This is a postprint of an article published in
Loessner, H., Endmann, A., Leschner, S., Westphal, K., Rohde, M., Miloud,
T., Ha?mmerling, G., Neuhaus, K., Weiss, S.
Remote control of tumour-targeted Salmonella enterica serovar
Typhimurium by the use of l-arabinose as inducer of bacterial gene
expression in vivo
(2007) Cellular Microbiology, 9 (6), pp. 1529-1537.
Remote control of tumor targeted Salmonella enterica serovar Typhimurium by
the use of L-arabinose as inducer of bacterial gene expression in vivo
running title : Remote control of tumor targeted Salmonella
Holger Loessner1)*, Anne Endmann1), Sara Leschner1), Kathrin Westphal1), Manfred
Rohde2), Tewfik Miloud3), Günter Hämmerling3), Klaus Neuhaus4), Siegfried Weiss1)
1) Molecular Immunology, Helmholtz Centre for Infection Research,
Inhoffenstrasse 7, 38124 Braunschweig, Germany
2) Dept. Microbial Pathogenesis, Helmholtz Centre for Infection Research,
Inhoffenstrasse 7, 38124 Braunschweig, Germany
3) Molecular Immunology, DKFZ, German Cancer Research Center, Im
Neuenheimer Feld 280, 69120 Heidelberg, Germany
4) Dept. Microbiology, ZIEL, Technische Universität München, Weihenstephaner
Berg 3, 85354 Freising, Germany
corresponding author : Helmholtz Centre for Infection Research, Inhoffenstrasse 7,
38124 Braunschweig, Germany, Tel.: +49-531-6181 5109, Fax.: +49-531-6181 5002,
keywords : Salmonella; remote controlled gene expression; bacterial vector; tumor
gene therapy, PBAD, L-arabinose
We have used Salmonella enterica serovar Typhimurium (S. typhimurium) which are
able to colonize tumors besides spleen and liver. Bacteria were equipped with
constructs encoding GFP or luciferase as reporters under control of the promoter
PBAD that is inducible with L-arabinose. Reporter genes could be induced in culture
but also when the bacteria resided within the mouse macrophages J774A.1. More
important, strong expression of reporters by the bacteria could be detected in mice
after administration of L-arabinose. This was especially pronounced in bacteria
colonizing tumors. Histology demonstrated that the bacteria had accumulated in and
close to necrotic areas of tumors. Bacterial gene induction was observed in both
regions. PBAD is tightly controlled also in vivo since gene E of bacteriophage ΦX174
could be introduced as inducible suicide gene. The possibility to deliberately induce
genes in bacterial carriers within the host should render them extremely powerful
tools for tumor therapy.
Bacteria and bacterial products have a long history as tumor therapeutics (Pawelek
et al., 2003). Recently, several strictly or facultative anaerobic bacteria have attracted
particular attention because they display the ability to target and accumulate in solid
tumors. Such bacteria were adapted to deliver therapeutic factors and were applied
alone or in combination with conventional therapeutics (Dang et al., 2001; Jain and
Forbes, 2001; Theys et al., 2003).
S. typhimurium, a motile, invasive Gram-negative bacterium, is able to colonize solid
tumors. Chemoattractive compounds produced by quiescent tumor cells have been
shown to contribute to this effect in vitro (Kasinskas and Forbes, 2006). In vivo, S.
typhimurium primarily colonizes hypoxic, nutrient rich, necrotic areas (Forbes et al.,
2003) but is also able to invade tumor cells (Avogadri et al., 2005; Pawelek et al.,
2002). Colonization results in retardation of tumor growth thus prolonging survival of
the mice (Clairmont et al., 2000; Low et al., 1999; Pawelek et al., 1997; Zhao et al.,
2006). In clinical trials, systemically administered bacteria colonized human tumors at
high frequency without causing adverse toxic effects (Nemunaitis et al., 2003).
S. typhimurium has been successfully used to deliver therapeutic molecules to
tumors such as prodrug-converting enzymes and antigens (King et al., 2002;
Nishikawa et al., 2006; Pawelek et al., 1997). Expression of these molecules was
driven by constitutive promoters. However, for the purpose of tumor therapy,
inducible systems would be more appropriate. A range of promoters have been
characterized that respond to various stimuli, e.g. particular substrates, temperature,
pH, ionic strength, hypoxia, radiation or other environmental conditions (Cases and
de Lorenzo, 1998). Most of such systems are excluded from in vivo use. We
concentrated therefore on the PBAD promoter from the arabinose operon of
Escherichia coli. This system can be activated by the sugar L-arabinose. The PBAD
promoter is widely used to positively control expression in bacterial cultures (Guzman
et al., 1995). Here we show that heterologous genes under control of PBAD can also
be induced in vivo when L-arabinose is administered systemically. S. typhimurium,
particularly when colonizing tumors, were highly susceptible to remote control. This
opens a great number of possibilities for improvement of the tumor-specific delivery
of therapeutic molecules by bacteria.
L-arabinose induced reporter gene expression in bacterial culture and host
To investigate the possibility of inducing bacterial gene expression under remote
control in the mammalian host, we first established sensitive reporter systems. We
constructed plasmids encoding green fluorescent protein (GFP) (Cormack et al.,
1996) or firefly luciferase (Fluc) under control of the PBAD promoter. SL7207, an
attenuated experimental vaccine strain of S. typhimurium (Hoiseth and Stocker,
1981), was transformed with these vectors. Strong induction of GFP or Fluc
expression could be observed when L-arabinose was added to cultures of such
transformants (Fig. 1A,B). Reporter gene expression ceased after 2 hrs (Fig. 1B),
most likely because the bacteria had consumed all the sugar at this time point
(Supplementary Fig. 1A). Of note, induction of reporter genes retarded bacterial
growth and in parallel bacteria appeared enlarged under such conditions
(Supplementary Fig. 1B and C, and data not shown). We then wanted to see whether
induction is also possible when the Salmonella transformants were residing inside
host cells. Therefore, GFP and Fluc transformants were used to infect J774A.1
mouse macrophages. Extracellular bacteria were killed with gentamycin and
subsequently L-arabinose was added to the cultures at different concentrations. After
incubation for 3 hrs, cells were lysed and bacteria were recovered from the lysate
and analyzed for GFP or Fluc expression. As shown in Figure 1C and D, increasing
concentrations of L-arabinose in such cultures resulted in increased expression of
GFP or Fluc by the bacteria. Furthermore, the number of GFP expressing Salmonella
increased with rising concentrations of L-arabinose (data not shown).
L-arabinose inducible expression of luciferase in tumor targeted S. typhimurium
Before testing the induction of gene expression in tumor targeted S. typhimurium in
vivo, we first determined whether sufficient L-arabinose can accumulate in the
circulation of mice to induce the PBAD promoter. Figure 2A shows that after
intraperitoneal administration, a high but transient concentration of the sugar could
be found in the blood of mice. The decrease after 4-6 hrs is probably due to
degradation by host enzymes (Seri et al., 1996).
S. typhimurium SL7207 carrying the Fluc reporter plasmid was therefore injected
intravenously into mice bearing a subcutaneous CT26 tumor of roughly 0.5 cm in
diameter. Three days later, L-arabinose was injected intraperitoneally and
colonization as well as Fluc expression by bacteria from tumor, spleen and liver was
determined. As can be seen from Figure 2B, more than 10-fold enrichment of
bacteria in tumors was observed. This was not as dramatic as previously observed
with other S. typhimurium strains (Bermudes et al., 2002). We also noticed during all
of these experiments that accumulation of SL7207 in tumors was rather variable
despite all tumors were colonized (Fig. 2B). Nevertheless, the low accumulation
factor and the variability should not influence the proof of principle of remote control
of bacterial gene expression in vivo.
When Fluc activity was determined in the three tissues, it became clear that by using
the PBAD promoter, it is possible to deliberately induce bacterial genes after
colonization of the mammalian host. Administration of L-arabinose resulted in a
strong induction of Fluc expression in bacteria from all three tissues (Fig. 2C).
Interestingly, the induction factor was higher in the tumor than in spleen and liver.
This is even more obvious when Fluc activity per CFU was compared (Fig. 2D).
Probably, the bacteria are more accessible to the sugar in the tumor or the
physiological state of tumor colonizing bacteria results in higher gene expression.
Induction of reporter gene expression was specific since administration of equal
amounts of D-glucose or L-rhamnose did not result in detectable Fluc activity when
bacteria from tumors and spleen were tested (data not shown). Similarly, a construct
where the Fluc gene was under the control of the T7 promoter could not be induced
by L-arabinose administration in tumor or spleen colonizing bacteria (data not
Non-invasive in vivo imaging of L-arabinose induced luciferase expression
The bioluminescence that is inducibly generated in bacteria containing Fluc, should
be detectable by non-invasive in vivo imaging. We therefore injected into tumor
bearing mice, SL7207 bacteria that carried plasmids encoding Fluc under the control
of PBAD as before. Induction of Fluc was monitored using the supercooled CCD
camera system IVIS-100 (Xenogen) after intraperitoneal injection of L-arabinose and
luciferin-substrate. As shown in Figure 3A, little Fluc activity is detectable before
induction. However, after application of L-arabinose strong Fluc activity becomes
readily visible. In concordance with the biochemical assays (Fig. 2C), mainly tumors
displayed Fluc activity.
To simplify the non-invasive in vivo imaging, we constructed a new S. typhimurium
strain that contained the lux operon of the bacterium Photorhabdus luminescens
under the control of PBAD as a single copy in the bacterial chromosome of SL7207.
No substrate needs to be administered for lux. When tumor bearing mice were
infected with this bacterial strain and L-arabinose was subsequently administered
intraperitoneally, tumor-specific bioluminescence could be detected (Fig. 3C). No
light was observed without induction (Fig. 3C,D). With this vector, the intensity of
bioluminescence was lower compared to Fluc but the maximal level of light emission
was reached earlier (Fig. 3B,D). The lower level of light emission is most likely due to
the single copy of the lux operon present in the carrier bacteria and a slightly different
emission wavelength of light generated by lux. Tumor colonization properties of lux
bacteria were similar to the strain carrying the fluc plasmid (data not shown).
Accessibility of tumor colonizing bacteria to the inducer L-arabinose
Tumor colonizing S. typhimurium are supposed to be found in necrotic and hypoxic
areas (Forbes et al., 2003; Pawelek et al., 1997). Whether this was also true for
strain SL7207 was tested by immune histology. As can be seen in Figure 4A-C,
bacteria could be found mainly in necrotic areas and to some extend in bordering
areas that are most likely hypoxic. To see whether L-arabinose can reach all bacteria
i.e. bacteria in necrotic areas as well as bacteria at the rim, we employed bacteria
that carried a low copy number plasmid encoding GFP under the control of PBAD.
These bacteria were injected into tumor bearing mice and after administration of L-
arabinose, tumor colonizing bacteria were examined for GFP expression by
histology. Figures 4D-F show that bacteria expressing GFP can be detected in both
regions of the tumor. Thus, L-arabinose probably perfuses all areas of the tumor
leading to local concentration that is sufficient to induce the PBAD promoter. No GFP
expression could be detected in any region of the tumor when inducer was omitted
(data not shown).
In necrotic areas, bacteria are most likely extracellular. This was confirmed by
histology (Fig. 4G). However, it was unclear whether in the rim areas SL7207 would
invade or reside outside tumor cells. As shown in Figure 4H, in such rim areas most
bacteria appear to reside in extracellular spaces and only a few seem to be
associated with cells. Transmission electron microscopy confirmed the extracellular
location of such bacteria (data not shown).
Whether the bacteria are specifically attracted to these parts of the tumor or whether
the presence of the bacteria causes cell death as a consequence of necrosis or
apoptosis, as indicated by the small nuclei, could not be decided (Fig. 4H). Most
likely the latter is true since after bacterial colonization the necrotic areas become
larger compared to uninfected tumors (data not shown). Nevertheless, tumor growth
was not significantly retarded. Cells in non-colonized and thus unaffected regions
apparently continued to grow unaltered (Fig. 4I).
Not all bacteria displayed in Figure 4D-F expressed GFP despite retention of the
expression plasmid. One explanation for the lack of GFP expression in some bacteria
could be the phenomenon of an "all-or-nothing" induction of the PBAD promoter.
Subsaturating concentration of L-arabinose inside the tumor could result in gene
activation in one but not the other bacterium (Siegele and Hu, 1997). Alternatively,
the physiological state of some bacteria might prevent the induction of the reporter
gene via PBAD. In concordance with this explanation, when we used bacteria
constitutively expressing GFP we also found non-fluorescent bacteria despite
retention of the expression plasmid in all bacteria.
L-arabinose inducible bacterial lysis in vitro and in vivo
As a test for tight control in vivo, we placed the lysin gene E of the bacteriophage
ΦX174 under control of PBAD. Already 100 molecules of this protein should lead to
loss of bacterial viability (Maratea et al., 1985). When the construct was integrated as
a single copy into the chromosome of the SL7207 bacteria, no interference with
growth or integrity in bacterial cultures was observed in the absence of L-arabinose
(Fig. 5A,B). However, upon addition of the inducer lysis occurred within 90 min (Fig.
5A) and only bacterial debris could be observed at this time (Fig. 5C).
Whether the same was true in vivo was tested with tumor colonizing bacteria. Figure
5D shows that the bacteria preferentially colonized tumors similar to parental
bacteria. Administration of L-arabinose resulted in preferential death of bacteria in the
tumor as compared to spleen and liver. This is in concordance with the preferential
induction of Fluc in tumor colonizing bacteria described before (Fig. 2C,D). Similarly,
only one third of the bacteria were lysed under these circumstances. Thus, the PBAD
promoter system is suitable for the introduction of inducible suicide genes into carrier
bacteria, which adds important safety features to tumor targeting bacteria.
Remote control of gene expression is a new strategy for in vivo use of tumor
targeting bacteria. To our knowledge, only one bacterial system has been described
so far for deliberate induction of genes within the host. It is based on the bacterial
response to ionizing irradiation (Nuyts et al., 2001; Theys et al., 2003). However,
irradiation itself results in damage of normal tissue and is undermined by low
compliance from patients. Thus, use of a low molecular weight biocompatible
compound like L-arabinose in combination with the corresponding molecular switch -
PBAD - as described in the present work, appears to be ideal for such purpose. PBAD is
derived from the E. coli arabinose operon and should be transferable to other
L-arabinose appears to fulfill all criteria of a versatile inducer. It is a biocompatible
food component and suggested for clinical use in other contexts. It not only reaches
bacteria that reside in the phagocytic vacuole as shown by our in vitro experiments
but can also reach bacteria that reside in niches that are provided by tumors. Remote
gene induction was indiscriminate of the areas of the tumor. After systemic
administration of L-arabinose the sugar only transiently circulates in the blood of mice
and is degraded by host enzymes (Seri et al., 1996). Accordingly, expression of
genes under control of PBAD is short-lived.
Bioluminescence in vivo imaging of light emitting tumor targeted bacteria is an
attractive approach in the diagnosis of tumors (Yu et al., 2004). Our approach of
transient, L-arabinose mediated induction of bioluminescence emission by the
bacteria is sufficient to localize the bacteria inside the tumor but liberates the bacteria
from the burden of constant expression of the luciferase.
Only a portion of the bacteria within the same location was induced. The
physiological state of bacteria may account for this observation. However, all-or-
nothing induction at sub-saturating inducer concentrations is an alternative
explanation. It can be attributed to fluctuating expression of the L-arabinose
transporter in an uninduced state of the endogenous arabinose operon (Siegele and
Hu, 1997). This might add to the versatility of the system, since induction concerns
only a portion of the bacteria and might provide lytic bacteria for repeated release of
therapeutic molecules. On the other hand, strains could be engineered that are able
to constitutively take up L-arabinose giving rise to homogenous in vivo induction
(Morgan-Kiss et al., 2002).
The regulation of PBAD is tightly controlled also in vivo. A suicide gene could be
integrated into the carrier bacteria without interference with growth, invasiveness and
tumor colonization. Together with homogenous in vivo induction this would introduce
an important safety feature into tumor targeting bacteria. In addition, bacterial lysis
can augment the release of therapeutic macromolecules from the bacteria (Jain and
Mekalanos, 2000), thus potentially increasing the efficacy of the bacterial vector.
The discovery of other remotely controllable promoters will eventually give rise to the
construction of multifunctional tumor targeting bacteria which can be manipulated in
several ways in vivo, thus allowing a safe, effective and individual tumor therapy.
Bacterial strains, plasmids and growth conditions
S. typhimurium strain SL7207 (hisG, ΔaroA) was kindly provided by Bruce Stocker
(Hoiseth and Stocker, 1981). The coding sequence of the firefly luciferase (Fluc)
gene from plasmid pGL3-basic (Promega) was placed under control of the PBAD
promoter regulator cassette, originally subcloned from Plasmid pBad18 (Guzman et
al., 1995), yielding plasmid pHL259 (pMB1 origin of replication). Similarly, gene
gfpmut2 encoding a bright variant green fluorescent protein (Cormack et al., 1996)
was placed under control of the PBAD promoter, yielding either the high copy number
plasmid pHL238 (pUC origin of replication) or the low copy number plasmid pHL239
(p15A origin of replication). The low copy plasmid pSL1 is a derivative of pSMART-
LCKan (Lucigen) and harbours the luxCDABE operon originating from Photorhabdus
luminescens (Winson et al., 1998). lux was placed under control of the PBAD promoter
and integrated into the chromosome of strain SL7207, yielding strain SL7207::HL289.
In brief, lux was linked to the PBAD promoter and the cassette subcloned into a
derivative of plasmid pUX-BF5 harbouring a Tn7 mini-transposon (Bao et al., 1991),
yielding plasmid pHL289. Transposon-mediated site specific integration into the
chromosome of SL7207 was carried out according to the method of Bao et al..
Similarly, gene E of bacteriophage ΦX174 was placed under control of the PBAD
promoter and integrated into the chromosome of strain SL7207, yielding strain
SL7207::HL260a. Bacteria were grown in Luria-Bertani medium (LB) supplemented
with 100 µg ml-1 ampicillin, 30 µg ml-1 kanamycin, 30 µg ml-1 streptomycin, 15 g l-1
agar where appropriate. Bacteria were mostly grown at 37°C, in liquid medium with
Infection of J774A.1 mouse macrophages and addition of L-arabinose to cell culture
14 Download full-text
J774A.1 mouse macrophages (ATCC TIB-67) were grown in IMDM medium (Gibco)
supplemented with 10 % (v/v) fetal calf serum, 2 mM l-glutamine and 10 mM HEPES.
2,5 x 105 cells were seeded in each well of a 6-well cell culture plate, incubated
overnight and subsequently infected with S. typhimurium at a multiplicity of infection
50:1 (bacteria : cells). Plates were centrifuged for 1 min at 1000 g and incubated 30
min at 37°C. Then, cells were washed with medium and extracellular bacteria were
killed by incubation with medium containing 50 µg ml-1 gentamycin (Biochrom). The
medium was replaced after 1,5 hr with medium containing 10 µg ml-1 gentamycin. L-
arabinose dissolved in medium was added at different concentrations into wells and
incubation continued for another 4 hrs. Cells were washed twice with phosphate-
buffered saline (PBS) and reporter gene expression was analyzed.
Flowcytometric analysis of bacterial GFP expression
Cultured bacteria or bacteria present within the lysate of infected J774A.1 mouse
macrophages were analyzed by flow cytometry. Infected J774A.1 cells were first
lysed in PBS containing 0,5 % (v/v) Triton X-100 and incubated at 37°C for 5 min.
Then, an equal volume of PBS containing 2 % (v/v) fetal calf serum and 2 mM EDTA
was added before samples were analyzed. Two color flow cytometry was applied in
order to differentiate GFP expressing bacteria from autofluorescent cellular debris as
described previously (Bumann, 2002). This method exploits the difference of
orange/green emission ratios of GFP expressing S. typhimurium and cellular debris
(Supplementary Fig. 1a,b). A threshold value was set to exclude bacteria expressing
no GFP and background autofluorescence from measurements. An appropriate
scatter gate was used to distinguish bacteria from large particles. Data were
analyzed with WinMDI 2.8 software.