Anti-Tumoral Effect of the Mitochondrial Target Domain of Noxa Delivered by an Engineered Salmonella typhimurium

Bacterial cancer therapy relies on the fact that several bacterial species are capable of targeting tumor tissue and that bacteria can be genetically engineered to selectively deliver therapeutic proteins of interest to the targeted tumors. However, the challenge of bacterial cancer therapy is the release of the therapeutic proteins from the bacteria and entry of the proteins into tumor cells. This study employed an attenuated Salmonella typhimurium to selectively deliver the mitochondrial targeting domain of Noxa (MTD) as a potential therapeutic cargo protein, and examined its anti-cancer effect. To release MTD from the bacteria, a novel bacterial lysis system of phage origin was deployed. To facilitate the entry of MTD into the tumor cells, the MTD was fused to DS4.3, a novel cell-penetrating peptide (CPP) derived from a voltage-gated potassium channel (Kv2.1). The gene encoding DS4.3-MTD and the phage lysis genes were placed under the control of PBAD, a promoter activated by L-arabinose. We demonstrated that DS4.3-MTD chimeric molecules expressed by the Salmonellae were anti-tumoral in cultured tumor cells and in mice with CT26 colon carcinoma.


Introduction
In cancer therapy, the p53-induced apoptosis pathway, a major mechanism in tumor suppression, has been extensively exploited because of its role in the induction of apoptosis in cancerous cells [1,2]. The deregulation of apoptosis leads to cancer development, proliferation, and treatment resistance [3,4]. The mitochondriamediated apoptotic pathway is largely regulated by the Bcl-2 family of proteins [5], which possess at least one of four conserved motifs known as the Bcl-2 homology domains (BH1 to 4). These domains have been divided into three subfamilies. One of these, the BH3-only proteins, including Bik, Bid, Bim, Bmf, Hrk, Bad, Puma, and Noxa, exhibit sequence homology only in the BH3 domain [5,6]. Noxa is a transcriptional target of p53 that mediates induction of apoptosis via activation of mitochondrial damage and the intrinsic apoptosis signaling pathway [7,8,9,10]. In recent studies, two functional domains in Noxa, the BH3 domain and the mitochondrial targeting domain (MTD), have been identified. Interestingly, the MTD was identified to be a prodeath domain, and shown to cause massive necrosis in vitro through cytosolic calcium increase; it is released from the mitochondria by opening the mitochondrial permeability transition pore [10,11].
Bacterial cancer therapy relies on the fact that several bacterial species preferentially accumulate and replicate within tumors [12,13,14,15]. In particular, several avirulent Salmonella typhimurium have been shown to be capable of reducing tumor mass [16,17,18,19]. This trait of bacteria could be enhanced by genetic engineering aimed towards enabling these bacteria to express therapeutic molecules selectively in the targeted tumors. The therapeutic molecules, however, do not readily pass through bacterial membrane. Various methods, such as active transport of recombinant proteins by fusion with signal peptides, have been used to enable bacteria to secrete foreign proteins [20,21]. One obvious limitation associated with this method is that secretion using signal peptides depends on the heteroprotein properties. To avoid these problems, a phage lysis system, most notably that of E. coli bacteriophage lambda, has been developed and applied to release heteromacromolecules of any kind [22,23,24]. Phage lambda encodes two proteins (S and R) which are sufficient to cause bacterial cell lysis [25,26,27]. Here, we employed lysis genes of a newly characterized Salmonella phage to release the heteromacromolecules from anti-tumoral Salmonella typhimurium [16], using MTD as an anti-tumoral cargo protein. This system was tested in vitro using cultured tumor cells and in vivo using tumor-bearing mice. Although the system was set up to express and release the therapeutic cargo out of the bacteria into the tumor tissue, the cargo also must subsequently be able to enter tumor cells. To overcome this problem, we constructed an MTD fused to a cell-penetrating peptide (CPP). CPPs are short, membrane-permeable, cationic peptides that are capable of both targeting intracellular proteins and carrying cargo proteins into tumor cells [28,29,30,31]. The best studied CPPs include those derived from the trans-acting activator of transcription (TAT) protein from the human immunodeficiency virus 1 (HIV-1) [32] and those derived from penetratin, a Drosophila Antennapedia homeoprotein [33]. DS4.3 (RIMRILRILKLAR) is a newly identified CPP peptide derived from S5 subunit of a voltagegated potassium channel (K v 2.1) (see below, manuscript in preparation [34,35]).In the present study, the DS4.3-MTD chimeric protein deployed by an attenuated Salmonella typhimurium was found to be effective both in vitro and in vivo against experimental tumor models.

Plasmids
The pLYS plasmid carried a 1.7 KB DNA fragment of the iEPS5 phage carrying lysis genes that had been cloned between the NheI and EcoRI sites in pBAD18 [40]. The pLYS plasmid and other plasmids used in this study carried the glmS gene from Salmonella typhimurium 14028s at the ClaI site in pBAD24. The glmS DNA was PCR-amplified using a pair of primers that contained the ClaI enzyme site and priming sequence with glmS: forward, ggatcgatatgtgtggaa ttgttggcgc, and reverse, ggatcgatttactctacggtaaccgatt (ClaI sites are underlined). pBAD24P BAD ::lacZ was constructed by cloning the lacZ gene such that it was under the control of P BAD , as follows. The lacZ DNA was PCR amplified from pRS415 [41] using the following primers: forward, aagaattcgtcgttttacaacgtcgtga, and reverse, ttgtcgacttatttttgacaccagacca, carrying EcoRI and SalI sites (underlined), respectively, and cloned between the two restriction sites of pBAD24 [40]. The pLYSP BAD ::lacZ plasmid was constructed by subcloning lacZ under the control of the P BAD system into pLYS in the PsiI site using PCR amplification by primers carrying SmaI sequences and P BAD ::lacZ priming sequences, with the pBAD24P BAD ::lacZ as the template.
The DNA of human Noxa with a DS4.3 cell penetration peptide sequence (cgcattatgcgtattctgcgcattctgaaactggcgcgt) was synthesized and cloned under P BAD [40] between the NcoI and PmeI sites in pBAD24. The DS4.3-MTD with P BAD system was PCR-amplified using the following pair of primers, containing SmaI enzyme sites, with P BAD ::DS4.3-MTD as a template: forward, aacccgggttatgacaacttgacggcta; reverse, aacccgggttatcaggttcctgagcaga (restriction sites are underlined). The amplified PCR product was digested with SmaI and ligated into the PsiI site of pLYS to generate pLYSP BAD ::DS4.3-MTD.

Growth conditions
Except when indicated otherwise, cultures were grown in LB medium (Difco Laboratories) containing 1% NaCl with vigorous aeration at 37uC. For solid support medium, 1.5% granulated agar (Difco Laboratories) was included. Antibiotics were from Sigma Chemical. When present, antibiotics were added at the following concentrations: ampicillin, 50 mg/ml; kanamycin, 50 mg/ml; chloramphenicol, 15 mg/ml. b-galactosidase assay b-Galactosidase assays were performed essentially as described by Miller [42]. For determinations of b-galactosidase levels in bacteria under different conditions of induction and noninduction, overnight cultured bacteria were diluted 1:50 in 50 ml LB media as described in the text. Cultures were further incubated at 37uC to log phase and induced with L-arabinose to a final concentration of 0.04%. The cultures were taken at the indicated times and were separated into supernatant and bacterial pellet fractions by centrifugation (30006 g, 10 min) and filtration (0.45 mm). Each sample was assayed in triplicate.

Cell lines and animals
Five-to six-week-old male Balb/c mice were purchased from Samtako Korea. CT26 mouse colon cancer cells, HeLa cells, and Hep3B cells, from the Waterborne Virus Bank (Seoul, Korea), were grown as a monolayer in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS and 1% antibiotic-antimycotic mixture (Gibco, Invitrogen). Tumor-bearing mice were generated by subcutaneous implantation of CT26 tumor cells (10 6

SDS-PAGE and Western blot analysis
For visualization of b-galactosidase and Noxa by Western, the cultures were taken at the indicated times and were separated into supernatant and bacterial pellet fractions by centrifugation and filtration. The samples were sonicated in the presence of SDS (0.1%) and protein concentrations were determined by Bradford method [43]. 50 mg of each sample was loaded on 8-15% SDS-PAGE gels and transferred to a nitrocellulose membrane (Bio-Rad, Hercules, CA, USA). After the transfer, the membrane was blocked with 5% skim milk and probed with 1:10,000-diluted mouse anti-b-galactosidase antibody (Santa Cruz, sc-65670) or 1:500-diluted rabbit anti-Noxa BH3 domain antibody (ABGENT, AP1316a) at 4uC overnight. Subsequently, the membrane was incubated with goat anti-mouse IgG-linked horseradish peroxidase (Santa Cruz, sc-2005) or goat anti-rabbit IgG-linked horseradish peroxidase (Santa Cruz, sc-2004) under the same conditions for 1 hr.

Optical bioluminescence imaging
Bioluminescence imaging was performed as previously described using an IVIS 100 system (Caliper) [13].

Cell death assay
The survival rate of the cancer cells was determined by 3-(3,4dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay as described [44]. Each cell line was seeded into 96-well plates at 10,000 viable cells per well. Twenty-four hours later, the cells were washed with PBS and placed in serum-free medium containing varying concentrations of chimeric Noxa (1, 10, 20, and 50 mM) or concentrated supernatants from bacterial lysates (the protein concentrations were 2, 20, and 200 mg). The filtrated bacterial supernatants was concentrated with a centrifugal filter (AmiconH Ultra-15, 3K), and the protein concentrations were determined by Bradford method [43]. Twenty-four hours later, the surviving cells were stained with MTT and quantified by absorbance at 540 nm. The MTT assay results were plotted with mean 6 S.D. of three experiments.

Histochemical and immuno-fluorescence Stain
Tumor tissue was collected from experimental mice that had undergone bacterial therapy and was embedded into an O.C.T. compound (Sakura, USA). Cryosections of 7 mm were obtained and mounted on ProbeOnTM Plus microscope slides (Fisher, USA). The b-Galactosidase Stain was performed with a b-Galactosidase Staining Kit (Mirus Bio LLC, USA). Briefly, the slides were washed with PBS and incubated in X-gal (5-bromo-4chloro-indolyl-b-D-galactopyranoside) staining solution at 37uC for 4 hrs in a humidified chamber, counter-stained with hematoxylin, and mounted with Faramount (Dako, USA). For immunofluorescence Stain, frozen section was fixed with cold acetone and Noxa was detected by rabbit anti-Noxa IgG (SantaCruz, SC-30209) and goat anti-rabbit IgG-FITC (SantaCruz, SC-2012) and nuclei was stained with 4,6-diamidino-2-phenylindole (DAPI). Stained samples were mounted using the SlowFade/Antifade Kit (Invitrogen, Carlsbad, CA, USA). The stained slides were examined using an Olympus BX51 microscope and an imaging program (analySIS LS starter).

Statistical analysis
Statistical analysis was performed using the SPSS 18.0 statistical package (SPSS Inc., Chicago, IL, USA). Statistical analysis was performed using the two-tailed Student's t test or two-way ANOVA. A P value of ,0.05 was considered statistically significant (* P,0.05, *** P,0.0001).

Characterization of bacterial lysis induced by a lysis system of novel phage origin
In an attempt to develop a system of delivery of anti-tumoral MTD into tumor tissue through Salmonellae, a bacterial lysis system of Salmonella phage origin was examined. The DNA of the iEPS5 phage, isolated from a field sample, was partially digested with the restriction enzymes NheI and EcoRI. Each of the fragments was then placed under the control of the P BAD promoter of the E. coli arabinose operon, which is induced by L-arabinose, and cloned into pBAD24 [40]. S. typhimurium 14028s was transformed with the recombinant plasmids and the library was examined for the lysis phenotype. The clones that showed the lysis phenotype upon addition of L-arabinose were selected. By sequencing the inserted fragments, a plasmid carrying a 1,856 bp DNA fragment (GenBank JF304023,5) with the lysis phenotype was selected and named 'Lys' [45] (Fig. 1). For subsequent studies, we used the highly attenuated DppGpp S. typhimurium [36], which also carries a mutation in the glmS gene, as the host. Thus, all the plasmids used in this study carried glmS (Fig. 1A, GlmS + p) so that the plasmids were maintained in the absence of selection pressure [37]. This balanced lethal host vector system using GlmS + p relies on the phenotype of the GlmS 2 mutant, which undergoes lysis when grown in the absence of N-acetyl-D-glucosamine (GlcNac) [37,46,47,48]. The 'Lys' fragment was subcloned into pBAD24 carrying glmS gene to generate pLYS (Fig. 1B).
First, the phenotype of controlled expression of the putative lysis function of pLYS was examined under in vitro growth conditions ( Fig. 2A). S. typhimurium defective in ppGpp synthesis carrying pLYS was grown in LB. When the A 600 of the culture reached ,1.0, at T = 2 hrs, L-arabinose was added to a final concentration of 0.04% (w/v), and the optical density of the culture (A 600 ) and total colony-forming units (CFU) were determined at the indicated times. The culture continued to proliferate in the absence of Larabinose addition, up to more than 5610 9 CFU/ml and A 600 ,5.0 at T = 4 hrs and thereafter. However, upon addition of L-arabinose to the culture, CFU reduced about 10 5 -fold, to about 1610 4 CFU/ml and an A 600 ,0.5 at T = 4 hrs. The optical density of the culture (A 600 ) changed accordingly: decreased upon addition of L-arabinose while increased in its absence. This result demonstrated the lysis function of the 'Lys' fragment cloned in pLYS, and also the control of its expression by L-arabinose.
Subsequently, the nature of the bacterial cell death following expression of the lysis genes was determined in vitro to verify that the lysis had resulted in the release of cellular contents into the media. SKS002 (DppGpp, DglmS::km) was transformed with pLYSP BAD ::lacZ (Fig. 1C) or pBAD24P BAD ::lacZ carrying lacZ (Fig. 1D) encoding b-galactosidase under the control of the P BAD promoter. The Salmonellae transformed with the pLYSP BAD ::lacZ expressed both lysis protein and b-galactosidase, and those transformed with the pBAD24P BAD ::lacZ only expressed bgalactosidase, upon addition of L-arabinose. When the A 600 of the culture reached ,1.0, L-arabinose was added and samples were taken 2 hrs after the induction. Supernatant was separated from cultured bacteria by centrifugation and filtration. The filtered culture media, bacterial pellet, and whole culture were examined for b-galactosidase activity [42]. As expected, in the case of Salmonellae carrying the pLYSP BAD ::lacZ strain, over 90% of the bgalactosidase activity was detected in the filtered culture media. By contrast, b-galactosidase activity in the Salmonellae carrying P BAD ::lacZ was detected mainly in the bacterial pellet (Fig. 2B). This was further verified by Western blot analysis using a bgalactosidase-specific antibody with the same samples (Fig. 2C). bgalactosidase-specific bands (,130 kDa) were detected in the whole-culture media of Salmonellae carrying either plasmid only upon L-arabinose addition. Most importantly, for the Salmonellae carrying pLYSP BAD ::lacZ, almost the same amount of b-galactosidase band was detected in the filtered media as in the whole culture, and little was detected in the bacterial pellet. Conversely, for the Salmonellae carrying pBAD24P BAD ::lacZ, the majority of the b-galactosidase was found in the bacterial pellet and little was detected in the filtered media. Taken together, this suggested that the majority of the proteins encoded in pLYSP BAD ::lacZ were released into the medium upon induction of the lysis function.

Control of lysis system in vivo
Next, we examined the controlled expression of the lysis system carried by S. typhimurium targeted to a tumor mass grafted in mouse model. This was because MTD is a highly cytotoxic protein that cannot be constitutively released without significant damage to nontumor tissue [10]. Thus, we determined biodistribution of the DppGpp Salmonella 3 days post infection (dpi) of BALB/c mice ectopically implanted with mouse CT-26 colon carcinoma cells, via intravenous route. Bacteria were counted in two major reticuloendothelial organs, liver and spleen, in addition to tumor tissue by CFU determination ( Table 2). The results demonstrate that DppGpp Salmonella preferentially accumulated in tumors over livers at a ratio of .48,000 : 1, and over spleen at a ratio of .65,000 : 1. All tumor-to-liver ratios were statistically significant (P,0.05). Previously, we have analyzed clinical parameters of CT26-bearing mice treated with the DppGpp Salmonella carrying cytolysin A (Cly A) under the control of inducible promoter, TetA promoter [49]. The analysis revealed a significant elevation of the markers of inflammation and hepatic function in the animals that were induced at 0 dpi but not in those induced at 3 dpi. Taken together, we concluded that it would be safe to induce lysis of intratumoral Salmonellae at 3 dpi. To monitor intratumoral bacterial lysis, we used Salmonellae carrying a bacterial luciferase (lux) operon in their chromosome [19], that enables non-invasive imaging of live bacteria using in vivo optical imaging system. These bioluminescent bacteria (SKS003), carrying pLYS, were injected intravenously into CT26-bearing (Fig. 3A). L-arabinose (60 mg) was administrated to animals by intraperitoneal injection at 3 dpi. As shown in Figure 3A, the bioluminescent signal decreased considerably compared with uninduced controls approximately 8 hrs after the administration of L-arabinose (Fig. 3A). Figure 3B shows the quantification of this result. At last, the number of live bacteria in the tumor tissue 8 hrs after the induction of the lysis system was determined by counting the CFUs from homogenized tumor tissues (Fig. 3C). Bacterial CFUs had decreased by over 99% at 8 hrs after L-arabinose administration, compared with the uninduced control group. This result coincided with the decreased bioluminescence signals seen after administration of L-arabinose in the tumor site ( Fig. 3A and B).
The release of the cytosolic contents of bacteria following induction of the lysis system was further verified in the mouse model using Salmonellae expressing b-galactosidase. The mice bearing CT26 were intravenously injected with Salmonellae (SKS002) carrying either pLYSP BAD ::lacZ or pBAD24P BAD ::lacZ. Three days after the bacterial injection, L-arabinose was administrated by intraperitoneal injection. Eight hours later, tumor samples were taken from the mice and the distribution of b-galactosidase following induction of bacterial lysis was analyzed by histochemical staining of the tumor tissue samples with X-gal (Fig. 3D). Blue pigmentation from b-galactosidase was detected in mice treated with Salmonellae carrying either plasmid, but only after L-arabinose administration. It should be noted that the scattering of blue pigmentation widely out of the area targeted by the Salmonellae carrying pLYSP BAD ::lacZ, but not in the tumor tissue targeted by the Salmonellae carrying pBAD24P BAD ::lacZ. In the latter case, the blue pigmentation was confined to the narrow area between necrotic and proliferative regions where Salmonellae were known to be localized [50,51]. Taken together, this was interpreted to mean that b-galactosidase induced in Salmonellae targeted to the tumor by L-arabinose, was released from the Salmonellae upon induction of the lysis system, and spread into neighboring areas, necrotic and proliferative tumor regions. Thus, the phage lysis system would be applicable to deliver proteinous antitumor cargo into tumor tissue.  Antitumor effect of the MTD Although the system was set up to express and release the MTD from the Salmonellae into the tumor tissue, the peptide needs to get into the tumor cells to be effective. The MTD by itself cannot penetrate tumor cells; however, the MTD fused to eight arginine residues (R8-MTD) has been shown to be able to penetrate into tumor cells and induce necrotic cell death [10]. In this study, therefore, we fused the MTD to different CPPs to improve the delivery into tumor cells further [29,30]. We synthesized peptides consisting of the eleven amino acids of the MTD and four C9 terminal amino acids fused to the C9 terminus of four kinds of CPPs with a trimeric glycine linker: TAT of HIV, penetratin of Drosophila, and, DS4.3 of K V 2.1.Pep-1 is an artificial CPP [52] (Fig. 4A). The CPP-MTD peptides were synthesized and tested for their abilities to induce cell death in vitro using various cell lines. We added 0-50 mM chimeric CPP-MTD peptides into Hep3B, HeLa, and CT26 cells ( Fig. 4A and B). After 24 hrs incubation, cell survival was determined by MTT assay. In all cases, the treatment with chimeric CPP-MTD caused cell death in a dosedependent manner. However, DS4.3-conjugated MTD induced cell death at the lowest concentration among all of the chimeric CPP-MTD peptides tested (Fig. 4B). Thus, the 30-amino acid DS4.3-MTD chimeric peptide was chosen as the cargo to be delivered by the Salmonellae using the phage lysis system. The DNA encoding the nucleotide sequence for DS4.3-MTD was synthesized and cloned under the L-arabinose-inducible P BAD in the pLYS plasmid background, generating pLYSP BAD ::DS4.3-MTD (Fig. 1E). Salmonellae (SK2002) transformed with this plasmid were cultured in LB to an A 600 of about 1.0 and induced for the expression of DS4.3-MTD as well as the phage lysis system by the addition of Larabinose. Samples of the bacterial culture were taken at the indicated times after the induction, and the media and bacterial pellets were divided, and analyzed for DS4.3-MTD expression by Western blot using an MTD-specific antibody (Fig. 4C). Two hours after induction with L-arabinose, DS4.3-MTD was detected in the supernatant media as well as in the bacterial pellet. Four and 8 hrs after the induction, DS4.3-MTD was detected mainly in the supernatant, suggesting that the lysis system functioned properly in this setting.
Subsequently, the DS4.3-MTD expressed in the Salmonellae and released into the media was tested for its cytotoxic effect on tumor cells (Fig. 4D). The supernatant of the culture taken after 4 hrs of induction was concentrated, added to cultured CT26 cells, and incubated overnight. Cell death was determined by MTT assay. The concentrated supernatant media of the Salmonellae carrying pLYSP BAD ::DS4.3-MTD induced with L-arabinose showed the induction of cell death in a dose-dependent manner. By contrast, the concentrated supernatant of the Salmonellae carrying pLYS showed no cell death effect. Thus, DS4.3-MTD expressed and released by Salmonellae was demonstrated to be effective in vitro.
Lastly, the antitumor effect of the Salmonellae expressing DS4.3-MTD was tested using BALB/c mice implanted with CT26 on the right thigh. A total of 1610 7 Salmonellae (SK002) carrying pLYS or pLYSP BAD ::DS4.3-MTD were injected into the tumor-bearing mice through the tail vein when the tumor size reached about 100-150 mm 3 . Three days after the injection, L-arabinose was intraperitoneally injected and changes in the size of the tumor were determined (Fig. 5A and B). The results showed a maximum anti-tumoral effect with Salmonellae carrying pLYSP BAD ::DS4.3-MTD induced by L-arabinose injection, followed by the same treatment without L-arabinose injection. We also observed an antitumor effect from the Salmonellae carrying pLYS induced for bacterial lysis, and this effect was considerably greater than that without bacterial lysis, which was about the same as that in untreated controls (the PBS-treated group).
Subsequently, we verified the expression and release of DS4.3-MTD in the above tumor tissue upon administration of Larabinose (Fig. 5C). Immuno-fluorescence staining revealed expression of DS4.3-MTD in the induced samples at the area between necrotic and proliferative regions where Salmonellae were known to be localized [50,51]. Evidently, DS4,3-MTD under the control of P BAD was induced as well as b-galactosidase (Fig. 3D) in the mouse model upon administration of L-arabinose. However, we observed the presence of MTD in the uninduced samples, although marginal, suggesting that DS4.3-MTD expressed at the basal level would be responsible for the anti-tumoral effect with Salmonellae carrying pLYSP BAD ::DS4.3-MTD without L-arabinose injection (Fig. 5A). The tumors from the mice treated above were extracted 5 days after the L-arabinose administration and analyzed by H&E staining (Fig. 5D). This histopathological analysis revealed that the cells died by necrosis. The extent of the necrosis was most serious in the tumor samples taken from mice treated with bacteria carrying pLYSP BAD ::DS4.3-MTD induced by L-arabinose administration (over 70%), and less so in those from uninduced mice and those from mice treated with Salmonellae carrying pLYS, following the order of antitumor effects of the treatments. Taken together, using a mouse model, the DS4.3-MTD expressed and released from Salmonellae by the phage lysis system was demonstrated to be effective in inducing necrosis of tumor cells, resulting in tumor suppression.

Discussion
The MTD has been shown to induce necrotic cell death indiscriminately both in tumor cells and normal cells [11]. Thus, unless a specific tumor-targeting signal is employed [10], MTD cannot be developed as an antitumor drug. Our bacterial delivery system provides an alternative means to deliver the MTD of Noxa selectively to the tumor site, by taking advantage of tumortargeting characteristics of the bacteria. However, the delivery of therapeutic materials requires release of the macromolecules from the bacteria after colonization of tumor tissue. Various bacteria delivery systems have been developed for this purpose. Although selective secretion using a signal peptide would be one way to overcome the barrier, finding a suitable signal peptide for a specific macromolecule is a cumbersome process. Alternatively, macromolecules can be released from bacteria by the lysis of the bacterial membrane with a phage lysis system as demonstrated in this study. The advantage of using this type of lysis system is that macromolecules of any kind would be delivered in this way into the tumor site; however, this requires controlled expression of the lysis gene so that the system will be turned on only when bacteria have predominantly accumulated in the tumor. An obvious disadvantage, however, using one inducible promoter, P BAD in this case, for induction of both tumorlytic gene and the lysis gene, would be sub-maximal expression of the tumorlytic gene, which might consequently diminish the antitumoral therapeutic effect. Thus, it would be desirable to use different inducible promoter systems, i.e. a combination of tet promoters [49] and P BAD , for induction of tumorlytic gene and the lysis gene.
Usually, 3-4 days after the injection of bacteria, we observed the greatest preferential accumulation of bacteria in tumors over endothelial organs. Interestingly, this ratio varied depending on the bacterial strains and model animals. We counted 0 CFU in the reticuloendothelial organs with E. coli (MG1655) in tumor-bearing mice [13]. With DppGpp Salmonella, we counted ,10 4 CFU, thus ratio of tumor-to-reticuloendothelial organs ranged in ,10 5 (Table 2), which was far better than the Salmonella VN20009, a widely used bacterial therapy strain, of which ratio ranged in ,10 3 [53]. Our recent publication reported biodistribution of DppGpp S. typhimurium in myocardial infarction (MI) rat models, in which DppGpp Salmonella selectively accumulated and proliferated in MI tissue [54]. Early after injection (12 hours), bacterial load was found primarily in the spleen and liver, most likely captured by phagocytic macrophages [55], with a relatively small bacterial  burden in the heart. After 24 hours, however, the number of CFUs in myocardial tissue increased dramatically, reaching a maximum on 3 and 5 dpi of ,10 6 CFU/g, whereas the bacterial burden in the liver and spleen declined over the same period of time to undetectable levels. No bacterium was detected in the liver and spleen on 5 dpi. No sign of serious local or systemic inflammatory reactions was noted following i.v. administration of DppGpp S. typhimurium: the levels of c-reactive protein (CRP) and procalcitonin (PCT) were not significantly changed. A similar result was obtained with DppGpp Salmonellae carrying ClyA [49]. Thus, 3 dpi would be the time point to express and release antitumoral macromolecules from the bacterial strain used in this study (Fig. 3). The P BAD system has provided an efficient means of controlling gene expression of bacteria in tumors remotely by intraperitoneal administration of its inducer, L-arabinose [19,56]. However, the evident basal level expression of the cargo proteins in the absence of inducer ( Fig. 2C and 5C) demands further improvement of the regulatory circuit.
In most cancers, blood vessel growth does not keep pace with tumor cell growth that leads to pronounced hypoxia often accompanied by necrosis [57]. Facultative anaerobes such as Escherichia coli and Salmonella spp. should be able to grow in viable as well as necrotic areas of the tumor. Previously, we have traced the E. coli that was i.v. injected in model mice bearing CT26, and reported that bacteria were detected initially towards the center of the tumor. As time passed, they appeared to spread to the border between the necrotic and viable areas of the tumor. Concomitantly, b-galactosidase and MTD expressed from intratumoral Salmonella were found at the area between necrotic and proliferative regions where Salmonellae were known to be localized ( Fig. 3D and Fig. 5C). The confinement of bacteria has been suggested to be due to host immune system that prevents bacterial dissemination throughout tumors [51]. In particular, neutrophils were shown to prevent bacteria from spreading from necrotic to viable tumor tissue. In fact, however, bacterial-induced destruction of viable cells would position bacteria at the interface with the necrotic region. At last, as tumor regresses, bacteria would be confined to the outermost part where scarcely any viable region remains. Consistently, a remarkable correlation between antitumor effect of bacterial treatment and the extent of necrosis was observed (Fig. 5D).
One of the main concerns raised regarding bacterial therapy is the possibility of infection caused by the bacteria used for the antitumor therapy. This, however, would not be a problem in this system, since the therapeutic bacteria are easily cleared using antibiotics, as demonstrated previously [54]. In this study, we showed that the Salmonellae equipped with the phage lysis system do undergo lysis upon its induction: over 99.9% of bacteria were lysed upon induction with L-arabinose in vivo and in vitro (Figs. 2A and 3C). In fact, we observed that at the end of the bacterial treatment with the phage lysis system (,two to three weeks after the injection), the number of bacteria counted at the tumor site was less than 100 (manuscript in preparation). This is probably because bacteria are no longer protected in immunologically privileged environment [58], as tumor regresses, and thereby cleared by host immune system. Thus, this feature could provide another level of safety by reducing the number of live bacteria. Taken together, the phage lysis system provides not only a means to release antitumoral macromolecules from the bacteria, but also an additional level of safety for bacteria-mediated cancer therapy.