Manipulating cytomegalovirus genomes by BAC mutagenesis: Strategies and applications

Abstract
The generation of CMV mutants has been a difficult task especially because of the large
genome size and the slow replication cycle of the CMVs. The recent cloning of CMV ge-
nomes as infectious bacterial artificial chromosomes (BAC) in E. coli opened new horizons
for the construction of mutant CMVs by utilizing the methods of bacterial genetics. This
chapter gives an overview of BAC cloning and the various mutagenesis techniques that al-
low targeted as well as random manipulations of CMV genomes. Selected examples give an
impression of the power of the reverse and forward genetic procedures. The new techniques
provide the basis for a comprehensive analysis of CMV gene functions as well for vaccine
development.
Introduction
Our knowledge about cytomegalovirus (CMV) gene functions is advancing rather slowly,
although the DNA sequence of the HCMV laboratory strain AD169 has been known
since 1990 (Chee et al., 1990) and those of rodent CMVs have also been available for
several years (Rawlinson et al., 1996; Vink et al., 2000). Reasons for the overall scarcity of
data are of course the large number of proteins encoded and the size of the genomes with
associated difficulties in generating viral mutants. CMVs have the largest genomes among
the herpesviruses, ranging in size from 221 (rhesus CMV) to 241 kbp (chimpanzee CMV)
(Davison et al., 2003; Hansen et al., 2003). Coding potentials of up to 208 and 254 pro-
teins were predicted for the HCMV laboratory strain AD169 and for clinical isolates of
HCMV, respectively (Chee et al., 1990; Murphy et al., 2003b). Based on a more stringent
evaluation, the genetic content of the AD169 strain was estimated at 145 genes and that of
wild-type HCMV at between 164 and 167 genes (Dolan et al., 2004). The majority of the
proteins encoded by HCMV has not yet been characterized. Even for those proteins about
which we have some functional knowledge, e.g. the envelope glycoproteins, a clear under-
standing of the mode of action is usually missing. A number of ORFs has been subcloned
in plasmid expression vectors and their functions studied independently of the viral infec-
tion and separately from other viral genes, for instance by transient transfection experi-
ments. However, many viral proteins have to interact with other viral or cellular proteins in
order to exert their effects and many genes display a distinct temporal expression pattern
during the infection cycle. Thus, it is desirable to study the role of genes in the context of
the viral infection.
Manipulating Cytomegalovirus
Genomes by BAC Mutagenesis:
Strategies and Applications
Wolfram Brune, Markus Wagner, and Martin Messerle
4

Brune et al.��  |
In order to assign defined properties of the virus to certain genes, attempts to generate
variants of CMVs with altered phenotypes were already being made in the seventies. Since
the replication machinery of herpesviruses is highly accurate, spontaneous mutants emerge
only rarely and chemical (treatment with nitrosoguanidine) or physical methods (irradia-
tion with UV light) had to be used to introduce mutations into CMV genomes (Ihara et
al., 1978; Yamanishi and Rapp, 1977). Restricted growth at an elevated temperature was
a convenient read-out and selection principle. A series of temperature-sensitive HCMV
and MCMV mutants were generated by these means and on the basis of their ability to
complement each other in cell culture they could be divided into groups (Akel and Sweet,
1993). Rescue experiments and the assignment of mutations to specific genes had to await
the cloning of subgenomic DNA fragments and the determination of the DNA sequences
of CMV genomes. But even then the connection of a mutation with a certain phenotype
turned out to be difficult, because the respective CMV genomes often contained several
mutations (Kumura et al., 1990).
The introduction of insertion and deletion mutagenesis by Spaete and Mocarski
(1987) paved the way to reverse genetics of CMVs. This was the method of choice for
the disruption of CMV genes that are non-essential for replication in cell culture and also
for generating recombinant CMVs expressing reporter genes in order to follow the course
of an infection in animal hosts (reviewed by Mocarski and Kemble, 1996). Although this
was a major step forward, the method was afflicted with a number of limitations that
precluded rapid genetic analyses of CMVs. For instance, non-homologous recombination
is usually favored in eukaryotic cells and homologous recombination events are relatively
rare. Accordingly, the mutagenesis method was rather inefficient and the formation of il-
legitimate recombinants was often observed. Separation of mutant and wild-type viruses
required lengthy and time-consuming plaque purification or limiting dilution procedures.
Isolation of viral mutants displaying a growth deficit remained an even more difficult task.
The application of selection methods, e.g. by employing the gpt, neo or puro gene, facilitated
the isolation of recombinant viruses (Abbate et al., 2001; Greaves et al., 1995; Wolff et al.,
1993) but did not overcome the intrinsic limitations of the approach.
The regeneration of CMVs from overlapping cosmids represented a new principle in
that recombination between the overlapping fragments led to the reconstitution solely of
mutant virus, obviating the need to select against non-recombinant wild-type virus (Ehsani
et al., 2000; Kemble et al., 1996;). The method was used successfully to generate interstrain
variants of HCMV (Kemble et al., 1996) as well as an HCMV ie1 mutant, which dis-
played an impaired growth phenotype and had to be grown on a complementing cell line
(Mocarski et al., 1996).
The genetic analysis of small DNA viruses profited from the cloning of the complete
viral genomes into plasmid vectors and the manipulation of the viral DNA sequences with
the powerful molecular techniques in vitro or in E. coli. For many years the sheer size of the
CMV genomes represented an insurmountable barrier, which prevented their cloning as a
single entity. Only the development of new cloning vectors, e.g. the E. coli F-factor-derived
bacterial artificial chromosome (BAC) that offered a cloning capacity of up to 300 kbp
(Shizuya et al., 1992), made this vision become real.

Manipulating Cytomegalovirus Genomes by BAC Mutagenesis |  �5
Cloning of CMV genomes as bacterial artificial chromosomes
At first sight cloning of CMV genomes may appear a trivial task, which can be achieved
easily by ligation of the linear viral genome into a BAC cloning vector followed by transfor-
mation into an E. coli host. In practice however, such an approach has not been successful.
This is probably due to the fact that large DNA molecules are prone to shearing and exten-
sive handling leads to their destruction. For this reason the viral DNA and the BAC vector
were joined the other way round, i.e. the BAC vector sequences were introduced into the
viral genome by homologous recombination utilizing the cellular recombination machinery
(Figure 4.1A). To this end, a recombination plasmid was constructed that contained the
BAC vector flanked by viral DNA sequences homologous to the intended insertion site in
the viral genome. The linearized recombination plasmid was transfected into fibroblast cells
followed by infection with the respective CMV strain (Borst et al., 1999; Messerle et al.,
1997). Recombinant viruses carrying the BAC vector integrated into their genomes were
amplified in the presence of mycophenolic acid and xanthine using the guanosine phos-
phoribosyl transferase (gpt) gene for selection as described by Greaves et al. (1995). Fol-
Figure �.1  (A) Cloning of a CMV genome as a BAC. The BAC replicon (blue) flanked by viral
sequences is transfected into fibroblasts, which are subsequently infected with CMV. Viral
genomes that have incorporated the BAC cassette by homologous recombination are isolated 
and transferred to E. coli. (B) The principle of generating mutant CMVs using BAC technology. 
The CMV genome maintained as a BAC in E. coli is used as a substrate to introduce a mutation 
(M, red). The mutant BAC is then transfected into fibroblasts in order to reconstitute mutant
virus.
BAC
BAC
BAC
BAC
BAC
M
E. coli
E. coli
Fibroblast
Fibroblast
A B

Brune et al.��  |
lowing amplification of recombinant viruses, circular replication intermediates of the viral
genomes were isolated from infected cells and electroporated into E. coli. Bacterial clones
harboring CMV BACs were selected by means of the antibiotic resistance encoded on the
vector backbone. The CMV BACs could be propagated stably in E. coli owing to the repE,
oriS, and parA, B and C elements of the BAC vector (Borst et al., 1999; Messerle et al.,
1997; Shizuya et al., 1992).
The initial attempts, which aimed at the cloning of the complete MCMV genome,
resulted in the generation of BACs that lacked approximately 8 kbp of viral sequences
and thus were not infectious (Messerle et al., 1997). Most likely, enlargement of the CMV
genome by insertion of the BAC vector sequences is not compatible with proper packaging
into viral capsids, favoring the encapsidation of genomes that compensated the enlargement
by spontaneous deletion of other viral sequences. In further attempts a few non-essential
CMV genes were therefore replaced with the vector sequences (Borst et al., 1999; Messerle
et al., 1997). This strategy turned out to be successful, allowing the generation of infectious
CMV BACs. Upon transfection into permissive cells, the CMV BACs gave rise to forma-
tion of plaques and infectious virions (Figure 4.1B). The CMV BACs displayed a remark-
able stability in E. coli: after days or even weeks of propagation in suitable E. coli hosts the
BACs did not show an altered restriction profile and remained infectious. This property
formed the basis for targeted manipulation of the BAC-cloned CMV genomes in E. coli.
Once cloned into E. coli, the missing viral sequences could easily be re-inserted into the
BACs to restore the full-length CMV genomes. In this way, MCMV as well as HCMV
BACs with full-length genomes were generated (Hobom et al., 2000; Wagner et al., 1999).
The resulting genomes are oversized, necessitating the development of strategies to elimi-
nate the vector sequences upon transfection into permissive cells in order to prevent the
occurrence of spontaneous deletions. For this purpose, the BAC vector was flanked with
loxP sites to allow recombinase Cre-mediated excision (Hobom et al., 2000; Messerle et al.,
1997). The very same strategy was also adopted by Yu et al. (2002). They managed to clone
the full-length HCMV AD169 genome by insertion of a BAC vector cassette between
ORFs US28 and US29. Notably, a virus that retained the BAC vector sequences in the
genome displayed delayed growth kinetics and produced a reduced yield. These growth
restrictions were abrogated following Cre-mediated removal of the BAC vector, supporting
the idea that oversized genomes are packaged inefficiently. By inserting an intron-contain-
ing Cre gene within the BAC vector sequences, Yu and colleagues adopted an elegant strat-
egy for the self-excision of the BAC vector in eukaryotic cells (Yu et al., 2002), which had
previously been developed for a pseudorabies virus BAC (Smith and Enquist, 2000).
In order to clone the RhCMV genome, Chang and Barry (2003) expanded the use
of Cre recombinase by performing Cre-mediated insertion of a BAC vector into a loxP-
tagged virus genome, a strategy originally developed by Smith and Enquist (2000). In the
first step, a loxP-flanked EGFP cassette was inserted into the intergenic region between
US1 and US2 of the RhCMV genome followed by the removal of the cassette via Cre/lox
recombination leaving a single remaining loxP site in the genome. In the second step the
BAC vector, which also carried a loxP site, was inserted with the help of the transiently
expressed Cre recombinase. After transfer of the BAC into E. coli the intron-containing Cre
gene was inserted into the vector backbone giving rise to a self-excisable RhBAC (Chang
and Barry, 2003).

Manipulating Cytomegalovirus Genomes by BAC Mutagenesis |  ��
During reconstitution of the full-length MCMV genome it was observed that flanking
the BAC vector with identical sequences led to its spontaneous excision in eukaryotic cells
through homologous recombination (Wagner et al., 1999). In combination with the selec-
tion-pressure against packaging of oversized genomes this principle was used to generate
MCMV genomes without the BAC vector. The resulting genomes were indistinguishable
from the wild-type MCMV genome, i.e. no traces of the BAC vector remained in the ge-
nome. Thus, this approach appears especially suitable for the generation of mutants that
will be used for in vivo studies and should differ from the wild-type virus by a specific
mutation only. Remarkably, the recombinant MCMV constructed in this way displayed
biological properties in vitro and in vivo that were virtually identical to those of the parental
MCMV strain, indicating that the MCMV genome was not altered during propagation as
a BAC in E. coli (Wagner et al., 1999).
Following the BAC cloning of MCMV, a series of other human and animal CMV
strains has been cloned (Table 4.1) and we are convinced that no principal problems will
prevent the cloning of the missing members of the CMV family and of the beta-herpesvi-
ruses HHV-6 and HHV-7. Of especial importance was the cloning of clinical isolates of
HCMV (Hahn et al., 2002; Murphy et al., 2003b), because after extended passage of such
strains in cell culture they tend to lose their wild-type characteristics, e.g. the property to
grow in certain cell types such as endothelial cells. Thus, cloning of clinical HCMV isolates
helps to preserve their properties and provides an opportunity to subject the specific char-
acteristics of those strains to genetic analysis.
Mutagenesis of BAC-cloned CMV genomes
The establishment of methods to manipulate the CMV genomes in E. coli was the next
step after successful BAC cloning. Due to the large size of the BACs simple modification
techniques, e.g. digestion of the BAC DNA with restriction enzymes followed by ligation,
could not be applied, simply because almost every enzyme cuts several times in the huge
Table �.1  Cytomegaloviruses cloned as BACs in E. coli
Virus, strain Full length BAC excisable Reference
HCMV, AD169 No No Borst et al. (1999)
HCMV, AD169 Yes Yes Hobom et al. (2000)
HCMV, AD169 Yes Yes Yu et al. (2002)
HCMV, Towne ATCC No No Marchini et al. (2001)
HCMV, Towne RIT No No Hahn et al. (2003b)
HCMV, Towne long No No Hahn et al. (2003b)
HCMV, Toledo No No Hahn et al. (2003b)
HCMV, FIX No No Hahn et al. (2002)
HCMV, PH No No Murphy et al. (2003b)
HCMV, TR No No Murphy et al. (2003b)
MCMV, Smith No No Messerle et al. (1997)
MCMV, Smith Yes Yes Wagner et al. (1999)
MCMV, K181 Yes Yes Redwood et al. (unpublished)
GPCMV, 22122  Yes No McGregor and Schleiss (2001)
RhCMV, 68-1 Yes Yes Chang and Barry (2003)

Brune et al.��  |
CMV BAC molecules. However, others had already established sophisticated methods for
the manipulation of large DNA molecules or the E. coli chromosome and these methods
could be adapted for the manipulation of CMV BACs. Fortunately, mouse and human
geneticists became interested in the manipulation of BACs around the same time, so one
could profit from their efforts. Meanwhile, there is a wealth of techniques that facilitate the
introduction of any kind of mutation into the BAC-cloned CMV genomes with high ef-
ficiency and within a very short time frame.
RecA-mediated allelic exchange
Manipulation of CMV genomes usually aims at the generation of different alleles of a gene,
which typically result in different phenotypes of the virus. The possible alterations range
from the complete removal or inactivation of a gene (null-allele) to the exchange of a single
nucleotide only (point mutation). The mutagenesis technique can also be used to intro-
duce additional genes or DNA sequences into CMV genomes (insertions). The mutant
allele is transferred into the BAC-carrying E. coli host cells with the help of a shuttle vec-
tor. Recombination between the mutant and the parental allele makes use of the general
recombination machinery of E. coli. The RecA protein and exonuclease V of E. coli (en-
coded by recBCD) are the minimal requirements for this recombination pathway. In many
E. coli strains the recA gene is inactivated, rendering the strains recombination-incompetent
and providing enhanced stability for cloned DNA fragments. Although recA expression
is required for mutagenesis it increases the risk of instability of the BAC-cloned CMV
genomes. Fortunately however, the CMV BACs seem to display remarkable stability in
recA+ strains (Borst et al., 1999; Messerle et al., 1997; Yu et al., 2002). Nevertheless, we
recommend keeping the exposure of the CMV BACs to recombination enzymes as short as
possible in order to minimize the risk of acquiring adventitious mutations. For this reason,
mutagenesis was initially performed in an E. coli strain that can be transiently rendered
recA+ (Kempkes et al., 1995; Messerle et al., 1997) and later the recA gene was provided on
the backbone of the shuttle plasmid (Borst et al., 2004; Hobom et al., 2000).
Homologous recombination between the mutant allele on the shuttle plasmid and the
parental allele on the BAC typically leads to the formation of a cointegrate between the two
molecules (Figure 4.2). Such recombination events are probably rare, but appropriate selec-
tion schemes allow the detection of bacterial clones that contain cointegrates. Two kinds
of shuttle vectors have been employed. One is based on the pSC101 replicon and offers a
temperature-sensitive mode of replication (Borst et al., 2004; Posfai et al., 1997) and the
other one contains the R6K replicon and thus replicates only in E. coli strains that express
the lambda π gene (Smith and Enquist, 1999; Smith and Enquist, 2000; Yu et al., 2002).
Following the transfer of pSC101-based shuttle plasmids, bacterial clones harboring coin-
tegrates can be selected by shifting the bacteria to a temperature that is non-permissive for
the replication of the shuttle plasmid and by selecting for the antibiotic resistances that are
encoded by the BAC and the shuttle plasmid, respectively (Borst et al., 2004). R6K-based
shuttle plasmids can be transferred either by transformation or by conjugation. Since they
cannot replicate in BAC-carrying E. coli, bacterial clones that survive under selection with
antibiotics must have integrated the DNA sequences of the shuttle plasmid into the BAC.
Resolution of the cointegrate occurs spontaneously by homologous recombination but
again this is a rather rare event. Accordingly, selection against bacteria that have retained

Manipulating Cytomegalovirus Genomes by BAC Mutagenesis |  ��
the cointegrates by utilizing the negative selection marker sacB proved to be highly use-
ful in identifying those clones that harbor mutated BACs (Borst et al., 1999; Chang and
Barry, 2003). Since the sacB gene has an intrinsic propensity for spontaneous inactivation,
a kan/lacZ cassette and blue-white screening of colonies in the presence of X-Gal were also
used to identify modified BACs (Yu et al., 2002). Depending on whether resolution occurs
via the same homologous sequence that was used during generation of the cointegrate or via
the sequence that flanks the mutation on the other side, the outcome is either a BAC with
the parental configuration or a BAC carrying the mutation (Figure 4.2). Final conforma-
tion of the successful introduction of a mutation has to be obtained by restriction analysis,
PCR, and/or sequencing.
This two-step recombination method worked highly reliably in our hands. Admittedly,
the construction of the shuttle plasmid may involve several cloning steps and can be time-
consuming. The homologous sequences flanking the mutation each have to be 1.5 to 2
kbp in size in order to achieve a reasonable chance of recombination. With respect to the
stability of the BAC it is reassuring that recombination only occurs at a low frequency and
BAC
M
BAC
M
BAC
M
BAC
BAC
M
SP
RecA
RecA RecA
Figure  �.�  BAC  mutagenesis  in  E. coli using a shuttle plasmid. A mutation (M) flanked by
viral sequences (hatched and crosshatched) is cloned into a shuttle plasmid (SP). The shuttle 
plasmid  usually  contains  a  selectable  and  a  counterselectable  marker  (e.g., kan  and  sacB). 
Introduction  of  the  shuttle  plasmid  into  the  E. coli  host  containing  the  CMV  BAC  leads  to 
homologous recombination and formation of a cointegrate.  In a second recombination step, 
the  cointegrate  is  resolved,  leading  either  to  a  mutant  (left)  or  an  unaltered  genome  (right). 
Homologous recombination is dependent on the E. coli RecA enzyme and requires homologies 
of at least 500 bp.

Brune et al.�0  |
requires relatively large regions of identical sequences. If several different mutations have to
be targeted to the same site in the viral genome or if a mutant virus with a specific mutation
is needed to answer a sophisticated question, we regard it as worthwhile to invest the time
to construct a shuttle plasmid.
Recombination with linear DNA fragments (ET mutagenesis)
The cloning effort associated with the shuttle plasmid approach prompted the search for
more rapid BAC mutagenesis techniques, especially for simple modifications, e.g. the in-
troduction of deletions. The idea was to use a linear DNA fragment carrying a selectable
marker (e.g. the kanamycin resistance gene) flanked by DNA sequences homologous to the
viral target site and to introduce the marker by two simultaneous recombination events, i.e.
by double cross-over (Figure 4.3A). Some recombination systems of bacteriophages such as
the recE and recT gene functions of the E. coli prophage Rac or the redα and redβ functions
of phage lambda utilize linear DNA substrates (Zhang et al., 1998). These recombination
markerFRT FRT
BAC
FRT FRTM homol.homol. marker
BAC
FRT FRTM marker
BAC
FRTM
BAC
M homol.homol. pos. + neg. marker
gene X
BAC
pos. + neg. marker
M homol.homol. gene X (+mutation)
BAC
gene X (+mutation)
Red αβγ
Red αβγ
Red αβγFlp
A B
Figure �.�  BAC mutagenesis using linear DNA fragments. (A) A selectable marker flanked by
FRT sites and a mutation (M) is amplified by PCR. About 50 bp of homologous sequences at
either end are introduced through the PCR primers. The linear PCR product is electroporated 
into  an  E. coli  host  that  expresses  the  red α, β,  and  γ  genes  of  bacteriophage λ.  The  Red 
proteins perform the homologous recombination. In a second step, the selectable marker can 
be removed using Flp recombinase. The mutation and a single FRT site remain in the mutant 
genome. (B) Two-step replacement strategy. A viral gene is first replaced by a selectable and a
counter-selectable marker cassette. In a second step, the marker cassette is replaced either by 
the wild-type or a mutated gene sequence.

Manipulating Cytomegalovirus Genomes by BAC Mutagenesis |  �1
proteins are highly efficient and need only 30 to 50 nucleotides of homologous sequences to
mediate recombination. Thus, the linear recombination substrates can easily be generated
by PCR with the homologous sequences being incorporated into the synthetic oligonucle-
otides used as primers for the amplification of the selectable marker (Figure 4.3A). Since
the homology arms and also the site of mutation can be freely chosen, the method is not
reliant on the availability of appropriate restriction enzyme sites within the viral genome
and obviates preparatory cloning steps. Thus, mutations can be introduced virtually at any
position of the viral genome. Since selection is straightforward and fast and recombination
is highly precise, recombinant genomes can be generated within a couple of days.
The method was first applied to BACs by Stewart and colleagues and designated
ET recombination (Muyrers et al., 1999; Zhang et al., 1998). Originally, the recombina-
tion functions were expressed by a multi-copy plasmid vector, but now low copy vectors
(Datsenko and Wanner, 2000) and defective λ prophages (Lee et al., 2001; Yu et al., 2000)
or mini-λ DNA (Court et al., 2003) integrated into the genome of E. coli strains seem to be
more common and help to express adequate levels of the recombination proteins. As linear
DNA is rapidly degraded in bacteria, the exonuclease inhibitor redγ from bacteriophage λ
must be co-expressed together with the recombination functions.
The redαβ recombination system is characterized by a seemingly higher efficiency in
comparison to the recABCD-based system. Since a number of short repeated sequences are
present in CMV genomes, there is a risk for unwanted recombination events and potential
instability of the viral BACs during mutagenesis. However, inducible expression systems
allow the recombination activity to be controlled and therefore the risk of instability can
generally be held well in check (Wagner and Koszinowski, 2004).
Initially the method was used preferentially to delete genes from the CMV BACs by
replacing viral sequences with a selectable marker, but by including additional sequences
between the marker and the homologous regions on the linear fragment, the insertion of
these sequences into the BAC-cloned CMV genomes can also be accomplished (Figure
4.3A). If the linear fragment is generated by PCR, the length of the insertion is defined by
the technical limitations that apply to the synthesis of long oligonucleotides. Thus, to insert
larger DNA segments the sequence of interest and the homologous sequences for recom-
bination first need to be cloned next to the selection marker (Borst et al., 2001). The linear
fragment required for recombination can then be obtained either by isolation from the
plasmid or by amplification via PCR. After mutagenesis, the selection marker (one kbp in
the case of the kanamycin gene) can be removed by Cre or FLP recombinase if the marker is
flanked with loxP or minimal FRT sites leaving only 34 bp of remaining sequences (Figure
4.3A) (Wagner et al., 2002; Wagner and Koszinowski, 2004). This method also means that
it is possible to reuse the removed selection marker and create double and triple mutants for
genes dispersed over the CMV genome as an alternative to the use of a different selection
marker for each mutation. One example is the generation of a set of single, double and triple
MCMV mutants for the MHC class I interacting genes m04, m06 and m152 (Wagner
et al., 2002). Liu and colleagues proved the versatility of the method in a comprehensive
screen for essential and non-essential HCMV genes (Dunn et al., 2003).
The introduction of single gene deletions is by no means the only application of this
powerful method. Using linear PCR fragments containing the kanamycin resistance gene
and homology arms complementary to regions of the viral genome as much as 119 kbp

Brune et al.��  |
apart, large deletions covering dozens of genes can be attained. As shown schematically
in Figure 4.4, a set of 16 viable MCMV mutants was generated with deletions covering
between 4 and 49 genes, confirming that the majority of MCMV genes are non-essential
for virus replication in cell culture (M. Wagner, unpublished).
The ability to add a few nucleotides within the oligonucleotide primers was also used
to C-terminally epitope tag viral genes in the context of the viral genome. An HA-tagged
viral protein can easily be detected by commercially available antibodies specific for the nine
amino acid HA peptide, thereby enabling the investigation of its temporal expression pat-
tern and intracellular localization, and of potentially interacting proteins by Western blot,
immunofluorescence and co-immunoprecipitation experiments (Atalay et al., 2002; Brune
et al., 2003; Bubeck et al., 2004; Kattenhorn et al., in press; Menard et al., 2003).
For many applications the few nucleotides remaining of the selection marker after
FLP or Cre recombinase-mediated excision are acceptable. However, if only one or a few
nucleotides are to be exchanged, e.g. in order to change one amino acid of a protein, the
concomitant introduction of additional sequences is not tolerable. Two ways have been
explored to solve the problem: mutagenesis can either be performed in the absence of a
positive selection marker or the selection marker can be replaced in a second round of
mutagenesis (Figure 4.3B). redαβ-mediated recombination with linear fragments or single
stranded oligonucleotides can occur with such a high frequency that the identification of
mutated BACs by PCR screening becomes feasible (Ellis et al., 2001). The construction of
a revertant of a MCMV m152 deletion mutant by this strategy proved its applicability for
CMV BACs (Holtappels et al., 2004). However, a frequency of successful recombination
Figure  �.�  Example for site-directed mutagenesis using linear PCR fragments: defining
essential genes of MCMV with a number of large deletions in the viral genome. Deletions of up 
to 43 kbp are introduced as shown in Figure 4.3A. The deletion mutants shown in this figure
were all viable, defining 101 of 169 predicted viral genes (60%) as nonessential for MCMV
replication. The 101 deleted genes include 82 genes of unknown function.
= ORF = immunomodulatory gene = deleted region= essential gene
N A B M H
C G F K
L I EJ O P
ori
m04m06 M33
M
45
M
78
m
131
m
129
m
138
m
152
m
157
M
36
m
41
Δm01–22 ΔM23–26 ΔM27-31 ΔM32–36 ΔM37-43
Δm144–158 Δm159–170ΔM118–121 ΔM128-139
HindIII genome fragments of MCMV
Δm01–17
M
43
m
139–141
m
144

Manipulating Cytomegalovirus Genomes by BAC Mutagenesis |  ��
events of about 0.1% meant that screening of several thousand bacterial clones by PCR was
required.
In the second approach, a positive and a negative selection marker (or a screening
marker) are first inserted into the CMV BAC at the target site. In a second step the mark-
ers are replaced by a fragment carrying the desired mutation. Britt et al. used the ampicillin
resistance gene in combination with the lacZ gene (Britt et al., 2004). Clones in which
replacement of the marker genes with a linear fragment occurred could conveniently be
detected as white colonies by white-blue screening, although just 0.1 to 0.8% of the clones
were positive. Dunn et al. (2003) used the E. coli rpsL gene as a negative selection marker
for the construction of revertant viruses. rpsL confers streptomycin sensitivity when intro-
duced into a streptomycin-resistant E. coli host strain. The loss of the negative selection
marker by replacement with a fragment providing the respective wild-type ORF of HCMV
was detected by selection for streptomycin resistance. Since the loss of the negative selec-
tion marker by unwanted rearrangements will also result in streptomycin-resistant clones,
it may be necessary to characterize a number of clones until a BAC with an intact genome
configuration and the desired mutation is identified.
Transposon mutagenesis
Transposons are mobile genetic elements that can relocate and insert at any genomic loca-
tion in a virtually random manner. In bacteria they have been utilized widely as random in-
sertion mutagens both at the genomic level and in the analysis of individual genes (reviewed
by Hayes, 2003). With their cloning as BACs, CMV genomes also became accessible to
transposon mutagenesis. The power of the technique lies in its simple and rapid application
and in particular in its capacity to generate a large number of mutants within a short time.
Several different transposon systems have been tested for their suitability to ma-
nipulate CMV genomes (Table 4.2). Because of the potential vulnerability of the large
BAC molecules to shearing in vitro, transposon mutagenesis in E. coli was preferred. The
Table �.�  Transposon systems used for BAC mutagenesis
Transposon Properties References
Tn1721 (TnMax) In vivo system, preferential insertion 
into BAC, high efficiency
Brune et al. (1999, 2001, 2002,
2003), Hobom et al. (200), Ménard 
et al. (2003), Yu et al. (2003), 
Zimmermann et al. (in press)
Tn5 In vivo system, preferential insertion 
into bacterial genome, requires re-
transformation
Smith and Enquist (1999), Yu et al.
(2003)
Tn5 a In vitro system, results in high 
percentage of incomplete BACs
McGregor et al. (2004)
Tn5 a
(transposome)
In vivo system, requires transposon-
transposase complex (transposome)
McGregor et al. (2004)
Tn7 a In vitro system, results in high 
percentage of incomplete BACs
McGregor et al. (2004)
Tn7 In vivo system, insertion into defined
attachment site (non-random)
Hahn et al. (2003a)
aCommercially available through Epicentre and NEB, respectively.

Brune et al.��  |
first transposon system for the mutagenesis of an entire herpesvirus genome was based
on Tn1721 (a member of the Tn3 family), which displays a preference for insertion into
negatively supercoiled plasmids (Brune et al., 1999). The transposon is provided on a donor
plasmid with a temperature-sensitive replication mode (Figure 4.5A). The transposition
control unit with the transposase gene is also located on the donor plasmid but separately
from the transposable element and is therefore not translocated. Accordingly, secondary
transposition events can be prevented by eliminating the donor plasmid after mutagenesis.
Following the introduction of the donor plasmid into E. coli harboring the BAC, the bac-
teria are kept at the permissive temperature and the transposon can jump into the CMV
BAC. Subsequently, the bacteria are shifted to the non-permissive temperature to eliminate
the donor plasmid. BAC clones with an insertion are identified by selection with those
antibiotics whose resistance is encoded by the transposon and the BAC vector. In this way
thousands of BAC mutants can be generated overnight. Transposition into the bacterial
chromosome occurred in only 5 to 10% of the clones, obviating a need for enrichment of
clones with mutant BACs (Brune et al., 1999; Hobom et al., 2000). Almost all of the BACs
contained a single insertion, which remained stable in the genome after reconstitution of
Figure  �.5  Transposon mutagenesis: different approaches. (A) A temperature-sensitive
transposon donor  (TD) plasmid containing a selectable marker gene within the transposable 
element is maintained in E. coli together with the CMV BAC at 30°C. Transpositions occur at 
a low frequency. Plating the bacteria at 42°C with appropriate selection results in the loss of 
the TD plasmid. In case of a Tn1721-derived transposon roughly 90% of the BACs contain a
transposon insertion as this transposon inserts preferentially into plasmids or BACs. (B) A λ pir-
dependent suicide TD plasmid is delivered to the BAC host by conjugation. The Tn5-derived 
transposon  inserts  more  often  into  the  bacterial  genome  and  less  frequently  into  the  BAC. 
Isolation of BACs  from numerous clones and  transformation of  fresh bacteria  facilitates  the 
retrieval of mutant BACs. (C) In vitro mutagenesis. The TD plasmid, the BAC, and transposase 
enzymes  are  combined  in  a  test  tube,  where  transposition  takes  place.  The  transposition 
reaction is then transformed into E. coli in order to isolate individual clones.
BAC
TD
BAC
BAC
TD
BACBAC
λpir
BAC BAC
30°C
42°C
BACTD
+ transposase
A B C

Manipulating Cytomegalovirus Genomes by BAC Mutagenesis |  �5
viruses. Primer binding sites at the respective termini of the transposon allowed the deter-
mination of the insertion site by direct sequencing. This method was first used to identify
essential and non-essential genes of MCMV (Brune et al., 1999). Later on, mutants of
whole families of CMV genes, namely of the HCMV envelope glycoprotein genes and
the MCMV US22 family members, were isolated from BAC libraries in order to study
their function (Hobom et al., 2000; Menard et al., 2003). Although the distribution of the
transposon insertions did not seem to be completely random, the libraries were of sufficient
complexity to find mutants of each member of the gene families.
Smith and Enquist (1999) adopted a transposon mutagenesis procedure based on the
Tn5 transposon, using a donor plasmid with an R6K origin of replication. Such plasmids
need the Lambda π protein for propagation and cannot replicate in the recipient bacteria
that contain the CMV BAC (Figure 4.5B). Upon transfer of the Tn5 donor plasmid by
conjugation or electroporation the transposon is inserted either into the BAC or the bacte-
rial chromosome. Since the latter occurs in a substantial proportion of the transposition
events, one needs to isolate the BACs and retransform fresh bacteria in order to enrich
for the mutant BACs. Yu et al. (2003) used both previously described transposon systems
(i.e. the Tn5- and the Tn1721-derived transposons) to create a comprehensive library of
HCMV AD169 mutants that enabled them to define non-essential and essential ORFs
of HCMV. In addition to the selection marker (kn), both transposons contained reporter
genes (GFP, lacZ) for eukaryotic as well as for prokaryotic cells. The lacZ marker facilitated
the generation of revertant HCMV BACs by allelic exchange and blue-white screening.
Another interesting property of one of the transposons is the ability to tag the truncated
proteins in four of the six possible ORFs in order to identify proteins that are expressed by
an ORF carrying the insertion (Yu et al., 2003).
One could speculate that in vitro transposition systems would be more efficient for the
mutagenesis of BACs, because the amount of transposase and donor plasmid can more eas-
ily be controlled and optimized if necessary (Figure 4.5C). McGregor et al. (2004) tested
two commercially available in vitro mutagenesis systems that are based on Tn5 and Tn7,
respectively. In principle both systems were considered to be suitable for the manipula-
tion of a guinea-pig CMV BAC. However, since handling the BAC molecules cannot be
avoided during the in vitro reaction and the subsequent electroporation into E. coli, the
majority of the resulting BACs either displayed deletions or rearrangements. This problem
could be elegantly overcome by the use of so-called Tn5 “transposomes”. Transposomes are
complexes between the transposon molecule and the transposase, which are formed in vitro
in the absence of Mg2+ ions. Once transferred into bacteria carrying the BAC, the complex
is activated by the Mg2+ ions inside the cells, resulting in the insertion of the transposon
into a target molecule. Since the complex is self-limiting there is no risk for secondary trans-
position events. Unfortunately, Tn5 is not able to discriminate between the BAC episome
and the bacterial chromosome and thus there remains a need to isolate the mutated BACs
and retransform them into new E. coli bacteria in order to separate them from clones that
received an insertion into the bacterial genome (McGregor et al., 2004; McGregor and
Schleiss, 2004).
The transposon Tn7 differs from other mobile elements in that it can insert with a high
frequency into a single specific site (attTn7) in the E. coli chromosome. Following engineer-
ing of attTn7 into a BAC and the removal or sealing of the natural attTn7 site in the E.

Brune et al.��  |
coli genome, Tn7-based transposons can be used with high efficiency to shuttle any desired
gene or DNA sequence into a BAC. Hahn et al. (2003a) used the system to demonstrate
that the growth of a UL122 (ie2) deleted HCMV can be complemented in cis by the re-
introduction of an ie2 cDNA.
In order to perform a comprehensive analysis of a single CMV gene, it would be desir-
able to generate a panel of mutants with the mutations distributed throughout the ORF. It
has indeed been shown that whole series of mutants affecting a single gene can be isolated
from CMV BAC libraries (Brune et al., 2001; Hobom et al., 2000; Menard et al., 2003).
However, since the bulky transposon insertions usually disrupt the function of the gene,
one can gain little insight into the structure-function relation of a protein by this approach.
To study the function of individual domains of a protein most of the transposon would
have to be removed, leaving behind only a small in-frame insertion within an ORF. This has
not yet been achieved in the context of an infectious CMV BAC, because it would require
modification of the BAC in vitro with the previously mentioned risk of its destruction. In
view of these limitations Bubeck et al. (2004) subcloned the MCMV gene M50, which is
involved in capsid egress, into a plasmid vector and subjected it to comprehensive mutagen-
esis using a modified Tn7-transposon system with a potential for random insertions. The
bulk of the transposon was subsequently removed by treatment with a restriction enzyme
and religation. The library of plasmid-based mutants was then transferred back into the
CMV BAC using a procedure catalyzed by a site-specific recombinase (see below). The
resulting series of viral BACs (104 mutants) made it possible to study the M50 gene in the
context of the viral replication process and to draw important conclusions regarding the
function of different domains of the M50 protein.
Considering the versatility of the transposon-based techniques and their relatively
simple, rapid and inexpensive application it becomes obvious that these techniques have a
tremendous potential for the analysis of CMV gene functions.
Site-specific insertions
Construction of a revertant virus by repair of the respective mutation is one of the controls
that is mandatory to prove that a specific phenotype is due to the disruption of a candidate
gene and does not result from an accidental mutation in another gene. However, as pointed
out by Yewdell and Hill (2002), restoration of the original genetic sequence and rescue of
a phenotype can be easily misinterpreted, because the repair of a gene may also restore the
expression of neighboring or not yet identified overlapping genes. They therefore proposed
that the insertion of a candidate gene elsewhere in the genome would be a better control.
The ectopic insertion of genes can be performed with several of the described tech-
niques. The use of Tn7-mediated transposition for the cis-complementation of an HCMV
ie2 deletion mutant has already been mentioned (Hahn et al., 2003a). As an alternative, ET
recombination was used to insert the gene for the small capsid protein at an ectopic posi-
tion in the HCMV genome in order to prove that this protein is essential for viral growth
(Borst et al., 2001)
Since FLP and Cre recombinases catalyze site-specific recombination in both direc-
tions, i.e. excision as well as insertion, it was examined whether the recombinases can be
used for rapid insertion of DNA sequences (Menard et al., 2003; Bubeck et al., 2004; E.
Borst and M. Messerle, unpublished). The candidate genes are cloned into a suicide vector

Manipulating Cytomegalovirus Genomes by BAC Mutagenesis |  ��
that carries an FRT site, the R6K replicon and a small antibiotic resistance gene (zeocin
or kanamycin). An FRT target site has to be anchored in the CMV BAC first (e.g. by
ET mutagenesis) and after transient expression of FLP recombinase in the bacteria the
suicide plasmid can be efficiently and rapidly integrated into the CMV genome. The vector
backbone is integrated together with the candidate gene, thus slightly enlarging the viral
genome, but the vector sequences encompass 1.5 kbp at most and this is usually tolerated.
Bubeck et al. (2004) have successfully used the method to generate hundred MCMV mu-
tants in order to analyze the function of the essential M50 gene.
Despite the success and the justification of the approach one has to add a word of
caution: even if a gene is transferred together with its promoter to an ectopic position, one
cannot necessarily expect its expression level to be unaffected. For instance, the ectopic re-
insertion of the MCMV m157 gene into an m157 deletion genome led to an only partial
rescue of the mutant’s phenotype, i.e. susceptibility of the resulting virus to NK cell control
in C57BL/6 mice (Bubic et al., 2004). Likewise, the transfer of the HCMV lytic origin of
replication to another genome position did not restore the full growth capacity (E. Borst
and M. Messerle, unpublished). Obviously, the influence of neighboring genetic elements
on the activity of the transferred genetic elements cannot be neglected and ectopic inser-
tion can therefore represent just one of several genetic methods used to investigate viral
functions.
Complementation of essential genes and conditional mutants
Only BAC-based mutagenesis techniques allow the disruption of essential genes, as the
construction of mutant genomes becomes independent of their biological properties. Trans-
fection of CMV genomes carrying lethal mutations is characterized by a failure in plaque
formation and aside from the information that the gene is essential one does not get any
further insight. In order to learn at which step of the replication cycle the gene is involved, it
would be desirable initially to complement the missing function and grow up the virus and
then to perform infection experiments under non-complementing conditions.
In one of the first analyses of essential CMV genes, transient co-transfection experi-
ments with a glycoprotein B (gB) knock-out genome of HCMV and gB expression plas-
mids were performed (Strive et al., 2002). Although the complementation was far from
satisfactory, leading to mini-plaques and the release of a limited number of infectious viri-
ons only, the experiments gave us at least an idea as to which domains of the glycoprotein
are important and which are dispensable. In the case of the MCMV ie3 deletion mutant,
a complementing cell line based on NIH 3T3 cells could be generated and the mutant
grown up (Angulo et al., 2000). Since the titers of the ie3 mutant on the complement-
ing cell line remained one to two orders of magnitude below those of the wild-type virus,
it appeared that either the amount of the trans-complementing protein or the temporal
expression pattern were inadequate for full complementation. A similar observation was
made with HCMV UL99 mutants. On complementing cells they produced substantially
reduced virus yields in comparison to the parental virus (Silva et al., 2003). Nevertheless,
sufficient quantities of the mutant viruses could be obtained to perform infection experi-
ments. The UL99 mutants were not impaired in DNA replication and gene expression but
capsids accumulated in the cytoplasm, suggesting that UL99 plays an essential role for the
envelopment of HCMV virions.