Engineering the gut microbiota to treat chronic diseases

Abstract

Gut microbes play vital roles in host health and disease. A number of commensal bacteria have been used as vectors for genetic engineering to create living therapeutics. This review highlights recent advances in engineering gut bacteria for the treatment of chronic diseases such as metabolic diseases, cancer, inflammatory bowel diseases, and autoimmune disorders.

INTRODUCTION

Studies with germ-free and antibiotic-treated animals reveal a critical role for commensal organisms of the mammalian gastrointestinal tract (the gut microbiota) in the regulation of host immune responses, metabolism, and behavior (Sudo et al. 2004; Smith et al. 2007; Sekirov et al. 2010; Rooks and Garrett 2016; Kennedy et al. 2018). In some cases, specific metabolites released by the gut microbiota that directly affect the host have been identified (e.g. short chain fatty acids) (Feng et al. 2018; Bilotta and Cong 2019). Such studies suggest the therapeutic potential of manipulating gut microbial metabolites to improve host health and treat disease. Approaches to alter microbial metabolites include the manipulation of diet, use of medications, or administration of bacteria known to release beneficial metabolites (e.g. probiotics) (Gerritsen et al. 2011; Choi and Cho 2016; Mukherjee et al. 2018). This review focuses on a key variant of the latter approach, which is genetically engineering specific bacterial species to produce beneficial metabolites and incorporating these bacteria into the host microbiota. The choice of metabolites include novel compounds, enzymes that degrade toxic compounds found in diet or produced endogenously, beneficial host small molecule metabolites, or even bacterial secondary metabolites (Table 1). In the latter case, engineered bacteria may offer greater therapeutic utility than native bacteria because they can be manipulated to produce higher yields of endogenous bacterial products and to have additional safety features such as passive containment systems. When long-term delivery of a therapeutic compound is required for maximal therapeutic effects, engineered bacteria with a high potential for colonization may be advantageous due to less need for frequent administration (Pinero-Lambea et al. 2015). Furthermore, the slow release of the beneficial metabolite targeted directly at the distal small intestine or colon may reduce off-target effects of a compound.

Table 1.

Engineered therapeutic bacteria categorized by their target disease and expressed cargo

Disease model	Expressed cargo	Strain	Route of administration	Model Animal	Refs
Atherosclerosis	heat shock protein-65	Lactococcus lactis NZ9000	Oral	mouse	(Jing et al. 2011)
	N-acyl phosphatidylethanolamine	Escherichia coli Nissle 1917	Oral	mouse	(May-Zhang et al. 2019)
Autoimmune encephalomyelitis	heat shock protein-65	Lactococcus lactis NCD02118	Oral	mouse	(Rezende et al. 2013)
Cancer (general)	cytosine deaminase	Salmonella typhimurium VNP20009 and VNP20047	Systemic	rat	(Mei et al. 2002)
		Bifidobacterium longum 105-A	-	In vitro	(Nakamura et al. 2002)
		Bifidobacterium longum	Intratumoral	rat	(Sasaki et al. 2006)
		Bifidobacterium infantis 2001	systemic	mouse	(Yi et al. 2005)
		Clostridium sporogenes	Systemic	mouse	(Liu et al. 2002)
		Clostridium sporogenes	Systemic	rat	(Nuyts et al. 2001a)
		Clostridium acetobutylicum DSM792 and NI4082	intratumoral	rat	(Theys et al. 2001)
		Salmonella enterica aroA strain SI7207-4S2	Intraperitoneal	mouse	(Royo et al. 2007)
		Bifidobacterium breve I-53-8w	Systemic	mouse	(Hidaka et al. 2007)
	mTNF-α	Clostridium acetobutylicum DSM792	-	In vitro	(Nuyts et al. 2001b)
	TNF-related apoptosis-inducing ligand (TRAIL)	Salmonella typhimurium VNP20009	Systemic	mouse	(Ganai et al. 2009)
	nitroreductase	Clostridium beijerinckii NCIMB 8052	Systemic	mouse	(Lemmon et al. 1997)
Cancer (colon)	shRNAs against catenin β-1	Escherichia coli BL21DE3	Oral or Intravenous	mouse	(Xiang et al. 2006)
	Cytolysin A	Escherichia coli MG1655	Systemic	mouse	(Jiang et al. 2010)
	mitochondrial targeting domain of Noxa fused with cell-penetrating peptide DS4.3 (DS4.3-MTD)	Salmonella typhimurium SHJ2037, SMR2130, SKS001, SKS002, SKS003	Systemic	mouse	(Jeong et al. 2014)
Cancer (Liver)	endostatin	Bifidobacterium adolescentis	Systemic	mouse	(Li et al. 2003)
Cancer (melanoma)	thrombospondin-1	Salmonella choleraesuis	Systemic	mouse	(Lee et al. 2005)
Diabetes	GLP-1(1-37)	Lactobacillus gasseri	Oral	rat	(Duan et al. 2015)
	proinsulin and IL-10	Lactococcus lactis	Oral	mouse	(Takiishi et al. 2012)
	IL-10 and glutamic acid decarboxylase-65	Lactococcus lactis	Oral	mouse	(Robert et al. 2014)
Inflammatory Bowel Disease	heat shock protein-65 fused with six tandem repeats of IA2P2 sequences	Lactococcus lactis	Oral	mouse	(Liu et al. 2016)
	penetratin-GLP-1 fusion protein	Bifidobacterium longum HB15	Ileum	rat	(Wei et al. 2015b)
	interleukin-10	Lactococcus lactis (LL-Thy12)	Oral	human	(Braat et al. 2006)
	α-melanocyte-stimulating hormone	Bifidobacterium longum HB15	Oral	rat	(Wei et al. 2015a)
	superoxide dismutase or catalase	Lactobacillus casei BL23	Intragastric	mouse	(LeBlanc et al. 2011)
	superoxide dismutase	Lactobacillus plantarum and Lactococcus lactis	Oral	rat	(Han et al. 2006)
		Lactobacillus gasseri NC1501	Oral	mouse	(Carroll et al. 2007)
		Bifidobacterium longum HB15	Oral	mouse	(Liu et al. 2018)
Obesity	N-acyl-phosphatidylethanolamine	Escherichia coli Nissle 1917	Oral	mouse	(Chen et al. 2014) (Dosoky et al. 2019) (May-Zhang et al. 2019)
Oral mucositis	Trefoil Factor 1	Lactococcus lactis (SAGX0085)	Topical	human	(Caluwaerts et al. 2010) (Limaye et al. 2013)
Phenylketonuria	phenylalanine ammonia lyase and L-amino acid deaminase	Escherichia coli Nissle 1917 SYNB1618	Oral	mouse and monkey	(Isabella et al. 2018)

To date, products engineered to be produced by therapeutic bacteria have included small molecules, vaccine antigens, enzymes, interleukins, and antibodies. In this review, we group the therapeutic bacteria by individual species engineered followed by target clinical condition. In theory, any species of bacteria typically found in the human gut could be engineered. Since the composition of the gut microbiota varies both longitudinally (e.g. ileum vs colon) and latitudinally (e.g. intestinal mucus versus lumen) (Sekirov et al. 2010), bacteria that localize to the most appropriate target region should be chosen for modification. However, the choice of vectors for in situ bacterial production of therapeutic compounds have historically seem to be driven by practical considerations such as the ease of genetic manipulation, large-scale culture, and low toxicity. For this reason, the most commonly used genetically engineered strains are lactic acid bacteria and Escherichia coli strains (Behnsen et al. 2013).

Engineered Lactic Acid Bacteria

Lactic acid bacteria represent a group of Gram-positive, non-spore-forming, nonpathogenic, non-invasive, facultative anaerobes that are mostly used in food fermentation. Lactic acid bacteria such as Lactococcus lactis are not part of normal human gut flora but are reported to survive in the environment of the human gut (Klijn et al. 1995). Many lactic acid bacterial strains, including L. lactis, L. casei, L. gasseri, L. bulgaricus and Streptococcus thermophilus, are designated as Generally Recognized As Safe (GRAS) by the United States Food and Drag Administration (Cano-Garrido et al. 2015; Cook et al. 2018; Goh et al. 2011). The role of engineered lactic acid bacteria in the treatment of disease has been recently reviewed (De Moreno De Leblanc et al. 2015; Charbonneau et al. 2020).

Engineered lactic acid bacteria and inflammatory bowel disease (IBD)

One early target of engineered lactic acid bacteria was IBD, which includes Crohn’s Disease and ulcerative colitis (Martin et al. 2013). Because interleukin-10 (IL-10) is critical to downregulating inflammatory cascades (Stordeur and Goldman 1998), L. lactis was engineered to secrete IL-10 (Steidler et al. 2000). In mice given dextran sulfate sodium (DSS) to model the chronic colitis, daily intragastric administration of L. lactis secreting IL-10 reduced colitis by 50% (Steidler et al. 2000). IL-10 secreting L. lactis also prevented development of spontaneous colitis in IL-10^−/− mice (Steidler et al. 2000). Milk fermented by IL-10 producing L. lactis effectively prevented 2,4,6-trinitrobenzene sulfonic acid (TNBS)-induced colitis in mice (Del Carmen et al. 2012). This modified strain also reduced dimethylhydrazine-induced colorectal cancer in mice (del Carmen et al. 2017). One mechanism for the anti-inflammatory effects of IL-10 secreting L. lactis is the induction of regulatory T cells (Tregs) via dendritic cells activation (Huibregtse et al. 2012). In a Phase I clinical trial with 10 individuals with Crohn’s Disease, administration of IL-10-secreting L. lactis reduced disease activity of 8 individuals without systemic side effects (Braat et al. 2006). For biological containment, the IL-10 transgene was inserted into the ThyA gene to create thymidine auxotrophy (Braat et al. 2006). This containment strategy appears to be effective, as engineered L. lactis was not detectable in stool two days after administration (Steidler et al. 2003). Despite promise of these Phase I studies in 2006, we are not aware of any publications reporting follow-up clinical studies.

Instead of using lactic acid bacteria directly secreting IL-10 as a therapeutic strategy, an alternative is to use lactic acid bacteria to deliver a plasmid containing a eukaryotic expression cassette gene into host cell (e.g. pValac) (Guimarães et al. 2009). In this strategy, the lactic acid bacteria were engineered to express surface proteins Listeria monocytogenes Intemalin A or Staphylococcus aureus Fibronectin Binding Protein A to enable internalization of the bacteria by mammalian epithelial cells (Guimarães et al. 2009)(del Carmen et al. 2015). L. lactis and Streptococcus thermophilus transformed with pValac::il-10 plasmid reduced injury in acute and chronic models of colitis (Del Carmen et al. 2013) (del Carmen et al. 2015) (Del Carmen et al. 2014).

Another potential treatment for IBD has been the oral administration of recombinant antioxidant enzymes such as superoxide dismutase (SOD) or catalase, since the fonnation of reactive oxygen species is a prominent feature of IBD (Harris et al. 1992). However, recombinant SOD has a half-life of only 5-10 min in circulation which restricts the utility of using the recombinant protein as a therapeutic (Turrens et al. 1984). Of interest is whether engineered bacteria that continuously produce SOD might overcome this problem. In a rat model of colitis induced by trinitrobenzene sulphonic acid (TNBS), administration of SOD-producing L. lactis reduces indicators of colitis including myeloperoxidase activity, nitro-tyrosine immunoreactivity, and colon injury (Han et al. 2006). Engineering of other strains of lactic acid bacteria to express SOD also produces similar effects. For instance, SOD-producing L. gasseri reduce the severity of colitis in IL-10^−/− mice (Carroll et al. 2007). SOD-producing L. casei BL23 reduce colitis in TNBS-treated and DSS-treated mice (LeBlanc et al. 2011). A catalase-producing strain of L. casei BL23 also reduces colitis in the TNBS model (LeBlanc et al. 2011). Similarly, catalase- or SOD-producing Streptococcus thermophilus were effective in preventing TNBS-induced colitis in mice (del Carmen et al. 2014). To our knowledge, no human clinical trials of SOD-secreting or catalase-secreting lactic acid bacteria have been performed.

Because administration of anti-tumor necrosis factor-α (TNFα) antibodies (e.g. Humira) is an effective treatment for IBD, L. lactis have been engineered to secrete single domain antibody fragments (nanobodies) that can neutralize TNFα in vitro (Muyldermans 2013). Oral administration of these nanobody-secreting L. lactis reduces inflammation in DSS-induced colitis in mice and improves established enterocolitis in IL-10^−/− mice (Vandenbroucke et al. 2010). To our knowledge, no human clinical trial of nanobody-secreting L. lactis have been performed.

Lactic acid bacteria engineered for immunotolerance

In addition to IBD treatment, another potential therapeutic use of engineered lactic acid bacteria is to induce immune tolerance in individuals with aberrant immune responses (allergic or autoimmune diseases). Presentation of antigens in the context of lactic acid bacteria appears to induce proliferation of Tregs that suppress immune responses (Huibregtse et al. 2007). Oral administration of L. lactis engineered to secrete ovalbumin stimulates immune tolerance to ovalbumin in ovalbumin T-cell receptor (TCR) transgenic mice (DOl 1.10) (Huibregtse et al. 2007). Autoimmune diseases that have been targeted using antigen secreting lactic acid bacteria include celiac disease (where anti-gliadin antibodies are produced in response to foods with gluten) and type 1 diabetes (where antibodies are produced against pancreatic peptides such pro-insulin or glutamic acid decarboxylase-65 [GAD-65] participate in killing pancreatic β-cells). Administration of L. lactis engineered to express gliadin peptide effectively reduces delayed hyperactivity in gliadin-sensitized nonobese diabetic (NOD) transgenic mice (Huibregtse et al. 2009). Orally administered L. lactis engineered to secrete both proinsulin and IL-10, stably revert type 1 diabetes in NOD mice which were also given low-dose anti-CD3 monoclonal antibodies (mAb) (Takiishi et al. 2012). Orally administered L. lactis secreting GAD-65 and IL-10, along with low-dose anti-CD3 mAb, reverses diabetes in recent-onset NOD mice (Robert et al. 2014).

More recently, induction of immunotolerance to heat shock proteins has been identified as a mechanism for treating inflammatory disease. Heat shock proteins are highly expressed in inflammatory diseases such as atherosclerosis and encephalomyelitis (Colaco et al. 2013), and evidence supports a role for autoimmune responses to heat shock proteins in contributing to the initiation and progression of atherogenesis (Xu 2002). To induce immunotolerance to heat shock proteins, L. lactis were engineered to express mycobacterial heat shock protein-65. These bacteria were administered orally to atherosclerosis prone low density lipoprotein receptor^−/− mice. Heat shock protein expressing L. lactis protects against endothelial damage and reduces atherosclerotic lesions (Jing et al. 2011). This treatment suppresses heat shock protein specific proliferation, increases IL-10 production and reduces interferon-gamma (IFN-γ) level. Similarly, oral administration of heat shock protein-expressing L. lactis inhibits the development of autoimmnune encephalomyelitis in C57BL/6 mice, as measured by reduced inflammatory cell infiltrate in the spinal cord, increased IL-10 production, and reduced IL-17 and IFN-γ levels (Rezende et al. 2013). In type 1 diabetes, destruction of pancreatic β-cells occurs when autoimmune T cells react against autoantigens such as insulinoma antigen-2 (IA2), human heat shock protein-60 (HSP60) and those mimicking mycobacterial heat shock protein-65. Oral administration of recombinant L. lactis-expressing heat shock protein-65 fused to six tandem repeats of IA2P2 (an artificial peptide based on sequences from IA2 and HSP60), to NOD mice reduces hyperglycemia, increases glucose tolerance, and decreases insulitis (Liu et al. 2016). Despite the early promise of these tolerance induction strategies, to the best of our knowledge, no clinical trials with tolerance-inducing engineered L. lactis have been performed.

Lactic acid bacteria engineered for other functionalities

Oral mucositis is a significant and coimnon complication found in cancer patients undergoing chemotherapy and radiation therapy. L. lactis producing the mucosal protectant human Trefoil Factor 1 (hTFF1) was formulated as a mouth rinse (AG013) for treating oral mucositis. Pharmacodynamics, pharmacokinetics, and safety studies conducted in hamsters show that AG013 is effective for treating radiation induced oral mucositis in these animals (Caluwaerts et al. 2010). Phase 1 studies of topically applied AG013 are now completed in patients with advanced head and neck cancer that received chemotherapy (Limaye et al. 2013). Preliminary efficacy data showed that AG013 reduces the duration of symptoms of oral mucositis by nearly one-third.

Engineered lactic acid bacteria is also a novel strategy for treatment of diabetes. In the past decade, the notion that cells that are not pancreatic β cells can be reprograimned to secrete insulin has been intensely studied. Incubating isolated intestinal cells from mouse embryos with glucagon-like peptide (GLP)-1(1-37) causes them to convert into cells that produce insulin in response to glucose (Suzuki et al. 2003). Similarly, injection GLP-1(1-37) into pregnant mice led to jejunal and ileal epithelial cells that produce insulin in response to glucose, while injection into adult mice led to much lower, but still significant insulin production by ileal epithelium (Suzuki et al. 2003). These observations led to engineering L. gasseri to secrete GLP-1(1-37). Remarkably, oral administration of GLP-1 secreting L. gasseri converts intestinal epithelial cells into glucose-responsive insulin-secreting cells which leads to enhanced glucose tolerance in diabetic rats (Duan et al. 2015). Cells surrounding the reprogrammed cells maintain normal functions. This study provided evidence for a safe and effective oral treatment for diabetes.

Lastly, lactic bacteria have been engineered to protect against human iimnunodeficiency virus (HIV). Most HIV transmission occurs on the mucosal surfaces of the GI or cervicovaginal tracts. Current attempts to develop an effective vaccine against HIV have been largely unsuccessful. Bacteria engineered to secrete antiviral compounds that can successfully colonize the host microbiota may be a more durable and practical approach to provide protection against HIV. Cyanovirin-N (CV-N) has been shown to block HIV infection of cervical cells (Buffa et al. 2009). A topical gel form of CV-N is effective in monkeys after a rectal challenge of experimental chimeric simian/HIV virus (Tsai et al. 2004). However, use of such topical gels may have limited efficacy in humans because of the need for application prior and after sexual intercourse to effectively block HIV transmission. Because Lactobacillus is already part of the bacterial biofilm coating the cervicovaginal mucosa, Lactobacillus engineered to secrete CV-N might serve as a living bioshield against HIV transmission. As a proof of this concept, L. jensenii engineered to express CV-N reduces S/HIV transmission by 63% and reduces viral load by 6-fold in colonized macaques (Lagenaur et al. 2011). Furthermore, the engineered L. jensenii was able to colonize and produce CV-N for 6 weeks after treatment, thereby providing a persistent protection against viral transmission. To the best of our knowledge, no clinical trials have been undertaken to test these therapeutic compounds.

Engineered Salmonella spp.

Salmonella are a genus of motile Gram-negative facultative anaerobes whose pathogenic strains cause intestinal infections. An important feature of Salmonella, along with a number of other gram-negative facultative anaerobes like E. coli, is the ability to home to and proliferate in solid tumors (Mi et al. 2019)(Van Dessel et al. 2015). After intravenous injection into tumor-bearing mice. Salmonella preferentially colonizes tumors compared to other organs by four orders of magnitude (Forbes et al. 2003). After entry into tumors through blood vessels. Salmonella migrate to regions of the tumor away from the blood vessels (Ganai et al. 2011). Salmonella appears to prefer necrotic regions for proliferation (Forbes et al. 2003), and its motility allows for penetration into these necrotic areas (Kasinskas and Forbes 2006). However, the importance of motility remains controversial, as studies with Salmonella mutants that lacked motility or are unresponsive to chemical gradients still showed tumor invasion and colonization (Crull et al. 2011)(Broadway et al. 2015)(Broadway et al. 2017)(Feigner et al. 2018). Even without engineering. Salmonella has been reported to induce tumor cytotoxicity and to enhance vulnerability of tumors to chemotherapeutics. Multiple mechanisms have been proposed for these effects: competition by bacteria for nutrients essential to tumor growth, secretion of metabolites/toxin lethal to tumors, induction of TNFα secretion by tumor macrophages, induction of cell cycling of previously quiescent tumor cells (making them vulnerable to chemotoxic agents), induction of connexin 43 (enhancing permeability of chemotherapeutic agents into tumor mass), and suppression of P-glycoprotein expression (reducing capacity of tumor cells to efflux chemotherapeutic) (Mi et al. 2019)(Rosenberg et al. 2002). Genetic engineering has been used to enhance the safety of administering Salmonella and the selective activation of therapeutics within tumors.

Any beneficial anti-tumor effect of administering Salmonella must be achieved at a dose that does not induce the potentially lethal septic cytokine storm typical of sepsis. To this end, the S. typhimurium strain VNP20009 has been engineered with deletions in purI and msbB, so that it fails to acquire external adenine and expresses a lipid A variant which poorly induces TNFα, respectively. Other mutant S. typhimurium strains with attenuated virulence tested for bacterial-mediated cancer therapy include AR-1 (an arginine and leucine auxotroph), ΔppGpp (with mutations in relA and SpoT leading to defective ppGpp synthesis) (Mi et al. 2019), and SF200 (with mutations in Lipid A and flagella synthesis) (Feigner et al. 2018). Administration of VNP20009 to nine different mouse transplantable tumor models resulted in significantly slower growth of tumor in some, but not all tumor types (Rosenberg et al. 2002). However, in a phase I clinical study with metastatic cancer patients, intravenous infusion of VNP20009 showed no antitumor effects since the maximum tolerated dose (3x10⁸ cfu/m²) was still inadequate to exert the desired effect (Toso et al. 2002). Higher doses of VNP20009 showed toxicity symptoms such as anemia, low platelet count, persistent bacteremia, high blood bilirubin, nausea, vomiting, diarrhea, hypophosphatemia, and elevated alkaline phosphatase (Toso et al. 2002).

The characteristic tumor-homing properties of Salmonella can be exploited to specifically deliver anti-cancer compounds to tumors and thereby overcome limitations of conventional chemotherapy such as poor tumor selectivity and limited penetrability (Forbes 2010). Ganai et al engineered VNP20009 to produce murine TNF-related apoptosis-inducing ligand (TRAIL) under the control of a γ-irradiation-inducible RecA promoter (Ganai et al. 2009). In vitro, the secreted TRAIL selectively induces a caspase-3-mediated apoptosis in 4T1 mammary carcinoma cells. Systemic administration of tins TRAIL-producing VNP20009 strain in a subcutaneous mammary carcinoma mouse model followed by induction using γ-irradiation. successfully reduces tumor growth and enhanced host survival (Ganai et al. 2009). Wen et al engineered ΔppGpp S. typhimurium to express tissue inhibitor of metalloproteinases 2 (TIMP-2) (Wen et al. 2018). When administered intracranially into BALB/c mice bearing orthotopic glioma brain tumors, these TIMP-2 expressing therapeutic bacteria reduce tumor size and increase survival by 60% compared to controls (Wen et al. 2018). Lee et al engineered S. choleraesuis to express thrombospondin-1, an endogenous angiogenesis inhibitor (Lee et al. 2005). Administration of thrombospondin-l expressing S. choleraesuis to mice previously inoculated with B16F10 melanoma cells inhibits tumor growth and enhances survival compared to controls (Lee et al. 2005). Jeong et al engineered ΔppGpp S. typhimurium to express the mitochondrial targeting domain of Noxa (MTD), a protein which can induce activation of the intrinsic apoptotic signaling pathway, fused with the cell-penetrating peptide DS4.3 (DS4.3-MTD) to enhance uptake by tumor cells (Jeong et al. 2014). To facilitate release of DS4.3-MTD from bacterial cells, the bacteria were also engineered to express two phage lysis genes under the control of an L-arabinose promoter (Jeong et al. 2014). Administration of these therapeutic bacteria to mice implanted with CT26 colon carcinomas, followed by a dose of L-arabinose, leads to almost complete killing of the tumors (Jeong et al. 2014). Although these pre-clinical studies appear highly promising, to the best of om knowledge, no clinical trials have been conducted using any of these therapeutic bacteria.

The tumor-homing properties of Salmonella can also be exploited to activate highly toxic chemotherapeutics selectively at tumor sites. King et al engineered VNP20009 to express cytosine deaminase (CD), which converts the pro-drug 5-fluorocytosine (5-FC) to its cytotoxic metabolite 5-fluorouracil (5-FU) (King et al. 2002). Administration of CD-expressing VNP20009 to mice previously implanted with murine B16-F10 melanoma cells and then treated with 5-FC show that 5-FU selectively forms at tumor sites (King et al. 2002). Administration of CD-expressing VNP20009 into three different mouse models of implanted colon carcinoma cells (C38, WiDR, and LoVo), followed by administration of 5-FC, dramatically reduces tumor bmden in all three models (King et al. 2002).

Tumor colonization and 5-FU conversion is also seen in tumor-bearing rats (Mei et al. 2002). Royo et al further optimized this approach by engineering an attenuated strain of S. enterica to express CD under the control of a salicylate promoter (Royo et al. 2007). CD-expressing S. enterica were then administered to mice previously implanted with murine F1.A11 fibrosarcoma cells to form palpable tumors and the mice treated with 5-FC and either salicylate (inducer) or tetracycline (non-inducer control). Mice treated with salicylate have ~50% greater reductions in tumor compared to tetracycline treated mice or mice that receive a non-CD expressing S. enterica strain (Royo et al. 2007). These results suggest that engineered Salmonella strains can be used to selectively activate cytotoxic drags within the tumor site, potentially increasing the therapeutic window for such drags.

In 2001, a Phase I clinical trial to administer Salmonella typhimurium expressing E. coli CD (TAPET-CD or VNP20029) to advanced or metastatic cancer patients in combination with 5-FC was proposed (Cunningham and Nemunaitis 2001). Three patients with refractory cancer received intratumoral injections of TAPET-CD (3×10⁶ - 3×10⁷ CFU/m² once every 28 days) coupled with 100 mg/kg/day of 5-FC orally on days 4 to 14 of each of the 28 day cycle for six cycles. This led to bacterial colonization for at least 15 days after initial injection and conversion of 5-FC to 5-FU in 2 of the 3 patients (Nemunaitis et al. 2003). To the best of om knowledge, no further clinical trials of TAPET-CD have been reported.

Further optimization of bacterially-directed chemotherapy has been used to make expression of the therapeutic recombinant protein independent of an administered inducer (e.g. L-arabinose or salicylate) by using inducible promoters that only respond to conditions that prevail in tumors (e.g. hypoxia). These promoters are crucial for the specificity of such delivery systems. Lesclmer and coworkers identified S. typhimurium promoters that exclusively control gene expression in cancer cells as well as the DNA motif responsible for tumor specificity (Leschner et al. 2012). They subsequently established the functional structure of these bacterial promoters as a combination of a weak basal promoter and a strong fumarate and nitrate reduction regulatory protein (FNR) binding site (Deyneko et al. 2016). Mengesha et al generated an engineered hypoxia-inducible promoter (HIP) to control expression of GFP and RFP in VNP20009 (Mengesha et al. 2006). When this HIP-reporter strain VNP20009 is administered to mice implanted with HCT116 human colorectal carcinoma cells, it leads to a 15-fold greater fluorescence in tumors compared to the control reporter strain under the control of a constitutive promoter. Furthermore, fluorescence for the HIP-reporter strain is almost exclusively seen in the tumor, while the constitutive reporter strain shows fluorescence in other tissues (Mengesha et al. 2006). Arrach et al identified other hypoxia-inducible promoters including the pflE and ansB promoters that could be used to selectively induce gene expression in a hypoxic tumor environment (Arrach et al. 2008). Swofford et al engineered VPN200010 using the luxI/luxR quorum-sensing system of Vibrio fischeri. They showed that with this quorum-sensing system, fluorescent reporter gene expression is only induced under conditions of high cell density. When administered to mice implanted with 4T1 mammary tumors, the fluorescence expression was limited to sites within tumors with high colony density (Swofford et al. 2015).

Engineered Escherichia coli

Escherichia coli is a Gram-negative facultative anaerobe found in the normal gut microbiota of wann-blooded mammals. Although several E. coli strains are key pathogens in infectious diseases, the majority are commensals. E. coli has been the organism of choice for many synthetic biology studies because of its well-characterized genetics and biochemistry, rapid growth, ease of genetic manipulation, and high-yield production (Huang et al. 2012). Although lab strains of E. coli such as MG1655 and BL21DE3 are commonly used for engineering, there has also been substantial use of E. coli Nissle 1917 (EcN), a highly colonizing probiotic strain previously used in Germany for treating ulcerative colitis, chronic constipation and intestinal infections (Behnsen et al. 2013). More recently, the probiotic E. coli strain Symbioflor-2 has been suggested as an alternative bacterial chassis for engineering due to enhanced safety profde relative to EcN (Kocijancic et al. 2016).

Engineered E. coli for cancer treatment

The use of engineered E. coli to treat cancer parallels that of Salmonella. Like Salmonella, E. coli strains systemically administered to mice efficiently colonize solid tmnors, especially targeting the necrotic regions of these tumors (Westphal et al. 2008; Weibel et al. 2008; (Kocijancic et al. 2016). Wildtype E. coli also colonizes tmnors, inducing significant fonnation of avascular necrotic tissue, expression of TNFα and matrix metallopeptidase 9, formation of granulation tissue by macrophages, and enhanced deposition of collagen IV (Weibel et al. 2008). In addition, E. coli engineered to express CD are able to selectively enhance conversion of 5-FC to the cytotoxic 5-FU at the site of the tumor (Lehouritis et al. 2013).

Several novel cargoes have been engineered into E. coli for the treatment of tmnors. One example is the production of short-hairpin RNAs (sliRNAs) to silence tumor-inducing genes of host cancer cells. Xiang et al. engineered E. coli BL21DE3 to produce sliRNA against CTNNB1 (catenin β-1), a key gene overexpressed or mutated in colorectal cancer (Xiang et al. 2006). Administration of these bacteria to nude mice xenografted with human colon cancer cells suppresses CTNNB1 expression in those cells (Xiang et al. 2006). Another example is E. coli MG1655 engineered to express cytolysin A (ClyA), a pore-forming hemolytic bacterial protein with a high capacity to kill mammalian cells (Jiang et al. 2010). A single-dose of ClyA-expressing E. coli MG1655 to BALB/c mice subcutaneously implanted with CT26 colon cancer cells slows the rate of tumor growth and caused extensive tumor necrosis (Jiang et al. 2010). When the CT26 cells are injected intravenously in mice to establish lung metastases, treatment with ClyA-expressing E. coli MG1655 significantly reduces the burden of lung metastases and extends survival (Jiang et al. 2010). Furthermore, when used in combination with low-dose radiation, ClyA-expressing E. coli MG1655 enhances the ability of radiation to halt growth of inoculated B16F10 and 4T1 tumor cells (Jiang et al. 2010).

Engineered E. coli for treatment of obesity, diabetes, cardiometabolic diseases

Engineered EcN incorporated into the gut microbiota have been recently used to treat or inhibit the development of obesity and related metabolic disorders. Studies in our laboratory have focused on increasing the bacterial production of N-acyl-phosphatidylcthanolamincs (NAPE), which are biosynthetic precursors of N-acylethanolamines (NAE). NAPE and NAE are endogenously produced bioactive lipids which play important roles in regulating food intake but their production is downregulated in chronic fat diet feeding (Gillum et al. 2008). Increasing NAEs has several beneficial effects in animal models including increased satiety, reduced inflammation, and reduced pain (Piomelli 2013). We have engineered EcN to produce higher levels of NAPEs by transforming phosphatidylethanolamine V-acyltransferase (Chen et al. 2014)(Dosoky et al. 2018). Administration of NAPE-expressing EcN in the drinking water to C57BL/6J mice maintained on a high fat diet significantly inhibited the development of obesity partially by reducing food intake. The anti-obesity effects of NAPE-expressing EcN persisted 4 weeks after ending treatment due to colonization in the intestine (Chen et al. 2014). In a follow-up study, administration of NAPE-expressing EcN for only two weeks without pre-treatment with antibiotics was sufficient to establish a persistent protection against weight gain (Dosoky et al. 2019).

Obesity also increases the risk for cardiometabolic diseases, which includes cardiovascular disease, non-alcoholic fatty liver disease, and diabetes. In a mouse model of atherosclerosis and fatty liver disease, NAPE-expressing EcN was able to significantly reduce body weight as well as reduce liver fat, inflammation, and signs of early fibrosis (May-Zhang et al. 2019). NAPE-expressing EcN reduced serum cholesterol levels and markedly decreased the extent of necrosis in atherosclerotic lesions, although it did not significantly impact total area of atherosclerotic lesions.

Other engineered EcN have been developed as a potential diabetes treatment. Duan et al engineered EcN to produce the insulintopic protein GLP-1 or pancreatic and duodenal homeobox gene 1 (PDX-1) (Duan et al. 2008). These proteins stimulate intestinal epithelial cells to synthesize insulin in response to glucose. Cell free medium derived from overnight cultures of engineered EcN stains stimulated insulin production in human Caco-2 cells. The follow-up study to these in vitro studies explored whether oral administration of GLP-1 (1-37) secreting bacteria can ameliorate hyperglycemia in rat model of diabetes. However, the authors did not continue with EcN but instead engineered Lactobacillus to secrete GLP-1, as described in the previous section (Duan et al. 2015).

Engineered E. coli for the treatment of other diseases

The versatility of E. coli has enabled its development for treatment of several other diseases. Phenylketonuria is a genetic disease characterized by an inability to metabolize phenylalanine (Phe), which leads to neurotoxicity. Standard treatment for phenylketonuria is a strict low-protein diet combined with amino acids supplemented with trace elements to prevent nutritional deficiencies. One potential alternative to a protein-restricted diet is the engineering of commensal bacteria to help metabolize Phe. One notable strain developed was a genetically modified EcN that produced a Phe-metabolizing enzyme (SYNB1618) that converts Phe to a more harmless compound, trans-cinnamic acid, which is further metabolized in the liver and excreted into urine as hippurate. Administration of SYNB1618 reduced blood Phe levels in a phenylketonuria mouse model and healthy Cynomolgus monkeys given an oral Phe challenge (Isabella et al. 2018). A recent phase 1/2a clinical trial evaluated the safety, tolerability, as well as clearance from the GI tract of SYNB1618 administered orally as a liquid formulation to healthy adult volunteers and phenylketonuria patients (ClinicalTrials.gov identifier NCT03516487). This trial showed that SYNB1618 was viable and able to convert Phe in the human GI tract, as evidenced by the presence of trans-cinnamic acid and hippurate in SYNB1618-treated healthy volunteers and PKU patients that was absent in subjects treated with placebo. SYNB1618 did not induce serious adverse events and was cleared within an expected time frame, indicating no lasting GI colonization. A larger efficacy study evaluating an improved solid formulation of SYNB1618 is expected to begin in 2020.

Similar to lactic acid bacteria, another application for engineered EcN is the protection against HIV. One such design is EcN engineered to secrete HIV-gp41-hemolysin A hybrid peptides, which can interfere with viral attachment, fusion, entry, or replication in target cells (Rao et al. 2005). These engineered EcN can also colonize the GI of mice and secrete peptides throughout the luminal mucosa epithelial surfaces for a period of weeks to months. This approach can not only prevent new HIV infections but also can potentially be used in combination with standard antiretroviral drug therapies as a means of preventing viral rebound in individuals who have ceased treatment.

So far, EcN constitutes a prospective vector for delivering therapeutic molecules to treat several human diseases. The expression of therapeutic molecules can be deliberately regulated in a temporal and quantitative manner through a strategy known as in vivo remote control of bacterial vectors. While the ability to increase therapeutic molecule production is important for effective treatment of diseases, the ability to control expression is essential to limit adverse effects. The choice of a suitable remote control depends on the colonized niche in the body and the specific requirements of intervention. One study examined the suitability of in vivo remote control of EcN engineered with novel promoters that can be chemically induced by L-arabinose, L-rhamnose, or anhydrotetracycline (Loessner et al. 2009). In rodent studies, each promoter displayed a specific induction profile depending on the bacteria niche of residence and the route of inducer administration. The study demonstrated the feasibility of in vivo remote control for EcN colonizing the GI tract as well as in tumors with the inducers given orally. Such promoters represent valuable somces for the design of controllable bacterial vector systems that can potentially overcome safety concerns and enable a flexible treatment system of various human diseases.

Engineered Bifidobacterium spp.

Bifidobacterium are Gram-positive non-pathogenic, non-spore-forming branched anaerobes found naturally in the lower gastrointestinal tract of mammals. Wildtype Bifidobacterium spp. provide a range of benefits to the host including enhancing the immune response, preventing carcinogenesis, and protecting against viral infections (Hidaka et al. 2007). Bifidobacterium delivered orally can stably colonize the intestine, provide long-term retention and increase efficiency of target drags. Similar to Salmonella and E. coli, Bifidobacterium can selectively colonize and grow in the hypoxic regions of solid tumors, making the bacterium a model candidate for targeting cancers. B. adolescentis engineered to produce cndostatin. which inhibits angiogenesis, markedly suppressed growth of Heps mouse liver cancer cells inoculated into BALB/c mice (Li et al. 2003). Other gene delivery systems included various Bifidobacterium strains (Nakamura et al. 2002) (Hidaka et al. 2007) (Sasaki et al. 2006) (Yi et al. 2005) engineered to deliver active CD enzymes to hypoxic regions of solid tumors in rodents which strongly inhibited tumor growth.

Bifidobacterium is also engineered for therapeutic use in IBD management. B. longum was engineered to deliver α-melanocyte-stimulating hormone, a tridecapeptide derived from pro-opiomelanocortintliat has potent anti-inflammatory properties. This engineered bacterium successfully colonizes the intestine and inhibits DSS-induced ulcerative colitis in rats (Wei et al. 2015a). Another therapeutic strategy is B. longum delivery of human manganese superoxide dismutase (rhMnSOD), an antioxidant enzyme that ordinarily has poor penetration and stability as a recombinant protein, through the fusion and co-delivery of a transporter peptide PEP-1 penetratin. This engineered bacterium successfully delivers rhMnSOD into the colon, inducing a potent anti-inflammatory effect and suppressing DSS-induced ulcerative colitis in mice (Liu et al. 2018). The concept of utilizing a fusion protein with penetratin to increase delivery was explored in another study that engineered B. longum to express bioactive penetratin-GLP-1 fusion protein for the treatment of type 2 diabetes (Wei et al. 2015b).

Engineered Clostridium spp.

Clostridium are Gram-positive obligate anaerobe, pathogenic strains of which are responsible for many intestinal infections. However, a number of studies show that these bacteria can proliferate in the hypoxic or necrotic areas of solid tumors and thus were investigated for anti-cancer applications (Lehouritis et al. 2013). Clostridia are sporeforming and thus are injected as spores, which travel to the tumor site and germinate only in anoxic regions. One of the first strains tested was C. histolyticum which induced significant tumor regression and lysis in mice (Parker et al. 1947). A more virulent strain of C. tetani was found to proliferate in and shrink tumors but consequently resulted in rapid death of tumor-bearing animals due to high toxicity (Malmgren and Flanigan 1955). Thus, it became apparent that Clostridium strains must be non-pathogenic for its safe use as treatments. One effective strain. C. novyi. was genetically modified to be non-pathogenic by deleting the gene encoding for the lethal α-toxin NT. This modified strain C. novyi-NT retained its abilities to target and destroy viable tumors (Dang et al. 2001). This strain also secretes liposomase, which has been exploited to enhance the release of liposome-encapsulated drags within tumors (Cheong et al. 2006) leading to the suggestion that its expression in other engineered bacterial strains could be beneficial (Cheong et al. 2007).

Like other bacterial species discussed, engineered Clostridium has also been developed as an alternative approach to cancer gene therapy in which the bacterium is engineered as a tumor-specific vector to deliver anti-tumor genes. For instance, C. beijerinckii was engineered to contain a gene for E coli nitroreductase, which activates a nontoxic prodrag CB1954 into a toxic anticancer drag. These nitroreductase-producing bacteria enhance the killing of tumor cells in vitro by CB1954 and specifically target tumors when injected intravenously into mice (Lemmon et al. 1997). The specificity of gene targeting to hypoxic tumors is further improved by the incorporation of radio-induced promoters, which can control gene expression spatially and temporally. Administration of CD-expressing non-pathogenic C. sporogenes under the control of radio-induced promoters to rats bearing rhabdomyosarcoma enhances anti-tumor responses and increases its specificity (Nuyts et al. 2001a). Similarly, a non-pathogenic CD-expressing C. acetobutylicum has been used to deliver the target protein in a similar tumor model. To further increase effectiveness of engineered C. acetobutylicum, the hypoxic and necrotic areas of the tumor were increased by adding combretastatin A4 phosphate which increases enzyme activity in tumors and induces vascular collapse (Theys et al. 2001). Intravenous administration of 5-fluorocytosine following the systemic delivery of CD-expressing C. sporogenes spores exclusively delivers the enzyme to tumor tissues in mice and results in a strong antitumor effect (Liu et al. 2002). In addition to CD delivery, TNF-α can be delivered locally to a tumor. Clostridia engineered with a radiation-inducible recA promoter was able to secrete TNFα after irradiation and enhance the therapeutic ratio in cancer treatment (Nuyts et al. 2001b). Thus the combination of radiotherapy with Clostridium-mediated protein delivery may provide new possibilities for cancer therapy.

We note that C. perfringens exerts substantial anti-tumoral effects but these beneficial effects are offset by a robust host inflammatory response that often causes significant lethality. Thus engineered attenuated strains of C. perfringens have been considered for therapeutic use (Kubiak and Minton 2015). While some success has been reported in grant abstracts (Plevy 2006), we are not aware of any peer-reviewed publications where engineered C. perfringens have been successfully used in animal models of disease. C. perfringens attenuated by insertion of the inflammation suppressive factor Panton-Valentine Leukocidin (PVL) gene to disrupt its Sod gene was originally reported to enhance survival of mice administered PANC02 tmnors (Li et al. 2008). However this report was subsequently retracted (Li et al. 2010).

Engineered Bacillus spp.

Bacillus subtilis is a Gram-positive aerobic endospore-fonning bacterium with GRAS status, and is used extensively as a model organism for genetics and cellular metabolism studies (Buescher et al. 2012; Koo et al. 2017). It is found in soil and also in the GI tract of humans and some mammals. Its GRAS status led to its use as a probiotic for both animals and humans (Hong et al. 2005). B. subtilis has long been favored as a microbial cell factory for the overproduction of industrially important enzymes, vitamins, and functional sugars. Recently, engineered B. subtilis biofilms were developed as smart living glues (Zhang et al. 2019). Major advantages of using B. subtilis include its high secretory function, production abilities, and low fermentation media requirements (van Dijl and Hecker 2013; Ÿztürk et al. 2016). Protease-deficient strains of B. subtilis have improved production yields of heterologous proteins (Westers et al. 2004). In 2003, a genome-reduced B. subtilis was developed by deleting 7.7% of its genome without affecting its growth or replication (Westers et al. 2003), and was then used to produce several heterologous proteins and biochemicals (Morimoto et al. 2008; Manabe et al. 2011; Juhas et al. 2014). More recently, Reuss et al. constructed B. subtilis 168 strains with a genome reduction of 36% (Reuß et al. 2017). When streptavidin was expressed on killed B. subtilis spores, it enabled targeting colon cancer cells through conjugating monoclonal antibodies to the spore surface (Nguyen et al. 2013). As a biological containment strategy of the genetically engineered B. subtilis spores, chimeric genes were inserted in the two thymidylate synthase genes, thy. I and thyB. This procedure makes the spores with mutated thymidylate synthase gene strictly dependent on thymine (or thymidine) and cannot survive in a thymine- or thymidine-deficient environment (Hosseini et al. 2018) (Duan et al. 2015)

Conclusion and future directions

Bacteria engineered to express therapeutic compounds show a remarkable versatility and potential for treating a wide variety of diseases. The ability of bacteria to home to tumors make them remarkable vehicles for cancer treatment, while the ability of some engineered bacteria to stably colonize the gut makes them efficient vehicles for treatment of chronic diseases. The relative ease of expressing various heterologous proteins with clear clinical relevance would suggest that engineered therapeutic bacteria could be a major new wave of medicine. Yet despite the large number of successful animal trials, very few clinical trials have been undertaken. A clear demonstration of the efficacy in humans is greatly needed to realize the therapeutic potential of engineered bacteria.

One obvious barrier to the initiation of large-scale human trials is the concern of whether current passive containment measures are sufficient to ensure that only individuals who have consented treatment are exposed to engineered bacteria. Although limiting the capacity of bacteria to spread among individuals can be done by the generation of auxotrophy through deletion of essential genes, regulatory agencies remain highly sensitive to concerns of the general public about genetically-modified organisms. To gamer public support and confidence, additional studies that are well-controlled, highly rigorous, and demonstrate effectiveness of passive containment are urgently needed.

Another barrier to both clinical trials and the widespread adoption of engineered therapeutic bacteria is the need to grow up sufficiently large (and uncontaminated) stocks of recombinant therapeutic bacteria in the absence of antibiotics for selection. Expression of heterologous proteins typically comes with a high fitness penalty. In the absence of a strong selective pressure for retention of the heterologous genes (e.g. antibiotic selection), their loss generally occurs rapidly. Although effective large-scale fermentation strategies for unmodified probiotic strains are well established, similar large-scale fermentation techniques for engineered bacteria without antibiotic selection still need to be fully developed. Given the demonstrated potential of engineered therapeutic bacteria to treat a variety of diseases, finding effective solutions to break down the barriers to clinical studies is critical.

KEY POINTS.

• Bacterial homing to tumors has been exploited to deliver therapeutics in mice models.

• Engineered bacteria show promise in mouse models of metabolic diseases.

• Few engineered bacterial treatments have advanced to clinical studies.