Salmonella typhimurium (hereafter S. typhimurium), as Gram-negative facultative anaerobic bacteria, are good candidates for cancer therapy and delivering therapeutic antitumor agents. However, it is necessary to reduce the virulence of such bacteria and enhance their tumor-targeting ability, and their immunostimulatory ability to induce tumor cell apoptosis. In this study, we constructed a S. typhimurium mutant named S634 harboring aroA mutation and additional mutations involved in modifications of lipid A. Upon intraperitoneal infection in mice, the aroA-deficient strain S634 showed greatly attenuated virulence and preferential accumulation within tumor tissue. We next investigated the ability of S636, the asd mutant derivative of S634, to deliver the anti-angiogenic agent “endostatin” (S636/pES) and to inhibit tumor growth in mouse CT26 colon carcinoma and B16F10 melanoma models. S636/pES-treated tumor-bearing mice showed suppressed tumor growth and prolonged survival, compared to mice treated with either the bacteria carrying empty plasmids or PBS intraperitoneally. Immunohistochemical studies demonstrated that, when tumor-bearing mice were infected with S636/pES, Salmonella colonization and endostatin expression were accompanied by the increase of apoptosis level and suppression of tumor angiogenesis within tumor tissues. Our findings showed that endostatin gene therapy delivered by attenuated S. typhimurium displays therapeutic antitumor effects in murine tumor models.
Conventional tumor therapies including chemotherapy, radiotherapy, and surgery have obvious limitations, such as the toxicity to normal tissues or cells, low tumor-targeting, and the disability of penetrating tumor tissue, and usually resulting in incomplete damage to the tumor . Therefore, it is necessary to explore more effective means for cancer treatments. Early studies have demonstrated that some obligate or facultative anaerobes have natural preference for tumor tissue, such as Clostridium , Bifidobacterium , Escherichia coli , and Salmonella [5, 6]. Especially, Salmonella, as Gram-negative and facultative anaerobic bacteria, have many potential advantages in tumor therapy. (i) Salmonella preferentially accumulate within tumor tissue compared to normal tissues , and motile Salmonella may disperse into different regions of tumor tissue [8, 9]. (ii) Salmonella can attack a variety of tumors, such as colorectal carcinoma, melanoma, breast cancer, and prostate cancer, leading to inhibition of tumor growth through multiple mechanisms . (iii) Salmonella can be developed as live attenuated bacterial vectors for delivering tumor therapeutic agents, including cytokines, anti-angiogenic agents, tumor antigens, apoptosis-inducing factors, and small interference RNAs . (iv) Salmonella with the immunomodulatory ability can elicit strong adjuvant activities [11, 12], thereby improving therapeutic benefits when combined with other cancer treatments [13, 14].
In the past two decades, therapeutic antitumor effects mediated by recombinant attenuated Salmonella have been widely studied in animal models and human clinical trials [15,16,17]. Notably, considering intrinsically pathogenic properties of Salmonella, efficient attenuation of bacterial virulence is the prerequisite for the application of Salmonella in preclinical and clinical studies. For example, the well-known S. typhimurium strain termed VNP20009  was lipid A-modified (msbB-) and auxotrophic (purI-), in which the gene purI is responsible for the adenine synthesis. The tumor-seeking auxotrophic strain named A1-R, which requires arginine and leucine for growth, has been tested for tumor-targeting and antitumor efficacy on a variety of murine tumor models [18,19,20,21]. Besides, since the study of SL7202 by Hoiseth and Stocker , the auxotrophic mutation of aroA-deficiency has been universally used to attenuate bacterial strains and generally considered safe .
Lipopolysaccharide (LPS) present in almost all Gram-negative bacteria is the major component of the outer leaflet of the outer membrane and a well-known inducer of the innate immune responses . It consists of the highly variable O-antigen polysaccharide, a short core oligosaccharide, and the anchor lipid A [25, 26]. Lipid A represents the most conserved portion of LPS. Extracellular lipid A is recognized by TLR4  while cytosolic caspase-11 will sense intracellular lipid A to promote the innate immune response [28,29,30]. The studies regarding structure-activity relationship of lipid A indicate that factors including the number, length, and symmetry of the fatty acid chains of lipid A govern its immunological activity [31, 32]. Among these factors, the total number of fatty acid chains is the most important one. Previous study indicated that the failure of clinical human trial using VNP20009 may be partially attributable to its generating penta-acylated lipid A, which is an antagonist for TLR4 and caspase-11. S633, the parent strain of S634 used in this study, was a S. typhimurium mutant with modifications of bacterial LPS (ΔpagL7 ΔpagP8 ΔlpxR9 ΔarnT2 ΔeptA3 ΔlpxT4), which harbors hexa-acylated lipid A that has been considered to possess the most potent immunomodulatory activity .
It has been widely recognized that angiogenesis, the formation of new capillaries from pre-existing vasculature, is critical for tumor growth and metastasis . Thus, inhibition of angiogenesis inside tumor tissue is another potential direction of cancer treatments. The anti-angiogenic strategy targets stable proliferative endothelial cells in the tumor vasculature rather than genetically unstable tumor cells and thereby reduces the probability of drug-resistance in the case of repeated dosing. Endostatin, a 20-kDa C-terminal fragment of type XVIII collagen, discovered in 1997 , is one of the most potential inhibitors of angiogenesis. Endostatin has been shown to bind to a variety of receptors on the surface of endothelial cells , blocking the proliferation and migration of endothelial cells  and inducing endothelial cell apoptosis on the cellular level . In detail, endostatin competitively inhibits the binding of vascular endothelial growth factor (VEGF) to endothelial cells, blocking VEGF-induced tyrosine phosphorylation of KDR/Flk-1, and thus affecting mitogenic activities of VEGF on endothelial cells . Endostatin competes with fibronectin and proangiogenic ligand to prevent binding to integrin α5β1, thereby disrupting the migration of endothelial cells . Endostatin specifically binds to nucleolin on the cell surface with high affinity and can be internalized and transported into cell nuclei of endothelial cells via nucleolin. In the nuclei, endostatin inhibits the phosphorylation of nucleolin, which is critical for the proliferation of endothelial cells . Animal studies also demonstrated that endostatin could suppress tumor growth by suppressing the neovascularization [41,42,43]. But unfortunately, phase II studies of recombinant endostatin in the United States showed significantly low anti-angiogenic potency [44, 45]. In 2005, Endostar, a novel recombinant human endostatin purified from Escherichia coli with an additional His-tag structure, was approved by the State Food and Drug Administration of China (SFDA) for the treatment of non-small cell lung cancer . In order to improve therapeutic antitumor efficacy of endostatin, different delivery vehicles were used, which include virus, plasmids, microspheres, and live attenuated bacterial vectors. Notably, endostatin gene therapy delivered by Salmonella [47, 48] and Bifidobacterium  could suppress tumor angiogenesis and tumor growth in murine tumor models.
In this study, we generated an aroA-deficient S. typhimurium strain derived from a previously constructed strain, which harbors additional mutations involved in modifications of its lipid A. Upon intraperitoneal infection in mice, our constructed aroA-deficient strain showed greatly attenuated virulence and preferential accumulation within tumor tissue. In order to test whether newly constructed strain is suitable for delivery of antitumor agents to elicit improved therapeutic efficacy, the strain was equipped with the prokaryotic expression plasmid of one well-known anti-angiogenic agent “endostatin”. The potential antitumor and adverse side effects of generated endostatin-expressing S. typhimurium aroA-auxotroph were evaluated in mouse models of CT26 colon carcinoma and B16F10 melanoma.
Materials and methods
The CT26 (mouse colon carcinoma) and B16F10 (mouse melanoma) cell lines were purchased from the cell bank of China Committee for Typical Culture Collection, China Academy of Sciences. Cells were grown in RPMI 1640 (CT26) or high-glucose DMEM (B16F10) supplemented with 10% FBS (fetal bovine serum) and 1% penicillin–streptomycin, and cultured at 37 °C in a humidified atmosphere of 5% CO2. Cells were counted using a Fuchs–Rosenthal counting chamber and seeded into 24-well plates (2 × 105 cells per well), six-well plates (5 × 105 cells per well), or 75 cm2 culture dishes (5 × 106 cells per dish). The culture medium was changed every 2 days until cells reached 80% confluence.
Bacterial strains and plasmids
The bacterial strains and plasmids used in this study are listed in Table 1. Suicide vector technology was used to generate precise deletion mutations. Two primer pairs, DaroA-1F/DaroA-1R and DaroA-2F/ DaroA-2R, were used to amplify approximately 400-bp DNA fragments upstream and downstream of the aroA gene. These two fragments were fused by PCR using primers DaroA-1F and DaroA-2R. The fused PCR product was cloned into pYA4278 to construct a new suicide plasmid (pSS262). The deletion of aroA was introduced into S. typhimurium strain S633 by allelic exchange using the suicide plasmid pSS262, thereby generating the aroA-deficient strain S634. Similarly, using the suicide plasmid pSS021, the gene “asd” of S634 was knocked out, generating the strain S636.
We utilized the plasmid-bacteria balanced-lethal system to ensure the stability of expression plasmids carried by Salmonella bacteria in vivo . Briefly, we first cloned the optimized complementary DNA of human endostatin into the open reading frame of the empty plasmid pYA3342, which is with the asd-complementary background, thereby constructing the endostatin-expressing plasmid pYA3342-endostatin (thereafter, “pES” for short). Then, by introducing the plasmid pES into the strain S636, the endostatin-expressing aroA-deficient strain was generated (S636/pES).
Preparation of Salmonella bacteria for in vitro and in vivo experiments
All bacterial strains were cultured on LB agar plates or in LB broth containing appropriate antibiotics. A single colony of strains was picked, inoculated into LB broth, and grown overnight in a shaking incubator (37 °C, 180 r.p.m.). The next day, the overnight culture was diluted 100-fold into fresh medium to grow to exponential phase with an optical density value at 600 nm (OD600) of 0.8–0.9. The bacterial cells were then collected by centrifugation (4000×g for 10 min), washed with phosphate buffer saline (PBS), quantified by spectrophotometry, and diluted in PBS to obtain the desired concentration of bacteria in an appropriate volume for the in vitro and in vivo experiments. The bacterial count was calculated as follows: 1.0 OD600 = 0.8 × 109 CFU.
Determinations of bacterial phenotypes
The phenotypes of bacterial strains including the lipid-A-modified strain S633, the aroA-deficient strain S634, and the wild-type S. typhimurium strain UK-1 were determined in vitro and assays were repeated at least twice. The growth rates of bacterial strains in LB medium in a shaking incubator (37 °C, 180 r.p.m.) were measured every 1 h, with an initial OD600 value of 0.03.
LPS profiles of Salmonella strains were examined by the method of Hitchcock and Brownusing cultures standardized based on the OD value at 600 nm .
Bacterial sensitivity to human complement was tested as described before . Briefly, 50 μL of bacterial suspensions containing about 2 × 107 CFU Salmonella were added to 50 μL of non-treated sera or heat-inactivated sera (HIS) (1:1) and the mixture was incubated for 30 min at 37 °C. Serial dilutions of the samples in PBS were plated on LB plates and incubated overnight at 37 °C. The CFU of alive bacteria were determined. Human blood was taken from volunteers and heat-inactivated serum was prepared for 2 h at 56 °C as a control. The rfc mutant S378, the LPS of which contains only one O-unit, was used as the positive control to prove the bactericidal activity conferred by complement components of non-treated sera.
Swarming motility was assessed on LB plates solidified with 0.3% agar (wt/vol) as described previously . In brief, 6 μL of bacterial suspension (~1 × 106 CFU) was spotted onto the middle of the semi-solid plates and subsequently incubated at 37 °C for 6 h. The diameters of the swarming colonies were measured. The plates containing 0.3% agar were allowed to dry at room temperature for 2 h and freshly grown bacteria were collected from LB agar plates followed by washing and dilution in PBS.
Invasion assays for determining bacterial infection in cancer cells were performed as described previously . The CT26 and B16F10 cells were seeded into the individual wells of 24-well plates 16 h prior to infection, to obtain a density of 5 × 105 cells per well. A total of 5 × 107 of CFU of bacterial strains prepared as described above were added to cancer cells to achieve the desired multiplicity of infection (100:1), and the mixture was incubated at 37 °C under 5% CO2 for 2 h. Following three times of washing with PBS, tumor cell line-optimal medium containing gentamycin (200 μg/mL; Sigma) was added and cells were incubated for additional 1 h to kill extracellular bacteria. Intracellular bacteria were then collected after washing and extraction with the lysis buffer (0.05% Triton X-100 diluted in PBS). The lysate was diluted in PBS and plated onto LB plates and the plates were incubation at 37 °C overnight before enumeration of the CFU of alive bacteria.
Indirect immunofluorescent assays were performed to determine the invasion of S636/pES in tumor cells and the expression of endostatin. Cells were seeded into six-well plates containing cover slips and cultured for 16–24 h before use. Bacterial strains were grown and prepared as described above. The collected bacteria were washed twice with PBS, diluted in serum-free medium, and added to cancer cells at a ratio of 100:1. After incubation of the mixture at 37 °C under 5% CO2 for 2 h, the cells were washed with PBS, and further cultured with gentamycin-containing medium for 1 h to kill extracellular bacteria. Then, the cells were washed gently with PBS, fixed in 4% paraformaldehyde, and stained with a rabbit anti-Salmonella polyclonal antibody (dilution, 1:500; catalog no. ab156656, Abcam) for 12 h at 4 °C. The goat anti‑rabbit immunoglobulin G polyclonal antibody conjugated with Alexa Fluor 488 (dilution, 1:100; catalog no. ab150077, Abcam) were added and the slides were incubated for 1 h at room temperature. For the staining of endostatin, mouse anti-endostatin monoclonal antibody and goat anti-mouse polyclonal antibody conjugated with Alexa Fluor 647 (dilution, 1:100; catalog no. ab150115, Abcam) were used. CytoPainter Phalloidin‑iFluor 555 Reagent (dilution, 1:1000; catalog no. ab176756, Abcam) and DAPI (dilution, 1:100; catalog no. R37606, Invitrogen) was applied to indicate cell boundaries and nuclei, respectively. The cells were observed under a confocal microscope.
Western blot analysis of the expression of endostatin by S636/pES
The expression of endostatin by S636/pES was analyzed by western blotting. Bacterial pellets were boiled for 5 min in loading buffer and subjected to 12% SDS-PAGE (sodium dodecyl sulfate-polyacrylamide gel electrophoresis) and then transferred to nitrocellulose membranes (Bio-Rad, China). The membranes were probed with a mouse anti-endostatin monoclonal antibody (dilution, 1:2000; catalog no. MA1-40230, Themofisher), followed by a horseradish peroxidase (HRP)-conjugated anti-mouse secondary antibody (dilution, 1:2000; catalog no. 1030-05, Southern Biotech). Immunoreactive proteins were detected using ECL western blotting substrate (catalog no. 32209, Themofisher) and visualized by an image reader machine.
Six-to seven-week-old female BALB/c and C57BL/6 mice (20–25 g) were purchased from Dashuo Biotechnology Co., Ltd. (Chengdu, China). Animal care, experiments, and killing were performed following the principles stated in the Guide for the Care and Use of Laboratory Animals. All efforts were made to minimize animal suffering during the experiments. The mice were acclimated for 7 days after arrival before experiments started.
For the establishment of tumor models in mice, CT26 (5 × 105) and B16F10 (2 × 105) cells were collected and suspended in 100 μL of PBS respectively prior to subcutaneous injection into the right back of each mouse. The CT26 and B16F10 cell lines were used to establish colon carcinoma models in BALB/c mice and melanoma models in C57BL/6 mice, respectively. When tumor volumes of mice reached about 200 mm3, tumor-bearing mice were grouped randomly and received scheduled treatments.
Determination of bacterial virulence in mice
The 50% lethal doses (LD50s) of Salmonella bacterial strains of S633, S634, and UK-1 were determined in BALB/c mice as previously described . Freshly grown bacteria were collected, washed, and diluted to the required inoculum density in PBS by adjusting the OD600 value of bacterial suspension. Groups of five mice each were infected intraperitoneally or orally with various doses of the wild-type S. typhimurium strain, or its derivatives used in this study (i.e., S633 and S634) in a volume of 100 or 20 μL, ranging from 1 × 102 to 1 × 109 CFU. The mice were observed for 4 weeks post infection and deaths were recorded daily. Experiments were repeated two times. The LD50s of different strains were calculated with the software SPSS.
Colonization of engineered Salmonella
In the colonization experiment, subcutaneous tumor models of colon carcinoma were established in BALB/c mice. When tumor volumes of mice reached about 200 mm3, 100 μL of bacterial suspension containing 5 × 106 CFU of S633 or S634 bacteria, or PBS was injected into each mice (day 0). Tumor-bearing mice were then killed at the indicated dpi (dpi 1, 3, 7, 14, 21), and tissues of spleen, liver and tumor were taken to determinate bacterial burden. Tissue samples were homogenized and diluted in PBS, and dilutions of 101 to 107 (depending on the tissues) were plated onto LB plates containing appropriate antibiotics to determine the number of viable bacteria. The bacterial burden within normal and tumor tissues was expressed as CFU per g tissue. Colonization experiments were repeated twice.
The possible therapeutic benefits of S636 carrying endostatin-expressing or empty plasmids were evaluated in CT26 colon carcinoma and B16F10 melanoma-bearing mice successively. Tumor-bearing mice were randomly divided into three groups (termed “S636/pES”, “S636/pEmpty”, and “PBS”, respectively) of eight mice each and 100 μL of bacterial suspension (5 × 106 CFU) or PBS was injected into each of mice intraperitoneally. The tumor volumes of mice of different groups were measured with a caliper every 2 days. Tumor volume (mm3) was estimated using the formula (L × W2 × π)/6, where L is the length and W is the width of the tumor in millimeters. Mice with CT26 tumors exceeding 1500 mm3 were scheduled for euthanasia. For the mouse melanoma model, tumor volumes and survival of mice were recorded until serious illness appeared. Meanwhile, body weights of tumor-bearing mice were measured as an indicator for general health status. Animal experiments for evaluation of therapeutic antitumor benefits of our construct were performed twice and the results were taken together for analysis.
Histological and immunohistochemical studies
During animal experiments, four CT26 tumor-bearing mice of each group were killed at 14 dpi and normal and tumor tissues were taken for histological and immunohistochemical studies. Tissue samples were fixed immediately with 4% paraformaldehyde. Standard hematoxylin and eosin staining of paraffin-embedded tissues was performed for pathological examination. For the immunohistochemical staining, heat-induced antigen retrieval was performed at temperature >95 °C in 10 mM sodium citrate buffer (pH 6.0) and followed by cooling down at room temperature. Endogenous peroxidase activity was quenched by incubating the sections with 3% hydrogen peroxide for 10 min. After that, the sections were blocked with blocking buffer (0.1% Triton X-100, 3% BSA, and 2% normal donkey serum) for 1 h, and incubated with following primary antibodies: anti-Salmonella (Abcam), anti-endostatin (Themofisher), anti-CD34 (catalog no. ab185732, Abcam), and anti-Caspase-3 (catalog no. ab13847, Abcam). After washing with PBS, HRP-conjugated secondary antibodies were added. Then, the sections were stained with a freshly prepared 3,3′-diaminobenzidine (DAB) chromogen and counterstained with hematoxylin. Photos were taken in five random fields of each sample under a digital microscope by using a bright-field illumination. For immunohistochemical sections, the integrated optical density of positive staining was tested by the software Image-Pro Plus 6.0.
Numerical data were expressed as means ± SEM. One-way analysis of variance followed by Dunnett’s or Bonferroni’s multiple comparison test was used to evaluate bacterial motility, invasion activities, colonization profiles, therapeutic benefits, and clinical chemistry parameters for multiple comparisons among groups. The Kaplan–Meier method was used for survival, and differences were analyzed by the log-rank test. All analyses were performed using GraphPad Prism 5.0. P < 0.05 was considered as statistically significant (*, #, or †); P < 0.01 as very significant (**, ##, or ††); and P < 0.001 as extremely significant (***, ###, or †††).
Construction of the aroA-deficient S. typhimurium strain and evaluation of its phenotypes
In this study, we aimed at generating an attenuated S. typhimurium strain with high tumor-targeting for cancer therapy. We used previously constructed strain S633 (ΔpagL7 ΔpagP8 ΔlpxR9 ΔarnT2 ΔeptA3 ΔlpxT4) (Table 1), of which the LPS was characterized in homogeneous hexa-acylated lipid A. For further attenuation, we introduced the deletion of aroA into S633 and thereby generated the strain S634, which was auxotrophic for aromatic amino acids and unable to grow in minimal salt media (data not shown). We evaluated the virulence of bacterial strains by determining their intraperitoneal LD50s in BALB/c mice, and revealed that the virulence of S633, whose LD50 is 5.3 × 103, was reduced by at least 20-fold compared to the wild-type strain UK-1, whose LD50 is <3 × 102, and S634, whose LD50 is 1.2 × 107, was >2000-fold attenuated when compared to its parent strain S633. The oral LD50s of S633 and S634 were both larger than 109, indicating these two mutants are avirulent compared to the wild-type strain UK-1 (Table 2).
Bacterial growth curves were evaluated in LB broth, and it was shown that S634 grew more slowly than its parent strain S633 and the wild-type strain UK-1 (Supplementary Fig. S1). All cultures grew to stationary phase after 8 h of shaking culture (37 °C, 180 r.p.m.), despite of slight differences among the OD values of different cultures in the platform period. The deletion of aroA did not seem to affect the LPS phenotype (Fig. 1a) or the sensitivity to serum complement (Fig. 1b). S634 showed decreased motility on semi-solid agar plates by about one-third compared to its parent strain, as determined by the diameter of swarming motility (Fig. 1c). To test the interaction between Salmonella bacteria and the tumor, cancer cells of the CT26 and B16F10 cell lines were incubated with S633, S634, and UK-1 in vitro. It was shown that the ability of the aroA-deficient strain S634 to invade the CT26 and B16F10 cells was decreased, compared to S633 and UK-1 (Fig. 1d). In addition, the declined growth rate and the decreased ability to invade cancer cells were at least partly recovered by introduction of aroA-complementary plasmids into the aroA-deficient strain (data not shown).
Engineered S. typhimurium specifically colonizes tumor tissue in vivo
Colonization experiments were conducted to determine the tumor-targeting ability of the aroA-auxotrophic strain S634 in subcutaneous tumor models of colon carcinoma. On the 1st, 3rd, 7th, 14th, and 21st day after tumor-bearing mice received intraperitoneal infection of bacterial strains, the samples of tumor, liver, and spleen were taken and bacterial burdens were determined. At dpi 1, between 1 × 104 and 1 × 106 CFU/g of S634 were observed in normal tissues (liver and spleen) and about 1 × 108 CFU/g in tumor tissue. After that, bacterial burdens in the liver and spleen tissues were decreased gradually (Fig. 2a, b), but remained relatively stable in tumor tissue (Fig. 2c). The strain S634 preferentially accumulated in tumor tissue compared to normal tissues, with tumor-to-normal tissue (liver) ratios ranging from 1500:1 to over 100,000:1 throughout the experiment (Fig. 2d). However, the parent strain S633 colonized tumor tissue with low specificity (tumor-to-liver ratio ≈ 10:1) before the mice succumbed to its infection (data not shown).
S636/pES can express endostatin and invade cancer cells in vitro
In this study, we attempted to test whether the strain S634 is suitable to deliver therapeutic antitumor agent “endostatin”, one of the most potent inhibitors of angiogenesis. We constructed the endostatin-expressing plasmid named “pES” (asd+) and introduced it into S636, the asd− derivative of tumor-targeting S634, thereby generating S636/pES, in which plasmids and bacterial strains form a balanced-lethal system. The expression of endostatin in Salmonella bacteria S636 was confirmed in vitro by western blotting (Fig. 3a). The immunofluorescence assay also demonstrated that S636/pES could invade both CT26 and B16F10 tumor cells (Fig. 3b) and express endostatin (Fig. 3c). Notably, to avoid the loss of expression plasmids carried by Salmonella bacteria in vivo, we utilized the plasmid-bacteria balanced-lethal system in this study , showing that the plasmid stability was nearly 99% in vitro and in vivo (data not shown).
Combination of S. typhimurium and endostatin inhibits tumor growth and improves the survival of tumor-bearing mice
Animal experiments were performed in CT26 colon carcinoma- and B16F10 melanoma-bearing mice to study the potential therapeutic benefits elicited by the our Salmonella mutant and endostatin. When tumor sizes reached about 200 mm3, tumor-bearing mice were grouped randomly and received 5 × 106 CFU of S636/pES, S636/pEmpty, or 100 μL of PBS intraperitoneally. In CT26 tumor models, S636/pES showed stronger suppressive effects on tumor growth compared with S636/pEmpty at dpi 12 (P < 0.05, S636/pES versus S636/pEmpty) and 14 (P < 0.01) (Fig. 4a). At 12 dpi, the mean tumor volume of the PBS group was about 1600 mm3, whereas those of the S636/pEmpty and S636/pES groups were 1100 mm3 (P < 0.01, versus the PBS group) and 600 mm3 (P < 0.001, versus the PBS group), respectively (Fig. 4a). Then, we established aggressive B16F10 melanoma models for further validation. Similarly, mice of the S636/pES group showed significantly superior therapeutic benefits compared to mice treated with S636/pEmpty at dpi 12, 14, 16, and 18 (P < 0.05, or P < 0.01, S636/pES versus S636/pEmpty) (Fig. 4b). After 2 weeks (at dpi 14), the mean tumor volume of the PBS group was about 2800 mm3, whereas those of the S636/pEmpty and S636/pES groups were about 1750 mm3 (P < 0.01, versus the PBS group) and 850 mm3 (P < 0.001, versus the PBS group), respectively (Fig. 4b). Moreover, the survival of B16F10 tumor-bearing mice was obviously prolonged by treatments with S636/pES and S636/pEmpty (P < 0.01 and P < 0.001, respectively, versus the PBS group) as shown in Fig. 4c. Mice treated with S636/pES survived for the longest period of time, which is nearly 31 days, followed by S636/pEmpty-treated mice for 24 days (P < 0.05, S636/pES versus S636/pEmpty), whereas PBS-treated tumor-bearing mice were alive only for 15 days on the average (Fig. 4c). Taken together, engineered S. typhimurium carrying endostatin expression plasmids (i.e., S636/pES) exerted significant antitumor effects on both mouse tumor models. Furthermore, S636 carrying empty plasmids (S636/pEmpty) could also obviously suppress tumor growth and prolong the survival of melanoma-bearing mice, though its antitumor efficacy was not quite as powerful as that of S636/pES.
S636/pES inhibits angiogenesis and increases apoptosis within tumor tissues
Tumor samples were taken for immunohistochemical studies 2 weeks after tumor-bearing mice received treatments. Corresponding to the colonization experiments, S636 carrying either empty or endostatin-expressing plasmids could colonize the tumor after systemic infection, which was indicated by immunohistochemical staining for Salmonella (Fig. 5a, b). Besides, the expression of endostatin within tumor tissues was detected when mice were treated with S636/pES (Fig. 5a, b).
Since suppression of tumor growth is always accompanied by activation of apoptosis within tumor tissue, we analyzed the apoptosis by immunohistochemical staining for a cleaved form of caspase-3 (Fig. 5a), which is a key enzyme activated in the apoptosis pathway . Compared with the PBS group, the S636/pES and S636/pEmpty groups showed significantly elevated apoptosis in tumor tissues, which was indicated by the increased level of activated caspase-3 (Fig. 5c). To determine if the expression of endostatin within tumor tissue elicited anti-angiogenic effects, we performed the staining for CD34, which is mainly expressed on small vessel endothelial cells and tumors of epithelial origin . The expression level of CD34 that indicated the microvessel density was significantly decreased within tumor tissues of the S636/pES group compared to the PBS and S636/pEmpty groups (Fig. 5c). Thus, it was plausible that the accumulation of Salmonella bacteria and the expression of endostatin within tumor tissue elicited effects of apoptosis induction and anti-angiogenesis respectively, which at least partly explained the superior therapeutic benefits elicited by S636/pES than S636/pEmpty.
Adverse side effects of engineered S. typhimurium
During animal experiments, we also monitored the general health of tumor-bearing mice, as indicated by the body-weight change of mice. The body weights of mice were decreased by about 10% within the initial 2 days after treatment with S636/pES and S636/pEmpty while gradually recovered later, which took about 1 week (Fig. 6). It seems plausible to speculate that intraperitoneal infection of the aroA-deficient Salmonella bacteria S636 did elicit severe toxicity to the mice, while the administration dose adopted (5 × 106 CFU) in this study was tolerable for those tumor-bearing mice. Meanwhile, we also took samples of blood and normal tissues including liver and spleen from tumor-bearing mice at 14 dpi for studies. Pathologic examination by hematoxylin and eosin staining showed that, treatment with S636/pES and S636/pEmpty caused pathological changes to some extent on normal tissues, which indicated by the swelling and degeneration of hepatocytes in the liver and the inflammation in the spleen (Fig. 7).
Bacteria-mediated tumor therapies aim at overcoming some of the shortages or limitations of conventional treatments, such as low tumor-targeting and damage to normal tissues or cells. S. typhimurium, as facultative anaerobic bacteria, are good candidates of therapeutic antitumor agents and have been commonly used in bacteria-mediated tumor therapy. As S. typhimurium with intrinsically pathogenic properties may cause serious toxicity to the body especially after systemic infection, it is necessary to attenuate such bacteria adequately before clinical studies. One of the strategies for bacterial attenuation is via auxotrophic mutations. aroA gene is responsible for the synthesis of aromatic amino acids, which are not freely available in the mammalian host. Thus, deletion of aroA results in Salmonella bacteria auxotrophic, leading to increased immunogenicity and adjuvant potential, and has been commonly used to attenuate such bacteria . In this study, we constructed a fine-defined aroA-deficient S. typhimurium strain, accompanied with additional mutations including ΔpagL7 ΔpagP8 ΔlpxR9 ΔarnT2 ΔeptA3 ΔlpxT4, which involved in modifications of lipid A moiety of bacterial LPS. Deletion of arnT, eptA, and lpxT genes will remove the ability of S. typhimurium to catalyze the addition of specific moieties (l-Ara4N, pEtN, and the second phosphate group at the 1-position), resulting in maximum exposure of two phosphate groups in lipid A to facilitate its interaction with TLR4 receptor and caspase-11 [refs. 56–58]. Moreover, deletion of pagL, pagP, and lpxR genes enable Salmonella to produce hexa-acylated lipid A regardless of growing in media in vitro or in the host organ in vivo, where adverse environment such as low pH, or Mg2+, will activate these genes to alter the number and the symmetry of acylated chains in lipid A [59, 60]. These six gene deletions will ensure the lipid A integrity and maintain its most potent stimulatory structure in live Salmonella bacteria during its invasion, colonization, and persistence in the host gut-associated lymphoid tissues and in tumor tissues.
Thus, a new live Salmonella strain was constructed for cancer therapy and as a vector to deliver antitumor agents. As expected, the virulence of the aroA-deficient strain S634 in mice was significantly reduced, as shown by the LD50 value of 1.2 × 107, over 2000-fold increase of LD50 compared with that of the strain S633 upon intraperitoneal infection, and S634 was avirulent by oral route due to mutations of aroA and arnT and eptA (Table 2). Besides the significant reduced virulence, a good Salmonella candidate for tumor therapy should possess other properties such as ability to migrate toward the tumor, penetrate into tumor tissue, and colonize the tumor in large amounts. An in vitro model of continuously perfused tumor tissues indicated that motility and bacterial chemotoxis mediated by chemoreceptors were critical for directing Salmonella to migrate toward the different microenvironment regions of tumor cells, and Salmonella with higher motility have enhanced penetration and colonization within tumor tissues [61, 62]. In our study, we revealed that the motility of S634 was decreased by about one-third compared to that of its parent strain (Fig. 1c), consistent with the recent observation . But this motility deficiency in our construct did not compromise its ability to accumulate within tumor tissue in mouse model because S634 still showed the good ability to invade tumor cells in vitro, and S634 also preferentially colonized tumor tissue relative to normal tissues in vivo with tumor-to-liver ratios ranging from 1500:1 to over 100,000:1 (Figs. 2d and 3), indicating that the motility may not play an essential role in mediating its migration toward tumor tissues. Recent studies using mouse tumor models showed that bacterial colonization and migration process within the tumor in mice was a passive mechanism dramatically influenced by inflammatory cytokines and bacterial metabolism of Salmonella [63,64,65]. This could explain that most of Salmonella mutants in random mutants screen showing the great ability to preferentially proliferate throughout tumor tissues were tied to the auxotrophic mutations mostly associated with requirement of aromatic amino acid, leucine and arginine synthesis [21, 66]. However, whether the motility mediated by flagellum is vital for directing Salmonella to migrate toward necrotic hypoxic/anoxic environment of tumor cells in animal model is still controversial, but it is convinced that over-expression of secreted heterologous flagellin would greatly benefit the therapeutic antitumor efficacy conferred by Salmonella, as demonstrated by recent work, in which, tumor-targeting Salmonella engineered to overexpress Vibrio vulnificus FlaB effectively suppressed tumor growth and metastasis and prolonged survival in murine colon and melanoma models and these therapeutic effects were mediated by TLR4 signaling and augmented this tumor-suppressive host reactions by TLR5 signaling [67, 68].
Tumor-targeted Salmonella is often administered by intravenous or intraperitoneal route to induce systemic infection in tumor models, and reaching up to ratios of more than 1000:1 in cancerous tissue compared to healthy tissues (Fig. 2). Many genes, which are critical for establishing an oral infection, such as phoP, invG sseD, and ssrB, are not essential for intravenous infection to target, invade, and colonize the tumor cells , and in our study, deletion mutations of arnT and eptA genes were also proved to be required for oral infection, but not for intraperitoneal administration  (Table 2). However, an entire cell envelope, especially LPS, is important for Salmonella combatting against bactericidal activity of human complement and neutrophil clearing before attenuated Salmonella migrate toward and colonize tumor tissues [70, 71], because deeply rough Salmonella mutants showed low virulence but exhibited weak intrinsic antitumor effects [52, 72]. Our results indicated that the strain S634 exhibited the similar complete LPS pattern as the wild-type S. typhimurium UK-1 (Fig. 1a), explaining the similar sensitivity of the mutants to human complement (Fig. 1b) and successful invasion of S634 within tumor cells (Figs. 1d and 3).
While S634 preferentially accumulated within tumor tissue to a significant high number, this mutant still persisted in healthy tissues including the spleen and liver with comparable number for over observed 21 days (Fig. 2), indicating that only one aroA mutation is not sufficient for restricting Salmonella to replicate in the healthy tissues, other mutations are required to further decrease bacterial survival in healthy tissues for the clinical safety. The mutations associated with leucine and arginine biosynthesis are of particularly interest because these Salmonella auxotrophs were selected from random mutants by the ability to grow in successive tumor xenografts, and were demonstrated to receive sufficient amounts of these amino acids from the tumor environment but lose the ability to persist in the normal tissue environment [21, 66, 73]. The mutations such as htrA, SPI-2, and STM3120, are also potential candidates to reduce fitness in normal tissues and retain their fitness in tumors . The novel strategy to achieve rapid elimination of bacteria from normal tissues while retaining adequate antitumor ability was developed by placing an essential gene under a hypoxia conditioned promoter , which can be considered for reducing bacterial load in normal tissues and increasing tumor-targeting specificity in the future work. Another interesting approach to reduce colonization in healthy organs while retaining advantages of systemic application is by an intra-tumoral route of infection, allowing the extensive dose of bacteria to bypass the killing by innate immunity encountered by systemic infection and directly target the tumor cells [76, 77].
Systemic administration of attenuated S. typhimurium deprives cancer cells of nutrients by competition and activates antitumor immunity, resulting in tumor regression, tumor shrinkage, and even complete tumor eradication [78, 79]. In this study, we observed therapeutic efficacy against CT26 and B16F10 tumors after intraperitoneal administration of S636/Empty and S636/pES in mouse models (Fig. 4), including inhibition of tumor growth and prolonged mouse survival, which was consistent with other studies on the antitumor effects of S. typhimurium and endostatin [13, 48]. But this efficacy was also accompanied by severe adverse side effects including body-weight loss and enlarged spleen of the mice, and pro-inflammatory immune responses, although the mice were recovered to normal condition later (Figs. 6 and 7). Abundant Salmonella persistence in the healthy organs for long time and the presence of the most potent stimulatory inflammatory lipid A in this Salmonella mutant are two major reasons to result in the severe side effects. Initial inflammatory response including production of tumor necrosis factor-α (TNF-α), interleukin (IL)-1β, and other cytokines induced upon systemic administration of the bacteria is a prerequisite for bacterial targeting and entrapment in tumors [7, 63, 64], subsequent accumulation and proliferation of Salmonella within tumors will elicit infiltration of immune cells, such as neutrophils and macrophages, to the local site of tumors tissues, and secrete pro-inflammatory cytokines such as TNF-α and IL-1β, enabling immune privileged tumor environments shift from immune suppressive to enhanced immunogenic [77, 80,81,82]. The inflammatory responses in these two steps appeared to be mainly mediated through TLR4 signaling activated by the lipid A of Salmonella, the evidences to support this notion are from the facts that macrophage and neutrophil infiltration of bacteria-colonized tumors in TLR4−/− mice was lower than in the wild-type mice, and administering lipid A concurrently with Salmonella significantly increased bacterial accumulation in both necrotic and viable tumor tissues and enhanced the antitumor capability [67, 83], and this could also partially explain the ineffectiveness of Salmonella VNP20009 in phase I clinical trials involving human cancer patients because of msbB deletion resulting in penta-acylated lipid A in Salmonella, an antagonist for human TLR4 while it was demonstrated its excellent antitumor capability in tumor-bearing mice model [6, 17, 84, 85]. Our results indicated that the most potent TLR4 stimulator of a hexa-acylated lipid A achieved by a deletion combination of ΔpagL7 ΔpagP8 ΔlpxR9 ΔarnT2 ΔeptA3 ΔlpxT4 in our constructed strains S633, S634, and S636, may not be an optimal structure used for bacteria-mediated cancer therapy because of its highest endotoxic activity, shown in Fig. 7. Considering that activation of TLR4 on the tumor cells by bacterial infection results in tumor proliferation and progression , the lipid A structure with decreased activity to stimulate TLR4 signaling, such as monophosphoryl lipid A, which has been used in human clinic as vaccine adjuvants to enhance vaccine efficacy, should be investigated in the future construct as bacteria-mediated cancer therapy [69, 87,88,89].
Besides attenuated Salmonella itself as a therapeutic antitumor agent, it can express and release cytotoxic proteins, immunoregulatory proteins, and apoptosis-inducing factors, and some essential enzymes for the conversion of nontoxic prodrugs into cytotoxic drugs [78, 90]. In this study, we also tested the ability of S636, the asd− derivative of S634 to deliver the therapeutic antitumor protein. It has been widely recognized that angiogenesis, the formation of new capillaries from pre-existing vasculature, is one key process for tumor growth and metastasis. Thus, inhibition of angiogenesis inside tumor tissue is another promising strategy for cancer treatments. Endostatin, a 20-kDa C-terminal fragment of type XVIII collagen, is one of the most potent inhibitors of angiogenesis. Endostatin has been shown to block the proliferation and migration of endothelial cells  and induce endothelial cell apoptosis , thereby producing the anti-angiogenic activity. To employ anti-angiogenic therapy by utilizing the bacterial vector of tumor-targeting Salmonella, we introduced prokaryotic expression plasmids of endostatin (pES) into S636 and thereby generated endostatin-expressing Salmonella strain named S636/pES. The potential antitumor effects of S636/pES were evaluated in mouse CT26 colon carcinoma and B16F10 melanoma subcutaneous tumor models successively. As shown in Fig. 4a, S636/pES was able to retard the growth of CT26 tumors. Moreover, upon infection with S636/pES, the growth of aggressive B16F10 melanoma was also significantly delayed (Fig. 4b) and the survival of melanoma-bearing mice was significantly prolonged (Fig. 4c).
Previous studies have showed that endostatin can induce endothelial cell apoptosis through binding to a variety of receptors on the surface of endothelial cells. As shown in immunohistochemical studies for endostatin and CD34, when tumor-bearing mice were treated with S636/pES, the expression of endostatin within tumor tissue was accompanied by the decreased level of CD34 (compared to the S636/pEmpty and PBS groups), indicating suppression of angiogenesis (Fig. 5). Moreover, the expression level of CD34 in tumor tissues of the S636/pEmpty group was also significantly declined compared to the PBS group (Fig. 5c). It has been previously shown that after colonization, Salmonella would destroy local tumor blood vessels . Thus, the anti-angiogenic effect of endostatin delivered by S636 probably accounted for the superior therapeutic benefits of S636/pES in tumor-bearing mice compared to S636 carrying empty plasmids.
In summary, in this study, an attenuated S. typhimurium mutant carrying metabolic deficiency and hexa-acylated lipid A was constructed without any genetic scars in the Salmonella chromosome in a modular manner, and in vivo tests demonstrated the antitumor efficacy conferred by live Salmonella itself and its ability to deliver the anti-angiogenic agent endostatin. Next, we will optimize this construct to maximize its capability to colonize and accumulate within tumor tissues while decreasing its adverse side effects caused by amounts of Salmonella bacteria in healthy tissues and endotoxic activity, and we will also optimize the delivery plasmid, enable it to express the targeted proteins under the controllable condition to maximize its therapeutic functions while reducing its toxic side effects to healthy cells.
All applicable international, national, and/or institutional guidelines for the care and use of animals were followed. The animal care protocol was approved by Sichuan Agricultural University. All efforts were made to minimize animal suffering during the experiments.
This work was funded by National Natural Science Foundation of China (grant numbers 31570928 and 31472179).
Q.K., K.L., and Q.L. initiated the research. K.L. and Q.L. led in vitro and in vivo experimental design, data acquisition and analysis, and manuscript preparation together. P.L., Y.H., X.B., and Y.T. aided in data acquisition. Q.K. participated in experimental design, data analysis, and manuscript preparation.