Replacement recombination events targeted at immunoglobulin heavy chain DNA sequences in mouse myeloma cells

J. Mol. Biol. (1990)213,415-435

Replacement Recombination Events Targeted at Immunoglobulin Heavy Chain D N A Sequences in Mouse Myeloma Cells A. J. H. Smith and B. Kalogerakis Medical Research Council Laboratory of Molecular Biology Hills Road, Cambridge CB2 2QH, England

(Received 22 May 1989; accepted 8 December 1989) The rate of gene conversions and double crossovers between transfected and integrated # heavy chain immunoglobulin genes was measured in myeloma cells. The assay relies on correction of an integrated and defective ~ heavy chain expression unit, present in a repeated head to tail array in the genome of the myeloma cell line J558L. Following electroporation of these cells with restriction fragments containing normal immunoglobulin sequences, targeted recombination events are identified by a complement-mediated haemolytic plaque assay measuring production of functional IgM. Recombination results in replacement of a segment of the target sequence with the exogenous sequence. Different crossover positions are possible, giving rise to alternative rearrangements of the target site. In the case of one of the recombinants we analysed, more than one of the repeated targets had undergone a conversion event. The efficiency of homologous recombination was shown to depend on the extent of homology between transfected and target DNA. A targeting efficiency of 1 x 10 -s to 2 x 10 -s was achieved when the exogenous DNA contained 10,000 bases of sequence homologous with the target.

into immunoglobulin genes and their flanking sequences within their natural chromosomal context, and therefore permits analysis of controlling elements that exert their effects over a long distance. In addition, this technology may offer a rapid alternative means for constructing chimaeric immunoglobulin genes for the purpose of antibody engineering (Neuberger et al., 1985). It would be desirable to devise a general strategy for obtaining homology-directed (targeted) integrations into immunoglobulin genes in any hybridoma, or other cell type, and a number of schemes are potentially applicable to this problem (Jasin & Berg, 1988; Mansour et al., 1988). Before designing DNA constructs for this purpose, we decided to assess the requirements for obtaining efficient targeted integrations in myeloma cells with particular regard to the extent of homology required between the transfected DNA and the chromosomal target. Therefore, we designed a system in which recombination is readily detected by correction of a defective chromosomal immunoglobulin heavy chain gene. Production of functional IgM in this experiment is a phenotypic change that can be identified easily in situ by complement-mediated lysis assays (Kohler & Milstein, 1976; Kohler et al.,

1. I n t r o d u c t i o n DNA transfected into mammalian cells is capable of integrating into homologous chromosomal sites {Smith & Berg, 1984; Lin et al., 1985; Smithies et al., 1985; Thomas & Capecchi, 1986, 1987; Thomas et al., 1986; Doetschman et al., 1987, 1988; Song et al., 1987) but, in general, this occurs at a much lower frequency than the apparently random integration resulting from non-homologous recombination. The ability to integrate transfected DNA at specific sites by homologous recombination, should allow the structure of genes in mammalian cells to be experimentally manipulated to understand features of their expression and the role of their gene products not amenable to analysis by present technology. In the case of immunoglobulin genes, informative studies have been based on the transfection of cloned genes, manipulated in vitro, into myeloma cells (Neuberger, 1983). However, because the size of the DNA fragments that can be introduced into cells is limited, this method does not allow the analysis of long-range interactions involved in immunoglobulin gene rearrangements and expression. Homologous recombination offers the possibility of introducing predetermined changes 0022-2836/90/110415-21 $03.00/0

415

© 1990 Academic Press Limited

416

A . J . H . S m i t h and B. Kalogerakis

1976). Targeted integration of D N A into an endogenous defective heavy chain gene has been reported by Baker et al. (1988). I n this case, gene correction occurred by an event t h a t can be described formally as a single crossover resulting in insertion of the transfected vector, reconstruction of a functional gene and duplication of the target sequence. In the experiments described herein, we detect replacement recombination events at the chromosomal target; i.e. rearrangements t h a t arise as a result of double crossovers or gene conversions, with a transfected restriction fragment. We show that, depending on the crossover positions, different extents of target site replacement are possible. We have used this system to demonstrate t h a t the frequency of recombination is influenced by the extent of homology between the incoming restriction fragment and the chromosomal target.

2. Materials and Methods (a) Plasmid constructions The structure of plasmids used in these studies is shown in Fig. 1. The components of all the plasmids were derived from pSV-Vpl (Neuberger, 1983), pSV2neo (Southern & Berg, 1982), and pM1-SVgpt (Smith & Berg, 1984), and the vector backbone was pUC-12 (Yanisch-Perron et al., 1985). Restriction fragments were purified using the method of Vogelstein & Gillespie (1979). The plasmid pUC-VNeCpneo (Fig. l(a)) was used to make a chromosomal target for site-specific integration. It consists of an immunoglobulin heavy chain expression unit containing a/~ constant region, rendered defective by the insertion of a promoterless neo gene at a position interrupting the p coding sequence. The neo gene is present as a 2"0 kbt B g l I l - B a m H I restriction fragment from pSV2nso, which lacks the simian virus (SV40) early promoter but contains the entire neo coding sequence and the SV40 small t intron and early region polyadenylation signal sequences. This is inserted into a B a m H I site in the 2nd exon of the/~ constant region, originally derived as a 4"3 kb XbaI restriction fragment from pSV-V/~I. The modified XbaI restriction fragment (6"3 kb) is contained in the pUC-VNpC/~neo plasmid adjacent to a variable region segment consisting of the immunoglobulin heavy chain promoter, the leader exon, an assembled variable region exon VDJ2, unrearranged J3 and J4 segments, and the heavy chain enhancer. This variable region segment, a 3"9 kb EcoRI restriction fragment from pSV-V/~I, had been derived from a combination of genomic and cDNA sequence from a hybridoma secreting an IgD, 21 antibody with specificity to 4-hydroxyl-3-nitrophenacetyl (NIP) (Neuberger & Rajewsky, 1981). The variable region, constant region and nso coding sequences are in the same orientation in this construct. Furthermore, expression of the neo gene is dependent on the variable region segment and is presumed to result from the fact that the neo coding sequence can, in principle, be contained as part of a spliced transcript in myeloma cells, initiated at the heavy chain promoter and formed from the leader exon, the VDJ 2 exon and g exons up to the position of the nso t Abbreviations used: kb, 103 bases or base-pairs; NIP, 4-hydroxyl-3-nitrophenacetyl; DMEM, Dulbecco's modified Eagle's medium; PBS, phosphate-buffered saline; SRBC, sheep red blood cell.

insert, with polyadenylation of the transcript occurring at the position of the SV40 polyadenylation signal sequences. In this dicistronic mRNA, the constant region reading frame from exon 2 extends across the join at the position of the neo insertion into a short region of 5' untranslated sequence on the neo fragment, where it is terminated at a TAG codon 12 nucleotides before the neo initiation codon. Therefore, translation of the neo sequence probably results from translational termination and reinitiation (Peabody & Berg, 1986) given the juxtaposition of the termination codon and the neo initiation codon. The plasmid pUC-C/~SVgpt2 (Fig. l(b)) was used to make restriction fragments for targeted recombination with the defective heavy chain gene. This consists of the original 4"3 kb XbaI restriction fragment containing the p constant region sequence modified by insertion of the SVgpt gene at a HindIII site in the intron sequence between exon 4 and the membrane exons. The SVgpt gene was obtained as a 2-2 kb B a m H I restriction fragment from pM1-SVgpt (Smith & Berg, 1984) and contains the SV40 early promoter, small t intron and polyadenylation signal sequences. The SVgpt gene and constant region coding sequences are in the same orientation. The plasmid pUC-VNpCpSVgpt (Fig. l(c)) was derived by insertion of the 3"9 kb g e o R I restriction fragment from pSV-Vpl containing the variable region sequence into the single EcoRI site in the polylinker of pUC-CpSVgpt2. The orientation of the variable region with respect to the constant region is the same as in pUC-VNpCpneo. This plasmid therefore encodes a complete p polypeptide. pUC-VA2CpSVgpt (Fig. l(d)) was also used to generate restriction fragments for targeted recombination, and was derived from pUC-VNpCpSVgpt by deleting 1-6 kb of sequence between the EcoFtI site at the junction with the pUC vector and a StuI site in the variable region. It therefore lacks the heavy chain promoter and the 5' portion of the variable region coding sequence. The EcoRI site at the position of the deletion remains, but the EcoRI site in the polylinker between the variable and constant regions was removed by cutting, filling-in and religation. This plasmid is incapable of producing a /~ polypeptide as a consequence of the deletion introduced into the variable region. (b) Preparation of D N A for electroporation Restriction fragments for electroporation were prepared by digesting approx. 100 pg amounts of plasmid DNA. The DNA was fractionated on 0"7~o (w/v) agarose gels (10 cm × 10 cm). After resolution of the fragment, the gel was rotated by a right-angle and the DNA electroeluted onto a small area (1 cm 2) of GFC paper with a dialysis membrane backing. The DNA was removed from the paper by centrifugaton and precipitated with ethanol. The concentration of the DNA was determined and portions of DNA for a single electroporation (see below) were re-precipitated, and finally dissolved at a concentration of 2/~g//ll in sterile conditions. (c) Cell culture and selection conditions The cell line used in this study was the mouse plasmacytoma J558L (Oi et al., 1983). This produces no heavy chain, and does not contain any endogenous/~ constant region sequence. J558L however produces a 21 light chain which contributes to an anti-NIP idiotype when the cells express an exogenous heavy chain gene derived from a hybridoma secreting anti-NP antibody (Neuberger et al., 1984).

Targeted Recombination in Myeloma Cells

417

J558L cells were grown in Dulbecco's modified Eagle's medium (DMEM) with antibiotics and 10% (v/v) foetal calf serum. Selection for G418 resistant (neoa) transformants was carried out using the above medium supplemented with 1 mg G418 (Gibco)/ml in 24-well plates at a density of 2 x l05 cells/well, gpt selection was carried out using DMEM with antibiotics and 10% tbetal calf serum supplemented with 250 pg xanthine/ml, 20 ~g hypoxanthine/ml and 5 pg mycophenolic aeid/ml in 24-well plates at a density of 2 x l0 s and 2 x l04 cells/well, or in soft agar (0"3% w/v) in 3"5 cm diam. wells at a density of 5x l0 s and 5 x l04 ceils/well.

activity of approx. 5 x l0 s cts/min per pg, and used at a concentration of 5 x 106 cts/min per ml. The gpt probe was the 0"93 kb BglII-ApaI fragment of pM1-SVgpt, the neo probe was the l'0 kb BglII-SmaI fragment of pSV2neo, and the Cg probe was the 1-21 kb HindIII fragment from pUC-Cpneo. In the case of the 2 probe, this was labelled to a spec. act. of 5 x 106 cts/min per pg and used at a concentration of 5 x 104 ets/min per ml.

(d) Electroporation conditions

The experimental system was designed to detect replacement recombination events in the mouse p l a s m a c y t o m a cell line J 5 5 8 L by identifying sequences within a chromosomal target, which are exchanged, or substituted by gene conversion, with sequences located between regions of homology on transfected restriction fragments. I t is based on the use of an artificial target derived from an integrated form of the plasmid pUC-VseCl~neo {Fig. l(a)), which contains an immunoglobulin h e a v y chain expression unit. This is comprised of a complete variable region, which includes the h e a v y chain promoter and enhancer, and exon sequences encoding a domain with binding specificity to the hapten 4-hydroxyl-3-nitrophenacetyl (NIP); and adjacent p constant region modified b y the insertion of 2"0 kb of neo sequence in its second exon. After introduction into J 5 5 8 L cells (which contain no endogenous C# sequences), the plasmid is therefore capable of producing only an incomplete /~ polypeptide, as a consequence of the neo insert. Correction of this defective constant region sequence provides the means of identifying homologous recombination events. The artificial target was made by obtaining stable integrations of this plasmid D N A in the J 5 5 8 L genome in an initial round of transfection, in which cells were selected for G418 resistance (neoR: Southern & Berg, 1982). Although the neo gene does not contain its own promoter, neog transformants can be obtained after electroporation of J 5 5 8 L cells at a frequency of a p p r o x i m a t e l y 10 -4, indicating t h a t it is efficiently expressed in this construct from transcription initiated at the h e a v y chain promoter. T r a n s f o r m a n t s containing an a r r a n g e m e n t of the pUC-Vm, Cgneo D N A in which the variable and constant regions have remained i n t a c t can be used for the recombination assay. A recombination event, which removes this neo sequence and restores the reading frame of the g gene, will allow expression of a complete # polypeptide t h a t can be identified easily in J 5 5 8 L cells b y haemolytic plaque assay (see below). The plasmids pUC-CgSVgpt2 and pUC-VA2C#SVgpt (Fig. l(b) and (d)) were designed to provide restriction fragments capable of homologous recombination with this chromosomal target, and contain contiguous /~ sequence at the site of the neo insertion. These plasmids also have a modified constant region in which a SVgpt gene is located between exon 4 and the m e m b r a n e exons. The SVgpt gene includes the SV40 early promoter,

Care was taken to maintain J558L cells in logarithmic growth to obtain high transfection efficiencies. Cells were harvested for electroporation at a density of 2 x l0 s to 4x 10S/ml, centrifuged at 1000 revs/min in an IEC Centra-7R centrifuge, and then resuspended in phosphate buffer saline (PBS). They were then centrifuged again as before, and finally resuspended at a density of l0 s cells/ml in PBS. Portions of cells (10 T in 100 pl) were dispensed to chilled sterile tubes and DNA was then added (15 to 40 Aug in a volume of less than or equal to 20 jul). Electroporation was carried out by delivering 3 discharges to the cell/DNA suspension in a sterile plastic cuvette, from an EMBL/Apelex apparatus set at a capacitance of 222 nF with a field strength across the electrode of 4 kV/cm. After electroporation, the cells were allowed to recover on ice for l h, and were then added to 25 ml of DMEM plus serum and grown for 36 to 48 h before selection or plating in soft agar. For some experiments, the cells were plated in soft agar directly after recovery from electroporation. Viable cells were counted by trypan blue exclusion.

(e) Haemolytic plaque assays Sheep red blood cells (SRBCs) were labelled using NIPcaproate-O-succinimide (Cambridge Research Biochemicals) dissolved in dimethyl formamide and used at a final concentration of 150 pg/ml. The coupling reaction was carried out at 5°C for 1 h, the SRBCs were then washed 3 times with PBS and finally resuspended in 4 times their packed volume. Portions (15/~1) of resuspended NIPcoupled SRBCs were then used with 0-15 ml of 0"25~/o (w/v) top layer agar per 3"5 cm diam. well. For the direct recombination assay, myeloma cells were plated in soft agar at 5 x 10S/well with NIP-coupled SRBCs, with addition of complement after 2 days. Stable gpt + transformants of myeloma cells were plated at l04, l03 and 200 cells/well and complement was added after 4 to 6 days. Clones that gave rise to haemolysis were picked and either transferred into wells or into soft agar with NIPcoupled SRBCs for immediate recloning. (f) Southern blot analysis (Southern, 1975) Approximately 15 #g amounts of cell DNAs with, when appropriate, 10 pg of plasmid standards were digested with restriction endonucleases and fractionated, together with HindIII-cut ]t DNA, on agarose gels run in Trisacetate buffer (pH 8-0). After exposure to short-wave ultraviolet light for 2 min, the DNA was transferred to Gene Screen Plus membrane (DuPont) by vacuum blotting essentially as described in the LKB Vacublot equipment manual, omitting the depurination step. Hybridization was carried out using 32P-labelled probes made by nick translation !Rigby et al., 1977) to a specific

3. Results (a) Design of experiments

A. J. H. Smith and B. Kalogerakis

418

(a) B

X

E,B,Sc,X

B Bt

I I I~,sc

....

X,S

I

===,,,=,=~ =-_

,inn I i, , LVDJ2J3 ,.14 ( ~ 1 2

VNp

(b)

K

2

I C~I

E,B,Sc,X

Bt

neo

3 4

3

M

I

B GK

C/J.

B

K

I 7~c~01 II 1 2

., il

I pUC

I

E

X,S

II

4

I

M

I

Cu.

!

SVgpf

E

B

X E,B,Sc,X

I

I I l~c~l

t

C/~

I

pUC

B

K

(c)

nn

i

m

l|

I

n I ,

VNP

E B

B GK

II

m mm n ~ , mmu ~

L VD~ J3 J4 ( ' ~

(d)

Bt

12

I

Cp.

I

II

n •

I,

I I

, in

3'VDJ'2 J3 J4 O

t

VZ~

1 2

I

H a

in 3

C/J.

I

M

II I I i#~1 •

I

il all

4

B GK

•

I I

~

Bt

m

X B,Sc,X

3

E

X,S

SVgpf

B

I

C/~

K

X,S

I

pUC

E

I I

1

II el

~

4

I

M

I

SVgpf

l

C/.L

I

pUC--I

Figure 1. Structure of plasmid constructs. Plasmids are shown arbitrarily linearized at an EcoRI site. Exon sequences are denoted by black boxes, neo sequence by a striped box, gpt sequence by a cross-hatched box. E denotes the heavy chain enhancer. Restriction enzyme sites are abbreviated as follows: B, BamHI; Bt, BstXI; E, EcoRI; G, Bg/II; K, KpnI; S, SalI; Sc, SacI; X, XbaI. (a) pUC-VNpCpneo contains a 3"9 kb EcoRI variable region fragment and a 6-3 kb XbaI constant region fragment inserted into the EcoRI and XbaI sites of pUC-12, respectively. The constant region contains a neo sequence (2"0 kb) inserted at the BamHI site of exon 2. (b) pUC-CpSVgpt2 contains a 6"5 kb XbaI constant region fragment at the XbaI site of pUC-12. The constant region contains a SVgpt gene present as a 2-2 kb BamHI fragment at a position originally defined by a H i n d I I I site. (c) pUC-VNpC/~SVgpt contains the 3-9 kb variable region fragment from pUC-VNrCI~neo inserted in the same orientation at the EcoRI site of pUC-CftSVgpt. (d) pUC-VA2CftSVgpt contains a deletion of 1-6 kb from the 5' end of the variable region segment of pUC-VNrCpSVgpt, but retains the EcoRI site at the pUC-variable region junction. The EcoRI site in the polylinker sequence between variable and constant regions is removed.

Targeted Recombination in Myeloma Cells

I o.g

II

1.s

II

1"9

::

-,,.

::

;2

=:

',;

ik

i"

ik

I

kb

Figure 2. Potential crossover positions between the pUC-VNpCltneo chromosomal target and exogenous XbaI restriction fragment from pUC-CltSVgpt2. The size (in kb) of regions of homoloigy between the transfected and target DNA is shown. The BamHI site in the XbaI restriction fragment is indicated to denote the presence of normal constant region sequence across the site of the neo disruption in the target. Recombination by double crossovers at 1 and 3 will replace the neo disruption in the target and integrate the SVgpt gene. Thus, the clone should acquire an IgM+gpt+ phenotype. Recombination by double crossovers at 1 and 2 will only replace the neo disruption and thus the cell should acquire IgM+gptphenotype. It should be noted that homology between a common sequence located at the 3' end of the neo and SV#pt genes could allow for an alternative crossover position giving rise to a clone with an IgM+ypt+ phenotype. This has not been indicated. Open bar, variable region sequence; filled bar, constant region sequence; striped bar, neo gene; cross-hatched bar, SV#pt gene. XbaI and BamHI restriction sites are denoted by X and B, respectively.

which is functional in myeloma cells, and therefore allows cells containing stable integrations of these DNAs to be obtained by gpt selection (Muliigan& Berg, 1981). pUG-C~SVgpt2 contains no variable region, and pUC-VA2C~SVgpt contains a deletion of the 5' segment of the variable region including some of the coding sequence of the VDJ 2 exon. Therefore, transfection with these plasmids, or restriction fragments derived from them, will not give rise to cells producing a complete # polypeptide with a variable region encoding binding specificity for the NIP hapten, unless a homologous recombination event occurs with the chromosomal target. A double crossover between a 6-5 kb X b a I restriction fragment derived from pUC-C#SVgpt2 and the pUC-VC~neo target is illustrated in Figure 2 and shows, in principle, how sequence contained in the transfected DNA could replace the region with the neo insert to give uninterrupted coding sequence and insertion of a SVgpt gene (if the crossovers occur at positions 1 and 3). The prediction is that if cells contain this arrangement of integrated DNA they will be represented in the pool of gpt + transformants obtained after transfection. The position of the SVgpt gene should not interfere with expression of the p gene in this arrangement, since IgM-secreting cell lines, such as J558L, polyadenylate heavy chain transcripts at the end of exon 4 and therefore no splicing event to the membrane exons is required.

419

Rapid detection of J558L cells expressing a heavy chain with this variable region sequence is possible, since they also produce an endogenous 21 light chain (but no heavy chain) with which it can associate to make a secreted IgM capable of binding NIP (Neuberger et al., 1984). In the case of a complete ~ polypeptide, the IgM secreted from the cells will be capable of fixing complement. These can be conveniently identified by complement-mediated lysis of SRBCs coupled with this hapten; this results in the formation of plaques if the cells are plated in soft agar. To evaluate the efficacy of this technique, and to confirm that the SVgpt gene does not interfere with expression of the /~ gene, the plasmid pUC-VNpCpSVgpt (Fig. l(c)) was introduced into J558L cells by electroporation. This plasmid contains the same heavy chain variable region sequences present in pUC-VNpC#neo joined to the modified constant region from pUC-C#SVgpt2. Stable gpt + transformants were obtained at a frequency of l0 -3 of the viable cells plated in selective medium (this relatively high transformation efficiency is thought to be the consequence of an elevated level of transcription of the SVgpt gene obtained as a result o f its proximity to the heavy chain enhancer element included in this construct). These were then plated in soft agar together with NIP-coupled SRBCs, and approximately 70~/o of the gpt + clones gave lysis on addition of complement. An example of the appearance of the plaques that are obtained is shown in Figure 3. A probable explanation for gpt + colonies that failed to secrete functional IgM is that these contained plasmid DNAs that have suffered disruptions of their immunoglobulin sequences during the integration event.

(b) Construction of a cell line containing a target for homologous recombination A requirement of the experimental design was the derivation of a cell line containing a complete and uninterrupted copy of the heavy chain gene from pUC-VseCltneo, which could then be utilized as a target for homologous recombination. Cleavage of plasmid DNA at a unique restriction site normally results in the ends of the linear duplex becoming linked to host DNA by non-homologous recombination, such that the internal vector sequences remain intact on integration. However, since there was no unique restriction site in the non-essential pUC component suitable for linearization of the vector, we reasoned that, by generating a tandem head to tail array of the plasmid before transfection, one or more complete copies should then be present after integration into the host genome. To facilitate this, the pUC-VNpCl~neo plasmid DNA was cleaved at a unique B s t X I site (located in the 3rd exon of the constant region) and was then ligated at a high concentration of DNA; only head to tail concatamers result from this ligation. Stable neo R J558L transformants obtained after electroporation with

420

A. J. H. Smith and B. Kalogerakis

Figure 3. Haemolytic plaque assay using IgM-producing J558L cells. A soft agar well containing NIP-coupled SRBCs, seeded with approx. 100 9pt + J588L cells obtained from electroporation with pUC-VNpCIaSVgpt, on addition of complement after 5 days.

this DNA were cloned, expanded into mass culture and their genomic DNA then analysed by Southern blotting. Clones containing a head to tail arrangement of integrated plasmid DNA could be distinguished readily by SalI digestion of genomic DNA and identifying which gave restriction fragment of the size of full-length linear pUC-VNpC/aneo. A number of such clones were identified and one of these, J558neoalS, which contained the smallest number of copies was selected for use in the experiments discussed here. The arrangement of pUC-VNpC/aneo DNA in J558neoR1S consists of three copies of the plasmid arranged in a head to tail configuration. This is shown in detail in Figure 4(a) and was determined on the basis of Southern blot analysis after digestion with several restriction endonucleases, some of which are shown in Figures 5 to 8. We arbitrarily define the boundary of each copy at the EcoRI site between the pUC sequence and the variable region sequence, and designate these copies R1, R2 and R3. The R1 copy is interrupted within the variable region sequence by one of the integration breakpoints with the host DNA, the R2 copy is an intact plasmid sequence and the R3 copy is interrupted by the other integration breakpoint at a position in the

constant region, in the intron sequence between exon 4 and the membrane exons. R3 still has the potential to form a functional/a gene on removal of the neo sequence since, as explained earlier, this interrupted intron sequence is not required for expression. Digestion of J558neoR1S DNA with XbaI endonuclease and hybridization of Southern blots with a probe (C/a) specific to constant region sequence therefore, gives rise to two bands representing fragments of 6"3 kb and approximately 8-0 kb with relative signal intensities of 2:1, respectively (see Fig. 5(b)). These correspond to the two intact constant region sequences (6"3 kb fragment) from R1 and R2 and the interrupted constant region sequence (approx. 8"0 kb fragment) from R3, the size of the latter representing the distance between the preserved XbaI site in the constant region and the nearest XbaI site in the flanking chromosomal DNA. Since J558L contains no endogenous constant region sequence, only these plasmid-derived sequences are detected. The J558neoalS cell line maintains the neoR phenotype in the absence of selection. It does not spontaneously give rise to clones capable of activating complement lysis at a detectable frequency (< 10-s).

Targeted Recombination in Myeloma Cells

(c) Detecting targeted events in gpt + transformants Initial attempts to detect integration into the tandemly repeated target in J558neottlS were carried out using the 6"5 kb XbaI fragment from pUC-C#SVgpt2. As illustrated in Figure 2, this contains the SVgpt gene flanked by a total of 4-3 kb of constant region sequence homologous to the target site, with normal coding sequence across the site of the neo insertion. In addition, there is 0"8 kb of homology between the neo and SVgpt'genes as a consequence of common SV40 sequences at their 3' ends. As discussed, a double crossover with the transfected DNA within the 0"9 kb and 1"9 kb stretches of homology (crossovers 1 and 3 in Fig. 2) should give a cell with a gpt + phenotype that produces an IgM that can activate complement (IgM + phenotype). The experimental strategy involved first selecting for gpt + transformants after transfection with this restriction fragment. Electroporation of l0 T J558neoR1S cells with 20 #g of XbaI restriction fragment gave gpt + transformants at a frequency of 10 -4 to 10 -s of the viable cells plated in selective medium. After ten days in selection, the transformants were pooled and then screened using the haemolytic plaque assay to measure the fraction established as a result of homologous recombination. Haemolytic plaques arose at a frequency of one per 200 gpt + colonies when pooled cells were plated in soft agar together with NIP-coupled SRBCs and complement was added after five days, although the frequency varied on occasions between extremes of one per 102 and one per 103 . Maintaining the cells longer than ten days in gpt selective medium gave a significantly lower fraction of gpt+IgM + cells, since many gpt+IgM + clones appear to grow more slowly and produce smaller colonies than gpt+IgM - clones. The frequency of IgM + clones among the gpt + transformants was also determined in an alternative procedure in which cells were plated 48 hours after electroporation in soft agar with gpt selective medium. Cells were plated at a density of 5 × l0 s per 3"5 cm diameter well and the number of colonies counted ten days later (the frequency ofgpt + transformation was 4 x 10-5). The wells were then overlaid with soft agar containing hapten-coupled SRBCs and complement was added one day later. On average, one colony in 150 gave rise to haemolysis, in agreement with the above frequency. Again, many of the IgM + clones were significantly smaller. Colonies that gave rise to plaques were replated in soft agar with NIP-coupled SRBCs and clones of gpt+IgM + cells were subsequently isolated and expanded into mass culture. Analysis of the DNA of these gpt + clones proved that they were established as a result of the anticipated homologous recombination event (see below). Some of the gpt + clones that did give rise to plaques were also expanded and analysed (data not shown) and, as expected, contained apparently random integrations of the

421

SVgpt gene, with consequently no alteration of the target site. (d) Direct detection of targeted events We also screened for homologous recombination events directly, using the haemolytic plaque assay method, without first obtaining stable gpt + transformants. This screen should detect recombination events that did not involve integration of the SVgpt gene and would therefore not have been isolated in the previous procedure. These could, in principle, arise if a double corssover occurred at positions 1 and 2 in Figure 2, removing the neo disruption in the target and replacing it with normal sequence from part of the exogenous DNA, but leaving the remainder unintegrated. In this experiment, l0 T cells were electroporated with 20 #g of restriction fragment as before, but after allowing them to recover in medium plus serum for 36 to 48 hours, they were plated at high density (5 x 105 viable cells/3"5 cm diam. well) in soft agar with NIP-coupled SRBCs. Complement was added after two days, since growth of the cells at this density prevents [ysis being easily distinguished after this time. Although the plaques produced were small, they were clearly visible and could be counted accurately. With the 6"5 kb XbaI fragment from pCU-Cl~SVgpt2, we obtained, on average, one plaque per 5 x 105 viable cells plated, which represents an efficiency of IgM + transformation of 2 × l0 -6 (see Table 1). The viability of the cells measured immediately after electroporation was 50 to 60O/o. Therefore, on the basis of this efficiency of IgM + transformation, we calculated that there were, on average, l0 to 12 independent events per l0 T cells electroporated. Additional experiments were carried out in which l0 T cells were electroporated as before with 20 #g of DNA but were then plated immediately in soft agar with NIP-coupled SRBCs. T h e number of plaques obtained varied between six and eight. Cells from the area of haemolysis were picked and replated at low density in soft agar with NIPcoupled SRBCs. Because of the high cell density in the area of the plaques, only a fraction of the colonies in the secondary plating subsequently gave rise to plaques. However, IgM + clones could be obtained from 90% of the original plaques. This frequency showed that most of the plaques identified in this direct assay represented stable events, and were not the result of transient expression. IgM + clones derived from this experiment were plated in gpt selective medium to assess the frequency with which these clones had acquired the gpt + phenotype. About two-thirds survived the gpt selection, indicating a relatively high frequency of co-integration of the unselected SVgpt gene. (e) Analysis of the recombination event To prove that the IgM + transformants were derived as a result of targeted integration and to

A. J. H. Smith and B. Kalogerakis

422

(a)

Bt

5-8

I

K

12.9

g

9.6

I

X

I K Bt

12-9

it 6-3

XX

Bt

12-9

Ii

I

I

KX

XX

12-9

g 6-3 Bt

g 3.1

m

I 8,0

i

I

K X

XX

Bt

I K Bt X

[_.

.... j I R1

(b)

Bt

I

K

I

3"8

X

II 2-5 U 6"5

I K Bt

15-1

13-1

II

7-3

R3

R2

XX Bt K

10.6

6-5

I

KX

XX BI

K

3.1

12-9

II 2-5 g

I

II

8.0

I

I

K X

XX

Bt

K Bt

X

J__

.... L l

R1

(c)

Bt

I

K

R3

R2

3.8

I

7.3

X

II

6-5

XX Bt

K

8-0 KX

XX

Bt

R1

Bt K

I I

X

5"8

K Bt

XX

X

R3

12"9

II

6-3 Bt

10-9

II 12"9

9-6 I

K Bt

!l..l

.--J,

(d )

3-1

12.9

II 2.5 II

I K Bt

15-1

II

I

I

KX

XX

R1

Bt

R2

Fig. 4.

10.9

II 6'3

II

I

K X

I

X X Bt

R3

3-1

'

I 6"0

K Bt

I

X

Targeted Recombination in Myeloma Cells

(e)

Bt K

5.8

I

9.6

I

10.9

U

423

10-6

II

II

6.3 X

I KBt

XX

I

Bt

5.3

II

2-5

I I

8-2 I

I

KX

XX

I

I Bt

K

K Bt X

I

R1 R3 Figure 4. Arrangement of plasmid-derived DNA in parental and recombinant cell lines. Arrangement of integrated plasmid sequences in the cell lines: (a) J558neoalS; (b) A4 (IgM+ffpt+ phenotype); (c) an IgM- segregant of A4; (d) 3B2 (IgM+gpt- phenotype); (e) 2C1 (IgM+gpt + phenotype). Only restriction sites and the fragment sizes they generate that are relevant to understanding the Southern blots in Figs 5 to 8 are shown. The cell lines consist of tandem head to tail repeats of the plasmids. These are distinguished as RI, R2 and R3, delineated at the junctions of pUC and variable region sequence, and at the junctions with chromosomal DNA as indicated. The short line beneath each constant region denotes sequence that hybridizes with the C/~ probe. Open bar, variable region sequence; filled bar, constant region sequence; striped bar, neo gene; cross-hatched bar, SVgpt gene; continuous line, pUC sequence; broken line, chromosomal sequence; diagonals, integration breakpoints. Symbols for restriction endonuelease sites are as for Fig. I.

understand the nature of the recombination event, the arrangement of plasmid-derived sequences in these cell lines was determined by Southern blot analysis. In the following discussion, three examples of IgM + transformants obtained after transfection with the 6"5 kb XbaI restriction fragment from pUC-C#SVgpt2 are described in detail. (i) A n IgM +gpt + transformant derived by recombination at the R1 and R2 copies The recombinant cell line A4 was identified from a pool of stable gpt + transformants. The presumption is therefore t h a t the cell line must have acquired an SVgpt gene, and that the IgM + phenotype resulted from deletion of the neo disruption. By use of a probe specific to neo sequence, rearrangements that involve a deletion are readily identified by loss of detection of a target site. Figure 5(a) shows a Southern blot analysis of genomic DNA from J558neoR1S and the A4 recombinant restricted with XbaI endonuclease and hybridized with a probe specific to neo sequence. In the J558neoR1S DNA, the probe detects a restriction fragment of size identical to the 6"3 kb fragment obtained from digestion of the pUC-VNpCl~neo plasmid standard, and a restriction fragment with a size of approximately 8"0 kb; these represent, respectively, the intact constant regions from repeats R1 and R2, and the interrupted constant region from the R3 repeat. The neo probe detects only the 8"0 kb restriction fragment in the A4 recombinant, and thus either the entire R1 and R2 repeats have been removed in this cell line, or the neo sequence has been deleted from within the repeats. The transfected XbaI restriction fragment is 0.2 kb larger than the 6"3 kb XbaI fragment derived from the R1

and R2 repeats, by virtue of the size difference between the SVgpt and neo sequences. A recombination event that removes the neo sequence and inserts the SVgpt gene at one of these targets should "therefore alter the size of the corresponding XbaI fragment to that of the transfected DNA. The same filter shown in Figure 5(a) was rehybridized with a probe derived from the constant region sequence (Fig. 5(b)), which therefore detects both target and transfected DNA sequences. Consistent with the result obtained with the neo-specific probe, the 6"3 kb XbaI restriction fragment is detected in J558neoR1S DNA, but not in the A4 DNA. However, a restriction fragment is detected in the A4 DNA of size identical to the 6"5 kb XbaI restriction fragment derived from the pUC-VNeCpSVgpt plasmid standard (lane 5). In both cell lines, as expected, the 8"0 kb restriction fragment is detected. Moreover, the 2 : 1 ratio in signal intensity between the bands representing the 6"3 kb and 8"0 kb restriction fragments in J558neoalS is the same as for the 6"5 kb and 8"0 kb restriction fragments in A4. The data, therefore, suggest t h a t the constant regions from R1 and R2 have both been converted to size consistent with the loss of neo sequence and gain of an SVgpt gene, while the constant region in. R3 has remained unaltered. Hybridization with a probe specific to gpt sequence (data not shown) detects only the 6"5 kb restriction fragment in agreement with this interpretation. Taken alone, these data do not provide conclusive evidence t h a t the anticipated recombination event has occurred, since other circumstances could conceivably give rise to a genomic XbaI restriction fragment containing constant region and gpt sequence identical in size to the transfected DNA.

2

(o)

3

4

5

6"Skb

3.0kb

2

3

(b)

4 5

--

6"5kb

Figure 5. Southern blot analysis of genomic DNAs of J558neoR1S, the IgM+gpt ÷ transformant A4, and an IgM-gpt ÷ derivative of A4, restricted with XbaI. DNAs (15/lg) were fractionated on a 0"7 % gel, transferred to a nylon membrane and (a) hybridized with the neo probe and (b) rehybridized with the Cg probe. Lane l, J558L DNA plus l0 pg of pUC-VNrCpneo plasmid; lane 2, J558neoR1S DNA; lane 3 A4 DNA; lane 4, IgM- A4 DNA; lane 5, J558L DNA plus l0 pg of pUC-VNpCpSVgpt. HindIII cut i DNA was run adjacent to lane 5 (fragment sizes; 23"1, 9"4, 6"6, 4"4, 2"3 and 2-0 kb).

6"3 kb "

8 " 0 kb

I

Targeted Recombination in Myeloma Cells The further evidence necessary to prove this is provided by restriction endonuclease digests with the enzymes KpnI and BstXI (Figs 6 and 7, respectively). Restriction fragments detected with the constant region specific probe (C/~ probe) in Southern blots of KpnI and BstXI-restricted A4 DNA (lane 3 of Figs 6(b) and 7(b)) can be accounted for as follows. A single KpnI site is located in the SVgpt gene. Hence, targeted integration creates a second KpnI site in each of the repeats R1 and R2, which together with the deletion of the neo sequence decreases the size of the parental KpnI fragments derived from these repeats by 2-3 kb. Thus, the KpnI restriction fragment of approximately 9"6 kb in the parental DNA, which represents the distance between the KpnI site in R1 and the nearest KpnI site ill the adjacent chromosomal DNA, is reduced ill size to approximately 7"3 kb. Similarly, one of the two copies of the KpnI restriction fragment in the parental DNA identical in size to the 12"9 kb KpnI linear pUC-VNpCgneo (in lane 1), which represents the distance between the single KpnI sites in R1 and R2, is reduced to a size identical to the 10"6 kb restriction fragment generated by KpnI cleavage of pUC-VNpCI~SVgpt DNA (in lane 5). Since there is no recombination at R3, the 12-9 kb KpnI restriction frgment in the parental DNA, which is derived from the KpnI sites in R2 and R3, is preserved in the A4 DNA. Thus, bands of equal intensity are obtained fi'om the A4 DNA in Figure 6(b) that correspond to one copy of each of these fragments. Hybridization with the neo probe (Fig. 6(a)) and the gpt probe (data not shown) confirms that the fragments of approximately 7"3 kb and of 10"6 kb contain gpt but no neo sequence, that the 12-9 kb fragment contains neo but no gpt sequence, and that there is a predicted 2"5 kb fragment containing gpt sequence. BstXI cuts between the positions in the constant region sequence at which the neo and SVgpt genes are located in the target and transfected DNA, respectively. Therefore, recombination at the R1 repeat decreases the size of the fragment of 5"8 kb in the parental DNA, which represents the distance between the BstXI site in R1 and the BstXI site in the adjacent chromosomal DNA, by the size of the neo squence to 3"8 kb. Recombination at the R1 and R2 repeats alters the distance between their BstXI sites in the parental DNA (12.9 kb) as a consequence of the gain of an SVgpt gene in R1 and deletion of the neo gene in R2, resulting in a net gain of 0"2 kb, to give a fi'agment identical in size to the 13"l kb full-length linear pUC-VNpC~tSVgpt (in lane 5). Recombination at R2 also increases the size of the parental 12"9 kb restriction fragment derived from the BstXI sites in R2 and R3, by the size of the SVgpt gene gained in R2, to 15"l kb. Since there is no recombination at the R3 repeat, the restriction fragment of 3"1 kb that represents the distance between the BstXI site in R3 and the adjacent BstXI site in the chromosomal DNA remains unaltered. Hybridizations with the neo probe (Fig. 7(a)) and the gpt probe (data not shown) confirm that the 15.1 kb restriction fragment contains neo and gpt

425

sequence, and that the 13"l kb restriction fragment contains gpt but no neo sequence. The restriction fragments of 5"8 kb and 3"8 kb are detected with the C# probe only weakly on account of their limited complementarity. The neo-specific probe, however, detects the 5"8 kb restriction fragment but not the 3"8 kb restriction fragment, as would be expected. On the basis of this analysis, we propose that the structure of the plasmid-derived DNA in the A4 cell line is as shown in Figure 4(b) with two of the three target sites containing identical rearrangements. These are consistent with the occurrence of double crossover events at these sites of the type shown in Figure 2 at positions 1 and 3. (ii) An IgM- revertant of A4 that has deleted the R2 copy Restriction enzyme digests from a derivative of the recombinant cell line A4 that has reverted to an IgM- phenotype are shown in Figures 5, 6 and 7 (lane 4). Colonies with this phenotype were identified using the haemolytic plaque assay technique with cells from the original A4 cell line, and from IgM + subclones derived from it. The frequency with which these occurred was approximately 10% in the population after expansion of the subclones through 30 to 40 generations. The structure of the plasmid-derived DNA in one of these IgM- segregants is shown in Figure 4(c). We postulate that this arrangement resulted from a secondary homologous recombination event between the R1 and R2 repeats (which have identical sequence), which deleted the R2 repeat (this could have been intrachromosomal or sister chromatid recombination). Thus, one copy of the 6"5 kb XbaI restriction fragment, the 10.6 kb KpnI restriction fragment and the 13"1 kb BstXI restriction fragment are lost in the IgM- clone. The initial characterization of the target DNA in the J558neoR1S cell line indicated that a substantial portion of the variable region segment of the R1 repeat had been lost as a consequence of the original integration event. Therefore, although the R1 repeat in the A4 recombinant contained a functional constant region, we speculated that, as a result, it was unlikely to express a # polypeptide. The fact that the secondary recombination event that deleted the R2 repeat correlated with the loss of the IgM + phenotype is consistent with this. We, therefore, believe that the IgM + phenotype of the A4 cell line was entirely the result of expression from the R2 repeat. (iii) An IgM + gpt- transformant derived by recombination at the R3 copy The recombinant cell line 3B2 was identified from the direct application of the haemolytic plaque assay after transfection. This clone did not grow in gpt selective medium, and it was therefore assumed to have resulted from a targeted event that either did not co-integrate the SVgpt gene or integrated it in a manner that rendered it non-functional. Figure 8(a) shows Southern blots of XbaI-restricted J588neoR1S and 3B2 cell DNAs hybridized with the

,

2

'o)

3

4

5

9 , 6 kb

r2.9 k b

I

2

(b)

3

4

5

7"3 kb

'.10.6 k

F i g u r e 6. Southern blot analysis of genomie DNAs of J558neoalS, the IgM÷gpt + transformant A4, and an IgM-gpt + derivative of A4, restricted with KpnI. DNAs (15 pg) were fractionated on a 0"7~o gel, transferred to a nylon membrane and (a) hybridized with the neo probe and (b) rehybridized with the C~ probe. Gel lanes as in Fig. 5. t H i n d I I I - c u t DNA was run adjacent to lane 1.

9.6kb -

2.9kb

I

(a)

.3

4

5

?

3-1kb

5"8 kb

15"1 kb 12.9 kb -

I

2

3

(b)

5

3-8 kb

15"lkb 13"1 kb

F i g u r e 7. Southern blot analysis of genomic DNAs of J558neoalS, the IgM+gpt + transformant A4, and an IgM-gpt + derivative of A4, restricted with BstXI. DNAs (15/~g) were fractionated on a 0"7 ~o gel, transferred to a nylon membrane and (a) hybridized with the neo probe and (b) rehybridized with the C# probe. Gel lanes as in Fig. 5. 2HindIII-cut DNA was run adjacent to lane 5.

5"Skb

2"9 kb

•

Io)

2

3

Ib)

4 5

4

.e

4"Skb

6'5kb

8 ' 2 kb

Figure 8. Southern blot analysis of genomic DNAs of J558neoR1S, the IgM+gpt - transformant 3B2 and an IgM+gpt + transformant 3C1, restricted with X b a I . DNAs (15/~g) were fraetionated on a 0-7 °/o gel, transferred to a nylon membrane and (a) hybridized with the neo probe and (b) rehybridized with the C~ probe. Lane l, J558L DNA plus l0 pg of pUC-VNpC/~neo plasmid; lane 2, J558neoR1S DNA; lane 3, 3B2 DNA; lane 4, 3C1 DNA; lane 5, J558L DNA plus 10 pg of pUC-VNpC/JSVgpt plasmid. ), H i n d I I I - c u t DNA was run adjacent to lane 5.

5..3 kb

8'Okb

2

Targeted Recombination in Myeloma Cells neo probe (lanes 2 and 3, respectively). The 6"3 kb restriction fragment is detected in the 3B2 cell line but not the fragment of approximately 8"0 kb derived from the R3 repeat in the parental cell line. Thus, either the entire repeat has somehow been lost, or its neo sequence has been deleted. Hybridization with the Cp probe (Fig. 8(b)), however, detects a new restriction fragment in 3B2 with a size of approximately 6"0 kb, in addition to the 6"3 kb restriction fragment. The relative intensity of the hybridization signals from these is 1 : 2. This novel restriction fragment detectable with the C# probe could have arisen from random integration of DNA, but this fragment is not detected with the gpt probe (data not shown). Furthermore its size, approximately 2"0 kb less than the XbaI restriction fragment derived from the R3 repeat, is consistent with the loss of the entire neo sequence from this repeat, without integration of the SVgpt gene. We therefore postulate that the arrangement of DNA in 3B2 is as shown in Figure 4(d), in which the R1 and R2 repeats have remained unaltered and the R3 repeat has been the target for a recombination event, which occurred as shown in Figure 2, with crossovers at positions 1 and 2. This arrangement is confirmed from Southern blot analysis of the fragment sizes found in KpnI and BstXI restriction endonuclease digests (data not shown). In each digest, a new 10"9 kb restriction fragment is detected resulting from the 2'0 kb deletion, and there is only a single copy of the 12"9 kb restriction fragment. In addition, the parental 3"1 kb BstXI restriction fragment is still present in 3B2, proving that apart from the deletion of the neo gene the remainder of the R3 repeat and its junction with the chromosomal DNA is unaltered. (iv) An IgM + gpt + transformant derived by recombination at the R3 copy The recombinant cell line, 3C1, was isolated using the direct assay, but was able to grow under gpt selection. Southern blot analysis of XbaI-restricted 3C1 DNA after hybridization with the neo probe is shown in Figure 8(a), lane 4. As in the case of the 3B2 cell line, only the 6"3 kb restriction fragment is detected, although with a reduced relative signal intensity. Hybridization with the Cp probe (Fig. 8(b)) detects the same 6"3 kb restriction fragment and a novel restriction fragment with a slightly larger size than the approximately 8"0 kb restriction fragment in the parental DNA. Southern blot analysis of BstXI-restricted 3C1 DNA hybridized with the Cp probe (data not shown) shows that the 3"1 kb restriction fragment derived from the 1~3 repeat and flanking DNA is absent, but that a novel restriction fragment of approximately 5-3 kb is present. These data are consistent with the R3 repeat having undergone a double crossover (as shown in Fig. 2 occurring at positions 1 and 3) that results in removal of the neo sequence and acquisition of the SVgpt gene. This would therefore increase the size of the 3.1 kb BstXI restriction fragment by 2-2 kb (the size of the SVgpt gene) and

429

the approximately 8"0 kb XbaI restriction fragment by 0"2 kb (the size difference between the SVgpt and neo genes). As in the 3B2 recombinant, a 10"9 kb restriction fragment is detected, but there is no BstXI restriction fragment with a size of 12-9 kb. This, taken together with the 1:1 ratio of signal intensities of the 6"3 kb XbaI and approximately 8"2 kb XbaI restriction fragments indicates that, in addition to the targeted event, there has been a deletion of an entire repeat similar to the situation in the A4 IgM- revertant. We postulate that the arrangement of plasmid-derived DNA in 3C1 is as shown in Figure 4(e), in which the R1 repeat has remained unaltered, the R2 repeat has been deleted, and the R3 repeat has undergone the recombination event described above. Southern blot analysis of KpnI-restricted 3C1 DNA hybridized with the Cp probe (data not shown) detects a 9"6 kb and 10-6 kb restriction fragment, which confirms this. (f) Relationship between homology lengths and targeting efftciency To evaluate the influence on recombination frequency of different lengths of homololgy between the transfected DNA and target sequence, a series of restriction fragments was used derived from the plasmids pUC-C#SVgpt2 and pUC-VA2C#SVgpt. These contained either a reduced amount of constant region sequence relative to the 6"5 kb XbaI fragment from pUC-CttSVgpt2, and/or additions of variable region and pUC sequence. The restriction fragments are shown in Figure 9 aligned with the pUC-VNpCttneo plasmid as present in the R2 repeat, to illustrate the lengths of homology and positions of sequence divergence with the target. The direct screening assay was used for these experiments, since it is rapid and gives more reproducible results than the procedure involving gpt selection. J558neoR1S cells (10 T) were e[ectroporated with equivalent molar masses of these restriction fragments (see Table 1) and, after leaving the cells in medium plus serum for 48 hours, they were plated in soft agar with NIP-coupled SRBCs. Figure 10(a) and (b) shows an example of the plaques that are obtained on addition of complement 48 hours after plating in soft agar cells from transfections with the 6"5 kb XbaI fragment and the 11"5 kb EcoRI of pUC-VACgSVgpt, respectively. Several transfections were carried out with each restriction fragment and the efficiency of targeting was calculated from the ratio of the total number of plaques identified to the number of cells plated. The results are shown in Table 1. The targeting efficiency increases from values of 3"0 x 10 -7 to 11.0 x 10 -7 obtained with the 5"9 kb SacI-XbaI and 8-6 kb SacI restriction fragments, to values of 0-7 x l0 -s to 2-5 x I0 -s obtained with the 8-8 kb SalI-EcoRI and 11"5 kb EcoRI fragments. An assessment of the contributions made by different segments of homology on the transfeeted DNAs to the efficiency of targeting is possible from the data shown in Table 1. A targeting index was calculated

A. J. H. Smith and B. Kalogerakis

430

Total homology

(kb)

T

(el

7-4

5.1

(b)

I

X

!

(c)

4.9

so! !

1,5

1.9

~,

r I! !

I X I

2"3

X,S I

E

I

0 •9 , ,1"1

4.6

wI

7.6

(d)

~:

:I

s¢

(e)

7.8

(t)

10.1 E

E

Figure 9. Restriction fragments from the plasmids pUC-C~SVgpt2 and pUC-VA2C~SVgpt. Restriction fragments are aligned with the pUC-VNeCpneo plasmid (as is present in the R2 repeat). (a) The 8"8 kb EcoRI-SacI fragment from pUC-VA2CpSVgpt; (b)6"5 kb XbaI fragment from pUC-CpSVgpt2; (c)5"9 kb SacI-XbaI fragment from pUCCgSVgpt2; (d) 9"0 kb SacI fragment from pUC-CpSVgpt2; (e) 9-2 kb EcoRI linear i~om pUC-C~SVgpt2; (f) 11-5 kb EcoRI linear of pUC-VA2CI~SVgpt. The regions of homo)ogy with the target and their size (in kb) is indicated. The total homology for each fragment is also given, which includes the additional homology between the neo and gpt sequences. There are 3 regions of sequence divergence between the restricton fragments and the target conferred by the 2'0 kb neo sequence, the 2-2 kb SVgpt sequence and the 4 base insertion between the 3' variable region and the constant region sequence. These are indicated by cross-hatched, striped and filled triangles, respectively. Open bar, variable region sequences; filled bar, constant region sequence; line, pUC sequence. Symbols for restriction endonuclease sites are as for Figure 1.

to correct the fluctuations arising from variations in the a m o u n t of DNA added to the cells, the competence of the cells, and their viability. The targeting index is based on the ratio of the efficiency of targeting to the efficiency of gpt + transformation in the same experiment. The targeting index provides a useful basis for comparison of experiments using restriction fragments derived from the same plasmid. 0 n l y an approximate comparison is possible between restriction fragments derived from pUCC]~SVgpt2 with those from pUC-VA2C~SVgpt, since the latter contains the immunoglobulin heavy chain

enhancer. In independent experiments, this was observed consistently to increase the efficiency of ~pt + transformation by 20 to 40-fold. Consequently, although restriction fragments derived from pUCVA2C~SVgpt give higher efficiencies of targeting, the ratio of this efficiency to the gpt + transformation efficiency (i.e. the targeting index) is lower than for restriction fragments derived from pUC-C~SVgpt2. The 5"9 kb SacI-XbaI restriction fragment from pUC-C~SVgpt2 contains only 0"2 kb of homology with the target sequence between one of its ends, at

Targeted Recombination in Myeloma Cells

431

(o)

(b)

Figure 10. Direct comparison of targeting efficiencies by haemolytic plaque assay. Soft agar wells containing NIPcoupled SRBCs, seeded with 5 x l0 s J558neoR1S cells electroporated with (a) the 6"5 kb XbaI restriction fragment and (b) the 11"5 kb EcoRI linear of pUC-VA2C/~SVgpt, on addition of complement 48 h later. The number of plaques on each well are 1 and 7, respectively.

A . J . H. S m i t h and B. Kalogerakis

432

Table

1

Values f o r the targeting index Targeting efficiency

Restriction fragment

% Plating efficiency

Corrected efficiency

ffpt + efficiency

Targeting index

5"9 kb SacI-XbaI (17/~g)

6"0x 4"0 x 9-0 x 7"0x

l0 -~ 10 -7 10 -7 10 -7

70 72 82 65

9'0x 6"0 x ll'0 x ll'0x AV = 9"0x

8'6 kb SacI (26/~g)

3-0 x l0 -7 5.0x l0 -v

86 80

3"0 x l0 -7 6 ' 0 x l0 -7 AV = 4.5x 10 -7

8-0 x l0 -e 7.0x l0 -6

3"8 x l0 -2 7.1 x l0 -2 AV = 5.4x 10 -2

6"5 kb XbaI (20/~g)

2.0x 1-4 x 2"8 x 1"7 x

2-2x 1-4 x 4'0 x 2"3 x AV = 2"5x

10 -6 10 -5 10 -5

3-5x 10 -5 3.2 x 10 - s 3'0 x 10 -5 3'1 x 10 - s

6-5x 4-4 x 9-3 x 5"5 x A V = 6-4x S.D. = 2"1 x

10 -2 10 -2 10 -2 10 -2 10 -2 l0 -2

4"4x 4'6 x 5"5 x 2-1 x A V = 4'1 x

l0 -5 l0 -5 l0 -6 l0 -6 10 -6

2'8x 2"4 x 7'0 x 6"0x

l0 -5 l0 - s l0 - s 10 - s

ll-I x 8-1 x 5-9 x 3"0x A V = 7"0 x S.D. = 3"4x

l0 -2 l0 -2 l0 -2 l0 -2 10 -2 l0 -2

9'0 x 10 -6 8"6 x 10 -6 11-5 x l0 -6 23-2x l0 -6 14"2 x l 0 - 6 A V = 1 3 ' 3 x l 0 -6

12"9 x 9'7 x 15'0 x 19"6x

10 -4 l0 -4 10 -4 l0 -4

4-8 x 5"1 x 6"1 x 9"6x

10 -3 l0 -3 10-3 10 -3

9"2 kb EcoRI (28/~g)

8"8 kb EcoRI-SalI (27/~g)

10 -5 10 -5 10 -5 10 - 6

89 103 70 73

3-1 x 1"9 x 4"1 x l'8x

10 -6 10 -5 l0 -5 10 -5

71 41 75 87

6"2 x 5"0 x 9-2 x 18"8x

10 -6 l0 -6 10 -6 l0 -6

69 58 80 81 80

I 1-4 x l 0 - 5

10 -7 10 -7 10 -7 10 -~ 10 -7

10 - 6

2"1 x 4"0 x 3-3 x 2-8x

l0 - s 10 -5 l0 -5 10 -5

10 - s

12-2 x l 0 - 4

2.9x 1.0 x 2.7 x 2.5x AV = 2.3x S.D. = 0.9 x

10 -2 l0 -2 10 -2 10 -2 l0 -2 10 -2

9'3 x 10 - 3

A V = 7 ' 0 x l 0 -a S.D. = 2"3 x 10 -3

11"5 kb EcoRI (35 ~ug)

13"4 X 1 0 - 6 8"2X 10 -6 6"2 X l0 -6 13"0 X 10 -5 14"0X 10 -6

54 65 90 79 74

24"8 X 10 -6 12"6X 10 -5 6'9X l0 -6 16"5 X i0 -5 18'9X l0 -6 A V = 15'gx 10 -6

7"3 X 10 -4 5"2X 10 -4 4"7 X l0 -4 8"2 X l0 -4 5"0X l0 -4

1'8 X 10 -2 l'6X lO -2 1'3 X l0 -2 1"6 X l0 -2 2"8X 10 -2 A I ' = l ' 8 x l0 -2 S.D. = 0"6 x 10 -2

11'5 kb uncut pUC-VA2CI~SVgpt (35/Jg)

8"0x 10 -7 12"0 x 10 -~

72 87

l l ' 0 x 10 -7 14'0 x 10 -7 A | I = 12-5x 10 -7

5"6x 10 -4 8"7 x 10 -4

i . 4 x l0 -3 1"4 x 10 -3 A V = 1.4x 10 -3

For each transfection, l0 T cells were electroporated with the a m o u n t of DNA shown, allowed to recover for 48 h. and the number of viable cells then determined (average approx. 2.0 x 10T). Cells were plated in soft agar with hapten-coupled SRBCs, and the targeting efficiency was determined from the ratio of the total n u m b e r of plaques identified after addition of complement 48 I1 later, to tbe total number of cells plated (18 wells at 5 x 105 cells/well). The plating efficiency was measured from seeding 200 cells/well in soft agar and counting the number of colonies obtained after l0 days. The corrected (targeting) efficiency was obtained after adjustment for the plating efficiency; the average value for each restriction fragment is given. The efficiency of gpt + transformation was measured for each transfection by carrying out gpt selection in soft agar and determining the ratio of the n u m b e r of colonies obtained after l0 days to the n u m b e r of cells plated (5 x l0 s and 5 x 104 cells/well). For each transfection, we calculate a value,, defined as the targeting index, t h a t is the ratio of the targeting efficiency to the gpt + transformation efficiency. The average targeting index (A V) for each restriction fragment (with standard deviation (s.D.) were appropriate) is given. The targeting index is a relative value only, since the plating efficiency in agar with gpt selective medium is not known. However, an estimate of this can be obtained from repeating the same experiment with SalI-cut pUC-VNpCI~SVgpt. This gives an average of 4 plaques for every colony from gpt selection. Correcting for this gives, ibr example, a ratio of targeted integrations to total detectable integration events of 4'5 x l0 -3 for the 11.5 kb EcoRI restriction fragment.

t h e SacI s i t e , a n d sequence divergence

t h e p o s i t i o n o f t h e 2"0 k b o f i n t r o d u c e d b y t h e neo g e n e i n

targeting index for this restriction fragment is significantly higher, with an average value approxi-

the target. One of the crossovers must occur in this region of homology to correct the/~ gene, and this restriction fragment gives reproducibly the lowest

m a t e l y t h r e e t i m e s t h a t o b t a i n e d w i t h t h e 5"9 k b r e s t r i c t i o n f r a g m e n t . T h e 8"8 k b E c o R I - S a l I r e s t r i c -

frequency of recombination. In the case of the 6"5 k b X b a I r e s t r i c t i o n fragment, the region of homology with the

on this side of the t a r g e t is e x t e n d e d

sequence divergence b y 0"7 k b , a n d t h e

tion fragment extends the homology at this end f u r t h e r , b y a n a d d i t i o n a l 2-3 k b , a n d o n t h e b a s i s o f its average targeting index and correction for the 20 t o 4 0 - f o l d - e f f e c t o n gpt + t r a n s f o r m a t i o n efficiency introduced by the enhancer, this gives a 6 to 12-fold

Targeted Recombination in Myeloma Cells

increase over the frequency of recombination obtained with the 5"9 kb restriction fragment. By comparison, extending the homology present on the 5.9 kb S a c I - X b a I fragment, by an additional 2.7 kb on the other side of the position of sequence divergence introduced by the neo gene, has a much less significant effect. This assessment is based on the average targeting index calculated for the 8"6 kb SacI restriction fragment, which is approximately twice the value for the 5"9 kb S a c I - X b a I restriction fragment. Thus, it appears that the location of homology is important and, for the restriction fragments with approximately the same total amount of homology with the target site, an arrangement that leaves only 0"2 kb of homology on one side of a region of sequence divergence severely limits the efficiency of targeting. Although the 9"2 kb EcoRI restriction fragment contains an extra 2-7 kb of pUC sequence homologous to the target, its targeting index is not significantly greater than the value calculated for the 6"5 kb XbaI restriction fragment. This is in contrast to a comparison of the values of the targeting index obtained for the I 1"5 kb EcoRI and 8"8 kb EcoI~I-SalI restriction fragments, which are derived from the pUC-VA2CpSVgpt plasmid and contain 2"3 kb of variable region sequence homologous to the target at one end. In this case, the extra 2"7 kb of pUC sequence present in the 11"5 kb EcoFCI restriction fragment and located at the opposite end from the variable region sequence, results in an approximate threefold increase in the average targeting index. The effect on recombination frequency of adding aditional sequence with homology to the target to one end of the exogenous linear DNA therefore appears to be influenced by the extent of homology on the other side of the position of sequence divergence with the target site. A requirement for linearized DNA in order to achieve high targeting frequencies was also observed; the targeting index obtained with uncut pUC-VA2CpSVgpt DNA, was tentbid lower than when the plasmid was cut with EcoRI. 4. Discussion

An experimental system was designed to determine the frequency of targeted integration of exogenous DNA in antibody-producing cells. Although the present experiments involve detecting integration at an artificial target, the basic design of the replacement vector should be applicable for targeting into endogenous constant region genes. Replacement events allow specific chromosomal alterations to be made without incorporation of unnecessary plasmid vector sequences, and duplication of the target site, which can result in instability of the inserted DNA from secondary homologous recombination events (Jasin & Berg, 1988). To achieve replacement recombination, the genetic marker used to detect the transformants is flanked on each side by sequences homologous to the target site, and in our construct is contained within an

433

intron sequence of the p constant region, which is not required in a splicing event in IgM-secreting cells. Incorporation of the SVgpt gene at this position in the # gene does not interfere with normal expression. The results of the experiments described here are consistent with previous studies of homologous recombination in different mammalian cell types (Smith & Berg, 1984; Lin et al., 1985; Smithies et al., 1985; Thomas et al., 1986; Doetschman et al., 1987, 1988; Song et al., 1987; Thomas & Cappechi, 1987; Baker et al., 1988). The ratio of targeted to random integrations (10 -2 to 5 × l0 -3) is within the same range as previously reported, and we have confirmed that the frequency of recombination is enhanced by linearization of the DNA, and that no random integration of DNA is apparent in cells in which a targeted event has taken place. However, it has not been previously demonstrated that a defined deletion can he introduced at a specific chromosomal site. In addition, we have shown that the replacement event that corrects the target gene by removing the neo disruption can also result in acquisition of new sequence (the SVgpt gene) at a separate position in the target site. This is necessarily the case in clones identified with the IgM + phenotype, which were obtained after first selecting for gpt + transformants. However, in IgM + transformants detected by the direct application of the haemolytic plaque assay, a large fraction of the clones, when subsequently screened, were found also to have acquired the gpt + phenotype, and analysis of DNA from one of these showed that the SVgpt gene was present at the target site that had lost the neo disruption. Thus, in these experiments, although crossovers occurred within the regions of homology directly flanking the position of the neo sequence, resulting in its removal without acquisition of the SVgpt gene, a significant fraction probably occurred in both terminal regions of homology and co-converted unselected sequence (i.e. the SVgpt gene). A surprising and novel result we have obtained concerns the structure of the target DNA in the recombinant A4. Two copies of the repeated target site contain identical sequence rearrangements, in which the neo disruption has been deleted and an SVgpt gene inserted, with the remaining copy left unaltered. This structure could be accounted for by two independent integrations, although we would anticipate the probability of such an event to be extremely low. Integration of two copies of a transfected DNA at a site of homology has been detected in a target integration event in embryonal stem cells (Thompson et al., 1989). However, in that case, these were insertion vectors, and it would be possible to explain this arrangement as a result of recombination between two copies of the transfected plasmid, producing a dimer, before integration into the chromosome. In the situation described here, any form of recombination between transfected restriction fragments before integration could not explain the structure of DNA in the A4

434

A. J. H. Smith and B. Kalogerakis

cell line. The alternative explanation is that there was a single integration event at one of the repeats and the functional constant region and SVgpt sequences were then donated to the neighbouring repeat by gene conversion. Gene conversion of contiguous blocks of sequence between adjacent homologous chromosomal sequences has been demonstrated by Liskay & Stachelek (1986) in mouse L cells, and in their experiments occurred at frequencies of approximately l0 -6 per generation. We can speculate that recombination occurred at the R1 repeat first, thereby giving rise to a cell with a gpt + phenotype, and that during its propagation in culture the functional constant region and SVgpt sequences were converted onto the R2 repeat. Since the R1 repeat is incapable of producing a functional p01ypeptide, only those cells in which this second event had taken place would have been detected in the screen for IgM production. We are currently analysing the structure of the target DNA in other recombinant cell lines to see if similar multiple recombination events can be identified. Thomas & Cappechi (1987) observed a marked increase in recombination frequency with increasing lengths of homololgy between the exogenous and target DNA sequence. The same effect can be seen in our experiments from comparisons made between some of the restriction fragments, shown in Figure 9, on the basis of their targeting index; the value of the targeting index gives a relative measurement of recombination frequency, which corrects for inherent variabilities in the transfections, and appears to be reproducible. It is apparent from our experiments, however, that the position and context in which the extra homology is incorporated into the exogenous DNA is also important. Thus, the contribution of additional homology is dependent on its location with respect to the position of sequence divergence between target and transfected DNA sequences, and on the amount of homology with the target sequence already present. However, these complicated effects deserve further investigation, and it is possible that they may only be peculiar to the nature of the target site we have used. The 2.3 kb of variable region sequence included in some of the transfected restriction fragments contains the heavy chain enhancer, which elevates the level of gpt transcription and therefore the gpt + transformation efficiency. Thus, although this provides additional homology and increases the efficiency of targeted recombination, the net result is a decrease in the targeting index. The decrease in the relative number of targeted events to gpt + transformants is also apparent when cells electroporated with the E c o R I linear of pUC-VA2C#SVgpt are plated in soft agar with gpt selective medium, and the gpt + colonies obtained after ten days overlaid with hapten-coupled SRBCs. Approximately one in 300 to 400 gpt + colonies are then identified as IgM +, compared to one in 100 to 200 gpt + colonies after transfection with the 6.5 kb X b a I restriction fragment. Therefore, although the efficiency of

targeting with the 6"5 kb X b a I restriction fragment is lower, transfection with this DNA gives an effective enrichment of targeted events among the 9pt + transformants. This effect can be accounted for by the lower level of transcription obtained in the absence of the enhancer; the consequence of this is that there are probably fewer random integrations of exogenous DNA that result in sufficient expression to give the gpt + phenotype, whereas the targeted event juxtaposes the SVgpt gene to a site proximal to the enhancer. For the purposes of targeting into endogenous genes, consideration of the extent of homology incorporated into the construct will therefore depend on whether there are transcriptional elements present that will raise the frequency of stable transformation resulting from random integrations. Our long-term objective is to use homologous recombination to make specific alterations to immunoglobulin genes in myeloma and hybridoma cell lines. This technology could be of use in the construction of novel immunoglobulin genes to produce chimearic antibody molecules, and in the manipulation of sequences to aid in the understanding of mutation and gene expression at immunoglobulin loci. The results presented here indicate that this approach will be feasible, and that unselectable sequence changes could be incorporated into chromosomal sites by the design of suitable replacement vectors, thereby facilitating extensive restructuring of genes. Although the frequency of co-conversion of an unselectable sequence into the chromosomal DNA will presumably depend on its distance on the exogenous DNA from the location of the genetic marker used to select the transformants, the high frequency of cointegration of the unselected SVgpt marker obtained in recombinants isolated with the direct assay is encouraging. This suggests that this distance can be of the order of several thousand bases, provided that there is an equivalent length of homology between its position and the end of the restriction fragment. The homologous recombination assay described herein is rapid and reproducible. We expect it will be of use in evaluating other aspects of vector design that may affect the frequency of targeted integration and gene conversion. The authors thank M. Neuberger for many helpful discussions and gifts of DNA, C. Milstein and R. Pannell for advice concerning haemolytic plaque assays, W. Ansorge and R. Peppercock for instruction and use of the EMBL/Apelex electroporation apparatus and J. Karn for critical reading of the manuscript.

References

Baker, M. D., Pennell, N., Bosnoyan, L. & Shulman, M. J. (1988). Proc. Nat. Acad. Sci., U.S.A. 85, 6432-6436. Doetschman, T., Gregg, 1%.G., Maeda, N., Hooper, M. L., Melton, D. N., Thompson, S. & Smithies, O. (1987). Nature (London), 330, 576 578.

Targeted Recombination in Myeloma Cells Doetschman, T., Maeda, N. & Smithies, O. (1988). Proc. Nat. Acad. Sci., U.S.A. 85, 8583-8587. Jasin, M. & Berg, P. (1988). Genes Develop. 2, 1353-1356. Kohler, G. & Milstein, C. (1976). Eur. J. Immunol. 6, 511-519. Kohler, G., Howe, S. S. & Milstein, C. (]976). Eur. J. Immunol. 6, 292-295. Lin, F. L., Sperle, K. & Sternberg, N. (1985). Proc. Nat. Acad. Sci., U.S.A. 82, 1391-1395. Liskay, R. M. & Stachelek, J. L. (1986). Proc. Nat. Acad. Sci., U.S.A. 83, 1802-1806. Mansour, S. L., Thomas, K. R. & Capecchi, M. R. (1988). Nature (London), 336, 348-352. Mulligan, R. C. & Berg, P. (1981). Proc. Nal. Aead. Sci., U.S.A. 78, 2072-2076. Neuberger, M. S. (1983). EMBO J. 2, 1373-1378. Neuberger, M. S. & Rajewsky, K. (1981). Proc. Nat. Acad. Sci., U.S.A. 78, 1138-1142. Neuberger, M. S., Williams, G. T. & Fox, R. O. (1984). Nature (London), 312, 604-608. Neuberger, M. S., Williams, G. T., Mitchell, E. B., Jouhal, S.S., Flanagan, J . G . & Rabbitts, T . H . (1985). Nature (London), 314, 268-270. Oi, V. T., Morrison, S. L., Herzenberg, L. A. & Berg, P. (1983). Proc. Nat. Acad. Sci., U.S.A. 80, 825-829. Peabody, D. S. & Berg, P. (1986). Mol. Cell. Biol. 6, 2695-2703.

435

Rigby, P, W. J., Dieckmann, M., Rhodes, C. & Berg, P. (1977). J. Mol. Biol. 113,237-247. Smith, A. J. H. & Berg, P. (1984). Cold Spring Harbor Syrup. Quant. Biol. 49, 17 l - 18 I. Smithies, 0., Gregg, R. G., Boggs, S. S., Koralewski, M.A. & Kucherlapati, R.S. (1985). Nature (London), 317, 230-234. Song, K. Y., Schwartz, F., Maeda, N., Smithies, O. & Kucherlapati, R.S. (1987). Proc. Nat. Acad. Sci., U.S.A. 84, 6820-6824. Southern, E. M. (1975). J. Mol. Biol. 98, 503-517. Southern, P. J. & Berg, P. (1982). J. Mol. Appl. Genet. 1, 327-341. Thomas, K. R. & Capecchi, M. R. (1986). Nature (London), 324, 34-38. Thomas, K. R & Capecchi, M. R. (1987). Cell, 51, 503-512. Thomas, K. R., Folger, K. R. & Capecchi, M. R. (1986). Cell, 44, 419-428. Thompson, S., Clarke, A. R., Pow, A. M., Hooper, M. L. & Melton, D. W. (1989). Cell, 56, 313-321. Vogelstein, B. & Gillespie, D. (1979). Proc. Nat. Acad. Sci., U.S.A. 76, 615-619. Yanisch-Perron, C., Vieira, J. & Messing, J. (1985). Gene, 33, 103-119.

Edited by N. Sternberg