Use of site-specific recombination as a probe of nucleoprotein complex
formation in chromatin
Micha Schwikardi and Peter Dro¨ge
Institute of Genetics, University of Cologne, Germany
DNA transactions in eukaryotes require that proteins gain
access to target sequences packaged in chromatin. Further,
interactions between distinct nucleoprotein complexes are
often required to generate higher-order structures. Here, we
employed two prokaryotic site-specific recombination sys-
tems to investigate how chromatin packaging affects the
assembly of nucleoprotein structures of different complex-
ities at more than 30 genomic loci. The dynamic nature of
chromatin permitted protein–DNA and DNA–DNA inter-
actions for sites of at least 34 bp in length. However, the
assembly of higher-order nucleoprotein structures on targets
spanning 114 bp was impaired. This impediment was
maintained over at least 72 h and was not affected by the
transcriptional status of chromatin nor by inhibitors of histone
deacetylases and topoisomerases. Our findings suggest that
nucleosomal linker-sized DNA segments become accessible
within hours for protein binding due to the dynamic nature
of chromatin. Longer segments, however, appear refractory
for complete occupancy by sequence-specific DNA-binding
proteins. The results thus also provide an explanation why
simple recombination systems such as Cre and Flp are
proficient in eukaryotic chromatin.
Keywords: chromatin; DNA reactivity; nucleoprotein com-
plex; site-specific recombination; transcription.
Alterations in chromatin structure are involved in the regu-
lation of DNA transactions such as transcription and site-
specific recombination. Recently, chromatin remodeling and

The second system employed in this study is derived from
the E. coli gd transposon-encoded resolvase. The resolvase
system is more complex than the Cre system. In the first step
leading to recombination resolvase binds to a recombination
sequence called res. A single res is composed of three
binding sites (I–III) for resolvase dimers which together
occupy 114 bp (Fig. 1B). Three dimers bind cooperatively
to res with comparable affinities towards sites I and II in
order to generate a recombinogenic complex, termed resolvo-
some [12]. Two resolvosomes then synapse by random
collision [13]. Two res must be present as direct repeats on
the same negatively supercoiled DNA molecule. Only this
site orientation leads to the formation of a functional
synaptic complex, termed synaptosome, which entraps three
(–)supercoils [14]. Strand exchange is catalyzed by dimers
bound at paired sites I, while those bound at sites II and III
serve accessory roles in synaptosome formation and in the
activation of strand cleavage therein (Fig. 1B). This rather
complex architecture imposes directionality on recombina-
tion, i.e. strand exchange always results in deletion of DNA
between two res.
Recently, we have transferred the gd system to higher
eukaryotes [10]. Two resolvases containing activating
mutations (E124Q or E102Y/E124Q) and a SV40-derived
nuclear localization signal (NLS) at their C-termini are
recombination-proficient on episomal DNA. Full res are still
Correspondence to P. Dro
¨
ge, Institute of Genetics, University of
Cologne, Weyertal 121, D-50931 Cologne, Germany.

on the accessibility of accessory sites in res, gd102NLS
employs two recombination pathways in eukaryotic cells
that can be distinguished by the resulting products.
In order to compare the activities of Cre and gd102NLS
on substrates packaged in chromatin, we have generated
reporter cell lines that carry target sites for both recom-
binases randomly integrated into the host genome. By
comparing the efficiencies of recombination on episomal
and on genomic targets, we have shown that the dynamic
nature of chromatin renders a site of at least 34 bp in general
reactive for recombination. However, the assembly of and/
or the interaction between more complex nucleoprotein
structures on 114-bp targets was significantly impaired in
chromatin. This impediment was not affected by the tran-
scriptional status of chromatin nor by inhibitors of histone
deacetylases and topoisomerases.
EXPERIMENTAL PROCEDURES
Vectors
Expression vectors for Cre, wild-type gd resolvase, and the
two resolvase mutants have been described previously [10].
pTRE-res-lox-hygromycin-res-lox-lacZ (-RLHRLZ) was
generated by PCR using the recombination cassette of
pCH-RLNRLZ as template [10]. The neomycin gene was
substituted by the hygromycin gene of pTK-Hyg (Clontech).
The entire cassette was introduced into the Bam HI site of
pTRE2 (Clontech). The derivate pTRE-SLHSLZ was gener-
ated by PCR using pTRE-RLHRLZ as template. The corre-
sponding recombined product vectors were generated through
transformation into E. coli strain DH5a or 294-Cre [16]. They
were subsequently characterized by restriction digestion and

6
cells.
Electroporation was in 800 mL RPMI medium without
phenolred and glutamate (Life Technologies) at 960 mF
and 280 V. Transfection efficiencies were determined by
FACS (FACS Calibur; Becton Dickinson) using the program
CELL QUEST and GFP as a marker. They were typically in
the range of 40–70%. Trichostatin A (ICN Biomedicals,
Germany), dissolved in ethanol, was added to culture
medium at 3 m
M final concentration. Butyrate (Sigma,
Germany), dissolved in sterilized water, was tested at 0,5
Fig. 1. Schematic representations of recombination pathways.
(A) Cre-loxP pathway. LoxP sites (arrows with open head) are present
as direct repeats on a circular substrate. Synapsis occurs by collision of
two loxP-bound dimers (filled circles). The two sites are aligned in an
antiparallel orientation. Strand exchange will then lead to deletion. (B)
Recombination on two full res by wild-type or mutant resolvase. The
res are depicted as direct repeats on a circular substrate. DNA super-
coiling, required for the reaction with wild-type resolvase, is omitted for
clarity. After all three cognate sites within res (I, II, and III) are bound
by resolvase dimers (filled circles) a synaptosome is generated. The
interaction between dimers bound at accessory sites II and III, and the
catalytically active ones at paired sites I that is required to trigger strand
exchange is indicated by arrows. Recombination will lead to deletion of
DNA between two res. (C) Recombination by mutant resolvases on
sites I of res. After resolvase dimers (filled circles) are bound to sites I,
random collision leads to two functional synaptic complexes. When
subsites I align in an antiparallel orientation (top), recombination leads
to inversion of DNA between sites I. If both sites align in a parallel

recombination substrates than in our previous study. In order
to use episomal substrates as controls it was therefore
necessary to re-investigate how Cre and gd102NLS perform
under these conditions. Further, it was necessary to analyze
recombination on linearized episomal substrates as they
better resemble the topology of targets placed in chromatin
than circular substrates used before.
The recombination substrate, termed pTRE-RLHRLZ,
contains a tetracyclin-responsive-element (TRE)-CMV pro-
moter construct placed upstream of a recombination cassette
(Fig. 2A). Transcription from this promoter is regulated in
CHO-AA8 Tet-Off cells by doxycyclin. The promoter is
active in the absence of drug, while its presence leads to
rapid transcriptional inactivation [18]. The cassette is com-
posed of two directly repeated copies of res and of loxP
sites. They flank the coding region of the hygromycin gene
which serves as the resistance marker for the generation of
stable reporter cell lines (see below). We placed the coding
region of the lacZ gene downstream of the cassette.
Transcription is initiated 127-bp upstream of the first
nucleotide defining site III of the promoter proximal res,
Fig. 2. Recombination on episomal targets. (A) Diagram of substrate
vector pTRE-RLHRLZ. Relevant genetic elements are marked and
explained in the text. Start of transcription within the TRE-CMV
promoter is at 1374. (B) Normalized b-Gal activities as reporter for
recombination on episomal pTRE-RLHRLZ. The activity of the
reporter is expressed in (%) relative light units (RLU) and normalized to
the amount of protein in crude cell extracts. The activity resulting from
the recombined product (pTRE-RLZ) cotransfected with an expression
vector for a phage l integrase mutant was set as 100%. In each case,

gd102NLS contain identical NLS [10].
Normalized b-Gal activities were determined in cell
extracts prepared 72 h after transfection. The results show
that Cre and gd102NLS efficiently recombine linearized
pTRE-RLHRLZ (Fig. 2B). In contrast to our previous study,
however, we found that recombination by gd102NLS is
reduced in this cell line to a level of 60% of that observed
with Cre. Identical results were obtained when (–)super-
coiled instead of linearized substrates were cotransfected
with recombinase expression vectors (data not shown).
We also analyzed recombination on a linearized deriva-
tive of pTRE-RLHRLZ, termed pTRE-SLHSLZ. This sub-
strate contains two isolated sites I of res as direct repeats,
instead of two full res. gd102NLS is also proficient to
recombine sites I in the absence of accessory sites. The
efficiency of this reaction is significantly reduced, however,
reaching 10% of that observed with Cre (Fig. 2C).
Recombination on genomic substrates
Hygromycin-resistant cell lines were generated with linear-
ized pTRE-RLHRLZ. Southern blot analysis, PCR, and
DNA sequencing revealed that they contain between one
and about 20 copies of the substrate vector at different
genomic locations. Hence, the vector integrated probably
randomly into the host genome (data not shown). We first
analyzed recombination in cell line TRE2/3 which contains
a single copy of the vector. The analysis was performed in
the absence of doxycyclin, i.e. the TRE-CMV promoter is
active and transcription proceeds through the entire
recombination cassette.
The expression vectors for Cre, wild-type resolvase

compared to a control lacking the drug. Further, the residual
activity detectable in doxycyclin-treated TRE2/3R cells was
only threefold to fivefold higher than that in parental TRE2/3
cells, indicating that the TRE-CMV promoter was efficiently
inactivated by the drug (data not shown). When we tested
recombination in doxycyclin-treated TRE2/3 cells, however,
quantitation of Southern blots revealed that Cre and
gd102NLS remained unaffected by the transcriptional status
of the recombination cassette (Table 1).
The entire set of experiments exemplified above with
TRE2/3 cells expressing either Cre or gd102NLS was
performed with five different cell lines, thus investigating
recombination on more than 30 genomic copies of pTRE-
RLHRLZ. While Cre reproducibly recombined between 40
Fig. 3. Cre, but not resolvase, efficiently recombines genomic
targets. Genomic DNA was prepared from TRE2/3 cells 72 h after
electroporation with recombinase expression vectors. DNA was
digested with Bam HI, separated on a 0.8% (w/v) agarose gel, trans-
ferred to nitrocellulose membrane, and hybridized to a probe derived
from the N-terminal region of lacZ. Bam HI-digested pTRE-RLHRLZ
and pTRE-RLZ were used as unrecombined and recombined controls,
respectively.
q FEBS 2001 Site-specific recombination in chromatin (Eur. J. Biochem. 268) 6259
and 80% of targets, gd102NLS exhibited only a residual
activity (Table 1). We conclude that Cre, irrespective of the
transcriptional status of loxP sites in chromatin, efficiently
recombines pTRE-RLHRLZ at the majority of genomic
loci. However, gd102NLS appears to be severely impaired
when targets are packaged into chromatin. Further, this
impediment is maintained irrespective of the transcriptional

ation play important roles in the regulation of chromatin
structure. In particular acetylation of the N-terminal tails
of histones are thought to render chromatin more accessible
for DNA-binding proteins. In fact, DNA transactions such
as V(D)J recombination and transcription are enhanced
when histone deacetylases are inhibited [19,20]. Further,
eukaryotic topoisomerases appear to be involved in
chromatin organization, perhaps through direct interaction
with histone deacetylases [21]. We decided therefore to test
whether inhibitors of histone deacetylases and eukaryotic
topoisomerases might render genomic res more accessible
for gd102NLS, which could then lead to a significant
increase in recombination activity.
Two reporter cell lines were incubated with histone
deacetylase inhibitors butyrate or trichostatin A (reviewed in
[22]) at 24 h after transfection of recombinase expression
vectors. Cells were treated for 24 h, after which drugs were
removed and cells were incubated for additional 24 h.
Untreated cells served as controls. Further, controls with
b-Gal-expressing TRE2/3R cells treated in the same way
with trichostatin A showed that b-Gal activity increased up
to fourfold. This might indicate that the TRE-CMV pro-
moter becomes more accessible for the transcriptional
machinery. However, Southern analysis and b-Gal assays
revealed that the efficiencies of recombination by Cre and
gd102NLS remain unaffected by these inhibitors. Following
the same protocol, treatment of TRE2/3 cells with topo-
isomerase type I inhibitor camptothecin and with the
flavonoid EMD50689, the latter inhibits both type I and type
II topoisomerases [23], also showed no effect on recom-

Cre gd102NLS
(–)Doxy (1)Doxy (–)Doxy (1)Doxy
TRE2/3 (single copy) 66.8 74.2 2.0 2.5
TRE3/2 (single copy) 42.3 39.1 2.2 1.3
TRE3/5 (single copy) 75.0 ND ND ND
TRE2/2 (. 20 copies) 50.2 50.2 2.5* ND
TREII/3 (. 10 copies) 84.0 76.2 2.2 2.3
6260 M. Schwikardi and P. Dro
¨
ge (Eur. J. Biochem. 268) q FEBS 2001
substrate, the loxP site and the site I of res may be located
either in the nucleosomal core or in the linker DNA; the
length of the latter can vary significantly in vivo. In addition,
productive encounters readily occurred between sites I of
res and between loxP sites separated by 1.2 kb, thus further
strengthening the view of a rather flexible chromatin struc-
ture [2,9]. Importantly, our results have shown that these
basic properties of chromatin are not significantly altered by
the transcriptional status of DNA.
In contrast to loxP sites and sites I of res, the reactivity of
episomal and genomic full length res differed markedly in
our analysis. Genomic res were about 30-fold less reactive
for recombination than their epsiomal counterparts. We
consider two explanations for this result. Firstly, the ordered
nucleosomal organization of chromatin prevents the co-
operative binding of gd102NLS to all three sub-binding
sites. Even the repeated passage of the transcriptional
machinery does not affect this accessibility limit. Further-
more, as a significant amount of gd102NLS is present in
CHO cells throughout the time course of our experiments

assemble multicomponent nucleoprotein complexes. This
requirement significantly increases the stringency with
which DNA transactions are regulated in higher eukaryotes.
ACKNOWLEDGEMENTS
We thank members of our and of K. Rajewsky’s laboratory for critical
comments on the manuscript. The flavonoid EMD50689 was a kind gift
of Dr J. Ko
¨
hrle, Wu
¨
rzburg, Germany. Special thanks go to K. Rajewsky
for support with cell culture facilities. This work was financed through
SFB 274 and Deutsche Forschungsgemeinschaft grant Dr187/8–2
(PD).
Fig. 4. Resolvase recombines at genomic res via random collision
of site I-bound dimers. (A) Normalized b-Gal activities as reporter
for recombination in pTRE2/3 cells. The graph shows the mean values
from five transfection experiments. The 100% reference was obtained
with crude extracts prepared from pTRE2/3R cells transfected with
pPGKCrebpa. Note that the RLU are plotted in a logarithmic scale.
(B) PCR to analyze deletion in TRE2/3 cells. Genomic DNA was
prepared 72 h after transfection. The PCR product indicative of deletion
is marked (del.). The product resulting from unrecombined genomic
pTRE-RLHRLZ is also indicated (unrec.). (C) PCR to analyze inversion
in TRE2/3 cells. Only one PCR product (inv.) is generated. Products
were analyzed on 0.8% agarose gels and visualized by UV after
ethidium bromide staining.
q FEBS 2001 Site-specific recombination in chromatin (Eur. J. Biochem. 268) 6261
REFERENCES
1. Kadonaga, J.T. (1998) Eukaryotic transcription: An interlaced

for the topology of episomal DNA. FEBS Lett. 471, 147–150.
11. Nagy, A. (2000) Cre recombinase: the universal reagent for genome
tailoring. Genesis 26, 99–109.
12. Salvo, J.J. & Grindley, N.D.F. (1988) The gd resolvase bends the
res site into a recombinogenic complex. EMBO J. 7, 3609 –3616.
13. Sessions, R.B., Oram, M., Szczelkun, M.D. & Halford, S.E. (1997)
Random walk models for DNA synapsis by resolvase. J. Mol. Biol.
270, 413–425.
14. Dro
¨
ge, P. & Cozzarelli, N.R. (1989) Recombination on knotted
substrates by Tn3 resolvase. Proc. Natl Acad. Sci. USA 86,
6062–6066.
15. Arnold, P.H., Blake, D.G., Grindley, N.D., Boocock, M.R. & Stark,
W.M. (1999) Mutants of Tn3 resolvase which do not require
accessory binding sites for recombination activity. EMBO J. 18,
1407–1414.
16. Buchholz, F., Angrand, P.O. & Stewart, A.F. (1996) A simple assay
to determine the functionality of Cre or FLP recombination targets
in genomic manipulation constructs. Nucleic Acids Res. 24,
3118–3119.
17. Lorbach, E., Christ, N., Schwikardi, M. & Dro
¨
ge, P. (2000) Site-
specific recombination in human cells catalyzed by phage l
integrase mutants. J. Mol. Biol. 296, 1175–1181.
18. Gossen, M. & Bujard, H. (1992) Tight control of gene expression in
mammalian cells by tetracyclin responsive promoters. Proc. Natl
Acad. Sci. USA 89, 5547–5551.
19. Kwon, J., Morshead, K.B., Guyon, J.R., Kingston, R.E. &