Construction of Gene Targeting Vectors from λ KOS Genomic Libraries

We describe a highly redundant murine genomic library in a new λ phage, λ knockout shuttle (λKOS) that facilitates the very rapid construction of replacement-type gene targeting vectors. The library consists of 94 individually amplified subpools, each containing an average of 40 000 independent genomic clones. The subpools are arrayed into a 96-well format that allows a PCR-based efficient recovery of independent genomic clones. The λKOS vector backbone permits the CRE-mediated conversion into high-copy number pKOS plasmids, wherein the genomic inserts are automatically flanked by negative-selection cassettes. The λKOS vector system exploits the yeast homologous recombination machinery to simplify the construction of replacement-type gene targeting vectors independent of restriction sites within the genomic insert. We outline procedures that allow the generation of simple and more sophisticated conditional gene targeting vectors within 3–4 weeks, beginning with the screening of the λKOS genomic library. INTRODUCTION Homologous recombination between exogenous DNA and chromosomal sequences in embryonic stem (ES) cells is a widely used and powerful approach to genetically manipulate the mouse germ line (6,12). Exogenous DNA transfected into ES cells can have a variety of different fates ranging from the intended homologous recombination (HR) event to random integration or simply no integration at all (2,5). Because the latter two outcomes usually predominate, several measures to improve and enrich HR have been devised. The use of isogenic DNA in the targeting vector, which is derived from the same strain as the target cell (17), and increasing the length of homology generally improves HR frequencies (8). Several enrichment strategies for HR have been designed, and the most generally applicable is positive/negative selection (11). In using positive/negative selection, one positively selects for cells that stably integrate the positive-selection marker of the targeting vector and selects against those cells in which a non-HR event results in the stable integration of a negative-selection cassette (11). Most gene targeting projects aim to ablate gene function. However, it has become apparent that the same gene can serve important functions in different physiological systems and in distinct developmental processes. A sophisticated strategy has been devised to gain tissueor development-specific information by selectively modifying gene function. The creation of a conditional mutation is a powerful approach that allows one to preselect certain cell types and developmental time points to mutate a gene (19). In conditional mutants, loxP or FRT sites, which flank one or several exons of a gene, are introduced into the murine germ line by HR. The genomically altered mice are usually phenotypically normal. These mice are subsequently crossed with mice that express CRE or FLIP recombinase in a temporal or tissue-specific manner. The construction of a typical conditional replacement-type targeting vector is a cumbersome task and requires multiple cloning steps in which several selectable markers, three genomic fragments and several recombination sites (for example, two FRT and two loxP sites) must be assembled in a single construct. We have recently described two murine genomic libraries, λKO and λPS, with large insert size (14) that were successfully used to create mutations in ES cells (13,14). Essentially, the λKO/λPS system reduced the number of cloning steps to a single insertion of a selection cassette when creating a replacement-type targeting vector. The λKO vector’s large regions of homology that are flanked with negative-selection cassettes increase the frequency of successfully isolating targeted ES cell lines. The λPS vector is used for mapResearch Report 1150 BioTechniques Vol. 26, No. 6 (1999) Construction of Gene Targeting Vectors from λKOS Genomic Libraries BioTechniques 26:1150-1160 (June 1999) ping the locus and the generation of probes to verify targeted recombination in the ES cell. A particular useful feature of the λKOor λPS-based libraries is the presence of two loxP sites that flank the genomic insert and facilitate plasmid creation by simple viral infection of CRE recombinase expressing bacteria (14). The major disadvantage of the λKO system was the inherent difficulty of finding unique restriction sites at ideal positions in the pKO plasmids at which to place the positive-selection cassette. This task becomes even more problematic for vectors designed to introduce conditional mutations that require multiple cloning steps. Yeast cells exchange genetic material preferentially by homologous recombination, whereby short stretches of homology are sufficient for the highly accurate recombination reaction (10). This property of yeast cells was efficiently exploited in the targeted disruption of yeast genes (10) and the subcloning of DNA fragments from yeast artificial chromosomes (3). Recently, this system was utilized to generate targeting vectors by recombination of a yeast-selection cassette containing cloned ends of homology into the respective genomic plasmid vector (16). Here, we present a system that combines the advantages of both the λKO vector and the efficiency of the yeast homologous recombination machinery. Based on a new λ phage vector, λ knockout shuttle (λKOS), we created a highly redundant murine genomic library to facilitate the very rapid construction of replacement-type gene targeting vectors. We outline procedures for the generation of λKOS-based, replacement-type targeting vectors for simple and conditional mutations in a highly efficient and timesaving manner. MATERIALS AND METHODS Construction of lKOS λKO DNA (14) was concatamerized and ligated at the cos end (15). The concatamers were subsequently digested with NotI and ligated with EagIdigested products encompassing the yeast TRP1 and 2micron genetic elements generated by a primary polymerase chain reaction (PCR) with primers 5′-ccatcggccggcgaacgaagcatctgtgcttc-3′ and 5′-ccatcggccgcggtattttctccttacgcat-3′ from pYAC4 template DNA (4) and a secondary PCR with primers 5′-gcggaccaccatcggccggcgaacgaa-3′ and 5′-gcggaccaccatcggccgcgggtattttc-3′ using the primary TRP1/2micron product as template DNA. The ligation reaction was packaged in vitro using Gigapack Plus Packaging Extract (Stratagene, La Jolla, CA, USA) and allowed to infect Escherichia coli C600 cells. The TRP1/2micron product was used as a radiolabeled probe in filter hybridization to identify recombinant λKOS clones. Single λKOS plaques were recovered and diluted in 500 μL λ phage storage medium (SM) supplemented with 20 μL chloroform (15). CRE-recombinase-mediated excision of the pKOS plasmid from λKOS was performed in vivo by λKOS infection of E. coli strain BNN132 (9). Restriction mapping and partial sequencing verified the expected structure of the pKOS plasmid (Figure 1C). To test the yeast/E. coli shuttle capability of pKOS, competent yeast cells, auxotrophic for tryptophan and uracil, were prepared (see below), transfected with pKOS plasmid DNA and plated on tryptophandeficient yeast minimal media plates (BIO 101, Vista, CA, USA). Colonies appeared only on plates with pKOStransformed yeast but not on plates from mock transformation. Plasmid DNA was prepared as described below and shuttled into E. coli strain TG-1 by electroporation. Bacterial plasmid preparations of pKOS before and after yeast transformation were compared by restriction mapping, and no rearrangements were observed (data not shown). Preparation of a Murine Genomic Library in λKOS λKOS phage DNA was prepared from liquid lysate (15) using E. coli strain C600 and concatamerized at their cos ends (15). The concatamers were then digested with SalI, dephosphorylated with calf intestinal phosphatase (CIP) and further digested with BamHI. After fractionization on sucrose gradients (15) to remove polylinker sequences and the stuffer fragment, the fractions containing the phage arms were identified, pooled and precipitated by the addition of 0.7 volumes of isopropanol. After several washes in 70% ethanol, the phage arms were resuspended in TE buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA) at a final concentration of 1 mg/mL. Highmolecular-weight DNA isolated from brain tissue of a male mouse of the LEX-1 129/SvEv substrain was partially digested with Sau3AI (15) and size-selected by sucrose gradient centrifugation (15). Fragments in the range of 8.5–11.5 kbp were pooled, precipitated, washed and resuspended in TE as described above. Phage arms and genomic insert DNA were ligated (15), packaged in vitro using Gigapack Gold III Packaging Extract (Stratagene) and allowed to infect E. coli strain C600 cells. A total of 94 plates (150-mm diameter) were prepared, and 35 000– 45 000 recombinant λKOS phages were plated on each plate. The phage from each plate were amplified as individual subpools by scraping the top agar containing the phage particles into 35 mL of SM and 2 mL chloroform to provide a permanent source of each subpool library. Because of the outlined stringent biochemical purification scheme used for phage arm preparation, control packaging reactions containing phage arms that were ligated in the absence of genomic insert DNA produced titers below 0.1%. One-microliter aliquots of each amplified phage subpool were transferred into 96-well plates, air-dried at room temperature, sealed and stored at -20°C. These plates serve to identify subpools containing λKOS phages with the desired genomic inserts by PCR. A panel of genomic PCR primers were tested. A total of 100 000–120 000 plaque forming units (pfus) from the subpools that gave a positive PCR signal were plated on four 150-mm plates using E. coli C600 cells and screened by filter hybridization (15). Preparation of Frozen Competent Yeast Cells A single colony of yeast strain HMS-1 (Matα, ura3-52, ∆his3-200, lys2-801, ∆trp1 and leu2) was grown overnight in 10 mL complete media (BIO 101) in a rotatory shaker at 30°C Vol. 26, No. 6 (1999) BioTechniques 1151 to saturation. One hundred microliters of this overnight culture were diluted in 300 mL complete media and grown to an OD (at 600 nm wavelength) of 0.45 under the same conditions. Cells were harvested by centrifugation at 4000× g for 5 min at room temperature, the cell pellet was gently resuspended in 10 mL of sterile distilled (d)H2O and the centrifugation was repeated. The cell pellet was resuspended in 1.5 mL of freezing solution [1× vol lithium acetate stock (1 M lithium acetate adjusted with dilute acetic acid to pH 7.5), 1× vol 10× TE, 4× vol 50% glycerol and 4× vol dH2O] and incubated with gentle agitation at 30°C for 1 h. Competent cells were slowly frozen (wrapped in several layers of paper) in 220-μL aliquots at -80°C. Freezing reduced the transformation efficiency by about 50%, with the final transformation efficiency of 6–7 × 104/μg pKOS vector. Transformation of Competent Yeast and Bacterial Cells After thawing, 200 μL of the competent yeast cells were incubated with a 20-μL DNA mixture containing 250 μg carrier DNA (BIO 101 ), up to 5 μg of the pKOS vector and up to 3 μg of the URA3/chloramphenicol acetyltransferase (CAT) product for 30 min at 30°C with gentle agitation. Freshly prepared polyethylene glycol (PEG) solution (8× vol 50% PEG 4000, 1× vol 10× TE buffer and 1× vol lithium acetate stock, as above) (1.2 mL) was added and gently agitated for 30 min. After a heat shock for 15 min at 42°C without agitation, cells were collected by centrifugation for 5 min at 3000× g at room temperature. The cell pellet was resuspended in 500 μL 1× TE buffer, centrifugated again and finally resuspended in 200 μL TE buffer before plating on appropriate (-TRP/URA) selection plates (BIO 101). After incubation at 30°C for 2–3 days, individual colonies were streaked on a single (-TRP/-URA) selection plate and again incubated at 30°C overnight. Plasmid DNA was isolated using a published glass bead protocol for yeast cells (1). Electrocompetent DH10 bacteria were electroporated in the presence of 1 μL of the yeast plasmid preparation using standard procedures (15). Transformed bacteria were plated on Luria broth plates supplemented with 34 μg/mL chloramphenicol and 100 μg/mL ampicillin and incubated at 30°C overnight to further enrich for recombinants. Due to the size of the pKOS plasmids, all bacterial work was carried out at 30°C, which effectively inhibited unwanted recombination of pKOS plasmids. Plasmid DNA was prepared from liquid cultures of individual colonies using an alkaline lysis and PEG purification protocol (15). Construction of Replacement-Type Targeting Vectors Plasmid pURA3/CAT was created by first cloning a PCR product of the CAT gene generated from plasmid pBeloBAC11 (Research Genetics, Huntsville, AL, USA) with primers 5′cgaactgcagtgagacgttgatcggcacg-3′ and 5′-cggggtaccggcctcgctggccggcgcgccatttaaataactgccttaaaaaaattacg-3′. The product was digested with PstI and KpnI and ligated into the respective sites of pBSSKII (Stratagene) to generate pBSSKII/CAT. Next, the URA3 gene was amplified by PCR with primers 5′-cgaagccggcatcattacgcccgagtaataac-3′ and 5′-ctgctctagaggccatagcggccggatcctcgaggcgcgccatttaaatccttggcagaacatatcc-3′ from pYAC4 (4), cut with XbaI and NaeI and inserted into XbaI and SmaI of pBSSKII/CAT to generate pURA3/CAT as shown in Figure 2A. The integrity of the final clone was verified by sequencing of the URA3/CAT insert. The URA3/CAT insert, including the flanking multiple cloning site (MCS) as indicated in Figure 2A, was amplified with primer 1: 5′(25–40)ccgctctagaggccatagc-3′ and primer 2: 5′-(25–40)ctatagggcgaattgggtac-3′ by PCR using the following conditions: 2 cycles of 94°C for 45 s, 75°C for 1 min and 72°C for 2.5 min; 2 cycles of 94°C for 45 s, 70°C for 1 min and 72°C for 2.5 min and 30 cycles of 94°C for 45 s, 65°C for 1 min and 72°C for 2.5 min. The 5′ appended sequences of primers 1 and 2 correspond to 25–40 bp of homology to the 5′ and 3′ end of the desired insertion of the URA3/CAT cassette into a genomic pKOS vector insert, respectively, by yeast-mediated homologous recombination. For constructing simple replacement-type gene targeting vectors, the URA3/CAT cassette was replaced by the desired ES cell-selection cassette using the rare-cutter restriction sites flanking the URA3/CAT cassette by using standard cloning procedures (15). Similar cloning steps were required for the generation of more complex targeting vectors for conditional mutations as detailed in Figure 3 and in the Results and Discussion section. RESULTS AND DISCUSSION λKOS Vector and Genomic Library Construction Previously, we have shown that it is possible to build representative genomic libraries of partially constructed replacement-type targeting vectors to apply positive/negative-selection strategies for homologous recombination in ES cells (14). The insertion of the positive-selection marker into, on average, 15 kbp of genomic insert DNA, which were already flanked by negative-selection cassettes, by standard molecular techniques often became a cumbersome, if not impossible, task. To overcome this limitation, we constructed the λKOS vector that exploits the yeast homologous recombination machinery to place genetic markers into the genomic inserts. The λKOS vector was generated by introducing the yeast 2micron origin of replication and the TRP1 gene into the λKO vector backbone, which allows for the propagation and selection in yeast of the circular plasmid pKOS after CRE-mediated conversion (see below). As shown in Figure 1A, the new λKOS vector contains a stuffer fragment that was replaced by partially digested and size-selected genomic DNA from the brain of a 129/SvEv male mouse, which is the most commonly used mouse strain for gene targeting in ES cells. Figure 1B shows a schematic of a putative genomic pKOS clone after conversion of λKOS clone into a plasmid vector by CRE-mediated recombination following infection of recombinase-expressing BNN132 bacterial cells. Figure 1C displays restriction digests of the pKOS plasmid with the enzymes indicated in Figure 1A. BamHI Research Report 1152 BioTechniques Vol. 26, No. 6 (1999) and SalI restriction digests release the 5.8-kbp stuffer fragment from the 9.3kbp vector backbone. NotI linearizes the pKOS plasmid and can be conveniently used to linearize completed targeting vectors before ES cell electroporation. The two EcoRI sites, internal to the HSV-tk promoter, flank the stuffer fragment and the genomic inserts of the λKOS genomic library and are useful to estimate the average insert size of the genomic library, which is approximately 9.0 kbp, as shown in Figure 1D. Approximately 35 000–45 000 primary recombinant λKOS phages were plated onto each of 94 plates. Phage from each plate were individually amplified as λKOS subpools. One-microliter aliquots from each amplified subpool were transferred into 96-well plates, along with a positive and negative control template, and amplified by PCR using primers specific for independent genomic loci. As summarized in Figure 2E, 3–17 positive phage pools were typically identified in 11 independent projects. An average of approximately 8.3 phage pools per locus is in agreement with the theoretical coverage of the murine genome represented by the λKOS library, based on the number of individually plated recombinants and their average insert size (7). An important advantage of the ability to perform PCR on a genomic library is the speed and the rate of recovery of independent recombinants even after the library is amplified. Individual clones appear in independent subpools of the λKOS library and can, therefore, be separately isolated. This is important because the average insert size of 9.0 kbp requires high numbers of recombinants for good genomic coverage, making it impractical to screen by filter hybridization. Compared to the previously described λKO genomic library (14), the 6-kbp smaller insert size simplifies mapping strategies but does not reduce overall targeting frequencies in ES cells (unpublished observation). Over 120 clones from 30 independent genomic loci have been analyzed, and neither rearranged nor chimeric clones have been detected. Generally, 5–8 of the PCR-positive library subpools were plated using C600 bacterial cells and were filter-hybridized with a gene-specific probe. Only one additional round of plating and filter-hybridization was usually necessary to single out the desired phage for automatic plasmid subcloning by infecting BNN132 bacterial cells. Figure 1E shows that the analysis of genomic sequences obtained with the exon-specific primers E1 and E2, which were used for library screening, and the pKOS-specific primers G and X, which flank all genomic inserts (see also Figure 1, A and B), in combination with sequences obtained with primers derived from the X and G sequences, provide sufficient information about the relative position of the independent pKOS clones to each other. A central positioned clone (Figure 1E, pKOS#1, 2 and 4) is generally used for vector construction. On average, two successive primer walking steps suffice to obtain the complete genomic sequence of the locus, thereby revealing exon/intron structure, presence of repetitive elements and restriction site information. This method avoids the sometimes lengthy and problematic conventional restriction mapping analysis. The pKOS clones that extend beyond the 5′ or 3′ end of the targeting vector are used as PCR templates to derive external probes (Figure 1E) for Southern blot analysis of ES cells using the restriction site information obtained from the sequence analysis. Simple Targeting Vector Construction To integrate a positive-selection marker into the genomic insert DNA of pKOS plasmids, independent from the presence of suitable genomic restriction sites, the λKOS vector system takes advantage of the yeast homologous recombination machinery. It has been reported that 38–50 bp of homology, appended by PCR to amplified fragments, can apparently mediate sitedirected integration by homologous recombination in yeast cells (10). Based on this report, it has recently been observed that a yeast-selection cassette flanked by short stretches of homology can specifically recombine with the genomic insert of a targeting vector (16). To circumvent the need of any additional cloning steps, we constructed the pURA3/CAT plasmid in which a Vol. 26, No. 6 (1999) BioTechniques 1153 yeast/bacterial-selection cassette is flanked by several rare-cutter restriction sites as indicated in Figure 2A. Figure 2, A and B, indicates the chimeric oligonucleotides 1 and 2 that have at their 5′ end a short stretch of sequence homologous to the genomic region that flanks the sequences to be replaced by the selection marker. The 3′ end of the oligonucleotides can bind to pURA3/CAT and prime the amplification of the selection cassette including the rare-cutter restriction sites. Cotransformation of the URA3/CAT product with the respective genomic pKOS clone into yeast cells, which require both TRP1 and URA3 gene activity for growth on dual-selection growth plates efficiently selected for insertion events of the URA3/CAT product into the pKOS vector. Analysis of over 13 independent vectors indicated that every recombination event occured at the predicted locations. We also tested different lengths of homology, appended to the URA3/CAT cassette by PCR, varying from 25–40 bp. We generally obtained several independent colonies under double selection using competent yeast cells that were stored in aliquots at -80°C. Because no obvious difference in recombination frequency was observed, we now routinely use 30 bp of appended homologous sequence. The efficiency of obtaining recombinant clones was estimated at approximately 0.05%– 0.1% when colony numbers were compared from co-transformants plated on (-TRP/-URA) vs. (-TRP) plates. When the yeast-derived pKOS plasmids were shuttled back into bacteria, a second round of selection was applied by plating on ampicillin-resistant (Amp)/CAT plates. CAT selection was important because pKOS plasmids with and without the recombined URA3/CAT cassette coexisted in double-selected yeast colonies, as expected. Incubation of bacterial plates and liquid cultures for plasmid preparations were carried out at 30°C to prevent the sporadically observed instabilities of the large highcopy plasmids when cultured at 37°C (data not shown). Figure 2C shows that the successful recombination of the URA3/CAT cassette placed several rare-cutter restriction sites (e.g., SfiI) that flank the Research Report 1154 BioTechniques Vol. 26, No. 6 (1999) Figure 1. Structure of λKOS and excised pKOS plasmid derivatives. (A) The λKOS vector was assembled as detailed in Materials and Methods. The elements for bacterial propagation, origin of replication (ori), and ampicillin resistance (amp) are shown. The corresponding elements for yeast are the 2micron origin of replication (2micron) and the selection marker TRP1. The orientation of the HSV-tk genes, the KOS-specific primer binding sites X and G and the loxP sites are indicated by arrows. The restriction sites for library construction, SalI and BamHI, mapping, EcoRI and vector linearization, NotI, are also shown. (B) A schematic of a genomic KOS clone after CRE-mediated recombination into a plasmid vector. Note the reduction from two loxP sites in the λKOS phage as shown in Panel A to one after CREmediated recombination into the pKOS plasmid. (C) Restriction analysis demonstrating the integrity of the recombinant pKOS after CRE-mediated recombination of the λKOS vector into a pKOS plasmid containing the stuffer element shown under Panel A. BamHI and SalI separate the 9.3-kbp vector backbone from the stuffer element. EcoRI digestion removes an additional 150 bp of HSV-tk promoter sequences from the vector backbone due to the position of the EcoRI sites in the HSV-tk promoter. (D) A panel of randomly isolated genomic clones from the λKOS library after CRE-mediated recombination of genomic KOS plasmids digested with EcoRI. All plasmids show the expected 9.0-kbp vector fragment (as in Panel C). The estimated average genomic insert size is approximately 9.0 kbp. (E) A schematic of 5 independent pKOS clones identified and isolated with exon primers E1 and E2 by PCR. Clone pKOS#1 represents the future targeting vector, and the arrows indicate position and orientation of sequencing primers derived from the X and G end sequencing. External 5′ and 3′ probes for Southern analysis are derived by PCR as indicated. URA3/CAT cassette into the pKOS vector. These rare-cutter restriction sites allow one to easily replace the cassette with any particular marker that is useful for selection in ES cells. Figure 2D, 1–4, presents four independent examples of successive steps during vector construction. None of the four pKOS genomic inserts contained any natural SfiI restriction sites. After the replacement of the sequence between the homology arms of the pKOS vector with the URA3/CAT cassette by homologous recombination, SfiI digestion releases the 2.1-kbp URA3/CAT cassette from the recombinant pKOS plasmid. In a single cloning step, SfiI digestion/ligation was used to replace the URA3/CAT cassette with a positive-selection cassette suitable for ES cells. Figure 2D indicates that these positive-selection cassettes can also be released from the completed targeting vector by SfiI digestion, as either a 5.2-kbp fragment (IRESβGAL/MC1-Neo cassette) for constructs 1, 2 and 4 and as a 1.8-kbp fragment (PGK-Neo cassette) for construct 3. Both positive-selection cassettes have been constructed so as to be flanked with the same multiple cloning sites as the URA3/CAT cassette for convenient exchange (data not shown). Figure 2E shows that we have replaced genomic sequences ranging from 10 bp to over 3 kbp and have observed no significant differences in the frequencies at which modified vectors were obtained. Targeting Vector Construction for Conditional Gene Modification Figure 3 schematically details a convenient λKOS-based procedure to construct the more sophisticated replacement-type targeting vectors necessary for conditional gene modification. By placing the homology arms of the URA3/CAT cassette into intronic sequences flanking a genomic exon, in this instance exon 2, to be conditionally removed by CRE-mediated recombination, unique restriction sites (Figure 3, A–C, sites indicated by an A or B) are placed into the genomic pKOS vector. Oligonucleotides 3 and 4, as shown in Figure 3D, that contain, from 5′ to 3′, restriction sites A or A and B, a loxP element and sequences sufficient to prime Research Report 1156 BioTechniques Vol. 26, No. 6 (1999) Figure 2. Construction of simple gene targeting vectors. (A) A schematic of the pURA3/CAT plasmid vector. Flanking the selection cassette are several rare-cutter restriction sites. The restriction sites for SfiI contain independent and non-palidromic overhangs and allow directional replacement of the yeast/bacterial-selection cassette with a suitable ES cell-selection cassette. Primers 1 and 2 are used to amplify the URA3/CAT cassette including the flanking rare-cutter sites. Both primers contain at their 5′ end 25–40 bp of homologous sequences flanking the genomic region to be deleted in the targeting vector. (B) A schematic of a targeting vector as in Figure 1B. A deletion of part of exon 2 is planned by the indicated position of homology between the genomic insert of the pKOS vector and primer 1 and 2. (C) A schematic of a successful recombination of the amplified URA3/CAT cassette into the expected region of the pKOS vector. Note that several rare-cutter restriction sites are conveniently placed to replace the the URA3/CAT cassette with any desired ES cell-selection marker in a single cloning step. (D) Two consecutive steps in the construction of a simple targeting vector. All four plasmids were digested with SfiI. None of the pKOS plasmids contained a SfiI recognition site and remained intact (lanes 1, 4, 7 and 10). After yeast-mediated recombination of the URA3/CAT cassette (U/C) into the pKOS vectors, SfiI was used to excise the cassette from the recombinant pKOS vectors (lanes 2, 5, 8 and 11). After excision, the URA3/CAT cassette was replaced by an ES cell-selection cassette like IRESβGal/MC1neo (βG/neo; lanes 3, 6 and 12) or PGKneo (lane 9). (E) Summary of the numbers of separate subpools identified by a PCR-based screening in 11 independent projects. The number of independent recombinants per given genomic locus contained in the λKOS library might be slightly higher since independent clones might be present in one subpool but would only be once in this table. Also shown are the sizes of genomic sequences that were successfully replaced by the URA3/CAT cassette. In case of project Nos. 4 and 10, two different targeting vectors were created by yeast recombination. internally to the replaced sequences in Figure 3C, are used to amplify exon 2. Figure 3E shows the product that is digested with endonuclease A and is used to replace the URA3/CAT cassette from the modified pKOS vector, as indicated in Figure 3F by a single cloning step. Exon 2 is placed back into its natural context, but flanked on both sides by a repeat of loxP elements. The insert was always sequenced to identify possible errors in oligonucleotide synthesis or polymerase infidelity. Additionally, a unique restriction site B is inserted into one intron and used to insert a selection marker, for example, the hypoxanthine-guanine phosphoribosyl transferase (HPRT) mini-gene, flanked by FRT sites as a final convenient cloning step. In summary, one yeast recombination and two successive simple cloning steps are sufficient to obtain an otherwise complicated targeting construct. Using an HPRT-deficient ES cell line, one can positively and negatively select for targeted events by HAT and FIAU selection, respectively. The FRT-flanked selection marker can be removed by FLIP-mediated recombination between the FRT sites. This is achieved by transient expression of FLIP recombinase in ES cells after electroporation and selection for the loss of HPRT function under selection with 6-thioguanine. This outlined strategy for the construction of sophisticated targeting vectors was successfully accomplished nine of nine times within 4 weeks after initiating the λKOS library screening. As expected, all of the over 30 independent vectors generated with the λKOS system targeted ES cells with frequencies ranging from 5%–85%. Recently, transposon-mediated random integration into plasmid vectors was used to generate targeting vectors (20). This approach is based on the assumption that the integration of a transposable element in vitro is essentially unbiased, and that any insertion event, in particular into exonic sequences, can eventually be identified by massive colony PCR approach. In contrast, a single PCR is sufficient to generate the URA3/CAT cassette in the λKOS system, and every clone precisely contains the intended alteration. The random insertion event usually involves no replacement of genomic sequences, as is possible by the outlined yeast-mediated recombination, and therefore, does not allow the described simple production of the sophisticated conditional mutation vectors. A λphage/plasmid recombination system was a recent alternative approach to obtain replacement-type gene targeting vectors by a nonrandom fashion in a bacterial system (18). In fact, some aspects are similar to the λKOS system. This system also allows for the generation of genomic libraries that are partially constructed targeting vectors in which a genomic fragment can be Research Report 1158 BioTechniques Vol. 26, No. 6 (1999) Figure 3. Construction of gene targeting vectors for conditional mutagenesis. (A–C) The amplification of the URA3/CAT cassette, co-transfection into yeast and shuttling back into bacteria, as described in Figure 2, except that the chimeric amplification primers’ homology regions are from flanking intronic sequences. Note that a rare-cutter restriction site A is placed to flank the selection cassette. (D) Primers 3 and 4 amplify, by PCR, the region deleted by yeast recombination from the wild-type pKOS plasmid and place loxP sites and the restriction sites A and B on either end of the amplified product. (E and F) As a first cloning step, the rare-cutter restriction sites A are used to replace the URA3/CAT cassette. (G) In a second cloning step, rare-cutter restriction site B is used to insert a positive/negative-selection marker (e.g., the HPRT minigene) flanked by a direct repeat of FRT sites to complete the targeting vector. replaced by a selection cassette. In contrast to the observations with the λKOS system, the frequency of recombination decreased by approximately two orders of magnitude when the distance between the flanking homology arms was increased. While short stretches of sequence homology were sufficient for successful recombination in the yeast system, the reported bacterial recombination system required 200 bp of homology. This might require up to several cloning steps to flank the marker with the homologous arms. Because the λphage/plasmid recombination system has a reduced efficiency when creating deletions, and the phage is not subcloned into a plasmid vector but is rather directly used for ES cell modification, complex recombinant manipulations are required to construct sophisticated targeting vectors. The presently described λKOS system provides a very rapid and efficient method of constructing simple and complex replacement-type gene targeting vectors. The system begins with a very efficient genomic library screen by PCR, a single yeast transformation and one or two simple cloning steps depending on the complexity of the desired targeting vector and provides several advantages over other methods of vector construction.