Direct injection of naked DNA has been shown to allow transgene expression in muscle 1. However, for most gene therapy applications, the need for an efficient and relevant vector for gene transfer is now widely acknowledged. Classically, these vectors are classified into non-viral and viral vectors. The non-viral vectors are usually chemically defined compounds such as liposomes that complex the genetic material and mediate gene transfer 2. Viral vectors, replicating or not, are recombinant viruses in which all or part of the viral genome has been replaced by the expression cassette encoding the therapeutic transgene. Gene transfer can also occur from bacteria to a very broad range of recipients that include yeast 3 and plants 4 and several laboratories have reported a relatively high frequency of functional gene transfer from bacteria to mammalian cells 5–11. From these experimental studies, a third type of vector has been proposed: bacterial gene transfer vectors. In this review, we will describe the mechanism of gene transfer and provide examples of in vitro and in vivo applications of this technology.
Mechanism of gene transfer by bacterial vectors
The first limiting step of gene transfer lies in the entry of the genetic material into mammalian cells. With bacterial vectors, this is achieved through entry of the entire bacterium into the target cells. When professional phagocytic cells such as macrophages or dendritic cells are targeted, this entry is likely to happen through phagocytosis but, in the case of non-phagocytic cells, there are two major strategies for bacteria to gain entry into a eukaryotic cell 12. For certain genera such as Salmonella or Shigella, contact between the bacteria and the host results in the secretion by the bacteria of a set of invasion proteins that triggers intracellular signalling events. This leads to cytoskeletal rearrangement, membrane ruffling, and bacterial uptake by pinocytosis. For other genera such as Yersinia or Listeria, binding of a single bacterial protein to a particular ligand on the host cell surface is necessary and sufficient to trigger entry by a zipper-like mechanism.
Once inside the cells, the bacteria are localized in the phagosomal vacuoles and are targeted for degradation. Therefore, in order for the plasmid to be delivered to the nucleus, escape from the vacuolar compartment must be achieved. In the case of bacterial delivery vectors that remain in the phagosome such as Salmonella typhimurium, the mechanism of plasmid DNA escape in the cytosol remains unclear 11. A recent study suggests that part of the eukaryotic gene expression reported with this type of vector may in fact be due to a substantial activity of some eukaryotic promoters such as the immediate-early promoter of cytomegalovirus or the Rous sarcoma virus promoter in the bacteria, leading to protein instead of gene transfer 13. By contrast, the process of phagosomal escape is particularly well understood for vectors based on bacteria replicating in the cytosol (Listeria monocytogenes) 8, 14. After internalization, these Gram-positive bacteria are able to escape into the cytosol thanks to the action of the pore-forming toxin Listeriolysin O (LLO). Once in the cytosol, wild-type L. monocytogenes will replicate 15. An attenuated L. monocytogenes strain has been engineered to undergo self-destruction in the cell cytosol by production of a phage lysine under the control of the promoter of actA, which is preferentially activated when the bacteria are in the cytosol 8. This bacterial vector was able to deliver a plasmid carrying the gene for a chloramphenicol acetyltransferase or part of the chicken ovalbumin gene (ova) under the control of the CMV promoter into the P388D macrophage cell line. Functional gene transfer was in the range of 0.2% and both chloramphenicol acetyltransferase activity and presentation of ova epitopes were detected. Plasmid DNA recovered from macrophage clones cultured in selective medium for more than 12 weeks was found to be integrated into the macrophage genome at a frequency of 10−7. Another bacterial vector taking advantage of the pore-forming activity of the LLO protein is the invasive E. coli9, 10. In addition to the hly locus from L. monocytogenes encoding the LLO protein, the inv locus encoding the invasin protein from Yersinia pseudotuberculosis has been inserted into this bacterium 9. Invasin binds to β1-integrin expressed at the surface of mammalian cells and this binding is necessary and sufficient for entry of the whole bacterium into the mammalian cell 16. Therefore, invasin expression restricts the tropism of these bacteria to non-phagocytic cells expressing β1-integrin 17, while invasin-negative recombinant E. coli can be used to target professional phagocytic cells 18, 19. Figure 1 shows Caco-2 cells invaded by these invasive E. coli. In these bacteria, LLO is engineered to be intra-bacterial as opposed to secreted in wild-type L. monocytogenes9, 10, 17, 20. Therefore, once in the cell, the bacterium is degraded in the lysosome. This process releases LLO in the lysosomal compartment and the bacterial content that includes the plasmid to be delivered can escape in the cytosol through the pores formed by LLO.
Finally, as there is no active mechanism of transport of the plasmid from the cytosol to the nucleus, a likely way for the plasmid to gain access to the nucleus is disruption of the plasma membrane during mitosis.
Applications of bacterial vectors: immunotherapy/vaccination
Delivery of plasmid-encoded antigens under the transcriptional control of eukaryotic promoters by live, attenuated bacteria is a strategy that has been used successfully in vaccination. The methodology relies on the delivery of the plasmid in which a eukaryotic promoter drives the expression of the antigen into macrophages and/or dendritic cells. At the same time, pathogen-associated molecular patterns (PAMPs) present in the bacteria will stimulate these antigen-presenting cells and trigger an innate immune response in the form of the production of reactive oxygen, pro-inflammatory cytokines, and nitrogen species, as well as up-regulation of co-stimulatory molecules. These responses promote the maturation and migration of dendritic cells to secondary lymph nodes 21 and, in this way, PAMPs amplify the immune response against the antigen and act as adjuvants. PAMPs exert their action through binding to receptors of the Toll-like receptor (TLR) family 21. PAMPs include compounds such as lipopolysaccharide (LPS) 22, 23, bacterial DNA containing unmethylated cytosine-phosphate-guanine (CpG) dinucleotides 24, flagellin 25 or bacterial lipoproteins 26.
Vaccination against bacterial antigens
Following the demonstration that strains of Shigella flexneri mutated in a gene essential for cell wall synthesis were able to deliver plasmids in vitro resulting in β-galactosidase expression in mammalian cells 5, nasal inoculation of these bacteria in mice led to the induction of β-galactosidase-specific humoral and cellular responses 27.
Oral administration in mice of an attenuated Yersinia enterocolitica carrying a eukaryotic expression plasmid encoding Brucella genes elicited humoral and cellular responses. This immune response, in conjunction with potential cross-reacting antibodies against LPS of both Yersinia and Brucella, induced protection against a Brucella challenge 28.
Another set of studies exploited S. typhimurium as a gene delivery vector 6. When β-galactosidase was used as a transgene, specific cytotoxic T-lymphocytes (CTLs) and T-helper (Th) cells, mainly of the Th1 type, as well as specific antibodies could be detected after a single oral vaccination 6. A very similar immune response was elicited when two virulence factors of L. monocytogenes (LLO and ActA) were encoded in S. typhimurium vectors 29. Protection of mice against a lethal challenge with L. monocytogenes was observed 6, 30. The route of administration appears to affect these responses. β-Galactosidase as well as a fusion antigen of the Pseudomonas aeroginosa outer membrane protein with the fimbriae was administered orally or nasally. Multiple nasal administrations were necessary to obtain a T-cell response in the spleen compared with a single oral one 11, 30. Specific IgGs were observed in the gut, saliva, and serum after oral administration, whereas nasal administration gave rise to antibodies in the lungs, saliva, and serum, with hardly any antigen-specific IgA detected 11, 30. Independently, partial protective responses against Chlamydia were obtained in the lungs of mice after oral administration of Salmonella encoding the major outer membrane protein of Chlamydia trachomatis11.
Salmonella-mediated vaccination has essentially been achieved with a strain of S. typhimurium auxotrophic for aromatic amino acids, strain SL7207 31, but intranasal vaccination with S. typhi was also reported 32. In this study, the guaBA-attenuated S. typhi strain (CVD915) defective in guanine biosynthesis carried a eukaryotic expression plasmid encoding tetanus toxin fragment C. This vaccination induced antigen-specific antibody responses that were higher than those induced by a similar bacteria carrying a prokaryotic expression plasmid in which the expression of the same antigen was inducible in vivo32.
Vaccination against viral antigens
Viral infections have also been targeted through DNA vaccination mediated by bacteria. Oral vaccination with S. typhimurium encoding a hepatitis B virus surface antigen (HbsAg) proved successful at inducing CTLs in Balb/c mice 33 and a single administration of these bacteria to transgenic mice expressing HbsAg in the liver led to loss of expression of the antigen in hepatocytes 34. This loss of HbsAg was accompanied by hepatic flare that subsided after 3 weeks, while the suppression of HbsAg expression continued in the absence of overt liver pathology for the remaining duration of the experiment (12 weeks). This single administration of the recombinant bacteria induced CTLs, Th1 T cells, and HbsAg-specific IgG2 subclass antibodies as further proof that immune tolerance against the viral antigen had been broken. A similar S. typhimurium was also used in bacteria-mediated DNA vaccination against the non-structural region 3 (NS3) of hepatitis C virus in HLA-A2.1 transgenic mice 35. A single oral administration induced A2.1-restricted CTLs, INF-γ-producing T cells, and resistance against a challenge with NS3-expressing vaccinia virus 35. Recombinant S. typhimurium administered orally was also reported to induce an immune response against herpes simplex virus 2 36 and human papillomavirus 16 37, 38.
Vaccination against HIV was attempted using attenuated strains of Shigella and Salmonella carrying eukaryotic expression plasmids encoding HIV env and gp120, respectively 39, 40. Oral administration of recombinant Salmonella as well as intranasal vaccination with recombinant Shigella led to the activation of antigen-specific CD8 T-cell responses. In addition, Shigella HIV gp120 DNA vaccination conferred some protection against challenge with a vaccinia virus expressing env39. A study comparing the ability of the Shigella strain to different Salmonella strains suggested that the induction of antigen-specific CD8 T-cell responses was superior with Shigella41. In the same study, intranasal application of Shigella was shown to induce a similar cellular response to intramuscular DNA vaccination but the mucosal, antigen-specific IgA response induced by the bacterial vector was superior 41. Finally, an attenuated Shigella defective in LPS-O-antigen synthesis and carrying an HIV gag DNA vaccine was used as a boost, after priming by intramuscular DNA injection 42. Gag-specific T-cell responses were detected in the spleen and lungs and the prime/boost strategy resulted in a dramatically increased T-cell response in the lungs 42.
Other DNA vaccinations mediated by bacteria include vaccination against measles virus 43 and influenza 44 with a recombinant strain of S. flexneri and vaccination against pseudo-rabies virus using either non-pathogenic E. coli administered intramuscularly 45 or a swine-adapted strain of Salmonella choleraesuis46.
Salmonella strains have essentially been used to deliver DNA for therapeutic applications in oncology. In the first instance, ‘model’ tumour antigens such as β-galactosidase or human gp100 (hgp100) were encoded in eukaryotic expression vectors carried by strains of Salmonella7, 47, 48. Oral administration of these bacterial strains protected the mice against challenges with fibrosarcoma 7, renal carcinoma 47, and melanoma 48 cells expressing the relevant model antigen.
The next series of experiments assessed the efficacy of oral administration of Salmonella DNA carrier against autologous tumour antigens also expressed in normal tissues. Successful protection was observed with transgenes including the murine gp100 (mgp100) fused to the invariant chain 49, epitopes of mgp100 and TRP2 fused to ubiquitin 50, a minigene encoding epitopes of the tyrosine hydroxylase (TH) fused to ubiquitin 51, the complete TH coding sequence linked to a virally derived post-transcriptional regulatory element 52, and human carcinoembryonic antigen (hCEA) in an hCEA mouse transgenic model 53. As all of these antigens were self-antigens, these observations strongly suggest that Salmonella-mediated DNA vaccination can break immunological tolerance. In many cases, the protection observed was strongly improved by co-administration of interleukin (IL)-2 49, 52, 53.
The vaccination against a tumour antigen can be amplified by co-delivery of another plasmid encoding a cytokine. This strategy was illustrated by the vaccination of mice with a Salmonella strain carrying eukaryotic expression plasmids encoding the transcription factor Fos-related antigen 1 (Fra-1) overexpressed in an aggressive murine breast carcinoma model and IL-18, a cytokine known to suppress angiogenesis and to stimulate INF-γ production by T and NK cells 54. This multifunctional DNA vaccine proved effective in protecting against growth and metastases of breast cancer by combining the action of immune effector cells with suppression of tumour angiogenesis 54.
Genetic vaccination mediated by Salmonella has also been used to target vascular endothelium growth factor receptor 2 55. This receptor is up-regulated on proliferating endothelial cells of the tumour vasculature and strong cellular immune responses were elicited against these cells by oral administration of the bacteria. This vaccination resulted in the suppression of tumour vascularization and the vaccinated mice showed protection against tumour challenge and reduced growth of metastases. This effect was enhanced when a strain of S. typhimurium carrying plasmids encoding the vascular endothelium growth factor receptor 2 and the murine IL-12 gene was used 56.
Applications of bacterial vectors: bacteriolysis of tumours
The preferential replication of bacteria in certain experimental animal tumours was initially reported in the 1960s when certain strains of Clostridia were shown to proliferate exclusively in the tumour. It was assumed that the anaerobic bacteria were replicating in the necrotic centres of these tumours, leaving the well-oxygenated normal tissues unaffected. This bacterial growth was associated with lysis of large tumours with a necrotic/hypoxic centre but had little effect on small metastatic lesions. Some animals became ill and died during the peak of oncolysis, presumably from a combination of the systemic effects of the bacterial inflammation and the release of necrotic tumour debris 57–59. These pre-clinical studies led to a clinical trial in humans, using spores of Clostridia incapable of producing toxins 60. Most patients showed no evidence of objective regression of the tumours but abscesses in the tumour masses were detected in some patients. Cultures of biopsies demonstrated the presence of the injected micro-organism but further studies were abandoned due to the lack of clinical efficacy.
More recently, investigators have attempted to use the tumour-targeting properties of Clostridium for the selective delivery of pro-drug-activating enzymes 61–64. In these studies, the E. coli enzyme cytosine deaminase 62, 64, 65 and nitroreductase 61 were expressed in Clostridium and were shown to convert the non-toxic pro-drugs 5-fluorocytosine and CB1954, respectively, into toxic compounds capable of diffusing in the tumours and killing the cancer cells through a bystander effect. Using a similar principle, a strain of Clostridia was engineered in which a radio-responsive promoter drove the expression of TNFα 66. In an attempt to optimize the obligate anaerobic strain used, a screen of 26 different types of bacteria was performed and Clostridium novyi appeared particularly promising 67. A strain of C. novyi devoid of its lethal toxin was then engineered (C. novyi-NT) and when its spores were administered with conventional chemotherapeutic drugs, necrosis of tumours often developed within 24 h, resulting in a significant and prolonged anti-tumour effect 67. Encouraging anti-tumour activities in pre-clinical tumour models were also observed when C. novyi-NT was administered in combination with anti-microtubule agents 68 or associated with radiation therapy 69. Finally, this bacteriolysis was shown to trigger a long-lasting immune response amplifying the bacteriolytic action of C. novyi-NT 70.
Based on the same principles, the Gram-positive anaerobic bacteria Bifidobacterium have been shown to colonize large tumours. In contrast to Clostridia, Bifidobacteria are non-pathogenic, non-spore-forming, and are found naturally in the digestive tract of humans and other mammals. The first study was performed in the 1980s, when Ehrlich ascites tumours were implanted in the thigh muscles of mice 71. Systemic injection of a suspension of Bifidobacteria led to the colonization of tumours. This effect was amplified by daily administration of lactulose, a sugar substrate metabolized by bacteria but not by mammalian cells. However, no anti-tumour effects or increases in survival were found in these studies. More recently, B. adolescentis carrying a plasmid encoding the endostatin gene was shown to target subcutaneously implanted liver tumours in Balb/c mice, leading to inhibition of angiogenesis and growth of the tumour 72. The same group has also reported that the oral administration of B. longum carrying the endostatin gene was efficient in the same tumour model. This effect was amplified by co-administration of selenium, thought to act through an improved activity of NK and T cells 73. Bifidobacteria may therefore represent a safer alternative to Clostridia.
Gram-negative Salmonella have also been proposed as oncolytic agents. In contrast to obligate anaerobic bacteria such as Clostridia and Bifidobacteria, Salmonella are facultative anaerobic bacteria and have the potential to colonize oxygenated small metastatic lesions as well as large tumours with a hypoxic centre. The first demonstration of the potential of Salmonella was provided in 1997, when Salmonella auxotrophs injected into tumour-bearing animals were shown to replicate preferentially in tumours. The ratio of bacteria in the tumour to bacteria in normal tissues exceeded 1000/1 and this accumulation was accompanied by a therapeutic effect 74. In a separate study, Salmonella were shown to inhibit melanoma metastases, leading to a significant reduction in the size and number of micrometastases 75. To reduce the possibility of lipopolysaccharide-induced septic shock in patients, lipid-A-modified (msbB) Salmonella auxotrophs (purI– ) were developed 76. These mutants showed attenuated toxicity in mice and swine, associated with reduced host TNFα induction. This was achieved without losing the targeting of the tumour and the therapeutic effect in mice 76. In terms of mechanism, the Salmonella pathogenicity island 1 (SPI1) does not appear to be necessary for the anti-tumour activity of the bacteria 77. Disabling SPI1 dramatically reduced tumour cell invasion in vitro, but did not alter the anti-tumour activity in vivo. This observation strongly suggests that invasion of the cancer cells is not involved in this anti-tumour effect. By contrast, disabling SPI2 led to loss of the anti-tumour activity after either intravenous or intratumoural injections 77. SPI2 has a crucial role in systemic growth of Salmonella in its host and is required for survival within macrophages and epithelial cells 78. However, an SPI2-negative Salmonella strain did not suppress tumour growth in CD18-deficient mice with defective macrophages and neutrophils, suggesting that the loss of effect of this bacterial strain was not solely a function of increased susceptibility to immune clearance. The precise molecular mechanisms therefore remain to be determined. The anti-tumour effect mediated by Salmonella was shown to be enhanced by radiation 79 and low doses of cisplatin 80. To amplify the anti-tumour effect, new strains of genetically engineered Salmonella armed with therapeutically relevant genes have been produced. They include strains capable of delivering the herpes simplex thymidine kinase protein 74, 76 as well as the genes coding for endostatin 81 and thrombospondin-1 82. In addition, diagnostic imaging is an unexpected and potentially powerful application of tumour-targeting Salmonella83, 84. The strategy involves the ability of bacteria expressing the herpes simplex thymidine kinase protein to phosphorylate radiolabelled nucleoside analogues. In this way, the radiotracer becomes trapped in the bacteria and as Salmonella accumulates in tumours, the radioactive signal that can be monitored by positron emission tomography 83 provides information on the localization of tumour sites.
Studies with direct clinical relevance
The anti-tumour activity of the Salmonella strain VNP20009 was tested in dogs with spontaneous neoplasia, in the context of a phase I dose escalation trial 85. VNP20009 is a genetically modified strain of S. typhimurium that includes genetically stable attenuated virulence (a deletion in the purI gene), reduction of septic shock potential (a deletion in the msbB gene), and antibiotic susceptibility 86. Intravenous administration of VNP20009 at doses with acceptable toxicity resulted in detectable bacterial colonization of tumour tissue and significant anti-tumour activity, with four complete responses out of 41 animals treated 85. The same strain was administered to 24 patients with metastatic melanoma and one patient with metastatic renal carcinoma 87. The results established a maximum tolerated dose of 3 × 108 cfu/m2 and dose-limiting toxicity was observed in patients receiving 109 cfu/m2, with symptoms that included thrombocytopenia, anaemia, persistent bacteraemia, hyperbilirubinaemia, diarrhoea, and vomiting. VNP20009 induced a dose-related increase in the circulation of pro-inflammatory cytokines. Tumour colonization occurred in three patients but no objective tumour regression was observed 87. Another clinical trial involved the intratumoural injection of attenuated Salmonella expressing the E. coli cytosine deaminase gene (3 × 106 − 3 × 107 cfu/m2) in three refractory cancer patients, followed by administration of 5-fluorocytosine 88. No significant adverse events related to the treatment were observed. Two patients had evidence of bacterial colonization of the tumour that persisted for at least 15 days after the initial injection. Conversion of 5-fluorocytosine to 5-fluorouracil as a result of cytosine deaminase expression was demonstrated in these two patients 88. These two clinical trials highlight the need for the generation of strains with reduced toxicities and improved tumour colonization properties.
Applications to gastro-intestinal diseases
Inflammatory bowel disease (IBD) is a significant health-care problem characterized by diarrhoea, pain, other intestinal symptoms, and life-long relapses. The pathogenesis is complex and relies on the interaction between three essential factors: genetic susceptibility, intestinal bacteria, and the gut mucosal immune response. Despite significant progress in drug therapy, most strategies that use immunomodulation share the same limitations, which include lack of organ specificity, that result in unpleasant side effects. To address these limitations, a recombinant Lactococcus lactis strain was engineered to produce the anti-inflammatory cytokine IL-10 89. Administration of this strain led to local production of the anti-inflammatory cytokine and prevented the development of colitis in different mouse models 89. This strain was also modified to carry a mutation disabling the thymidylate synthase, resulting in a strain dependent on the availability of thymine or thymidine in the local micro-environment, which was less likely to accumulate in the environment 90. Following the demonstration of proof of principle, further studies are now focusing on characterization of the best possible transgene (reviewed in ref 91).
Another bacterial strain used in the treatment of experimental IBD is the invasive E. coli9, expressing invasin from Yersinia and LLO from Listeria and carrying a eukaryotic expression plasmid in which the constitutive CMV promoter drives expression of the immunomodulatory cytokine TGF-β1 92. Oral administration of these bacteria, which is known to allow gene transfer in vitro9, 17, led to a significant reduction of the severity of experimental colitis in mice, with vector-specific transcripts detected in colonic and extra-colonic tissues such as the lungs, liver, and spleen. To avoid expression in these extra-colonic tissues, the CMV promoter driving the expression of TGF-β1 was replaced by the inflammation-inducible IL-8 promoter. This substitution led to the restriction of expression of TGF-β1 in the inflamed colon without affecting the therapeutic effects 92, demonstrating that targeted gene expression and therapeutic benefits can be obtained using bacterial gene transfer in the gut.
Applications of bacterial vectors: emerging technologies
Delivery of large DNA molecules/artificial chromosomes
One of the problems encountered when trying to transfect mammalian cells with large DNA molecules is the possibility of mechanical breakage of these large molecules during the purification process. In that context, the utilization of bacteria to transfer large DNA molecules would simplify the procedure. The delivery of bacterial artificial chromosomes was first demonstrated into HeLa cells using an invasive E. coli93. Direct DNA transfer of up to around 1 Mb was demonstrated and as the bacterial vector is equipped with an inducible recombination system, modifications of the bacterial artificial chromosome sequences should be possible. More recently, efficient transfer of an alpha-satellite DNA cloned into a P1-based artificial chromosome was stably delivered into the HT1080 cell line and efficiently generated human artificial chromosomes de novo94. In the same report, a 160 kb construct containing the cystic fibrosis transmembrane conductance regulator (CFTR) gene was transferred into the same cells, where it was transcribed and correctly spliced 94. In a study in mice, large DNA molecules carrying the viral genome of the murine cytomegalovirus (MCMV) were transferred using E. coli and S. typhimurium as delivery vectors 95. This transfer led to a productive virus infection that resulted in elevated titres of specific anti-MCMV antibodies, protection against lethal MCMV challenge, and strong expression of additional genes introduced into the viral genome. Thus, the reconstitution of infectious virus from live attenuated bacteria presents a novel concept for multivalent virus vaccines launched from bacterial vectors.
Delivery of small interfering RNA (siRNA)
The utilization of double-stranded RNA to silence target genes (RNA interference: RNAi) has potential therapeutic applications that are widely acknowledged 96. Bacteria-mediated induction of RNAi was established by demonstrating target-specific gene silencing after transfer of double-stranded RNA from E. coli in the nematode Caenorhabditis elegans97. RNA stability in the bacterial vector and transfer into eukaryotic cells were enhanced when the E. coli vector was rendered deficient in RNAse III 98. In the light of these results obtained in the nematode, the delivery of double-stranded RNA or eukaryotic expression plasmids encoding siRNAs into mammalian cells can be envisaged in the near future.
The utilization of bacteria in gene therapy is a recent strategy but pre-clinical studies have already demonstrated the potential of this approach in infectious diseases, where they can be used as vectors for vaccination; in oncology, where they have potential in immunotherapy and tumour targeting; and in gastroenterology, where they could be used as vectors for topical delivery of immunomodulatory cytokines. Some of these bacteria have been tested in humans and their safety profile is acceptable. However, improvement in their tolerability will have to be made to increase the dose injected and achieve real efficacy.
The research in the author's laboratory is supported by Cancer Research UK.