A Highly Efficient Escherichia coli-Based Chromosome Engineering
System Adapted for Recombinogenic Targeting
and Subcloning of BAC DNA
E-Chiang Lee,* Daiguan Yu,† J. Martinez de Velasco,* Lino Tessarollo,* Deborah A. Swing,*
Donald L. Court,† Nancy A. Jenkins,* and Neal G. Copeland*,1
*Mouse Cancer Genetics Program and †Gene Regulation and Chromosome Biology Laboratory, National Cancer Institute–Frederick,
Frederick, Maryland 21702
Received November 1, 2000; accepted November 16, 2000
R ecently, a highly efficient recombination system
for chromosome engineering in E scherichia coli was
described that uses a defective ? prophage to supply
functions that protect and recombine a linear DNA
targeting cassette with its substrate sequence (Yu et
al., 2000, Proc. Natl. Acad. Sci. USA 97, 5978–5983).
Importantly, the recombination is proficient with DNA
homologies as short as 30–50 bp, making it possible to
use PCR -amplified fragments as the targeting cassette.
Here, we adapt this prophage system for use in bacte-
rial artificial chromosome (BAC) engineering by trans-
ferring it to DH10B cells, a BAC host strain. In addi-
tion, arabinose inducible cre and flpe genes are
introduced into these cells to facilitate BAC modifica-
tion using loxP and F R T sites. Next, we demonstrate
the utility of this recombination system by using it to
target cre to the 3? end of the mouse neuron-specific
enolase (E no2) gene carried on a 250-kb BAC, which
made it possible to generate BAC transgenic mice that
specifically express Cre in all mature neurons. In ad-
dition, we show that fragments as large as 80 kb can be
subcloned from BACs by gap repair using this recombi-
nation system, obviating the need for restriction en-
zymes or DNA ligases. F inally, we show that BACs can be
modified with this recombination system in the absence
of drug selection. The ability to modify or subclone large
fragments of genomic DNA with precision should facili-
tate many kinds of genomic experiments that were dif-
ficult or impossible to perform previously and aid in
studies of gene function in the postgenomic era.
Bacterial artificial chromosomes (BACs) (Shizuya et
al., 1992) have become the tool of choice for generating
long-range physical maps, positionally cloning disease
genes, and whole-genome sequencing due to their abil-
ity to be stably propagated. However, for them to be-
come an equally powerful tool for functional genomics,
easy methods for precisely manipulating BAC DNA are
required. A significant advance in our ability tomanip-
ulate DNA was provided by Zhang et al. (1998), who
showed that it is possibletomodify DNA in Escherichia
coli by homologous recombination using the RecET
proteins of E. coli. This methodology makes it possible
to target PCR-amplified linear DNA fragments with
short regions of homology (?60 bp) at their ends to
virtually any target DNA such as a high-copy plasmid,
the E. coli chromosome, or a BAC (Muyrers et al., 1999;
Zhang et al., 1998).
Recently, an alternative approach for chromosome
engineering in E. coli was described that makes use of
a defective prophage to supply functions that protect
and recombine the electroporated linear DNA (Yu et
al., 2000). This defective prophage carries a deletion
between cro and bioA (Yu et al., 2000). The PL operon
encoding gam and the red recombination genes, exo
and bet, is under the tight control of the temperature-
sensitive ? repressor (allele cI857). Recombination
functions can thus be transiently supplied by shifting
the cultures to 42°C for 15 min. This recombination
system does not require RecA function and depends
primarily on the expression of Exo, Beta, and Gam.
Gam inhibits the E. coli RecBCD nuclease from attack-
ing the electroporated linear DNA, while Exoand Beta
generate recombination activity.
? recombination functions expressed from a plasmid
havebeen used previously tomodify a BAC (Muyrers et
al., 1999). However, the ? prophage system appears to
be at least 50- to 100-fold more efficient than this, or
the RecET system, where recombination functions are
expressed from plasmids (Muyrers et al., 1999; Nara-
yanan et al., 1999; our unpublished results). Express-
ing the ? recombination functions from plasmids also
creates other problems not presented by the prophage.
1To whom correspondence should be addressed at the Mouse
Cancer Genetics Program, Building 539, Room 229, National Cancer
Institute–Frederick, Frederick, MD 21702. Telephone: (301) 846-
1260. Fax: (301) 846-6666. E-mail: email@example.com.
Genomics 73, 56–65 (2001)
doi:10.1006/geno.2000.6451, available online at http://www.idealibrary.com on
Copyright © 2001 by Academic Press
All rights of reproduction in any form reserved.
For example, drug resistance is used to maintain plas-
mids in the cell. For experiments using gene replace-
ment, drug-resistant markers often become rate-limit-
ing. A plasmid-based system also precludes using that
particular plasmid type for recombination cloning
studies because of incompatibility. Importantly, plas-
mid expression systems are leaky; thus, Gam and Red
functions are always present at some level. The pres-
ence of Gam causes a RecBCD defect, a condition that
results in plasmid instability and loss of cell viability
(Feiss et al., 1982; K. Sergueev, unpublished results).
Gam and Red can also cause BAC instability.
Here, this prophage system is modified for use in
BAC engineering, and the utility of this system for
recombinogenic targeting and subcloning of BAC DNA
is demonstrated. The ability tomodify large fragments
of genomic DNA with precision should facilitate stud-
ies of gene function in the postgenomic era.
MATERIALS AND METHODS
DH10B, weremaintained at 32°C becauseof thetemperature-inducible
prophage. DY303 was constructed by infecting DH10B cells (Gibco)
with a ? phage carrying recA (?cI857 recA?) (a gift from F. W. Stahl),
and lysogens were selected. Strain EL11 was constructed by replacing
thetetgeneof DY380 with a cassettecontaining thecat and sacB genes.
EL11 cells are TetS, CmR, and sensitive to 2% sucrose. Strain EL250
was constructed by replacing the cat–sacB cassette of EL11 cells with
araC and the arabinose promoter-driven flpe recombinase gene
(PBADflpe). EL250 cells are resistant to 2% sucrose. Strain EL350 was
constructed in a similar manner except that cre replaced flpe.
All of the strains used in this study, except
Construction of plasmids.
(kan)–FRT targeting cassette was PCR-amplified from pICGN21,
which was constructed by subcloning a 1.9-kb HindIII/AccI-digested
and filled-in FRT–kan–FRT fragment from pFRTneo into the NotI/
BclI-digested and filled-in cloning site of pIRESeGC. The FRTneo
was constructed by amplifying the kan gene along with the ?-lacta-
TGGCACTTTCGGG and 5?CTCAGAAGAACTCGTCAAGAAGG. The
amplified fragment was then targeted between theFRT sites in pNeo?-
gal (Stratagene). The pIRESeGC was generated by inserting the 2-kb
NheI/MluI-digestedandfilled-in eGFPcrefragment frompEGC intothe
3.5-kb BamHI-digested and filled-in cloning site of pNTRlacZPGKne-
oloxP (Arango et al., 1999). The pEGC was generated by subcloning a
1.05-kb EcoRI/KpnI PCR fragment containing the cre gene from
pGKmncre (a gift from P. Soriano) intothe EcoRI/KpnI site of pEGFP-
C1. This PCR fragment was generated by amplifying the cregene from
CGTACACC, which contain EcoRI and KpnI cleavage sites, respec-
tively, at their 5? ends.
To construct the pTamp vector, the amp-targeted pBeloBAC11
was first generated by replacing the loxP site in pBeloBAC11
(Shizuya et al., 1992) with the PCR-amplified amp gene from pEGFP
(Clontech). The primers used for amplification are 5?GCAAG-
TTTTGGTC, which are homologous to the amp gene of pEGFP (in
roman type) and to sequences flanking the loxP site in pBeloBAC11
(in italics). A 2.4-kb PCR fragment amplified from amp-targeted
pBeloBac11 with primers 5?GCAGGATCCAGTTTGCTCCTGGAGC-
GACA and 5?TGCAGGTCGACTCTAGAGGATC was then cloned
intothe XhoI/XbaI and filled-in site of pCS (Stratagene) tocreate the
pTamp vector. The 2.4-kb amp cassette containing an amp gene
along with 920 bp of 5?, and 370 bp of 3?, pBel0BAC11 vector se-
quence flanking the loxP site can be released by BamHI digestion
and used directly toreplace the loxP site in any pBeloBAC11-derived
BACs with amp.
The pKO4 vector containing the cat–sacB targeting cassette is a
derivative of pKO3 (Link et al., 1997) in which 605 bp between cat
and sacB had been deleted.
pBADflpe, which was constructed by subcloning a 1.4-kb PstI/KpnI
fragment from pOGFlpe (Buchholz et al., 1998) intopBAD/MycHis-A
(Invitrogen). The araC–PBADcre targeting cassette was amplified
from pBADcre, which was constructed by introducing a 1.2-kb
HindIII/NcoI fragment from pGKmncre into pBAD/His-C (Invitrogen).
cassette was amplifiedfrom
Amplification primers for targeting or GAP repair cassette DNAs.
For all primers listed below, nucleotides in italics arehomologous tothe
targeted sequence, while those in roman type are homologous toampli-
fication cassettes. TheTetRcassetteusedfor targeting cro–bioin DY330
was amplified from Tn10 with primers 5?TGGCGGTGATAATGGTTG-
CGACATCTTGGTTACCG. The cat–sacB cassette used for replacing
the tet gene in DY363 was amplified from pKO4 with primers
TGACGGAAGATCACTTCG and 5?GGCGCTGCAAAAATTCTTTGTC-
araC–PBADflpe and araC–PBADcre cassettes used for replacing the
cat–sacB in E L11 were amplified from pBADflpe and pBADcre
with primers 5?TGGCGGTGATAATGGTTGCATGTACTAAGGAGG-
AGGAGAGCGTTC. The IRES–eGFPcre–FRT–kan–FRT
used for targeting the Eno2 locus was amplified from pIGCN21 with
GTAGTGAGGA. The oligonucleotides used totarget the flag cassette
into the 5? end of the Sox4 gene were annealed and polymerase-
extended using primers5?GCGAGCGTGTGAGCGCGCGTGG-
AATCCATGGCCCCC. The linear pBR322 derivative used to sub-
clonethe25-kb fragment from themodified Eno2 locus was amplified
with primers 5?CTCTCCATGCCTGTCTGGGTGAGGGTGGCCCA-
TGGCAC (Eno2-C-L1) and GCAATGCAGAGAAGCCTTGTACTGG-
GAT GACAGAGACGGAGGGGAAGAGGAGGCGGCCGCGAT A-
CGCGAGCGAACGTGA (Eno2-C-R1/2). The amplification primers
for the other experiments were as follows: 48-kb modified fragment,
C-L2/3/4) and Eno2-C-R1/2; 60-kb modified fragment, Eno2-C-L2/3/4
(Eno2-C-R3); 80-kb modified fragment, Eno2-C-L2/3/4 and 5?CAT-
TGACAGGCGGCCGCGATACGCGAGCGAACGTGA (E no2-C-R4).
These primers contained 5? regions homologous to the target
sequence and 3? regions homologous to pBR322. PCR products
were purified using a Qiaex II gel extraction kit (Qiagen) and
digested with DpnI to remove contaminated template.
Preparation of electrocompetent cells and generation of recombi-
nants.For BAC modification, overnight cultures containing the
BAC were grown from single colonies, diluted 50-fold in LB medium,
and grown to an OD600? 0.5–0.7. Ten-milliliter cultures were then
induced for Beta, Exo, and Gam expression by shifting the cells to
42°C for 15 min followed by chilling on icefor 20 min. Cells werethen
centrifuged for 5 min at 5500g at 4°C and washed with 1.5 ml of
ice-cold sterile water three times. Cells were then resuspended in 50
E. coli-BASED SYSTEM FOR BAC ENGINEERING
?l of ice-cold sterile water and electroporated. For BAC transforma-
tion, the induction step was omitted.
Cell transformation was performed by electroporation of 100–300
ng linear DNA into 50 ?l of ice-cold competent cells in cuvettes (0.1
cm) using a Bio-Rad gene pulser set at 1.75 kV, 25 ?F with a pulse
controller set at 200 ohms. Onemilliliter of LB mediumwas addedafter
electroporation. Cells wereincubatedat 32°C for 1.5h with shaking and
spread on appropriate selective or nonselective agar media.
Production of transgenic mice.
subclone DNAs were purified using cesium chloride gradients as
described (Antoch et al., 1997). The 25-kb subclone DNA was linear-
ized by NotI digestion before microinjection. BAC DNA (1 ?g/ml) and
25-kb subclone DNA (2 ?g/ml) were microinjected into the pronu-
cleus of (C3H/HeN-Mtv?? C57BL/6Ncr)F2 zygotes. Transgenic
founders were subsequently identified by Southern analysis using a
cre probe or by PCR using primers 5?CTGCTGGAAGATGGCGAT-
TCTCG and 5?AACAGCAGGAGCGGTGAGTC that flank the 3? in-
Modified BAC and the p25-kb
Histochemical analysis of ?-galactosidaseexpression.
5 weeks of age were sacrificed in CO2and perfused with 4% para-
formaldehyde in PBS (pH 7.3). The brains, spinal cords, and eyes
were removed and postfixed for 3 h. Vibratome sections (20 ?m) of
brains were mounted on slides and used directly for X-gal staining or
for immunocytochemistry. For spinal cords and eyes, cryostat sec-
tions (20 ?m) were used that were made by cryoprotecting tissues in
30% sucrose in PBS overnight and embedding the tissues in freezing
compound (OCT, Sakura). Before X-gal staining, samples on slides
Miceat 4 to
F IG. 1.
int. PL and PR denote the left and right promoters, respectively. The gam and red genes, exo and bet, are under the control of PL, which is
repressed by the temperature-sensitive repressor cI857 at 32°C and derepressed at 42°C. tet replaces the segment from cro–bioA in DY380
cells. The araC–PBADflpecassette or the araC–PBADcrecassette replaces the segment from cro–bioA in EL250 or EL350 cells, respectively. The
promoter of the araBAD operon (PBAD), which can be induced by L-arabinose, controls the expression of the flpeor cregenes. Thick black lines
designate the prophage while thin lines represent E. coli sequence. Angle brackets define the ends of the cro–bioA region that was replaced
with tet, araC–PBADflpe, or araC–PBADcre. (B) The relative position of the Eno2 gene in the fully sequenced 250-kb BAC, 284H12, and the
different steps used to introduce Cre into the last exon of Eno2. In the targeting cassette, FRT sites are denoted by orange ellipses, the kan
gene is represented by a red rectangle, and the GFPcrefusion gene is represented by a blue rectangle. The green boxes represent Eno2 exons.
General strategy for BAC engineering. (A) The defective prophages used for BAC engineering contain the ? genes from cI857 to
T ABL E 1
Bacterial Strains Constructed in T his Work
DH10BF?mcrA ?(mrr–hsdRMS–mcrBC) ?80dlacZ?M15
?lacX74 deoR recA1 endA1 araD139 ?(ara,
leu)7649 galU galK rspL nupG
DH10B [?cl857 recA?]
W3110 ?lacU169 gal490 [?cl857 ?(cro–bioA)]
W3110 ?lacU169 gal490 [?cl857 (cro–bioA) ?? tet]a
DH10B [?cl857 (cro–bioA) ?? tet]
DH10B [?cl857 (cro–bioA) ?? cat–sacB]
DH10B [?cl857 (cro–bioA) ?? araC–PBADflpe]b
DH10B [?cl857 (cro–bioA) ?? araC–PBADcre]
a(cro–bioA) ?? tet indicates substitution of cro–bioA with tet.
bPBADrepresents the promoter of araBAD.
LEE ET AL.
werepostfixed with 0.25% glutaraldehydein PBS and briefly washed
with rinsesolution (0.1 M phosphatebuffer, pH 7.3, 0.1% deoxycholic
acid, 0.2% NP-40, and 2 mM MgCl2). X-gal staining was performed
by incubating samples in staining buffer (2.5 mg/ml X-gal, 5 mM
potassium ferricyanide, and 5 mM potassium ferrocyanide in stain-
ing buffer) for 2 h at 37°C followed by counterstaining with 0.25%
ABC Vectastain kit (Vector Labs) on 20-?m vibratome sections. Sec-
tions were blocked with PBS (pH 7.3; containing 0.2% Triton X-100,
1.5% bovine serum albumin, and 5% normal goat serum) at room
temperature for 2 h and incubated with primary Eno2 antibody, a
polyclonal rabbit anti-Eno2 antiserum (Chemicon) at 1:100 dilution in
PBS solution. After incubation with a secondary biotinylated antibody
and the ABC reagent, peroxidase was reacted with 0.05% diamino-
benzidine tetrahydrochloride and 0.003% hydrogen peroxide.
Immunostaining was carried out using the
Transfer of the Defective ? Prophage
into DH10B Cells
None of the E. coli strains carrying the defective ?
prophage generated in previous experiments can be
efficiently transformed with BAC DNA (Yu et al.,
2000), precluding their use for BAC engineering. To
facilitate its use for BAC engineering, the prophage
was introduced into DH10B cells, a BAC host strain.
The prophage could not, however, be transferred by
standard genetic methods because DH10B is recA?. To
circumvent this problem, a tet selectable marker was
introduced between cro and bioA in the defective ?
prophagecarried in strain DY330 (Table1). A P1 lysate
was then prepared on this new strain (DY363; Table 1)
and used to infect DY303 cells, a DH10B strain that
carries a ? cI857recA? lysogen (Table 1). By selecting
cells that are recA?based upon their UV sensitivity, it
was possible to obtain cells in which tet replaced the
recA?gene in the ? cI857recA? lysogen. This new
strain, DY380 (Table 1), can be transformed with
BAC DNA at efficiencies of 10?6to 10?4(data not
shown). The targeting efficiency of DY380 cells is
also similar to that of the original recA?DY strain (Yu
et al., 2000).
F IG. 2.
primers used toamplify pBR322 for subcloning by gap repair are shown as thick black arrows. Each primer alsocontains 20-nt segments at
its 3? end toprime pBR322. NotI and SalI cleavage sites were included in these primers tofacilitate release of the subcloned fragments from
the plasmid backbone. The location of SpeI restriction sites near Eno2 is also shown. SpeI restriction sites are not present on the linear
amplified pBR322 vector. (B) Gap repair intermediate showing pairing between a typical amplified pBR322 targeting cassette and the
modified Eno2 BAC. amp, amp-resistance gene; ORI, origin of replication. (C) (L eft) Representative CHEF gel results of NotI/SalI-double
digested BAC DNA from the different subcloning experiments. A 2.8-kb vector fragment and a rescued fragment can be seen in each lane.
M1, ?/HindIII marker. M2, midrange marker II (New England Biolabs). The SpeI-digested restriction pattern of each BAC is shown in the
right panel. As expected, an 8.9-kb band containing the 3? end of the Eno2 gene is seen in all the lanes. M3, 1-kb ladder (Gibco).
Fragments as large as 80 kb can be subcloned from BACs by gap repair. (A) The location of the 5? homologies on the amplification
E. coli-BASED SYSTEM FOR BAC ENGINEERING
Creation of DY380 Derivatives Containing Arabinose-
Inducible Cre or Flpe Genes
BAC targeting often makes use of a selectable
marker to introduce the targeting cassette into the
targeted locus. The selectable marker can, however,
interfere with the subsequent function of the targeted
locus. If the selectable marker is flanked with FRT or
loxP sites, it can be removed from the targeted locus by
Flp or Cre recombinases, thus eliminating this prob-
lem. To facilitate this operation in E. coli, two new
strains, EL250 and EL350, were constructed, by re-
placing the tet gene in the prophage carried in DY380
cells with araC and the arabinose-inducible flpe and
cre genes, respectively (Fig. 1A, Table 1). tet is located
between cl857 and bioA in the DY380 prophage. flpeis
a genetically engineered flp that has a higher recom-
bination efficiency than the original flp gene (Buchholz
et al., 1998). Thus, both strains have homologous re-
combination (the ? red genes) and site-specific recom-
bination (flpe or cre) functions, with the former con-
trolled by temperature and the latter by arabinose.
This dual regulation allows both selective targeting by
recombination and the subsequent removal of the se-
lection marker from the targeted locus by site-specific
The General Strategy for BAC Engineering
To test the efficiency of this prophage system for
BAC engineering, an IRES–eGFPcre–FRT–kan–FRT
cassette was targeted to the Eno2 locus carried on a
250-kb BAC (284H12, Research Genetics) (Fig. 1B).
The Eno2 gene is located in the middle of this fully
sequenced BAC (Ansari-Lari et al., 1998). The Eno2
gene was targeted because it is neural-specific and
expressed in most mature neurons (Marangos and
Schmechel, 1987). The goal was tocreate a BAC trans-
genic line that expresses Cre in all mature neurons for
use in conditional knockout studies. A BAC approach
was used since conventional transgenes often lack im-
portant regulatory sequences required for proper gene
The Eno2 BAC was modified using previously de-
scribed methods (Yu et al., 2000). First, the Eno2 BAC
was electroporated into EL250 cells, and six chloram-
phenicol-resistant (CmR) colonies were selected. Diges-
tion of BAC DNA from six CmRcolonies with EcoRI or
HindIII showed that one had an abnormal digestion
pattern. However, in other BAC electroporation exper-
iments involving the analysis of more than 76 addi-
tional colonies, noabnormal BACs were identified (our
unpublished results). These results indicate that BAC
rearrangements during electroporation are rare. The
PCR-amplified from a template plasmid, pIGCN21, us-
ing chimeric 63-nt primers. The 3? 21 nt of each primer
was homologous to the targeting cassette used for am-
plification, while the 5? 42 nt was homologous to the
last exon of Eno2 where the cassette was tobe targeted
by recombination. The primers were designed totarget
precisely the cassette downstream of the Eno2 stop
codon and upstream of its poly(A) site. EL250 cells
carrying the BAC were then shifted to 42°C for 15 min
to induce Exo, Beta, and Gam expression. The cells
were then electroporated with 300 ng of the amplified
cassette, and kanamycin-resistant (KmR) colonies were
selected. Approximately, 5200 KmRcolonies were ob-
tained from 108electroporated cells for a targeting
efficiency of ?10?5. No colonies were obtained from
Whole-cell PCR analysis of 24 selected colonies using
primers that flanked the targeted locus indicated that
all were correctly targeted. Sequencing of the targeted
region from six colonies, however, showed that three
carried point mutations. To determine whether these
point mutations were introduced during PCR amplifi-
cation or during homologous recombination, thetarget-
ing was repeated. This time, however, the PCR-ampli-
subcloned into the SmaI site of pBluescript by blunt-
end ligation before targeting, and plasmids carrying
wildtype amplified cassettes were identified by DNA
sequencing. These cassettes were then released from
the plasmid by BamHI digestion and used for target-
ing. Using this two-step method, all 12 targeted BACs
that were subsequently sequenced contained wildtype
IRES–eGFPcre–FRT–kan–FRT cassettes. These re-
sults indicate that the point mutations were intro-
duced during PCR amplification of the targeting cas-
sette rather than during targeting. Surprisingly, the
targeting efficiency using this two-step approach was
similar to direct targeting even though the BamHI
fragment contains 5?-protruding ends and extra bases
from the pBluescript polylinker that bear no homology
to the targeted region. This does not mean that sub-
cloning and sequencing should be performed routinely
as the targeted BACs must still be sequenced to verify
that they carry wildtype targeted cassettes.
Next, the kan selectable marker was removed to
prevent its possible interference with Cre expression.
Overnight cultures from single KmRcolonies were di-
luted 50-fold in LB medium and grown till OD600? 0.5.
Flpe expression from the EL250 cells was then induced
by incubating the cultures with 0.1% L-arabinose for
1 h. The bacterial cells were subsequently diluted 10-
fold in LB medium, grown for an additional hour, and
spread on chloramphenicol plates (12.5 ?g/ml). The
next day, 100 CmRcolonies were picked and replated
on kan plates (25 ?g/ml) to test for loss of kanamycin
resistance. All colonies were Kmsand contained a sin-
gle FRT site at the targeted locus. The high recombi-
nation efficiency likely reflects the tight control of Flpe
expression afforded by the PBADpromoter and the fact
that the FRT sites are located in cis rather than in
trans to one another.
Finally, the loxP site contained in the BAC vector
backbone, pBeloBAC11 (Shizuya et al., 1992), was re-
moved by a final round of gene targeting. This was
LEE ET AL.
performed to prevent any concatemerized BAC trans-
genes from being excised from the mouse germline by
Cre recombinase. To facilitate the removal of this loxP
site, a new plasmid, pTamp, was constructed that con-
tains an amp gene flanked by 920 bp of pBeloBAC11
sequence located 5? of the loxP site and 370 bp of
pBeloBAC11 sequence located 3? of the loxP site (Shi-
zuya et al., 1992). This amp insert can be released from
pTamp by BamHI digestion and used to replace the
loxP site in the BAC transgene by gene targeting (see
Materials and Methods). This targeting reaction is
very efficient due to the large amount of homology
between the amp cassette and the pBeloBAC11 vector
(56,200 colonies per 108electroporated cells).
Subcloning by GAP Repair
This recombination system can also be used to sub-
clone fragments from BACs without the use of restric-
tion enzymes or DNA ligases. In this case, subcloning
relies on gap repair to recombine the free ends of a
linear plasmid vector with homologous sequences car-
ried on the BAC (Fig. 2B). The linear plasmid vector
with an amp selectable marker and an origin of repli-
cation carries the recombinogenic ends. The vector is
generated by PCR amplification using two chimeric
primers. The 5? 45–52 nt of each primer is homologous
to the two ends of the BAC sequence to be subcloned
while the 3? 20 nt is homologous to plasmid DNA.
Recombination generates a circular plasmid in which
the DNA insert was retrieved from the BAC DNA via
gap repair. Circular plasmids are selected by their
To determine the maximum-sized fragment that
can be subcloned from BACs using this method, sev-
eral different pairs of primers were generated in
which the homology segments were located 25, 48,
60, or 80 kb apart in the E no2 BAC DNA (Fig. 2A).
Rare-cutter NotI and SalI restriction sites were also
incorporated into these primers so that the sub-
cloned fragments could be released from the recom-
binant clones intact. Using pBluescript as the clon-
ing vector, it was possible to subclone the 25-kb
fragment, but it was impossible to subclone larger
fragments (data not shown). Perhaps subclones con-
taining larger fragments on a high-copy vector are
toxic to the cell.
Todetermine whether this is the case, pBR322 was
used as the cloning vector with its copy number
control element intact. As shown in Fig. 2C (left
panel), fragments as large as 80 kb could be sub-
cloned with this lower copy number vector. Unlike
targeting, however, not all subclones had the correct
inserts as determined by restriction enzyme pattern
analysis (Fig. 2C, right panel). Some subclones
lacked inserts while others contained inserts with
aberrant restriction patterns (Table 2). It is impor-
tant therefore to fingerprint enough subclones so
that subclones with wildtype inserts are identified.
The ability to subclone large fragments of genomic
DNA by gap repair should facilitate many studies in
genome research that were difficult or impossible to
Production of Transgenic Mice
Todetermine whether the modified BAC contains all
of the regulatory sequences needed for neural-specific
Cre expression, it was injected into (C3H/HeN-Mtv??
C57BL/6Ncr)F2zygotes. A BAC transgenic line carry-
ing approximately two copies of the transgene was
then established. While eGFP was included in the orig-
inal Cre targeting cassette to provide a visual marker
for Cre expression, none of the subsequent Cre-ex-
pressing transgenic lines expressed detectable levels of
eGFP (data not shown). The reason for this is unclear
as cultured embryonic fibroblast cells transfected with
this same targeting cassette expressed readily detect-
able levels of eGFP (data not shown). As a control, two
transgenic lines carrying the 25-kb subclone were also
established (Fig. 2). The 25-kb subclone contains the
entire modified Eno2 coding region as well as 10 kb of
5? flanking sequence and 5 kb of 3? flanking sequences
(Fig. 2A). One transgenic line, 25kbp-1, carries approx-
imately four copies of the transgene, while the second,
25kbp-2, carries approximately five copies of the trans-
To assess Cre activity, the transgenic mice were
crossed toROSA26 reporter mice, which contain a lacZ
reporter that can be activated by Cre recombinase (So-
riano, 1999). Double heterozygotes were subsequently
analyzed by X-gal staining at 4 weeks of age. Several
different tissues were examined for X-gal expression
including the brain, spinal cord, eye, lung, heart, in-
testine, muscle, liver, spleen, and kidney. Blue-stained
cells were found only in neural tissue in the three
transgenic lines, indicating that both the BAC and the
T ABL E 2
Subcloning F ragments from the Modified BAC
by Gap R epair
No. Correct recombinants/
No. colonies examinedb
8/12 (1, 3)c
4/12 (0, 8)
aInduced EL250 competent cells carrying the modified BAC were
electroporated with 300 ng of the different recombinogenic amplified
pBR322 cloning vectors. The total number of AmpRcolonies from
each experiment is shown. The number of surviving colonies without
drug selection was ?108for each experiment.
bTwelve or 24 AmpRcolonies were randomly picked and analyzed by
Southern analysis following NotI/SalI digestion. Correct recombinants
contained a 2.8-kb cloning vector and the expected size insert.
cThe first number in parentheses indicates the number of colonies
in which the amplified cloning vector ends were joined without
undergoing recombination. The second number in parentheses de-
picts colonies with inserts that are aberrant in size.
E. coli-BASED SYSTEM FOR BAC ENGINEERING
25-kb subclone contain the regulatory elements needed
for neural-specific expression. The pattern of Cre ac-
tivity was, however, different in the three lines. Vi-
bratome sections of the brain from the BAC transgenic
mice showed blue-stained cells throughout the gray
matter but not in the white matter, indicative of Cre
activity in most neurons but not in glial cells (Fig. 3A).
In contrast, X-gal staining in the 25kb-1 and 25kb-2
transgenic mice was present in only a subset of neu-
rons, and expression was variable between the two
different lines (Fig. 3A).
Higher power magnification of the cerebellum of
the BAC transgenic mice showed that Cre was ex-
pressed in virtually all neuronal cells. This included
Purkinje cells in the Purkinje cell layer, granule and
Golgi cells in the granular layer, basket cells and
F IG. 3.
4-week-old animals carrying the Rosa26R reporter gene in addition to the BAC transgene, the 25kb-1 transgene, or the 25kb-2 transgene.
X-gal staining is indicative of Cre activity. (B) A section from the superior colliculus region of the brain of a BAC transgenic animal was
immunostained with an anti-Eno2 antibody (Chemicon) followed by X-gal staining for Cre activity. Blue staining is indicative of Cre activity,
while brown staining indicates Eno2 protein-positive cells.
Cre activity in the brains of transgenic mice carrying the Rosa26R reporter gene. (A) X-gal-stained sections from the brains of
LEE ET AL.
stellate cells in the molecular layer, and neurons of
the deep cerebellar nuclei (Fig. 4A). In contrast, in
the 25kbp-1 line, Cre was expressed in only a subset
of Golgi cells in addition to a few cells in the granule
and Purkinje cell layers (Fig. 4B). Glial cells of white
matter also expressed Cre, indicative of leaky ex-
pression. In the 25kbp-2 line, Cre expression was
limited to the gray matter and included a variety of
neuronal cell types, including most basket cells, stel-
late cells, Purkinje cells, and neurons of the deep
cerebellar nuclei (Fig. 4C). In contrast, few granule
cells and Golgi cells in the granule layer expressed
Higher power magnification of the hippocampus
and cortex showed similar results. In the hippocam-
pus of BAC transgenic mice, virtually all neurons in
the cornu Ammonis (CA) region and the dentate
gyrus (DG) expressed Cre (Fig. 4D). The same was
true in the cortex, where all five layers of the cortex
that contained neurons (layers II–VI) expressed Cre.
In contrast, the hippocampus of 25kbp-1 transgenic
mice showed reduced Cre expression in the DG (Fig.
4E ) and layers II and III of cortex (Fig. 4H). The
25kbp-2 transgenic mice showed even lower levels of
Cre expression in the DG (Fig. 4F). The CA1 and CA2
regions of the CA also failed to express Cre (Fig. 4F).
Cre expression was also greatly reduced in the cor-
tex, with layers II and III showing the most reduc-
tion (Fig. 4I).
Cre activity in the spinal cord, dorsal root ganglion,
and retina of the transgenic mice was alsoexamined to
determine whether Cre was expressed in mature neu-
rons within the peripheral nervous system. Similar to
what was observed for the central nervous system, Cre
was expressed in most mature peripheral neurons in
the BAC transgenic mice, while fewer peripheral neu-
rons expressed Cre in the two 25-kb transgenic lines
(data not shown).
To determine whether Cre was expressed in all
E no2 protein-positive neurons, a section from the
brain of a BAC transgenic animal was immuno-
stained with an anti-E no2 antibody followed by X-gal
F IG. 4.
from the cerebellum (A, B, and C), hippocampus (D, E , and F ), and cortex (G, H, and I) of a BAC transgenic animal (A, D, and G), a 25kbp-1
transgenic animal (B, E , and H), and a 25kbp-2 transgenic animal (C, F , and I). The Purkinje cell layer (P), granular layer (G), molecular
layer (M), deep cerebellar nuclei (DCN), and white matter (WM) of the cerebellum, the three subregions of the CA (CA1, CA2, and CA3), the
dentate gyrus (DG) of the hippocampus, and the six layers (I to VI) of the cortex are labeled.
Cre activity in the cerebellum, hippocampus, and cortex of transgenic mice. Higher power magnification of X-gal-stained sections
E. coli-BASED SYSTEM FOR BAC ENGINEERING
staining for Cre activity. As shown in Fig. 3B, virtu-
ally all E no2-positive neurons were active for Cre.
Targeting without Selection
Thehigh level of efficiency of recombination obtained
with this prophage system suggested that it might be
possible to perform targeting without drug selection.
Direct targeting would facilitate genomic experiments
in which the presence of a selectable marker, or even a
FRT or loxP site, is undesirable. Todetermine whether
targeting can be achieved without drug selection, a
24-bp flag tag was targeted tothe5? end of theSRY-box
containing gene 4 (Sox4) carried on a 125-kb BAC. For
these experiments, a 114-bp targeting cassette was
generated in which two 45-bp arms homologous to the
Sox4 gene flanked the 24-bp flag sequence. This DNA
fragment was created by synthesizing two 79-bp oligo-
nucleotides that overlapped at their 3? ends by 44 bp.
These overlaps were annealed and filled in by Taq
polymerase. Following electroporation of the flag-
tagged cassette into induced DY380 cells carrying the
Sox4 BAC, the cells were spread on LB plates to a
density of ?2000 cells per plate. Colonies containing
the flag tag were subsequently identified by colony
hybridization using a 30-bp flag-specific oligonucleo-
tide probe (24-bp flag tag and 3 bp on each side that
was homologous to the Sox4 targeted site). Among
3800 colonies screened from uninduced cells, no flag-
positive colonies were identified. In contrast, 7 flag-
positive colonies were identified in 4210 colonies ob-
tained from induced cells for an overall targeting
frequency of 1.7 ? 10?3. PCR amplification and direct
sequencing showed that each of the seven flag-positive
colonies was correctly targeted.
In the studies presented here, we describe a highly
efficient recombination system for manipulating BAC
DNA in E. coli that uses a defective ? prophage to
supply functions that protect and recombine the elec-
troporated linear DNA targeting cassettewith theBAC
sequence. Because the recombination functions are ex-
pressed from a defective prophage rather than a plas-
mid, the recombination functions are not lost during
cell growth as often happens with plasmid-based sys-
tems. Another advantage of this prophage system is
that the ? gam and red recombination genes are under
the control of the temperature-sensitive ? repressor
that provides a much tighter control of gam and red
expression than can be obtained on plasmids. This
tight regulation, combined with the strong ? PL pro-
moter, which drives gam and red expression to very
high levels, makes it possible toachieve recombination
frequencies that areat least 50- to100-fold higher than
those obtained with plasmid-based systems (Nara-
yanan et al., 1999; Muyrers et al., 1999; our unpub-
Theability tomanipulatelargefragments of genomic
DNA precisely, independent of the location of appro-
priate restriction enzyme sites, has many applications
for functional genomics, both in the mouse and in other
organisms. As shown here, Cre can be introduced into
thecoding regions of genes carried on BACs facilitating
the generation of Cre-expressing transgenic lines for
use in conditional knockout studies or for use in con-
ditional gene expression studies. Genes can also be
epitope-tagged and microinjected into the germline of
mice carrying a mutation in the gene. If the epitope-
tagged transgene rescues the mutant phenotype, the
epitope-tagged protein is functional, and the epitope
tag can serve as a marker for expression of the gene.
Likewise, a genecarried on a BAC can bereplaced with
another gene, and the function of the “knock-in” muta-
tion can be assayed in transgenic mice.
This recombination system also facilitates the gen-
eration of complicated conditional targeting vectors.
While the generation of such vectors often used totake
several months, it can now be performed in a only few
weeks. The ability to express reversibly Cre or Flpe
recombinases in E. coli speeds this process even fur-
ther. A selectable marker flanked with loxP or FRT
sites can now beintroduced intoan intron of a geneand
then removed by transient Cre or Flpe expression,
leaving behind a solo loxP or FRT site in the intron. A
limitation of this approach at the present time is the
lack of a BAC-based mouse physical map and the pau-
city of mouse genome sequence information. This
should all dramatically change, however, next year, as
the draft sequence of the mouse comes on-line and the
BAC physical map is completed.
The high recombination efficiency offered by this
recombination system also makes it possible to manip-
ulate BAC DNA without drug selection. Point muta-
tions, deletions, or insertions can now be engineered
intoany gene on a BAC in the absence of a confounding
linked drug selection marker or a loxP or FRT site. In
cases where the gene is mutated in human disease, the
exact disease-causing mutations can be engineered on
the BAC, and the effect of these mutations can be
analyzed in transgenic mice.
This recombination system also makes it possible to
subclonefragments as largeas 80 kbfrom BACs by gap
repair. Targeting vectors or transgenic contructs gen-
erated by BAC engineering can now be subcloned with
ease, and virtually any region of the engineered BAC
can be included in the final subclone. Subcloning by
gap repair should also facilitate the identification of
regulatory elements or locus control regions that may
be located at some distance from a gene. Many such
potential regulatory elements will be identified over
the next few years by comparative genome sequencing.
The ability to modify precisely these regulatory se-
quences on BACs, combined with the ability to include
or exclude them during the subcloning process, should
make it possible to dissect the function of these se-
LEE ET AL.
quences in the whole animal at a level not previously
We thank P. Soriano for the Rosa26 cre reporter line and the
pPGKmncre vector and R. Behringer for the pNTRlacZPGKneoloxP
vector. The National Cancer Institute, DHHS, supported this work.
All vectors and strains described in this article are freely available
for basic research purposes. To request these vectors and strains,
please send an e-mail to N. G. Copeland at firstname.lastname@example.org.
Ansari-Lari, M. A., Oeltjen, J . C., Schwartz, S., Zhang, Z., Muzny,
D. M., Lu, J ., Gorrell, J . H., Chinault, A. C., Belmont, J . W., Miller,
W., and Gibbs, R. A. (1998). Comparative sequence analysis of a
gene-rich cluster at human chromosome 12p13 and its syntenic
region in mouse chromosome 6. Genome Res. 8: 29–40.
Antoch, M. P., Song, E. J ., Chang, A. M., Vitaterna, M. H., Zhao, Y.,
Wilsbacher, L. D., Sangoram, A. M., King, D. P., Pinto, L. H., and
Takahashi, J . S. (1997). Functional identification of the mouse
circadian Clock gene by transgenic BAC rescue. Cell 89: 655–667.
Arango, N. A., Lovell-Badge, R., and Behringer, R. R. (1999). Tar-
geted mutagenesis of theendogenous mouseMis genepromoter: In
vivo definition of genetic pathways of vertebrate sexual develop-
ment. Cell 99: 409–419.
Buchholz, F., Angrand, P. O., and Stewart, A. F. (1998). Improved
properties of FLP recombinase evolved by cycling mutagenesis.
Nat. Biotechnol. 16: 657–662.
Feiss, M., Siegele, D. A., Rudolph, C. F., and Frackman, S. (1982).
Cosmid DNA packaging in vivo. Gene 17: 123–130.
Link, A. J ., Phillips, D., and Church, G. M. (1997). Methods for
generating precise deletions and insertions in the genome of wild-
type Escherichia coli: Application to open reading frame charac-
terization. J . Bacteriol. 179: 6228–6237.
Marangos, P. J ., and Schmechel, D. E. (1987). Neuron specific eno-
lase, a clinically useful marker for neurons and neuroendocrine
cells. Annu. Rev. Neurosci. 10: 269–295.
Muyrers, J . P., Zhang, Y., Testa, G., and Stewart, A. F. (1999). Rapid
modification of bacterial artificial chromosomes by ET-recombina-
tion. Nucleic Acids Res. 27: 1555–1557.
Narayanan, K., Williamson, R., Zhang, Y., Stewart, A. F., and Ioan-
nou, P. A. (1999). Efficient and precise engineering of a 200 kb
beta-globin human/bacterial artificial chromosome in E. coli
DH10B using an inducible homologous recombination system.
Gene Ther. 6: 442–447.
Shizuya, H., Birren, B., Kim, U. J ., Mancino, V., Slepak, T., Tachiiri,
Y., and Simon, M. (1992). Cloning and stable maintenance of
300-kilobase-pair fragments of human DNA in Escherichia coli
using an F-factor-based vector. Proc. Natl. Acad. Sci. USA 89:
Soriano, P. (1999). Generalized lacZ expression with the ROSA26
Cre reporter strain. Nat. Genet. 21: 70–71.
Yu, D., Ellis, H. M., Lee, E. C., J enkins, N. A., Copeland, N. G., and
Court, D. L. (2000). An efficient recombination system for chromo-
some engineering in Escherichia coli. Proc. Natl. Acad. Sci. USA
Zhang, Y., Buchholz, F., Muyrers, J . P., and Stewart, A. F. (1998). A
new logic for DNA engineering using recombination in Escherichia
coli. Nat. Genet. 20: 123–128.
E. coli-BASED SYSTEM FOR BAC ENGINEERING