aBhaskar Thyagarajan
Juergen Hoelters
3
and
1
Department of Genetics
Stanford University School of
Medicine, Stanford, CA
95305-5120, USA
2
Poetic Genetics, LLC, 863
Mitten Road, Suite 102
Burlingame, CA 94010, USA
3
ChirurgischesForschungslabor
Chirurgische Klinik und
Poliklinik-Innenstadt, Ludwig-
Maximilians-Universitat
Munchen, Munchen, Germany
potential utility.
carries out site-specific recombination between a mammalian genomes, at either inserted att sites or
native sequences.
4,5
Genomic integration mediated
by fC31 integrase in unmodified cells takes place
doi:10.1016/j.jmb.2005.11.098w40 bp attP recognition site in the phage genome
and a similarly-sized attB recognition site in theq 2005 Elsevier Ltd. All rights reserved.
Keywords: site-specific integration; fC31 integrase; pseudo site; gene
therapy; genome engineering
*Corresponding author
Introduction
The integrase enzyme from phage fC31 integrase
Streptomyces host genome (Figure 1(a)).
1–4
Recent
studies have shown that fC31 integrase can
also mediate efficient site-specific integration into0022-2836/$ - see front matter q 2005 E
† T.W.C. & J.L.P. contributed equa
Abbreviations used: HMM, hidde
FISH, fluorescence in situ hybridiza
E-mail address of the correspond
calos@stanford.edu2
, Patrick J. Kirby
1
, Robert T. Hillman
1
Michele P. Calos
1
*
The site-specific integrase from bacteriophagefC31 functions inmammalian
cells and is being applied for genetic engineering, including gene therapy.
The fC31 integrase catalyzes precise, unidirectional recombination between
its 30–40-bp attP and attB recognition sites. In mammalian cells, the enzyme
also mediates integration of plasmids bearing attB into native sequences that
have partial sequence identity with attP, termed pseudo attP sites. Here, we
analyzed the features of fC31-mediated integration into pseudo attP sites
in the human genome. Sequence analysis of 196 independent integration
events derived from three cell lines revealed w101 integration sites: 56%
of the events were recurrent integrations distributed among 19 pseudo
attP sequences. Bioinformatics analysis revealed a w30-bp palindromic
consensus sequence motif shared by all of the repeat occurrences and most
of the single occurrence sites, verifying that fC31-mediated integration into
pseudo attP sites is significantly guided by DNA sequence recognition. The
most favored unique sequence in these cell lines occurred at chromosome
19q13.31 and accounted for 7.5% of integration events. Other frequent
integration sites were in three specific sequences in subfamilies of ERVL and
L1 repetitive sequences, accounting for an additional 17.9% of integration
events. Integrations could occur in either orientation at a pseudo attP site,
were often accompanied by small deletions, and typically occurred in a
single copy per cell. A number of aberrant events were also described,
including large deletions and chromosome rearrangements. fC31 integrase-
mediated integration only slightly favored genes and did not favor
promoter regions. Gene density and expression studies suggested
chromatin context effects. An analysis of the safety of integration sites in
terms of proximity to cancer genes suggested minimal cancer risk. We
conclude that integration systems derived from fC31 integrase have greatThomas W. Chalberg
1
†, Joylette L. Portlock
1
†, Eric C. Olivares
2Integration Specificity of Ph
in the Human Genomelsevier Ltd. All rights reserve
lly to this work.
n Markov model;
tion.
ing author:ge fC31 Integrase
J. Mol. Biol. (2006) 357, 28–48when a vector bearing attB is introduced into cells
along with fC31 integrase, resulting in integration
at native sequences that resemble the wild-type
attP site, termed pseudo attP sites (Figure 1(a)).
5,6
The integration activity of fC31 integrase is
useful to establish and engineer cell lines,
5,7,8
d.

Integration Specificity of fC31 Integraseconstruct transgenic organisms,
8–10
and for ex
vivo
11–14
and in vivo gene therapy,
15–18
and is being
used by a growing number of investigators.
However, the profile of native integration sites
Figure 1. Site-specific integration and analysis of
integration sites. (a) In nature, fC31 integrase recombines
the attP site in the phage genomewith the attB locus of the
Streptomyces genome. In the context of mammalian cells,
fC31 integrase is used to mediate integration of a
plasmid bearing attB and a transgene into native pseudo
attP sites. This orientation is more effective than using an
attP sequence to find pseudo attB sites.
5
(b) Identification
of integration sites with PCR-based methods and plasmid
rescue. PCR-based rescue is shown at the left. Genomic
DNA bearing an integration is digested with restriction
enzymes. A short linker is ligated to the digest that allows
amplification only following amplification from vector-
specific primers. The products have only one side of the
integration junction. Plasmid rescue is shown at the right.
Genomic DNA bearing an integration is digested with
restriction enzymes that do not cut in the integrated
plasmid. These pieces are then self-ligated and trans-
formed into bacteria for analysis. The products usually
have both sides of the junction (attL and attR) intact.used by the enzyme is not well understood. This
information is particularly important in the context
of gene therapy and genetic modification of stem
cells, where the adverse effects of randomly
integrating vectors have led to cancers in a recent
clinical trial.
19
A previous study identified 31 fC31
integration sites in human 293 cells, including a few
sites that were seen multiple times, particularly an
apparent integration hotspot at chromosome 8p22.
5
Several integration sites detected in human kerati-
nocytes were also reported, including the 8p22
site.
11,12
The pseudo attP sites discovered in the
human genome shared only partial sequence
identity with wild-type attP.
5,11
Despite limited
sequence identity, there seemed to be a restricted
number of pseudo attP sites in the human and
mouse genomes.
5,15,16
It appeared likely that factors
in addition to DNA sequence identity, such as
chromatin configuration and transcriptional con-
text, could play a role in integration site selection.
A more thorough investigation of the features
characterizing phage fC31 integrase-mediated
genomic integration was desirable.
Recent studies have characterized the integration
specificity of various viral and transposon systems,
including avian leukosis virus,
20–22
human
immunodeficiency virus,
23
murine leukemia
virus,
20,24
adeno-associated virus,
25
and Sleeping
Beauty transposon.
26,27
Differences exist in inte-
gration profiles between these vector systems, but
some common trends have emerged. Retroviral
systems often favor integration into actively
expressed genes. While host genomic factors likely
influence these common features, viral biology
plays a role as well. For example, human immuno-
deficiency virus integrates with equal frequency
along genes, whereas murine leukemia virus shows
preference for the 5
0
end of genes, supporting a
model of viral tethering to host transcription
factors.
20,24,28
While they integrate infrequently,
vectors based on adeno-associated virus also favor
integration into transcriptionally active regions.
25
A study examining integration of the Sleeping
Beauty transposon revealed a short consensus
sequence, but no preference for integration into
genes.
26,27
All of these integrating systems
have little sequence specificity, so integration is
considered quasi-random, with very large numbers
(O10
7
) of integration sites possible in the genome.
Because of the more extensive sequence recognition
requirements of fC31 integrase, a more restricted
pattern of integration sites was expected and has
been observed.
In this study, we characterized fC31 integrase-
mediated integration into a variety of human cell
lines in culture. Plasmid rescue and PCR-based
approaches were used to characterize sites of
integration at the DNA sequence level, and these
sites were analyzed to explore the role of DNA
sequence motifs and genomic context. Physical
29mapping and gene expression analysis of inte-
gration sites were also carried out. Together, these
data provide a greatly improved understanding