Cell mass-dependent expression of an anticancer protein drug by tumor-targeted Salmonella

Bacterial cancer therapy relies on the properties of certain bacterial species capable of targeting and proliferating within solid malignancies. If these bacteria could be loaded with antitumor proteins, the efficacy of this approach could be greatly increased. However, because most antitumor proteins are also toxic to normal tissue, they must be expressed by bacteria that specifically target and exclusively localize to tumor tissue. As a strategy for treating solid malignancies, we recently evaluated L-asparaginase (L-ASNase) delivered by tumor-targeted Salmonella. In this system, L-ASNase was expressed under the control of the araBAD promoter (PBAD) of the E. coli arabinose operon, which is induced by injection of L-arabinose. Here, we further improved the performance of recombinant Salmonella in cancer therapy by exploiting the quorum-sensing (QS) system, which uses cell mass-dependent auto-induction logic. This approach obviates the necessity of monitoring intratumoral bacterial status and inducing cargo protein expression by administration of an exogenous compound. Recombinant Salmonella in tumors expressed and secreted active L-ASNase in a cell mass-dependent manner, yielding significant anticancer effects. These results suggest that expression of a therapeutic protein under the control of the QS system represents a promising engineering platform for the production of recombinant proteins in vivo.


INTRODUCTION
Bacterial cancer therapy relies on the natural propensity of certain bacteria, including Escherichia coli [1], Salmonella [2], Clostridium [3], and Listeria [4], to accumulate and replicate in various types of tumors [5]. However, the strains must be modified to eliminate factors responsible for bacteria-related toxicity, i.e., products of virulence genes and obvious endotoxins. In the case of Salmonella typhimurium, the first such strain (VNP20009) used for cancer therapy carried a mutation in msbB, which altered lipid A structure, and purI, to confer purine auxotrophy [6][7][8]. Another strain used in this context is the A1-R mutant, a leucine-arginine auxotroph with elevated antitumor activity that arose during selection by tumor passage [9][10][11][12][13][14]. This strain exerts extraordinary antitumor
In our studies, we have used the ΔppGpp strain of Salmonella typhimurium, which lacks both genes encoding ppGpp synthesis enzyme, relA and spot [27][28][29], because this strain is capable of targeting all solid tumors tested but is almost a million-fold less virulent than the wild-type strain due to defective expression of the genes encoded by Salmonella pathogenicity islands [30,31]. Recent studies have attributed tumor suppression by ΔppGpp Salmonella to a marked increase in production of proinflammatory cytokines, namely, TNF-a and IL-1b, by macrophages and dendritic cells infiltrated into the tumor mass [32,33].
Although tumor targeting by bacteria alone results in a reduction of tumor mass [5,9,[34][35][36][37], bacteria expressing therapeutic proteins are clearly more efficacious [30,38,39]. Because most anticancer proteins are to some degree also toxic to normal cells, their expression must be confined to tumor tissue. Following intravenous (i.v.) injection in mice, Salmonella rapidly accumulate in reticuloendothelial (RE) systems, liver, and spleen, but to a lesser extent in tumor tissue [38,39]. Several days later, the Salmonella gradually clear from the RE systems but increase in number in tumor tissue at a rate that depends on the strain. This accumulation of Salmonella is due in part to the immune-suppressive environment in tumor tissue, which allows proliferation of intratumoral Salmonella to densities greater than 10 9 colony-forming units (CFU)/g of tissue [40,41]. Using ΔppGpp Salmonella, we observed that controlled expression of cytotoxic proteins such as cytolysin [42] or Noxa [43] 3 days post-infection (dpi) did not cause notable systemic toxicity.
L-asparaginase (L-ASNase, EC2) of E. coli origin is widely used in therapy of acute lymphoblastic leukemia (ALL) [44]. L-ASNase primarily catalyzes the deamination of asparagine to aspartate, and to a lesser extent the conversion of glutamine to glutamate [45]. In tumor cells defective in asparagine synthetase, depletion of asparagine leads to inhibition of global protein synthesis [46] due to a failure to resupply asparagine [47,48], resulting in apoptotic cell death. L-ASNase administered via the intravenous (i.v.) route [47] is efficacious against ALL, but not against solid tumors, due to the limited distribution of the drug: L-ASNase injected i.v. rarely accumulates in grafted solid tumors in mouse models [49,50]. Moreover, i.v. administration of high doses of L-ASNase to tumor-bearing mice results in morbidity and mortality, and the side effects associated with L-ASNase include anaphylactic shock, coagulopathies, and hepatic and pancreatic toxicity [50,51].
Previously, we engineered Salmonella to express L-ASNase selectively within solid tumors using a remote gene control system based on the araBAD promoter (PBAD) of E. coli, which can be induced by intraperitoneal (i.p.) administration of L-arabinose. In mice harboring grafts of solid tumors, we demonstrated the antitumor efficacy of targeted L-ASNase using Salmonella transformed with this construct. Daily administration of L-arabinose was required to maintain continuous expression of L-ASNase. However, remote gene control via diffusion from the injection site had some intrinsic problems, including off-target effects and inhomogeneity. Frequent injection was also stressful for the animal. These inherent limitations of the PBAD-induced L-ASNase system mandated the development of a system in which L-ASNase is induced in a cell mass-dependent manner. To this end, we turned to a bacterial auto-induction system.
Certain species of bacteria exhibit cooperative behavioral patterns mediated by an environmental sensing mechanism termed the auto-induction system. This system allows bacteria to monitor their own population density by producing, secreting, and sensing a diffusible compound called the auto-inducer, whose concentration in the environment correlates well with local cell density during growth. At low cell densities, the auto-inducer is present at low concentrations, whereas at high cell densities it accumulates to the critical concentration required for activation of its target genes. For example, luminescence in the marine bacteria Vibrio fischeri and Vibrio harveyi is auto-induced by acyl-homoserine lactone (AHL). In this quorum-sensing (QS) system, the LuxI and LuxR proteins control expression of the luciferase operon (luxICDABE), which encodes the proteins required for luminescence. LuxI catalyzes the synthesis of the auto-inducer AHL [52,53], whereas LuxR is the cytoplasmic AHL receptor that functions as a transcriptional activator upon ligand binding [54]. After synthesis in vivo, AHL freely diffuses in and out of bacterial cells, and its concentration increases with local cell density [55]. When AHL reaches a critical threshold concentration, the LuxR-AHL complex activates transcription of the operon encoding luciferase [56].
In this study, we engineered S. typhimurium harboring an auto-inducible recombinant plasmid expressing a gene encoding L-ASNase under the control of the Vibrio QS expression system, and demonstrated the bacterial density-dependent expression of L-ASNase and its antitumor effects in a mouse graft model.

Expression of reporter protein under the control of the QS promoter system
The QS system (luxR, luxI, and luxR-luxI intergenic sequence including the luxR and luxI promoter) used in this study was deduced from the whole genome sequence of Vibrio fischeri and specifically amplified by PCR using primers listed in Table 1. The resultant DNA fragments were assembled by overlapping PCR to exclude incorporation of Oncotarget 8550 www.impactjournals.com/oncotarget additional bases, which occurs during typical subcloning procedures using restriction enzyme recognition sites. For efficient co-expression of transcriptionally coupled proteins (mCherry and L-ASNase) with LuxI, a wellknown ribosome-binding sequence (RBS) from pQE30 was also included ( Figure 1A). mCherry was deployed as a reporter to test the QS elements. Using this method, we constructed three different plasmids carrying the cell density-dependent expression system. pQSRI-mCherry consisted of luxR and the luxR-luxI intergenic sequence, including both promoters and mCherry transcriptionally coupled with luxI. pQSI-mCherry was the same except that it lacked luxR. pQSR-mCherry carried the mCherry gene under the direct control of luxI promoter, and thus lacked luxI, but retained intact luxR. The glmS gene was incorporated into these recombinant plasmids to generate a balanced-lethal host vector system [57] that complemented the phenotype of the GlmSmutant, which undergoes lysis unless its catalytic product N-acetyl-D-glucosamine is provided or complemented by a GlmS + -containing plasmid. These plasmids were transformed into an avirulent ΔppGpp Salmonella typhimurium, 14028s.
Salmonella carrying these plasmids were cultured in Luria-Bertani (LB) media, and samples were collected at the indicated times (h) to determine viable cell count (CFU, Figure 1B, left panel) for analysis of mCherry expression by fluorometry ( Figure 1B, right panel) and for Western blotting using mCherry-specific antibody ( Figure 1B, bottom panel). The results revealed that Salmonella carrying pQSRI-mCherry began to show fluorescence at ~3 h, which saturated at 9 h as the culture density rose above 10 9 CFU/ml, whereas those carrying pQSI-mCherry or pQSR-mCherry expressed no fluorescence. The expression of mCherry by Salmonella carrying pQSRI-mCherry was visualized by fluorescence microscopy ( Figure 1C). Bacterial cells expressing mCherry fluorescence began to be detected starting at ~4 h and increased in number as time passed. Figure 1D shows quantitation of the fluorescent Salmonella shown in Figure 1C. Salmonella carrying pQSR-mCherry failed to express mCherry because they lacked LuxI and were thus unable to produce auto-inducer [52]. Accordingly, addition of AHL to the media restored mCherry expression ( Figure 1E), albeit only partially, as expected. Taken (E) Expression of mCherry by ∆ppGpp Salmonella carrying pQSRI-mCherry or pQSR-mCherry supplemented with or without AHL in LB, as determined by fluorometry. Salmonella carrying pQSR-mCherry were included as a control. Asterisks ( * ) in panels B and D indicate significant differences compared to pQSI-mCherry ( *** P < 0.0001).
Oncotarget 8551 www.impactjournals.com/oncotarget together, these observations demonstrate that a cargo gene transcriptionally coupled with luxI in the presence of LuxR can be expressed in a cell density-dependent manner.
Subsequently, we determined the expression and distribution of mCherry expressed by tumor-targeted Salmonella carrying pQSRI-mCherry in C57BL/6 mice bearing MC38 tumors ( Figure 2). Salmonella (1 × 10 7 CFU) were injected through the tail vein when the tumor implanted in the right lateral thigh reached ~120 mm 3 . Mice were sacrificed on 1 and 5 dpi (days post-infection). The excised tumors were subjected to immunofluorescence staining and analyzed for Salmonella and mCherry expression. The number of Salmonella in tumor tissue was in the range of 10 7 -10 8 CFU/g in tumor tissue on 1 dpi, and a bit higher on 5 dpi, whereas the number in liver and spleen was considerably lower, between 10 3 and 10 4 CFU/g tissue in liver during the same time period (Figure 2A). mCherry expression was abundant in tumor tissue at 5 dpi, mostly overlapping with Salmonella, but much less abundant at 1 dpi. By contrast, in the reticular endothelial system, very little mCherry was detected ( Figure 2B). These results confirmed QSdependent mCherry expression in an animal system.

Expression and secretion of ASNase from Salmonella carrying pQSRI-ASNase
To develop a strain useful for treating solid malignancies using a cell density-dependent inducible QS system, the mCherry gene in the above constructs was replaced with the L-ASNase gene of E. coli B (BL21) [58]. To this end, we prepared three constructs, pQSRI-ASNase, pQSI-ASNase, and pQSR-ASNase ( Figure 3A), and introduced them into the host bacteria. These engineered Salmonella were grown in LB media, and samples were taken at the indicated times. The cells were harvested by centrifugation and analyzed for L-ASNase expression by Western blotting using a specific antibody ( Figure  3B). L-ASNase was induced in a cell density-dependent manner only in Salmonella carrying pQSRI-ASNase. The culture medium of Salmonella carrying pQSRI-ASNase was also harvested, and secreted protein in the medium was analyzed by Western blotting. As shown in Figure 3B (bottom row), we detected a distinct band corresponding to L-ASNase. These results indicated that L-ASNase was induced at an appropriate cell density, and that a significant fraction (~55%) was secreted into the media, as demonstrated previously [50]. The cell density-dependent expression of L-ASNase was verified by determining the enzyme activity in Salmonella growing in LB medium ( Figure 3C).
Previously, we engineered Salmonella expressing L-ASNase under the control of PBAD of the E. coli arabinose operon, which can be induced with L-arabinose (P BAD ::ASN), and demonstrated its effects on various solid tumors in a mouse model [50]. In this study, we compared the expression of L-ASNase from pQSRI-ASNase with that from P BAD ::ASN. For this purpose, Salmonella carrying pQSRI-ASNase were grown in LB medium, and cells carrying P BAD ::ASN were grown in LB media with Oncotarget 8552 www.impactjournals.com/oncotarget or without 0.2% L-arabinose. Equal amounts of bacterial cultures were harvested and separated into two fractions, bacterial pellet and media supernatant, and L-ASNase levels were analyzed by Western blotting (Figure 3D upper image). Addition of L-arabinose significantly induced L-ASNase expression in Salmonella carrying P BAD ::ASNase at levels comparable to those expressed from Salmonella carrying pQSRI-ASNase grown in LB medium for 9 h. In both cases, more than half of L-ASNase was secreted out of the bacterial cells, which can be ascribed to the extracellular nature of the protein [59]. This was further corroborated by determination of L-ASNase enzyme activities in the same bacterial medium: the levels were very low in Salmonella carrying P BAD ::ASN without induction, and the levels were higher (and comparable) in Salmonella carrying P BAD ::ASN induced by L-arabinose and those carrying pQSRI-ASNase ( Figure 3D lower graph).

Antitumor effect of L-ASNase expressed by Salmonella carrying pQSRI-ASNase
Next, we determined the antitumor effect of L-ASNase expressed and secreted by Salmonella carrying pQSRI-ASNase in vitro, using cultured mouse MC38 colon cancer cells. Salmonella carrying pQSR-ASNase or pQSI-ASNase were cultured for 9 h, and then the media were separated and concentrated. Concentrated media were added to cultured MC38 cells at final concentrations of 0.2 mg/ml, and then incubated for 24 h. MC38 cell viability was determined by MTT assay ( Figure 4A). Media from Salmonella carrying pQSRI-ASNase killed almost all (99%) cells, whereas media from bacteria carrying other plasmids had little effect on viability. The cytotoxic effect of L-ASNase was also examined by TUNEL assay (Figure 4B and 4C). Over 90% of MC38 cells incubated with media from Salmonella carrying pQSRI-ASNase were TUNEL-positive; again, the other bacterial strains had minimal effect. Taken together, these data indicate that L-ASNase expressed and secreted under the control of the QS system is highly toxic to MC38 cancer cells in vitro.
Oncotarget 8553 www.impactjournals.com/oncotarget effect than the phosphate-buffered saline (PBS)-treated control, which could be ascribed to the anticancer effect of Salmonella itself [9,36]. Tumors reached a mean size of 1,000 mm 3 by day 4 following PBS treatment and by day 8 following treatment with Salmonella carrying either pQSR-ASNase or pQSI-ASNase. By contrast, tumors in mice treated with Salmonella carrying pQSRI-ASNase were less than 500 mm 3 at the end of the experiment (day 14). Consistent with this, survival was considerably longer in animals that received Salmonella carrying pQSRI-ASNase than in the other groups ( Figure 5C).
Analysis of L-ASNase levels in tumor tissues from mice treated with Salmonella carrying pQSRI-ASNase ( Figure 6) revealed little L-ASNase at 1 dpi, but a considerable level at 5 dpi, as assessed by Western blotting with a specific antibody ( Figure 6A, n = 4) and immunofluorescence microscopy ( Figure 6B). Notably, L-ASNase co-localized with Salmonella in a densitydependent manner, i.e., L-ASNase was detected in areas with dense Salmonella populations ( Figure 6B). Based on these findings, we conclude that L-ASNase expressed in a bacterial density-dependent manner effectively suppressed solid tumor growth in our mouse model.

DISCUSSION
Successful application of cancer therapy using bacteria expressing a genetically engineered anticancer protein requires that the active agent be conditionally expressed in tumor tissue, leaving the reticular endothelial system undamaged. In this study, we developed and employed an expression system that is activated in a bacterial density-dependent manner, causing the therapeutic protein under its control to be expressed only when bacterial cell density rises above a threshold level (at least 10 8 CFU/ml, Figure 1B). Bacterial cell densities in healthy liver and spleen remained below 10 4 CFU/g. In vitro experiments showed that luxI promoter-driven expression of the cargo gene was activated as the cultures entered stationary phase, as bacterial cell density rose above ~10 9 CFU/ml in liquid culture ( Figure 1B). This induction was dependent on AHL produced by LuxI and its cognate binding protein LuxR. This theory was corroborated by restoration of induction by addition of exogenous AHL into the culture medium of Salmonella carrying pQSR-mCherry ( Figure 1E).
Examination of the expression of mCherry and L-ASNase in intratumoral Salmonella at 1 and 5 dpi by IF microscopy (Figures 2 and 6) revealed greater abundance at 5 dpi than 1 dpi, but significantly more mCherry/L-ASNase expression at 5 dpi in the area where Salmonella had accumulated heavily (Figures 2B and 6). This pattern of expression suggested that AHL reached the critical concentration only in these areas of high Salmonella density. Consistent with this, little mCherry expression was detected in the reticular endothelial system. A similar study using Salmonella carrying an episomal luxI promoter-luxR-GFP-luxI construct [60] concluded that GFP expression by intratumoral Salmonella colonies depended on the distance between neighboring bacteria; no colonies expressed GFP when the average distance to neighbors was greater than 155 µm. Thus, this QSdependent induction system is far superior to the PBAD system of E. coli, which provides a remote control system by which a tumor-toxic protein can be induced by i.p. administration of L-arabinose [37,42,50]. Consequently, sustained expression of the cargo protein by intratumoral bacteria requires repeated administration of L-arabinose. As demonstrated in this study, QS-dependent induction of the L-ASNase by intratumoral Salmonella provides a simple method for sustained expression of a protein drug without requiring supplementation with an exogenous Oncotarget 8554 www.impactjournals.com/oncotarget inducer, and achieved notable retardation of tumor growth and delay of mortality in tumor-bearing mice ( Figure 5).

Bacterial strains, plasmids, and culture conditions
The E. coli strain used in this study for gene manipulation was XL1-Blue. The QS expression module of Vibrio fischeri was derived from the sequence deposited in GenBank (AF170104.1). The expression plasmid pBAD24 (Invitrogen, USA) was chosen as a scaffold for construction of the QS expression system in Salmonella. The Salmonella strain used in this study, the ppGpp-defective strain SHJ2037 (relA::cat, spoT::kan), was derived from the wild-type Salmonella enterica serovar Typhimurium 14028s, and was described previously [27,61]. This strain is also GLmS -, and therefore requires N-acetylglucosamine for growth unless complemented by a GLmS + -containing plasmid [57].
Recombinant Salmonella were grown at 37°C in LB medium containing 1% NaCl. For solid support medium, 1.5% Bacto agar was included. All media were supplemented with an antibiotic (ampicillin, 100 µg/ml) and/or AHL (0.05-1 mM) as needed.

Construction of recombinant plasmids with various genes under the control of the QS expression system
To express target genes under the control of the QS system, the expression vector pQS was constructed using pBAD24 as a scaffold. The primer pair LuxRF/ LuxROLR (Table 1) was designed to amplify the gene encoding LuxR (accession no. AAD48473.1); the primer pair luxRI proOLF/LuxIR was used to generate a construct that fuses the luxR-luxI intergenic sequence with luxI gene (accession no. AAD48474.1). As a functional reporter, the gene encoding mCherry (Takara Bio, USA) was also amplified using primer pair mChOLF/mChR. The three amplicons were assembled by overlapping PCR, and then subcloned into the BamHI/SalI site of pBAD24 to yield pQSRI-mCherry ( Figure 1A). The recombinant plasmid pQSI-mCherry was prepared as follows: primers luxRI proF and LuxIR were used to generate a construct that fuses the luxR-luxI intergenic sequence with the luxI gene. This DNA fragment was assembled by PCR using the same mCherry gene described above. The resultant DNA fragment was subcloned into the BamHI/SalI site of pBAD24. pQSR-mCherry was also constructed by overlapping PCR with two constructs, a DNA fragment containing luxR and the luxR-luxI intergenic sequence amplified by PCR with primers LuxRF and LuxRI proR, and a DNA fragment containing mCherry amplified with primers mChOL-RIF and mChR. The resultant construct was also subcloned into the BamHI/ SalI site of pBAD24. For stable maintenance in vivo without exogenous inducer, the glmS gene was amplified with primers glmSF and glmSR primers from genomic DNA of Salmonella and inserted into the vector at the ClaI site.
To evaluate these constructs as a genetically encoded drug expression system in vivo, the three plasmids pQSRI-ASNase, pQSI-ASNase, and pQSR-ASNase, in which the mCherry gene was replaced with the gene encoding ASNase, were constructed using the procedure described above ( Figure 3A). These constructions used DNA fragments amplified with primer pairs ASN OLF/ASN R and ASN OL-RIF/ASN R. All constructs were confirmed by DNA sequencing and functionally tested in both E. coli XL1-Blue and DppGpp S. typhimurium.

Preparation of L-ASNase
To monitor the intracellular expression level of L-ASNase, recombinant bacteria were harvested by centrifugation at 5,000 × g for 5 min and lysed by sonication in lysis buffer (10 mM lysozyme, 10% sodium dodecyl sulfate). The resultant solution was further centrifuged at 13,000 × g for 30 min, and the supernatant was collected, filtered (0.22 µm pore size), and concentrated using a Centricon device (Amicon ® Ultra, 10K pore filter, Millipore). The proteins were precipitated with 10% trichloroacetic acid (TCA) for 1 hr at 4°C, and the pellets were re-dissolved in PBS for analysis of secreted L-ASNase.
Cell viability was assessed using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) colorimetric assay. Briefly, 100 μl cell suspension was inoculated in each well of a 96-well plate at a density of 2 × 10 4 cells/well. After 24 hr of culture, the medium was removed by aspiration and replaced with 100 μl experimental medium, and the cells were cultured for an additional 24 hr. MTT solution (10 μl) was added to each well, and the plates were incubated in the dark for 4 hr at 37°C. The number of remaining viable cells (i.e., cells with MTT dye) was determined by measuring the optical density at 490 nm in an enzyme-linked immunosorbent assay reader.

Animal experiments
Five-to six-week-old male mice (20-30 g body weight) were purchased from Samtako (Korea). When the tumor reached ~120 mm 3 , 1 × 10 7 CFU ∆ppGpp S. typhimurium transformed with the indicated plasmids were suspended in PBS and injected through the tail vein. All animal care, experiments, and euthanasia were performed in accordance with protocols approved by the Chonnam National University Animal Research Committee. Animals were anesthetized with isoflurane (2%) for imaging or a mixture of ketamine (200 mg kg -1 ) and xylazine (10 mg kg -1 ) for surgery. Mice carrying subcutaneous tumors were generated as follows: tumor cells cultured in vitro were harvested, suspended in 100 µl PBS, and injected subcutaneously into the right thigh (1 × 10 6 cells for MC38). Tumor volumes (mm 3 ) were estimated using the formula (L × H × W)/2, where L is the length, W is the width, and H is the height of the tumor in millimeters.

L-ASNase activity assay
Bacterial supernatants containing L-ASNase were prepared in concentrated forms for the enzymatic assay. Tumor tissue samples were extracted with four volumes of assay buffer from the Asparaginase Activity Assay kit (Abcam, ab107922), and insoluble materials were removed by centrifugation. Levels of L-ASNase enzymatic activity in bacterial supernatants (100 µg) and tumor tissues (100 µg) were assessed using the Asparaginase Activity Assay kit.

Statistical analysis
Statistical analysis was performed using the SPSS 18.0 statistical package (SPSS Inc., Chicago, IL, USA). The two-tailed Student's t-test was used to assess the statistical significance of tumor growth and differences in survival between treatment groups. P < 0.05 was considered statistically significant.