Engineering microbial diagnostics and therapeutics with smart control

Abstract

Microbes have become an increasingly powerful chassis for developing diagnostic and therapeutic technologies. While many of the earlier engineering efforts used microbes that expressed relevant proteins constitutively, more microbes are being engineered to express them with region-selectivity and disease-responsiveness through biosensors. Such “smart” microbes have been developed to diagnose and treat a wide range of disorders and diseases, including bacterial infections, cancers, inflammatory disorders, and metabolic disorders. In this review, we discuss synthetic biology technologies that have been applied to engineer microbes for biomedical applications, focusing on recent reports that demonstrate microbial sensing by using animal models or clinical samples. Advances in synthetic biology will enable engineered microbes to significantly improve the medical field.

Introduction

The microbiome is a topic of increasing interest, with microbes linked to many health outcomes [1-4]. There is an immense opportunity for these microbes to be engineered to diagnose and treat a wide range of health issues. Early works developed therapeutic microbes for diseases, including cancer, bacterial infections, and colitis [5-8]. A common theme of these foundational works is constitutive expression of therapeutic biomolecules. Many microbes are still being engineered to constitutively express therapeutic agents that efficiently treat disease. One recent microbe was engineered to constitutively express an enzyme that converts metabolites obtained from cruciferous vegetables into an anti-cancer compound [9]. This strain reduced the growth of colorectal carcinoma in a murine model. A second microbe was engineered to constitutively express and secrete self-assembling nanofibers that promote mucosal healing [10]. This strain ameliorated inflammation in a dextran sodium sulfate (DSS) model of murine colitis.

To improve engineered therapeutic microbes, it is increasingly common to limit expression of the therapeutic agents to precise locations and times by incorporating region- and disease-responsive biosensors. Many existing sensors, with inputs including various chemical and environmental signals, have the potential in microbial engineering for medical applications. Controlling the expression of therapeutics with biosensors improves the safety and genetic stability of these microbes by minimizing potential off-target effects of the therapeutics and by reducing metabolic burdens on the microbes, respectively. In addition, biosensing allows microbes to be engineered as diagnostics, further enhancing the potential of engineered microbes.

In this review, we provide a snapshot of the state-of-the-art microbes engineered with biosensors for biomedical applications. Specifically, we focus on recent work that has successfully demonstrated microbial biosensing by using animal models or clinical samples. Other recent reviews have covered complementary topics of microbiome and microbial engineering for medical applications [11-16].

“Smart” diagnostic and therapeutic microbes

Bacterial infections

Bacterial infections can be diagnosed and treated with microbes by utilizing sensors derived from the pathogens of interest (Table 1A and Figure 1A). For example, the signaling domain of the cholera autoinducer 1-senstive two-component system (TCS) from Vibrio cholerae was fused to the signal transduction domain of a Lactococcus lactis TCS to create a novel cholera sensor in L. lactis [17]. By linking the sensor to a colorimetric reaction-catalyzing enzyme, cholera infections were detected in mice after isolating the engineered L. lactis strain from fecal samples. Interestingly, L. lactis cells containing the sensor construct were less effective cholera prophylactics than a non-engineered wild type strain, indicating a possible metabolic “cost” of the sensor. In a therapeutic effort, the Pseudomonas aeruginosa quorum sensing system was transplanted to E. coli Nissle 1917 (EcN) [18]. This quorum sensing module controlled the expression of an anti-biofilm hydrolase, a Pseudomonas-specific lytic protein, and a self-lytic protein. The engineered EcN strain was able to treat P. aeruginosa infection in C. elegans and to treat and prevent infection in mice. In another therapeutic effort, only antibiotic-resistant V. cholerae was killed in zebrafish and Artemia salina models through conjugation of a plasmid expressing an intein-split toxin and its cognate antitoxin [19]. Following conjugation, a pathogenicity-associated transcription factor activated expression of the toxin, and an antibiotic resistance-associated transcription factor repressed expression of the antitoxin. Thus, this system functioned as an AND logic gate, and cell death occurred only in pathogenic and antibiotic-resistant V. cholerae cells. While an interesting technology, this strategy will have to overcome low conjugation efficiencies to make a significant therapeutic impact. The efforts described in this section establish the utility of incorporating infection-specific biosensors in microbial diagnostics and therapeutics.

Table 1.

Recent “smart” microbial diagnostics and therapeutics validated in vivo or with human clinical samples.

Diagnostic or Therapeutic	Disease	Model System	Strain	Sensor Input	Sensor Output	Ref.
A) Infection
Diagnostic	V. cholerae infection	Mouse	L. lactis	Cholera autoinducer 1	Colorimetric enzymatic reaction	[17]
Therapeutic	P. aeruginosa infection	C. elegans; mouse	EcN	P. aeruginosa 3OC12- HSL	Lytic protein; anti- biofilm hydrolase	[18]
Therapeutic	Antibiotic resistant V. cholerae infection	Zebrafish; A. salina	E. coli β3194	Transcription factors associated with antibiotic resistance and pathogenicity†	Cellular toxin	[19]
B) Cancer
Therapeutic	Melanoma; colorectal cancer	Mouse	S. typhimurium	Arabinose	Immune cell-recruiting proteins	[20]
Therapeutic	Liver cancer	Mouse	S. typhimurium	3OC6-HSL‡	Autolytic protein; lytic protein*	[21]
Therapeutic	Liver cancer	Mouse	E. coli NiCo21 (DE3)	3OC6-HSL‡	Autolytic protein*	[22]
Therapeutic	Liver cancer	Mouse	EcN	3OC6-HSL‡	Autolytic protein*	[23]
C) Inflammatory Disorders
Diagnostic	Inflammation	Human clinical samples	E. coli DH5αZ1	Nitrogen oxides	Fluorescence	[26]
Diagnostic	Inflammation	Mouse	E. coli NGF-1	Tetrathionate	Colorimetric enzymatic reaction	[27]
Diagnostic	Inflammation	Mouse	EcN	Tetrathionate; thiosulfate	Fluorescence	[28]
D) Metabolic Disorders
Diagnostic	Diabetes	Human clinical samples	E.coli DH5αZ1	Glucose	Fluorescence	[26]
Therapeutic	Phenylketonuria	Mouse; primate	EcN	Low oxygen	Phenylalanine transporter and degrading enzyme	[29]
Therapeutic	Hyperammonemia	Mouse	EcN	Low oxygen	Ammonium-utilizing enzyme	[30]
E) Other
Diagnostic	Intestinal bleeding	Pig	EcN	Heme	Luminescence (detected by microelectronics)	[31]
Diagnostic	Fever	Mouse	E. coli NEB10β	High temperature	Fluorescence	[32]

Abbreviations: 3OC6-HSL, 3-oxohexanoyl-homoserine lactone; 3OC12-HSL, 3-oxododecanoyl-homoserine lactone; EcN, E. coli Nissle 1917; GM-CSF, granular-macrophage colony-stimulating factor.

^†A plasmid containing a toxin-antitoxin system is transferred to V. cholerae via conjugation; if that stain is antibiotic resistant and pathogenic (two inputs for the AND logic gate), then a toxin is expressed and its cognate antitoxin is not expressed from the conjugated plasmid.

^‡3OC6-HSL is produced using positive feedback in the therapeutic strain as a quorum sensing molecule.

^*The therapeutic component of the engineered strain is constitutively expressed.

Figure 1. Microbes have been engineered to diagnose and treat many diseases and conditions.

A) Microbes have been engineered to sense pathogen-associated chemical signals. After recognizing these signals, the engineered microbes can report on the presence of the pathogens or produce proteins that induce lysis of the pathogens. B) Tumors create microenvironments with low oxygen content and low pH. Various microbes can natively sense and colonize these microenvironments with high selectivity. Colonization in this microenvironment is accompanied by the production of proteins that recruit tumor-killing immune cells, directly inhibit tumor growth, or cooperate with systemically administered anti-cancer drugs to kill the tumor cells. C) Toxins and microbes are able to pass through leaky epithelial barriers in the intestines and enter the blood stream. This passage leads to the recruitment of a variety of immune cells that induce inflammation. Microbes can be engineered to sense the gut microenvironment and inflammation-specific metabolites to diagnose this inflammatory response. These microbes can also be engineered to produce cytokines that inhibit the inflammatory response of the immune system. D) Metabolites are transported across the epithelial barrier between the blood stream and gut lumen. As such, engineered microbes in the gut can sense and interact with dysregulated metabolites. In addition, these microbes can be engineered to sense environmental signals in the gut microenvironment and to respond by producing a variety of enzymes that can help maintain healthy levels of specific metabolites.

Cancer

Scientists have equipped natively tumor-colonizing bacteria with biosensors to create efficacious cancer therapeutics (Table 1B and Figure 1B). For example, attenuated Salmonella typhimurium (aSt) was engineered to express a heterologous flagellin in the presence of arabinose [20]. Upon mouse tumor colonization following intravenous injection and arabinose supplementation, the engineered aSt strain increased immune cell recruitment in tumors and improved mouse survival rate as compared to the wild type strain. In another work, a positive-feedback quorum sensing system was linked to the expression of cancer therapeutics and an autolytic protein in aSt [21]. Upon reaching the threshold population level in a colonized solid tumor, most engineered microbes lysed to release their therapeutic payload, while the non-lysing minority seeded a new generation of therapeutic bacteria. Tumor-bearing mouse survival was substantially increased by orally dosing the mice with these therapeutic bacteria. In follow-up efforts, similar quorum sensing-lysis circuits were incorporated into E. coli platforms constitutively expressing anti-CD47 nanobodies [22], or anti-PD-L1 nanobodies, anti-CTLA-4 nanobodies, and granulocyte-macrophage colony-stimulating factor (GM-CSF) [23]. The survival of mice bearing tumors was also increased upon injecting tumors with these bacterial strains. Together, these results demonstrate that engineered microbes that deliver immunotherapeutic agents directly at the tumor site are effective for cancer therapies, while tumor targeting can be further improved by utilizing surface-displayed proteins that enable tumor binding [24,25].

Inflammatory disorders

Researchers have engineered microbes to sense and report on inflammatory markers in clinical samples and in vivo (Table 1C and Figure 1C). For example, sensors for nitrogen oxides (inflammation biomarkers) were linked to memory switches in E. coli to detect serum nitrogen oxides in human clinical samples [26]. Other efforts focused on detecting gut inflammation in vivo. In one case, a tetrathionate sensor controlled a genetic memory switch in a strain of E. coli with a long residence time in the gut [27]. By using an enzyme-driven colorimetric reaction, gut inflammation of mice could be diagnosed by engineered microbes for 6 months after bacterial administration. In another report, both an alternative tetrathionate sensor and a thiosulfate sensor were linked directly to green fluorescent protein (GFP) in EcN whose expression was measured by using a flow cytometer [28]. These strains were able to detect thiosulfate, but not tetrathionate, in the gut of mice with DSS-induced colitis.

Metabolic disorders

“Smart” microbes can aid in the diagnosis or treatment of a metabolic disorder through recognition or consumption of an overabundant metabolite (Table 1D and Figure 1D). For example, a glucose biosensor was linked to E. coli-based memory switches to report on the presence of glucose in human diabetics’ urine samples via a fluorescent reporter [26]. More recently, a low-oxygen sensor was used to express a phenylalanine transporter and phenylalanine ammonia lyase [29] or N-acetylglutamate synthase [30], to treat phenylketonuria in primate and mouse models and hyperammonemia in a mouse model, respectively. While not directly related to the disease state, the low-oxygen sensor limited enzyme production until the engineered microbe entered the microaerobic gastrointestinal tract, allowing for region-specific responses.

Other conditions

Microbes have also been engineered to diagnose other conditions (Table 1E). In one instance, EcN was engineered to express a luminescent reporter in response to heme to diagnose intestinal bleeding in a porcine model [31]. A unique quality of this diagnostic microbe was that no fecal processing was needed to analyze the results. Instead, prior to administration, the engineered microbes were loaded into an ingestible capsule with luminescent-detecting microelectronics. During passage through the gut, the bacteria sensed bleed-associated heme and produced a luminescent signal. The microelectronics detected the signal and transmitted it wirelessly to an external receiver. This is the first report of microbe-based real-time diagnostics.

Another strain of E. coli was engineered to sense temperature [32]. Protein thermosensors were evolved to respond to physiologically relevant temperature ranges. Microbes containing an evolved thermosensor variant were used as a diagnostic for mouse fever. In response to a housing temperature increase, the engineered microbes produced a GFP signal detected by in vivo fluorescence imaging. The same temperature sensor was also used to demonstrate localized fluorescent protein expression in mice in response to targeted temperature increases from focused ultrasound treatments. This concept of precise spatial targeting has many potential therapeutic applications where systemic administration is undesirable.

Conclusions and future perspectives

Microbes have been individually engineered to diagnose and treat a variety of diseases and disorders, including bacterial infections, cancers, inflammatory disorders, and metabolic disorders, with biogeographical or disease-state selectivity. These accomplishments have provided a foundation for future developments. Alternatively, new technologies can be applied to collectively alter microbiomes in situ. One report of in situ microbial engineering showed that gut microbes from 19 genera and 4 phyla could receive plasmids via conjugation from an engineered E. coli strain [33]. Although the rates of gene transfer may need to be improved, this work provides an alternative method for generating diagnostic and therapeutic microbes.

Several technologies can improve or complement future microbial diagnostics. For example, genetic memory modules with long-term stability can be harnessed to develop diagnostic microbes with reliable and readily accessible readouts [34-36]. Alternatively, cell-free diagnostic systems can be implemented to complement microbial diagnostics, providing highly sensitive, reproducible, and selective diagnostic methods [37,38]. For example, a paper-based diagnostic platform was created by controlling the expression of a fluorescent reporter with toehold switches [37]. In this platform, the target RNA acts as a trans-acting trigger RNA that binds to the toehold switch, thus freeing the ribosome binding site and allowing translation of the reporter. Using this technology, the authors were able to identify the presence of specific species and detect inflammation-associated RNAs and Clostridium difficile infections. A second system used a similar paper-based toehold switch to detect viral RNA in serum samples by employing a colorimetric reporter as an output [38].

Future microbial therapeutics can be improved through the use of multi-input genetic circuits in which at least one sensor responds to a disease-specific signal. Nearly all the therapeutic microbes described in this review rely solely on a single input, including a pathogen signaling molecule [18], an exogenous inducer [20], a self-produced quorum signal [21-23], or a biogeographical signal [29,30] to complete their therapeutic function. Importantly, most of the sensors employed so far respond to signals that are not directly related to the disease state. For example, the low-oxygen sensor [29,30] can activate the therapeutic protein in any microaerobic sites, in addition to the gastrointestinal tract. If the microbial population reaches the quorum threshold [21-23], the lysis can occur in any locations, releasing bacterial contents and therapeutics. The use of those sensors may lead to non-specific expression. Moving forward, therapeutic strains can incorporate sensors for disease-specific compounds and logic gates with multiple inputs (e.g. biogeography ‘AND’ disease state) to express the therapeutic protein. Integrating multiple biosensors into therapeutic microbes will reduce the possibility of false positive expression in complex body environments, allowing for highly specific control [39,40].

To develop the best possible engineered microbes, it will be crucial to understand their behavior in the complex in vivo environments [41,42]. For example, an oscillatory circuit was used to study the heterogeneity of genetically identical microbes in the gut [42]. This study showed that gut inflammation led to more growth heterogeneity than a noninflamed control, as seen by a bigger loss of oscillator synchronization. In addition, new technologies are being developed that will facilitate future studies on microbial dynamics. These technologies include novel sensors for gut compounds [43], recorders of horizontal gene transfer events [44], and reporters that allow for precise spatial resolution [45]. Furthermore, recent and future clinical trials will provide insights into the effective use of biosensors in autonomous microbes.

A variety of in vitro systems can mimic the complex gut environment. These systems, which can facilitate the growth of both mammalian and microbial cells, can be applied to testing and evaluating engineered microbes in complex gut-like environments prior to moving the organisms into animal models [46-48]. For example, a microfluidics-based system was developed to mimic the intestine by housing a layer of epithelial cells, a physiologically relevant oxygen concentration gradient, and a mixture of aerobic and anaerobic microbes [47]. Complex communities can also be simulated computationally, enabling high-throughput assessment of engineered microbes in a variety of disease-associated environments [49].

Maximizing the safety of engineered microbes is essential. One or multiple methods of viability control, including biocontainment genetic circuits [32,50], auxotrophic systems [29,30], and reliance on non-standard amino acids [51,52], should be incorporated into any microbe engineered for real world applications, in order to prevent its spread and persistence in the environment. However, these biocontainment systems can also impose a metabolic burden or reduce the functional residence time of the microbes in the body if they are loosely controlled. As such, tight viability control should be implemented by considering the differences between the body conditions and the surrounding environmental conditions when sensors are developed for biocontainment. Additionally, the inability of most microbes to permanently colonize the mature gut will require their repeated administration. However, this characteristic may be an advantage in terms of human safety because the long-term consequences of engineered colonizing microbes are poorly understood. Unlike abiotic therapeutics, live microbes have these unique challenges to overcome.

There are additional unique aspects to consider for the development of engineered microbes for medical applications. While microbes can be engineered for regional selectivity, they cannot be targeted at normally sterile sites, including the brain and bones. Furthermore, different microbes inherently possess distinct characteristics, including regional preferences, different residence times, metabolic pathways, and regulatory pathways that make them optimal hosts for specific applications. However, only a few microbes, such as the model species E. coli, currently have enough genetic tools available to be engineered as smart diagnostics and therapeutics. Expanding the toolbox for new microbes will greatly improve the outlook of smart engineered microbes. For example, new sensors are being developed for non-model organisms to reduce this barrier [53,54]. Microbes are also able to evolve, creating potential issues with genetic instability. Methods for enhancing genetic stability should be developed to reduce the probability of loss-of-function mutations.

Significant progress has already been made in engineering microbes for gut-centric medical applications. However, as the field continues to progress, microbes can be engineered to diagnose and treat diseases pertaining to other microbiomes, including the lung, oral, skin, and vaginal microbiomes. To make these advances, microbes will require precise location sensing and programmable therapeutic expression. In addition, genetic tools and sensors will need to be developed for microbes that are native to these expanded target locations. Through these advances, engineered microbes have the potential to make vast impacts in medicine, with many new potential applications on the horizon.