CHAPTER 4
Engineering of Hybrid phiC31-Zinc Finger Integrases
with Altered Site-Specificity
In this study, I performed all experiments and constructed all plasmids used except for the
inversion assay plasmids, which were built in collaboration with Joylette Portlock.

95
ABSTRACT
The modular domain structure of site-specific serine-recombinase family
members invites the notion that domains can be swapped to engineer a change in
sequence recognition. Here, we show that by fusing the catalytic domain of phiC31
integrase to the mouse transcription factor Zif268, hybrid recognition sites are
effectively recombined. This is the first demonstration of a functional serine integrase-
zinc finger hybrid in a mammalian cell context. The most notable zinc finger hybrids to
date have relied upon the FokI endonuclease, which creates uncontrolled double strand
breaks, as the catalytic domain. As phiC31 recombines strands via a controlled double
strand break reaction, a potentially safer alternative to the FokI endonuclease is now
available. We also show evidence in mammalian cells that the phiC31 catalytic domain
has the ability to recombine its attB/attP recombination sites without the need for its
DNA binding domain. Our results suggest that site-specificity could be engineered by
mutating the catalytic domain of phiC31 in addition to swapping its DNA binding
domain.

96
INTRODUCTION
Precisely targeted genetic modification is a highly desirable feature for
therapeutic gene therapy. Site-specific integration of a donor gene of interest and a
locus within the human genome would help to alleviate some of the safety concerns
associated with gene therapy. In the case of whole-gene additions, it would possibly
create a more stable and long-lasting effect, devoid of silencing. The holy grail of gene
therapy would be genetically modifying a single nucleotide of an affected gene with
site-specificity through homologous recombination. To date, the most successful gene
therapy trials have utilized viruses. While viruses have proven to be well adapted to
delivering therapeutic genes, they fail to provide a site-specific integration profile. Site-
specific recombinases are non-viral alternatives, capable of catalyzing a recombination
event between two strands of DNA through cleavage and the rejoining of ends. While
more precise than viral vectors for gene therapy, these recombinases will still have to be
engineered to gain the desired specificity for gene therapy.
There are two large families of recombinases, catalyzed by tyrosine and serine.
The former family includes the Cre and Flp resolvases, which have been well described
(1, 2, 3). Briefly, Cre protein is capable of recombining two identical loxP sites. LoxP
sites are 34 bases in length and consist of an asymmetric 8 base-pair core flanked by
two 13 base-pair (b.p.) palindromic stretches. Flp protein is analogous to Cre. It
recombines two identical FRT sites, which are 34 bases in length, including an 8 base
core. While both Cre and Flp are efficient at catalyzing the recombination between their
respective DNA recognition sites, LoxP and FRT, those sites do not natively exist in the
human genome. For practical gene therapy, an alternative is to mutate and evolve Cre or
Flp towards a locus in the genome. This has been accomplished to some degree with
both enzymes through a variety of methods (4, 5). Most of this success, however, has
been limited to bacterial assays. It is believed that this is due to the single module-like
structure of the tyrosine recombinases; the catalytic and binding domains are

97
intertwined (4). This makes mutating the binding domain to alter site-specificity a great
challenge. Additionally, both Cre and Flp recombine identical half-sites, lending to the
occurrence of back-reactions and excision of any integrated DNA.
In contrast, the serine recombinase family is structurally different than the
tyrosine family. Family members include resolvases, invertases and integrases (6). Like
their tyrosine counterparts, they catalyze a recombination event between two target
sequences. Unlike Cre and Flp, this family has shown promise in swapping DNA
binding domains in order to alter the enzyme’s natural recognition site (7, 8, 9, 10). This
is due to the structural nature of serine recombinases, which have two distinct modular
domains consisting of an N-terminal catalytic domain and a C-terminal DNA binding
domain. PhiC31 integrase, in particular, may be best suited to such a substitution of its
C-terminal DNA recognition domain.
PhiC31 consists of an approximately 170 amino acid N-terminal catalytic
domain fused via a short coiled-coil domain to a much longer 430+ amino acid DNA
binding domain (6). PhiC31 is capable of recombining non-homologous recognition
sequences termed attP and attB. By placing the 34 b.p. minimal attB onto a donor
plasmid, any plasmid can be unidirectionally recombined with another plasmid or
genome harboring the 39 b.p. minimal attP site. The attP site does not natively exist in
mammalian genomes; however, sites with partial homology to attP, termed pseudo attP,
are capable of recombining with attB. It has been previously demonstrated that a
phiC31 integrase variant, with an inactivated catalytic domain, is still capable of
binding a target (11). We reason then that swapping the phiC31 DNA binding domain
with another recognition domain could lead to altered site-specificity. Such domain-
swapping experiments amongst members of the serine recombinase family have been
demonstrated (7, 8, 9), although with limited success.
An alternative to fusing another serine family member recognition domain to the
catalytic domain is to use the powerful DNA binding capabilities of zinc finger

98
transcription factors. By fusing zinc finger binding domains to endonucleases, termed
zinc finger nucleases (ZFNs), homologous recombination in mammalian cells has
improved several thousand-fold (12). Additionally, these ZFNs are capable of being
redesigned to bind to native genomic sequences at high frequencies. One current
drawback is that the ZFNs are typically fused to an endonuclease that creates a double
strand DNA break (DSB) and relies upon host co-factors to complete the ligation.
Uncontrolled DSBs are known to be contributors to oncogenesis and translocations
(13). This is especially dangerous when the DSB occurs in an off-target site, as has been
reported (14). Unlike endonucleases, phiC31 catalyses a four-strand staggered DNA
break and covalently attaches to the 5’-ends in a transitory recombination step. The
phiC31 integrase then rotates one of the strands 180 degrees and completes the ligation
(11). By combining a designed ZFN with the catalytic domain of phiC31, precise and
safe site-specific integration may be achieved. This has already seen some proof of
principle success with Tn3 resolvase, a distant serine recombinase family member, by
fusing its catalytic domain to the mouse transcription factor Zif268 (10). These
chimeras were capable of recombining two truncated Tn3 res I sites, which were
flanked by inverted repeats of the Zif268 recognition site. At the time of embarking on
the study reported here, this functionality had only been demonstrated in a bacterial
assay context.
Barbas and colleagues were able to demonstrate zinc finger-recombinase
activity in 293 human embryonic kidney cells with various invertases (15). These
invertases, along with the zinc-finger Tn3 resolvase chimera, are hyperactive mutants
insensitive to sequence orientation. This leads to back reactions in which the integrated
product can be excised. Therefore, a unidirectional reaction, such as that promoted by
phiC31, would be an improvement upon the current tool-kit available for gene therapy.
Here, we describe the first hybrid fusion constructs involving the putative phiC31
catalytic domain and the mouse transcription factor, Zif268. We show that a set of these

99
hybrids are capable of resolving two plasmids harboring the phiC31 catalytic core
flanked by the Zif268 DNA binding sites. Moreover, we demonstrate the utility of these
hybrids in a mammalian cell context.

100
MATERIALS AND METHODS
PhiC31-Zif268 chimeras. Plasmid pCMVInt (16) was cut with SpeI and XhoI to
liberate the wild-type phiC31 coding fragment from the vector pCMV. This fragment
was then used as the PCR target for producing truncated phiC31 versions from either
the first 170 or 182 amino acids, each with a 3’ stop codon and a designed AflII
restriction site just 5’ of the stop codon. These fragments were then cloned back into
vector pCMV with SpeI and XhoI to create pCSI-170-stop and pCSI-182-stop. Next,
the mouse transcription factor Zif268 was manufactured from the PCR of mouse
genomic DNA. To create pCSI-170-Zif and pCSI-182-Zif, primers 5’-Afl2Zif and 3’-
NheIZif were used on mouse genomic DNA. For pCSI-170-L6-Zif, primers 5’-
AFL2L6Zif and 3’-NheIZif were used on mouse genomic DNA. These were then cut
with AflII/NheI and ligated together with similarly cut pCSI-170 or pCSI-182. To
create control plasmid pCSZif, primers 5’-SpeI-start-Zif and 3’-NheIZif were used and
the product ligated into vector pCMV. Lastly, plasmid pCS (16), which was created by
removing the encoded phiC31 from pCMVInt, was used as a negative control. Primers
to obtain the Zif268 fragments were designed as follows: 5’-Afl2Zif was 5’-
gcctaaggaacgcccatatgcttgccctgtc; 3’-NheIZif was 5’-gctagcctagtccttctgtcttaaatggattttg;
5’-AFL2L6Zif was 5’-
gccttaagggcagcggaggcagcggcggatccggcggcagcggcaccagcgaacgcccatatgcttgccctgtcgagtc
ctg; and 5’-SpeI-start-Zif was gcactagtatggaacgcccatatgcttgccctgt. Genomic DNA was
obtained via retrooribital bleed of an 8-week old BALB/c mouse (Jackson Labs, Bar
Harbor, Maine), followed by DNA isolation with a DNEasy kit (Qiagen, Germany).
This animal was housed in a pathogen-free environment at the Stanford Animal Facility

101
under guidance from the Administrative Panel on Laboratory Care (A-PLAC). All
constructs were verified with sequencing and analytical digests.
Inversion assay plasmids. The three inversion assay plasmids used, pBPZG-A,
pBPZG-B, and pBPZG-C, were synthesized as follows: The attP and attB phiC31
recognition sites and an inverted CMV promoter were removed from pBPGreen (17) by
cutting with SpeI and AgeI. An MCS containing EcoRI and XhoI ends was cloned into
the SpeI and AgeI sites to create an intermediate cloning vector. Plasmid pCMVInt was
used as a target for copying the CMV promoter flanked by the various hybrid attB and
attP recognition sites. pBPZG-A- Primers used: Z0attB 5’-
cggaattcgcgtgggcgggagcccaagggcacgcgcccacgcgagtccggaggctggatcggtcccgg-3’ and
Z0attP 5’-cggaattcgcgtgggcggaactcaaaggttaccgcccacgcgtgcaggtcgttacataacttacggtaa-3’.
The 3’-end of the CMV promoter was primed with primer Z0attB containing 20 bases
of CMV at the far 3’-end, the central 16-nt from the phiC31 attB core (underlined)
surrounded by the 9-nt inverted repeat of the Zif268 recognition site (italicized) and an
EcoRI restriction site for cloning. The corresponding 5’-end primer, Z0atP, contains 16-
nt from the central core of attP (underlined) flanked by the same Zif268 inverted
repeats (italicized), the first 20 bases of the 5’ CMV promoter and an EcoRI cloning
site. pBPZG-B – constructed similarly with addition of two linkers (capitalized)
between the attB/P core and Zif sites. Primers Z6attB 5’-
cggaattcgcgtgggcgACGAggagcccaagggcacgTGCATcgcccacgcgagtccggaggctggatcggtc
ccgg-3’ and Z6attP 5’-
cggaattcgcgtgggcgACGAagaactcaaaggttacTGCATcgcccacgcgtgcaggtcgttacataacttacggt
aa-3’. pBPZG-C – similar to pBPZG-A, but with a 22-nt attB/P core. Primers Z22attB
5’-cggaattcgcgtgggcgcggggagcccaagggcacgccccgcccacgcgagtccggaggctggatcggtcccgg-
3’ and Z22attP 5’-
cggaattcgcgtgggcggagagaactcaaaggttacccccgcccacgcgtgcaggtcgttacataacttacggtaa-3’.

102
Resolving assay plasmids. Plasmid pCB contains the phiC31 306 b.p. attB
directly 3’ of the CMV promoter. It was created by digesting pNC-attB (17) with SpeI
and NsiI to remove CMV-eGFP, blunted and then religated. Plasmid pattP-Green
contains a 255 b.p. attP and eGFP gene. It was created by first digesting pBPGreen with
SpeI and AgeI to remove the attP and attB and create cloning vector pGreen-SpeI/AgeI.
PCR of this digested product for the 255 b.p. attP with SpeI and AgeI ends was then
performed and ligated into the pGreen vector. Plasmid pCZ6B contains a CMV
promoter upstream of a 16-nt attB core being flanked by inverted repeats of the Zif268
binding site and linkers ATGCA and TCGT between the core and Zif268 sites. This
was created by digesting plasmid pBPZG-B with MluI to excise the attP-Zif268 site
hybrid and then religating the cut vector to itself. Finally, plasmid pZ6PGreen contains
the eGFP gene and the 16-nt attP core flanked by the same linkers and Zif sites as in
pCZ6B. It was created by digesting pBPGreen with SpeI and XbaI to give the cloning
vector pGreen-SpeI/XbaI. Plasmid pBPZG-B was then used as a template to PCR for
the attP-Zif-linker hybrid with compatible SpeI and XbaI ends to clone into vector
pGreen-SpeI/XbaI.
Plasmid pCZ6B-d is the same as pCZ6B, but with the two Zif sites situated as
direct repeats. It was created by digesting pCZ6B with BbsI and EcoRI to remove the
hybrid attB core. A double stranded oligo with the entire 16 b.p. attB core flanked by
direct repeats of the Zif site was then ligated into the cut vector with compatible
EcoRI/BbsI sticky ends. Plasmid pZ6PGreen-d has both Zif sites situated as direct
repeats surrounding the 16 b.p. attP core. It was created by using pZ6PGreen as a PCR
template for primers Mlu-Zif-attP 5’-gtacgcgtcgcccacgcatgcaaacctttgag and MluI-green
5’-gtacgcgttaagatacattgatga. Mlu-Zif-attP contains the direct repeat Zif site just 5’ of
the 16 b.p. attP core. Plasmid pZ6PGreen was digested with MluI to remove the entire
hybrid attP-Zif core and eGFP. The PCR for the direct repeat gave a 1088 b.p. product,
which was then digested with MluI and ligated into the MluI vector, pZ6PGreen.

103
Recombination assays. 293 human embryonic kidney cells (ATCC, Manassas,
VA) were grown in Dulbecco’s Modified Eagle Medium (Invitrogen, Carlsbad, CA)
and supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. The
day before transfection, cells were seeded in 60 mm dishes to reach 60-80% confluency
upon transfection. FuGENE6 (Roche, Indianapolis, IN) reagent was used to transfect
cultures. For the inversion assays, 500 ng of a donor plasmid (pBPZG-A/B/C) was co-
transfected with 500 ng of pCMVInt, the phiC31-Zif268 hybrid constructs, or control
constructs. For the resolving assays, 250 ng each of pCB and pattP-Green were co-
transfected with pCMVInt as a positive control. 250 ng each of pCZ6B and pZ6PGreen
were co-transfected with one of the various phiC31-Zif268 hybrid constructs or by
themselves. Seventy-two hours after transfection, cells were resuspended and analyzed
for green fluorescence with the Guava bench top flow cytometer (Guava Technologies
Inc, Hayward, CA).
Recombination sequence analysis. Twenty-four hours prior to harvesting,
transfected cells were supplemented with 50 U/ml DnaseI (Invitrogen, Carlsbad, CA) to
reduce the background of any untransfected DNA. At 72 hours after transfection, low
molecular weight DNA was recovered as describe by Hirt (18). 100 ng of purified Hirt
extract was then amplified with PCR using primers that produce a 789 b.p. product on
any plasmids that underwent successful recombination. Primer flipdiag4 was 5’-
gctgaccgcccaacgaccc and primer GFPrev1 was 5’-cggtggtgcagattgaactt. This PCR
product was gel-purified (Qiagen, Germany) and cloned into a TOPO TA cloning
vector (Invitrogen). Ligated product was transformed into electrocompetent DH10B E.
Coli and plated overnight at 37°C onto KanR agar plates. White colonies were picked,
minipreped (Qiagen), and sent out for sequencing with the M13 Reverse primer.
Secondary structure analysis. Primary amino acid sequences for serine-
recombinase family members were analyzed with the secondary structure prediction
program JPRED (http://www.compbio.dundee.ac.uk/Software/JPred/jpred.html).

104
RESULTS
PhiC31-Zif268 chimeras are unable to recombine plasmids bearing hybrid
recognition sites. A previous study that fused the first 144 amino acids of Tn3 resolvase
to the DNA binding domain of Zif268 showed high recombination frequencies of
hybrid recognition sites on a single plasmid (8). In that study, a galK marker was
surrounded by two Tn3 catalytic cores each flanked by the Zif268 recognition site.
Successful Tn3-Zif hybrids were able to excise the galK marker thus making E. coli
transformants turn pale yellow. We sought to mimic this recombination in a mammalian
cell context with hybrid phiC31-Zif268 recombinases. To quantify recombination
activity, an inversion assay was developed. In this assay, eGFP is downstream of an
inverted CMV promoter. Flanking the CMV promoter are hybrid versions of the phiC31
catalytic core and Zif268 recognition site. At one end, the catalytic core is a shortened
attB site, which contains the TT dinucleotide core necessary for recombination (19). At
the other end of the inverted CMV is a shortened attP site, which also contains the TT
dinucleotide catalytic core. Each truncated att site is flanked by two 9 b.p. motifs
recognized by the DNA binding domain of Zif268. A functional hybrid integrase will
recombine the hybrid cores, thereby inverting CMV back to its proper orientation for
driving GFP expression. We used the best linker from the Tn3 study, Z+6, to fuse in
between the phiC31 catalytic domain and the Zif268 DNA binding domain. Two
additional hybrid recognition sites were also designed for use in the inversion assay.
Figure 4-1a illustrates the three constructs used in the inversion assay. For each assay
plasmid, we chose to use a minimum core size of 16 b.p. or 22 b.p. for both attB and
attP. This core length was based upon the known minimal length attB and attP sites
(20) and the Tn3 chimera study (10).
The putative catalytic domain of phiC31 integrase spans approximately amino
acids 1-170 (6). Therefore, we built a phiC31 catalytic-Zif268 DNA binding domain
hybrid that used the first 170 amino acids of phiC31 and the 88 amino acid Zif268 DNA

105
binding domain. A second hybrid was created using the first 182 amino acids of
phiC31. By inserting a 14 amino acid linker, L6, between the Tn3 resolvase and Zif
binding domain, Stark and colleagues were able to increase recombination frequencies
with their hybrids (10). Therefore, we built a third hybrid that contained this same
linker, L6. In addition, we built and tested two controls that contained just the 88 amino
acid Zif binding domain followed by a stop codon or the first 182 amino acids of
phiC31 followed by a stop codon. As a positive control, pCMVInt, which encodes wild-
type phiC31, and pBPGreen, which contains the full-length attB and attP cores flanking
an inverted CMV promoter, were used. Each of these hybrid constructs is illustrated in
figure 4-2.
The hybrid integrases and inversion plasmids were co-transfected, in various
combinations, into 293 human embryonic kidney cells. The level of GFP expression
was measured via flow cytometry 72 hours after transfection. Our results suggested that
phiC31-Zif268 hybrids, as built for this study, were unable to catalyze the inversion of
hybrid recognition sequence sites (Fig. 4-3). Curiously, some inversion was seen with
wild-type phiC31 on each of the three assay plasmids, pBPZG-A, pBPZG-B, and
pBPZG-C. This suggested that the minimal attB and attP lengths may in fact be shorter
than what was previously reported as 34 b.p. and 39 b.p. for attB and attP, respectively
(20). However, GFP expression from these hybrid inversion plasmids was significantly
below pBPGreen containing the full-length attP and attB. Also of note is that the
control integrase encoded by pCSI-182-stop was able to invert the hybrid assay
plasmids at least as efficiently as the full-length phiC31 integrase.
The phiC31 catalytic domain provides basal DNA recognition and
recombination activity. It is an interesting finding that the phiC31 catalytic domain
might be able to recognize its DNA binding substrate at some level. The observed
recombination activity of pCSI-182-stop, which contains only the N-terminal phiC31
catalytic domain, goes against previously published results of the serine recombinase

106
family (6, 21). Therefore, to confirm recombination activity, we transfected 293 human
embryonic kidney cells with the following groups: pCS + pBPGreen, pCMVInt +
pBPgreen, pZif-stop + pBPGreen, and pCSI-182-stop + pBPGreen. 500 ng of each
plasmid were transfected, to give a total of 1 µg DNA per group. Seventy-two hours
after transfection plasmid DNA was isolated and PCR was performed to detect
inversion events. We were able to detect PCR bands in both the pCMVInt and pCSI-
182-stop groups, but not in the pCS or pZif-stop groups (Fig. 4-4). The PCR products
from the pCMVInt and pCSI-182-stop groups were then sequenced to determine how
accurately the inversions had been completed. Wild-type phiC31, encoded by
pCMVInt, created perfect attB/attP recombination events at the dinucleotide core, as
previously reported (20). In contrast, sequencing results of the truncated 182 amino acid
phiC31 showed numerous insertion and deletion events within 100 bases flanking both
sides of the TT core. This suggests that the phiC31 catalytic domain was capable of
basal levels of recombination, but that the DNA binding domain was a requirement for
reliable recombination. Inefficient recombination was probably due to an inability to
form proper synapse complexes between DNA strands, a trait normally performed via a
concerted effort of both the catalytic and DNA binding domains (11).
While the large serine integrases, such as phiC31, are greater than 500 amino
acid residues in length, Tn3 resolvase and are considerably shorter at ~184 residues.
The catalytic domains for the shorter Tn3 resolvase and invertase lie within the first
~140 amino acids, while the remaining ~43 amino acids are responsible for DNA
binding (22). We used a secondary structure prediction program, JPRED, to determine
if there was any homology between phiC31 and the DNA binding regions of Tn3 and
(Fig. 4-5). Secondary structures for 27 other serine recombinase family members were
similarly predicted. Each of the larger serine recombinases seemed to correlate well up
until amino acids 170-190. This region was terminated by an elongated helix (in blue)
and coiled region (yellow). The same ~170 amino acid region seemed to be expanded,

107
but with a similar secondary structure to the ~140 amino acid catalytic domains of Tn3
and . In contrast, the 43 amino acid DNA binding domains of Tn3 and showed no
similarity with any section of phiC31’s secondary structure. The structural predictions
and biochemical results were suggestive of a model in which the phiC31 catalytic
domain retained DNA sequence recognition, apart from the mechanism occurring in
shorter serine recombinases such as Tn3 and . Adding to this model are mutants with
point mutations in the phiC31 catalytic domain that resulted in higher attP specificity
(23).
PhiC31-Zif268 chimeras can resolve two plasmids bearing hybrid recognition
sites. The inversion assay plasmids, pBPZG-A/B/C, seemed to suffer from marginal
background GFP expression (Fig. 4-3, pCS control). Therefore, we separated the CMV
promoter and eGFP gene into two plasmids, pCZ6B and pZ6PGreen (Fig. 4-1b). As
pBPZG-B resulted in the highest GFP expression in the inversion assay, it was chosen
as the model for pCZ6B and pZ6PGreen. For this assay, any functional hybrid
integrases should be able to resolve the pCZ6B and pZ6PGreen into a single monomer.
Hybrid recognition sites were oriented such that the CMV promoter would recombine
upstream of eGFP. Hybrid integrase activity could then be quantified as GFP expression
via flow cytometry. While the positive control group, pCB + pPGreen + pCMVInt,
demonstrated resolution, the hybrid constructs failed to show any significant GFP
expression (Fig. 4-3). It was possible that this was, in part, due to the orientation of the
Zif recognition sites flanking attB/attP. Therefore, we modified the Zif recognition sites
such that they were situated as direct repeats, rather than inverted, flanking the 16 b.p.
attB and attP cores (Fig. 4-1b). In addition, we constructed similar hybrids that
contained just the first 144 amino acids of phiC31. This served two purposes: first, the
successful Tn3 study (10) had used only the first 144 amino acids of Tn3. Second, we
were curious to study the effects of truncation on recombination activity. By truncating

108
the pCSI-182-stop hybrid down to the first 144 amino acids, a better understanding of
phiC31’s catalytic domain DNA recognition capability might be achieved.
Figure 4-6 details the GFP expression results from the assay plasmids modified
with Zif-site direct repeats. Noticeably, pCSI-170-Zif was significantly higher than the
other hybrids and the full-length phiC31 (pCMVInt, p-value: 4.15x10
-4
). This indicated
that the Zif DNA binding domain was necessary for efficient recombination between
hybrid sites. The significance was even greater between pCSI-170-Zif and the construct
coding for just the Zif binding domain (p-value: 2.94x10
-5
). The positive control for this
assay, wild-type phiC31 (pCMVInt) plus plasmids bearing full-length attB/P sites, had
the highest GFP expression, as expected. Sequence analysis of the recombined assay
plasmids in the pCSI-170-Zif group showed that the crossover had occurred within 6
b.p. of the hybrid TT dinucleotide core (table 4-1). The same recombination event was
observed with wild-type phiC31 integrase resolving the two assay plasmids.
Several of the hybrids appeared to actually suppress recombination when
compared to the donor only control (pCZ6B-d + pZ6PGreen-d). These suppression
hybrids included the truncated pCSI-144-Zif and pCSI-144-L6Zif. One explanation is
that the pCSI-144 hybrids were binding to their target sites, but blocked recombination
because of a loss in the ability to rotate the DNA strands into a proper orientation.
Another possible explanation is that catalysis via the catalytic serine in phiC31 was
blocked due to tertiary structural changes from the truncation. Future experiments
involving the substitution of amino acids 145-182 with random amino acids could
elucidate whether inhibition was due to a loss in length or because of structural changes.

109
DISCUSSION
This study demonstrates that phiC31 integrase is a modular enzyme capable of
undergoing domain swapping to alter its specificity. Moreover, the substituted DNA
binding domain need not be a related protein, as zinc finger proteins are not serine
recombinase family members. Previously, Tn3 resolvase, another serine recombinase
family member, was successfully engineered to behave as a zinc finger hybrid in E. coli
(10). We hypothesized that a phiC31-zinc finger chimera would be capable of
recombining hybrid recognition sites in human cells. It is an important finding, then,
that such chimeras are functional in higher eukaryotic cells. In human gene therapy, off-
target integration is a valid concern. Zinc finger nucleases have already shown great
promise in human cells (24). However, previous work in human cells has relied upon
endonucleases, such as FokI, which create an uncontrolled double strand break prior to
insertions. The results shown here suggest that phiC31 integrase is a viable alternative
when a more controlled break is required. In addition, the unidirectional nature of
phiC31 potentially avoids a back reaction that could occur with resolvase and invertase
zinc finger hybrids in human cells (15). Further testing is needed to determine if any
change from the wild-type phiC31 synapse and recombination mechanism has occurred
in these new hybrids.
We showed that the pCSI-170-Zif hybrid outperformed four other hybrids in a
mammalian cell assay. Unlike the Tn3 “Z-resolvase” chimeras, phiC31 Zif268 hybrids
showed higher recombination frequencies without the L6 14-amino acid linker. How the
L6 is hindering recombination is unknown. It could be blocking proper synapse
formation due to an extension in length between the phiC31 catalytic and Zif binding
domains. Alternatively, it may be hindering catalytic cleaveage at the TT dinucleotide
core. And lastly, binding to the Zif sites might be disrupted by the added length that
could be “pushing” the Zif binding domain further out from the TT core. The problem is
probably not due to the sequence of the L6 linker, as the same observations were made

110
with the pCSI-182-Zif hybrid, which is two residues shorter than the pCSI-170-L6Zif
hybrid.
The differences between the hybrids highlight the need to test additional hybrid
recognition sites, something that we did not fully explore in this study. We tested three
hybrid recognition sites. Of the three, Z+6 performed the best in the initial inversion
assay, which was then used for subsequent testing. The purpose of the Z+6 filler was to
provide room for both the catalytic and DNA binding domains to recognize and cleave
their attachment sites. It is quite possible that altering the length of either the Z+6 filler
or the attB/attP cores would lead to higher recombination activity.
In light of the evidence that the 182 amino acid catalytic domain is capable of
recombination (Figs. 4-3, 4-4), we may need to redefine the N-terminal and C-terminal
boundaries. For the large serine-recombinases, the N-terminus has been defined as the
catalytic domain, while the C-terminus is responsible for DNA recognition and initial
binding. In the current model from in vitro studies, the C-terminal domain will bind to
sequences flanking the att core. It has been shown that the DNA recognition by the C-
terminus occurs at base pair positions +/- 15 away from the TT core (19). Mutating
those base pair positions disrupts the formation of a stable synapse in vitro. Since the
182 amino acid truncation is able to complete recombination in mammalian cells,
phiC31 integrase may be working under a different model. Mammalian cell co-factors
may be playing a role in helping to form the required synapse prior to strand cleavage.

111
Table 4-1. Recombination junctions from resolving assay
Sequence attL-Zif hybrid recombination site attB attP
Perfect
gcgtgggcgatgcacgtgcccTTgagttcatcgtcgcccacgc
0 0
pCMVInt
gcgtgggcgatgcacgtgcccttgggCTcatcgtcgcccacgc
+5 +6
pCSI-170-Zif
gcgtgggcgatgcacgtgcccttgggCTcatcgtcgcccacgc
+5 +6
Shown are sequences from recombination between the hybrid attB-Zif and attP-Zif sites on the two
resolving assay plasmids (Fig 4-1b). The hypothetical perfect crossover is shown at the top, pCMVInt just
below, and then the top functioning hybrid from this assay is last, pCSI-170-Zif. The crossover core is
capitalized and bolded. In a perfect crossover the left most 8 b.p. of the 16 b.p. attB core are adjacent to the
right most 8 b.p. of the 16 b.p. attP core. This recombination junction is termed attL. Flanking attL are 4
b.p. and 5 b.p. space sequences (bolded), which are then flanked by direct repeats of the 9 b.p. Zif
recognition sequence (underlined). The crossover distance from the theoretical perfect crossover is
indicated in the attB and attP columns. A plus (+) indicates that the crossover occurred downstream of the
TT dinucleotide core.

112
Figure 4-1. Schematic diagrams of plasmids used in demonstrating phiC31-Zif
hybrid recombination. A. The inversion assay plasmid is shown with the inverted
CMV promoter flanked by hybrid attB-Zif or attP-Zif recognition sites as follows:
pBPZG-A contained just 16 bases of the attB or attP cores surrounded by the 9 base
pair Zif268 binding site shown as white arrows facing the 16 b.p. core in the shaded
box. pBPZG-B has an additional 4 base pair and 5 base pair spacer sequence between
the attB/attP cores and Zif268 site. pBPZG-C is similar to pBPZG-A, but with an
additional 6 base pairs of the attB/attP cores. A positive control, pBPGreen, contains
pCBZ6
pCBZ6-d
pZ6PGreen
pZ6PGreen-d

113
the full-length attB/attP recognition sequences flanking CMV. The attB and attP cores
face away from each other as shown by arrows flanking CMV in the plasmid schematic.
The attB/attP core sequences are shown with the catalytic TT dinucleotide cores
capitalized in bold. A second plasmid encoding either the hybrid or wild-type phiC31
was co-transfected with the inversion plasmid. The positive control, pBPGreen, was co-
transfected with wild-type phiC31. Functional hybrid integrases will invert CMV back
to the proper orientation for driving eGFP expression.
B. The resolving assay donor plasmids. The CMV promoter and the eGFP
encoding gene are on two plasmids. Successful recombination results in CMV
recombining upstream in proper orientation to eGFP. The first assay used inverted Zif
sites in both the CMV and eGFP plasmids (pCZ6B and pZ6PGreen). The second
resolving assay modified the Zif binding sites to be situated as direct repeats on both
plasmids (pCZ6B-d and pZ6PGreen-d). A third plasmid encoding either the hybrid
phiC31-Zif integrases or wild-type phiC31 was co-transfected with the two plasmids
above. Plasmids pCB and pPGreen, which contain the full-length attB/attP cores were
co-transfected with wild-type phiC31 as a positive control.

114
Figure 4-2. Schematic diagrams of five phiC31-zinc finger hybrids and controls.
Each hybrid or control was driven by the CMV promoter and co-transfected with either
the inversion plasmids or resolving plasmids. The black bar represents the 88 amino
acid Zif268 DNA binding domain. The gray bars represent the 613 amino acid version
of phiC31 intgrase, which has been truncated to 144 aa, 170 aa, or 182 aa. pCMVInt
encodes the full 613 amino acids of phiC31. The narrow black bar is the 14 amino acid
long L6 linger. These hybrids, truncations or controls were all cloned into the same
Amp
R
vector backbone, as shown above to the right.

115
Figure 4-3. Inversion assay of hybrid phiC31-Zif integrases on various hybrid
recognition sites. Graph shows GFP expression measured in mean fluorescence 72
hours after transfection with one of three inversion plasmids (pBPZG-A/B/C) and
various hybrid phiC31-Zif integrases. None of the assay plasmids showed significant
expression with the hybrid integrases, though pBPZG-B demonstrated a modest
advantage. Noticeably, pCMVInt, which encodes the wild-type phiC31, was able to
recombine attB/attP cores as small as 16 base pairs. Also of note was the negative
control pCSI-182-stop, which was able to recombine the hybrid recognition sites. A
postive control using pCMVInt and an inversion plasmid with full-length attB/attP sites
(pBPGreen) performed as expected. Four replicates of each group were tested in
separate transfections. Standard error bars are shown.

116
Figure 4-4. PCR to detect recombination activity of a truncated phiC31 integrase.
293 human embryonic kidney cells were co-transfected with pBPGreen and wild-type
phiC31 (pCMVInt) or one of three controls. Seventy-two hours after transfection, low
molecular weight DNA was isolated and amplified for the presence of the attL
recombination site on pBPGreen. A. Schematic diagram of pBPGreen with correct
orientation of the CMV promoter and arrows indicating direction of the PCR primers.
B. The visible bands corresponded to 789 base pairs. PCR products were gel purified
and sequenced to determine how accurately recombination had occurred by pCMVInt
and the truncated 182-residue version. Wild-type phiC31 was perfect at the TT core,
meaning attB and attP recombined without any deletions or insertions. In contrast,
pCSI-182-stop had numerous single base pair insertions and deletions surrounding the
TT core.
A. B.

117
Figure 4-5. Secondary structure prediction of serine recombinase family members.
We used the JPRED secondary structure prediction program to determine how similar
phiC31’s catalytic domain was with Tn3 resolvase. An alpha helix is represented as
blue, beta folds as red, and coiled regions as yellow. All of the larger serine family
members, which does not include Tn3 and , share similarity with phiC31 up until
approximately residue 180. The phiC31 catalytic domain appeared to be an expanded
version of the Tn3 and catalytic domains, which span residues 1-140. The DNA
binding domains of Tn3 and , residues 140-184+, do not appear to share similarity
with any part of the phiC31 catalytic domain shown to be capable of recombination
(Figs. 4-3, 4-4). Tn3 and do seem to share similarity with many of the other large
serine-recombinases, with three alpha helices extending out to residue ~250.

118
0
10
20
30
40
50
60
70
80
pCMVInt +
attP/B
Donor only pCSI-144-
Stop
pCSI-182-
stop
pCSZif-stop pCMVInt pCSI-170-
Zif
pCSI-170-
L6Zif
pCSI-182-
Zif
pCSI-144-
Zif
pCSI-144-
L6Zif
M
e
a
n

F
l
u
o
r

V
a
l
u
e
Figure 4-6. Resolving assay of phiC31-zinc finger hybrid recombination activity.
293 human embryonic kidney cells were co-transfected with the two resolving plasmids
(Fig. 4-1b) and the constructs listed above. Seventy-two hours after transfection, GFP
expression was quantified via flow cytometry. Hybrid pCSI-170-Zif recombined the
hybrid phiC31-Zif substrates at the highest frequencies (p-value: 4.15x10
-4
compared to
wild-type phiC31, pCMVInt). pCMVInt was also tested with its full-length attP and
attB recognition sites as a positive control (far left measurement). All groups were
tested in triplicate. Standard error bars are shown.

119
REFERENCES
1. Stark WM, Boocock MR and Sherratt DJ. 1992. Catalysis by site-specific
recombinases. Trends Genet. 8:432-9.
2. Sauer B. 1994. Site-specific recombination: developments and applications. Curr.
Opin. Biotechnol. 5:521-7.
3. O'Gorman S, Fox DT and Wahl GM. 1991. Recombinase-mediated gene
activation and site-specific integration in mammalian cells. Science 251:1351-5.
4. Coates CJ, Kaminski JM, Summers JB, Segal DJ, Miller AD, and Kolb AF.
2005. Site-directed genome modification: derivatives of DNA-modifying enzymes
as targeting tools. Trends Biotech. 23:407-19.
5. Hartung M and Kisters-Woike B. 1998. Cre mutants with altered DNA binding
properties. J. Biol. Chem. 273:22884–91.
6. Smith MC and Thorpe HM. 2002. Diversity in the serine recombinases. Mol.
Microbiol. 44:299-307.
7. Avila P, Ackroyd AJ, and Halford SE. 1990. DNA binding by mutants of Tn21
resolvase with DNA recognition functions from Tn3 resolvase. J. Mol. Biol.
216:645-55.
8. Grindley ND. 1993. Analysis of a nucleoprotein complex: the synaptosome of
gamma delta resolvase. Science. 262:738-40.
9. Akopian A, He J, Boocock MR, and Stark WM. 2003. Chimeric recombinases
with designed DNA sequence recognition. Proc. Natl. Acad. Sci. U S A. 100:8688-
91.

120
10. Smith MC, Till R, Brady K, Soultanas P, Thorpe H, and Smith MC. 2004.
Synapsis and DNA cleavage in phiC31 integrase-mediated site-specific
recombination. Nucleic Acids Res. 32:2607-17.
11. Schneider F, Schwikardi M, Muskhelishvili G, and Dröge P. 2000. A DNA-
binding domain swap converts the invertase gin into a resolvase. J. Mol. Biol.
295:767-75.
12. Porteus MH and Baltimore D. 2003. Chimeric nucleases stimulate gene targeting
in human cells. Science. 300:763.
13. Agarwal S, Tafel AA, and Kanaar R. 2006. DNA double-strand break repair and
chromosome translocations. DNA Repair. 5:1075-1081.
14. Beumer K, Bhattacharyya G, Bibikova M, Trautman JK, and Carroll D. 2006.
Efficient gene targeting in Drosophila with zinc-finger nucleases. Genetics.
172:2391-403.
15. Gordley, RM, Smith JD, Graslund T, and Barbas CF. 2007. Evolution and
programmable zinc finger-recombinases with activity in human cells. J. Mol. Biol.
367:802-13.
16. Olivares EC. 2003. Phage integrases for gene therapy: from concept to application.
Doctoral Thesis. Stanford University Publishing. Chapter 6:149.
17. Thyagarajan B, Olivares EC, Hollis RP, Ginsburg DS and Calos MP. 2001.
Site-specific genomic integration in mammalian cells mediated by phage phiC31
integrase. Mol. and Cell Biol. 21:3926-34.

121
18. Hirt B. 1967. Selective extraction of polyoma DNA from infected mouse cultures.
J. Mol. Biol. 26:365-9.
19. Gupta M, Till R, Smith MC. 2007. Sequences in attB that affect the ability of
phiC31 integrase to synapse and to activate DNA cleavage. Nucleic Acids Research
35:3407-19.
20. Groth AC, Olivares EC, Thyagarajan B, and Calos MP. 2000. A phage integrase
directs efficient site-specific integration in human cells. Proc. Natl. Acad. Sci. USA
97:5995-6000.
21. Ghosh P, Pannunzio NR, and Hatfull GF. 2005. Synapsis in phage Bxb1
integration: selection mechanism for the correct pair of recombination sites. J. Mol.
Biol. 349:331-48.
22. Rimphanitchayakit V, Hatfull GF and Grindley ND. 1989. The 43 residue DNA
binding domain of gamma delta resolvase binds adjacent major and minor grooves
of DNA. Nucleic Acids Res. 17:1035-50.
23. Keravala A, Lee S, Thyagarajan B, et al. 2007. Mutational derivatives of phiC31
integrase with enhanced efficiency and specificity. Mol. Therapy [in press].
24. Urnov FD, Miller JC, Lee YL, Beausejour CM, Rock JM, Augustus S,
Jamieson AC, Porteus MH, Gregory PD, and Holmes MC. 2005. Highly
efficient endogenous human gene correction using designed zinc-finger nucleases.
Nature 435:646-51.