colonies.
ABI Prism (PE Applied Biosystems) sequencing reactions were carried out according to the manufacturer's instructions.
Reactions for Southern analysis were carried out as described above except that the volumes were doubled and contained 460 ng of donor molecule (0.8 pM) and 5 μg of pUC19 as the target plasmid (2.8 pM). Following the 1-h incubation period, the reaction was stopped by the introduction of EDTA, to a final concentration of 63 mM, and 1 μg of proteinase K (EM Science) per ml in 10 mM Tris-HCl (pH 8.0) with 1 mM EDTA (TE) and incubated at 37°C for 30 min. The DNA was extracted with phenol-chloroform-isoamyl alcohol (25:24:1) and ethanol precipitated. Following precipitation, the pellet was air dried and resuspended in 30 μl of TE. Three microliters of the resuspended DNA was digested with either StyI or PacI (New England Biolabs). Undigested controls were incubated in restriction enzyme buffer in the absence of enzyme. Following incubation at 37°C for 1.5 h, DNA molecules were resolved electrophoretically on a 1% agarose gel at 0.16 V/cm2 for 14 h. The donor fragment was labeled with [α-32P]dCTP by Megaprime (Amersham-Pharmacia) and used as a probe in Southern hybridizations. Phosphorimaging was performed using a Storm860 and ImageQuant software, version 1.1 (Molecular Dynamics, Sunnyvale, Calif.).
Integration experiments were performed to compare recombinant IN and VLPs in the genetic assay with respect to integration efficiency and target site preference (Fig. (Fig.1).1). To estimate the efficiency of recombinant IN and VLPs (Fig. (Fig.1A),1A), equal volumes of protein were added to the reaction mixture (x axis a). Although the total protein in the recombinant IN extract (x axis b) was less than that in the VLP extract (x axis c), the estimated IN concentration in VLPs (x axis d) was similar to that of recombinant IN. This estimate was based on an immunoblot comparison of samples of recombinant IN of a known concentration to VLPs containing an unknown concentration of IN. ImageQuant analysis indicated that about 1/16 of total VLP protein consisted of IN.
This integration experiment showed VLPs to be approximately fivefold more active than recombinant IN. One possible explanation for this difference is that the nucleocapsid component of the VLP, which is encoded by TYA1, might confer conformational stability to IN. If so, it is possible that the efficiency of Ty1 IN could be enhanced by adding TYA1 to the reaction. To test this hypothesis, a mutant TYB1Δ plasmid, which lacks the genes encoding protease, reverse transcriptase, and IN, was overexpressed and VLP particles consisting of TYA1 only were purified. The addition of up to 3 μg of TYA1 particles to the integration reaction mixture increased the integration efficiency by approximately twofold (Moore and Garfinkel, unpublished results). Since overexpressed TYA1 forms particles which can be pelleted by ultracentrifugation (6), we also determined if IN binds to TYA1 particles in vitro. Although IN copelleted with TYA1 particles, this association was apparently nonspecific, since the particles also copelleted with bovine serum albumin, HIV-1 IN, and Rac1, a human GTPase (Moore and Garfinkel, unpublished results).
To determine if TMP- plus TET-resistant colonies arose from correct IN-catalyzed insertion of donor molecules into plasmid targets, representative colonies were clonally purified and sequenced outward from both ends of the donor. This sequence showed the donor-target junction and 5-bp TSD as well as the insertion site on the target. Plasmids which showed the donor molecule to be flanked by a perfect TSD and to have both ends joined at the same location in the target (concerted integration) were classified as correct integration events. Of the 61 IN-derived plasmids, 51 (84%) were correct integration events. Fifty-seven of the 66 (86%) VLP-derived plasmids were correct events. Thus, both INs showed the same proportion of correct integration events. Incorrect integration events consisted primarily of imperfectly matched TSDs and/or nonconcerted integration events in which the two ends of the donor were joined at different sites in the target (Table (Table1).1).
Sequencing data were also used to determine target sites for both VLP and recombinant IN (Fig. (Fig.1B).1B). The insertions appeared randomly distributed over the nonessential region of the plasmid, and no hot spots were evident.
To determine if preincubation of donor and IN or VLPs at 0°C enhanced integration efficiency, reaction mixtures were assembled without the target plasmid and incubated on ice before adding the target plasmid and shifting the reaction temperature to 30°C. Preincubation for up to 5 h showed no increase in TMP- plus TET-resistant colonies for either IN or VLPs (Moore and Garfinkel, unpublished results).
Using an electrophoretic analysis method for concerted integration, Braiterman and Boeke (4) have demonstrated that Ty1 VLPs are more active in the presence of Mg2+ than in the presence of Mn2+. We quantitated the integration efficiency of recombinant IN in the presence of Mg2+ or Mn2+ (Fig. (Fig.2A)2A) and found that the overall integration efficiencies were similar. However, sequence analysis of 39 integrant plasmids from Mg2+ colonies and 38 integrant plasmids from Mn2+ colonies revealed that in this experiment 74% of the Mg2+-derived products were correct integration events while only 45% of the Mn2+-derived products were correct. The most frequent class of aberrant integrations observed with either cation was integration of each end at separate sites in the target.
The observation that Mn2+ catalyzes a nonspecific nuclease activity of HIV-1 IN (15) led us to examine whether Ty1 IN has a similar activity. Such an activity could influence incorrect target site selection by providing random precleaved insertion sites. Additionally, adjacent nicked sites could result in a linearized target. Ty1 VLPs have been demonstrated to produce a significant proportion of integration events in which the donor is inserted near the ends of a linearized target (4). Events of this type might also lead to a higher proportion of incorrect integration events in reactions in the presence of Mn2+ rather than Mg2+. To determine whether Ty1 IN exhibited a more pronounced nuclease activity in the presence of Mn2+ compared to Mg2+, supercoiled pUC19 was incubated in the presence or absence of 1.2 μg of IN in buffer containing either Mg2+, Mn2+, or no cation under standard reaction conditions to monitor conversion of supercoiled circular DNA (RFI) to nicked circles (RFII) or linear forms. The results showed no differences in IN-mediated nuclease activity with either cation (Fig. (Fig.22B).
To confirm that integrants arose from an in vitro IN-mediated reaction rather than a bacterial cell recombination or DNA repair mechanism, integration products were visualized by Southern analysis using pUC19 as the plasmid target and the donor fragment as a hybridization probe (Fig. (Fig.3).3). After terminating the reaction, aliquots were digested with either StyI or PacI, which recognizes a single site at bp 267 or 132, respectively, in the donor molecule but no sites in the target plasmid. An additional aliquot was incubated in restriction enzyme buffer in the absence of enzyme.
In vitro integration resulted in two types of concerted products in addition to the nonconcerted events that were detected in this assay (Fig. (Fig.3A).3A). One type of concerted product resulted from a single donor molecule joined at each end to a unique site in the target (unimolecular concerted insertion), resulting in an RFII circle. Another type of concerted integration product which occurred in vitro arose from the coordinated integration of one end of each of two donor molecules at the same site in the target. This event resulted in linearization of the target molecule, with a donor molecule at each end (bimolecular concerted integration) (37). If all the donor ends were equally likely to be integrated, four orientations of donor molecules could occur. Digestion of an integrant population containing all four orientations with either StyI or PacI resulted in three products, since two of the orientations were indistinguishable from each other.
Reactions with Ty1 IN showed two major products in the undigested samples (Fig. (Fig.3B).3B). The slower migrating product was consistent with an RFII circle. This product was confirmed as a single donor insertion into a target by using a marker consisting of a partial restriction digest of pUC19 into which the donor fragment had been cloned (Moore and Garfinkel, unpublished results). The faster migrating product was consistent with a linearized, bimolecular concerted integration product whose predicted size is 4,414 bp. Although both products were digested with either restriction enzyme, only the bimolecular product gave rise to the three characteristic smaller fragments shown in Fig. Fig.3A.3A. The unimolecular insertion was reduced to a linear fragment of 3,550 bp. The apparent amount of these products increased with increasing IN concentrations. Products which migrated more slowly than the RFII likely represent other species of integrants such as multiple donor insertions or unimolecular half-site events which we have not yet characterized.
Using the oligoduplex assay for IN activity, we have previously characterized the ability of molecules with mutated ends to serve as substrates for Ty1 IN (30). Although an in vitro concerted integration reaction is not identical to an in vivo transposition event, it does resemble in vivo integration more closely than the oligoduplex assay. Wild-type (WT) U3 and mutated donors (Table (Table2)2) were used in the genetic selection assay with both IN and VLPs. Because the U5 sequence has been shown by a physical assay to be inhibitory for VLP-catalyzed integration (5), a donor molecule containing a 4-bp WT U3 end and a 4-bp WT U5 end was tested to determine if the U5 sequence affected integration efficiency. Vora et al. (35) have reported that the fifth position of the U5 RSV LTR inhibits activity of AMV IN and RSV IN, and that when this position is mutated to become a U3-like end, IN activities increase. To determine if Ty1 LTR sequences beyond the 4-bp termini altered donor utilization, an 8-bp U3/U5 donor was tested. The TG→CC U3 substitution is of particular importance because this mutation prevents in vivo transposition of an overexpressed Ty1 element (32). The terminal sequences involving rearrangements such as TG→GT reversal, TG→GG U3/GT→CC U3, T→A U3/T→A U3 strand flip, and all A-T ends were compared to the results obtained with these termini in the oligoduplex assay (30), and one substrate with nonphosphorylated ends was tested as a comparison to a similar substrate used previously with VLPs (14).
The results of these experiments indicated that most mutant ends were recognized to varying degrees by both recombinant IN and VLPs (Table (Table3).3). Replacement of one of the 4-bp U3 termini with a 4-bp U5 sequence did not reduce integration efficiency compared to the U3/U3 4-bp WT donor. Inclusion of an additional 4 bp of LTR sequence at each end (U3/U5 8-bp WT) also did not diminish integration efficiency, which would be expected if additional U5 sequences were inhibitory.
The TG→CC U3 mutation has been tested in an in vivo transposition assay (32). This mutant is defective for transposition, although it does undergo cDNA recombination with homologous targets. In the present study, this mutation coupled with a WT U5 end showed no reduction in integration efficiency compared to the U3/U3 4-bp WT donor or to the U3/U5 4-bp WT donor.
The TG→GT end surprisingly demonstrated a deficiency of 120-fold for IN and 374-fold for VLPs. Few integration events were recovered and most arose from aberrant insertions. In a previous study using an oligoduplex assay, we reported that molecules with G-C ends are poor substrates and that this mutation yields no product (30). Another donor having a G-C end, the TG→GG U3/GT→CC U3 mutation, which also fails to give rise to strand transfer products in the oligoduplex assay, resulted in only a ninefold reduction in integration efficiency for IN and a fourfold reduction for VLPs.
Because A-T ends are utilized in the oligoduplex assay (30), two donors were tested in which either the terminal sequences were flipped (T→A U3/T→A U3) or internal sequences were modified to create all A-T ends. These substrates showed reduced utilization compared to WT both for efficiency and accuracy of correct integration. The T→A U3/T→A U3 donor was reduced 1.7-fold for IN and 3-fold for VLP, while the more severe mutant comprised of four A-T pairs showed a 16-fold reduction for IN and a 26-fold reduction for VLPs.
Previous studies indicate that nonphosphorylated substrates yield WT levels of integration by VLPs (14). Our results with the U3/U3 WT nonphosphorylated substrate also indicated that 5′ phosphorylation is not necessary for IN-mediated integration.
Products generated by integration of selected mutant donor molecules were also examined by Southern analysis using a 32P-labeled donor as previously described (Fig. (Fig.4).4). Although donors carrying mutant ends gave rise to the same types of products as the WT substrate, the TG→GT U3 mutant was lower overall in amount of product generated and was extremely deficient in the RFII product. This result, combined with the low integration efficiency, suggests that the RFII product is primarily, if not exclusively, responsible for TMP- plus TET–resistant colonies recovered in Escherichia coli. The U3/U3 4-bp WT donor showed the same pattern as observed previously (Fig. (Fig.3B).3B). Interestingly, the TG→CC U3/U5 WT substrate, which is defective for in vivo transposition, showed product formation similar to that observed with the U3/U3 WT substrate. The U3/U5 4-bp WT substrate, while exhibiting both the RFII and linearized products in the undigested reaction, showed an altered digestion pattern from the U3/U3 WT donor. Although the StyI 3,880-bp fragment and the PacI 4,150-bp fragment were diminished, the 3,550-bp fragment of each digest was present as well as the smaller fragments—StyI, 3,220 bp and PacI, 2,950 bp. This result suggests that the U3 end was preferentially utilized for bimolecular integration.
Our results demonstrate that purified recombinant Ty1 IN can integrate an exogenously added linear donor into a supercoiled target plasmid correctly. Most integration events are concerted and contain a 5-bp TSD. We have used both a genetic selection and a physical assay to characterize integration products. The genetic assay allows recovery of integrant plasmids for further analysis while the physical assay provides a direct visualization of products without the intervening step of introducing the products into bacterial cells.
The genetic assay has been used to compare IN and VLPs with respect to quantitative efficiency, frequency of correct integration events, and target site selection. A comparison of similar concentrations of recombinant IN and VLPs shows VLPs to be fivefold more active than recombinant IN in the genetic assay (Fig. (Fig.1A).1A). The interpretation of this result, however, is not straightforward since it compares activities of Ty1 INs which are in different microenvironments and which have been purified by different methods. Improvements in protein purification methods may eventually reduce this difference in activity. The observation that addition of TYA1 to recombinant IN enhances integration activity by twofold suggests that TYA1 plays a role in IN stability but is not entirely competent to perform this function when added in trans. This is not surprising considering that IN within the VLP is associated with TYA1 during protein maturation, whereas recombinant IN is expressed outside the context of the particle. Additionally, we cannot rule out an as-yet-unidentified accessory factor which may copurify with VLPs. Sequencing analysis of plasmids recovered from antibiotic-resistant colonies shows about the same percentage of correct integrations regardless of whether recombinant IN or VLP-IN is used. The sequencing analysis (Fig. (Fig.1B)1B) also shows that highly preferred sites for either IN are absent when purified DNA is used as a target. However, Ty1 transposition in vivo shows target site selection near genes transcribed by RNA polymerase III (12). A naked DNA target which lacks chromatin structure as well as accessory proteins associated with RNA polymerase III transcription might not be expected to exhibit target site preferences.
We have characterized the divalent cation preference of Ty1 IN (Fig. (Fig.2).2). The presence of a divalent cation, Mn2+ or Mg2+, is required for catalytic activity of INs (15, 24, 25). Additionally, divalent cations have been shown to influence structural conformation of IN (2, 39). Although initial characterization of retroviral INs using the oligoduplex assay have been reported to require Mn2+ for optimal activity (10, 15, 24), Engelman and Craigie (15) have shown that HIV-1 IN can utilize either cation under varied reaction conditions. Vora et al. (37) and Fitzgerald et al. (17) characterized the cation requirement of AMV IN and demonstrated that, although Mn2+ efficiently promotes donor-to-donor insertions and unimolecular half-site insertions into a circular target, Mg2+ is more efficient for bimolecular concerted integration into a circular target.
Here, we report that correct concerted integrations catalyzed by Ty1 IN are more efficient with Mg2+ than with Mn2+. Although similar numbers of TMP- plus TET-resistant colonies result when either cation is present in the reaction mixture (Fig. (Fig.2A),2A), sequencing analysis reveals that only 45% of the Mn2+-derived integrants are correct integrations, compared with 74% when Mg2+ is present. Engelman and Craigie (15) have found that HIV-1 IN displays more nonspecific nuclease activity in the presence of Mn2+ than with Mg2+. Such an activity might play a role in nonconcerted integration by creating precleaved insertion sites in the target or by linearizing the target plasmid. VLP-catalyzed integration yields a significant class of products in which the donor is integrated near the end of a linear target (4). We have tested recombinant Ty1 IN for an alteration in nuclease activity in the presence of both cations under our standard reaction conditions and have observed no profound difference in the conversion of RFI to RFII circles with either cation (Fig. (Fig.2B).2B). Consequently, the increased proportion of nonconcerted Mn2+-promoted events does not appear to be due to nuclease activity on the target plasmid. However, we cannot rule out that Mn2+ interacts with the target in a different way or that Mn2+ interacts with other components of the reaction, permitting less stringent integration.
To visualize IN-catalyzed products directly, we used Southern analysis with a radiolabeled donor fragment (Fig. (Fig.3).3). This analysis revealed two predominant products. The faster migrating product is consistent with a 4,414-bp linear fragment resulting from a bimolecular integration (Fig. (Fig.3A).3A). The slower migrating product is consistent with an RFII circle containing one donor fragment per target plasmid, as indicated by an RFII electrophoretic marker. A comparison of the TG→GT U3/TG→GT U3 donor with the WT U3/U3 donor by using genetic and Southern analyses suggests that this product represents primarily bimolecular concerted events. This mutated donor shows a severe deficiency in both TMP-resistant colony formation and in RFII product in the Southern analysis. Other mutated donors which are used proficiently in the genetic assay exhibit near-WT amounts of the RFII product. This result suggests that the RFII product is primarily responsible for colony formation in the genetic assay. Although half-site insertions also result in RFII products, these events seem unlikely to give rise to colonies which resemble IN-mediated insertions in the genetic assay.
The possibility also exists that colonies arise from electroporation of the bimolecular concerted linearized product. Since the donor fragments initially insert in a concerted manner, sequence analysis of this fragment might still reveal a target site duplication. However, since either end of each donor can insert into the target, the sequencing analysis should have revealed some plasmids which lack a donor-target junction. If the linear molecules recircularized in the bacterial cells, sequencing would have revealed donor-donor junctions rather than donor-target junctions. No aberrant sequences of this type were detected. Even though we used a recA strain of E. coli as the bacterial host in the genetic assay, we cannot rule out the possibility that other recombination pathways convert bimolecular linearized products into a form resembling RFIIs. However, the results shown in Fig. Fig.44 compared with the results of the genetic assay suggest that this is not a common event. We base this conclusion particularly on the relative amount of bimolecular linear product observed with the TG→GT U3/TG→GT U3 substrate compared to the U3/U3 4-bp substrate. Although the amounts of linear product are similar, the results of the genetic assay show the TG→GT U3/TG→GT U3 substrate to be reduced by greater than 2 orders of magnitude compared to the U3/U3 4-bp substrate. If a significant number of recombination events yielded colonies in the genetic assay, we would have expected to observe more quantitative similarity between these two substrates.
In a similar assay using exogenous donor and Ty1 VLPs, Braiterman and Boeke (5) have evaluated several terminal and subterminal donor mutations. Their results show that VLP-mediated integration is tolerant of a wide range of mutations. In addition, their results suggest that the U3 terminus is preferred and that the U5 terminus is inhibitory. Although our results for the genetic assay show no reduction in integration efficiency when one of the U3 termini is replaced by 4 bp of U5 sequence, Southern analysis indicates that the U3 end is preferred in bimolecular concerted events. U3 is also the preferred end for AMV (16, 21, 36, 37), whereas U5 is the preferred end for HIV-1 (7, 20, 26, 33), human foamy virus (31), and feline immunodeficiency virus (34). Fitzgerald et al. (17) have suggested that the least effective terminus is that which is also involved in another element function and therefore is constrained from evolving into the most effective substrate for integration. The demonstrated U3-over-U5 preference of Ty1 is consistent with this hypothesis since the U5 terminus is also part of the TYA1 coding sequence.
Vora et al. (35) have shown that the fifth position of the RSV U5 LTR is responsible for a three- to fivefold preference of U3 ends over U5 ends by RSV IN and AMV IN. Since our U3 WT/U5 4-bp WT donor showed no reduction in utilization, we also analyzed a U3 WT/U5 WT donor containing 8 bp of each LTR. Additional LTR sequences did not result in further impairment of U3 WT/U5 WT donor utilization by either IN or VLP. Rather, the 8-bp LTR donor showed a two- to threefold enhancement in integration efficiency compared to the 4-bp donor. These differences in subterminal donor preferences may be attributed to two differences between the RSV LTR and the Ty1 LTR. The fifth position in the RSV LTR is the first nonidentical U3/U5 nucleotide, whereas the third position of the Ty1 LTR is the first nonidentical position. Additionally, the RSV LTR undergoes IN-mediated 3′ dinucleotide cleavage prior to integration (23), whereas Ty1 LTRs do not (30). Consequently, the critical fifth position of the RSV LTR may be analogous to the third position of the Ty1 LTR. If so, then the important nonidentical nucleotide is included in the 4-bp donor, and enhanced activity of the 8-bp donor may be due to additional subterminal LTR sequences. Braiterman and Boeke (5) have shown that subterminal mutations have a profound effect on VLP-mediated integration. Wilhelm et al. (38) reported that a subterminal mutation in the U3 LTR that changes the WT TGG at positions 4 to 6 to CAT abolishes Ty1 transposition in vivo.
Several of the termini examined in this study have been evaluated in an oligoduplex assay (30) which showed that Ty1 IN does not utilize oligoduplexes having G or C termini. The data presented here are broadly consistent with those results. However, the difference between G-C termini and A-T termini is not as distinct in the genetic assay as in the oligoduplex assay. Mutations that exhibit more severe effects in the oligoduplex assay than in a concerted integration reaction or in vivo integration have also been reported for retroviruses (1, 35).
Our results indicate that recombinant Ty1 IN is necessary and sufficient for catalyzing correct integration in vitro. Both recombinant IN and VLPs show a wide range of tolerance for non-LTR termini of an exogenously added donor, including a terminal mutation which has been shown to be ineffective for in vivo transposition (32). This result suggests that mechanisms which define specificity for correct ends for transposition may act differently with exogenous donors than with the endogenous element. Experiments which utilize the endogenous element in a two-ended integration assay may indicate whether IN is only promiscuous with exogenous donor or whether an element or host-encoded specificity factor is required to fully reconstitute donor preference in vitro.
We thank Lori Rinckel for Rac1, Patrick Clark for HIV-1 IN, Michael Hall for plasmid pMC1871, Duane Grandgenett for helpful discussions, Ellen Frazier for preparation of figures, and Dwight Nissley for critically reading the manuscript.