Engineering Phages to Fight Multidrug-Resistant Bacteria

Abstract

Facing the global “superbug” crisis due to the emergence and selection for antibiotic resistance, phages are among the most promising solutions. Fighting multidrug-resistant bacteria requires precise diagnosis of bacterial pathogens and specific cell-killing. Phages have several potential advantages over conventional antibacterial agents such as host specificity, self-amplification, easy production, low toxicity as well as biofilm degradation. However, the narrow host range, uncharacterized properties, as well as potential risks from exponential replication and evolution of natural phages, currently limit their applications. Engineering phages can not only enhance the host bacteria range and improve phage efficacy, but also confer new functions. This review first summarizes major phage engineering techniques including both chemical modification and genetic engineering. Subsequent sections discuss the applications of engineered phages for bacterial pathogen detection and ablation through interdisciplinary approaches of synthetic biology and nanotechnology. We discuss future directions and persistent challenges in the ongoing exploration of phage engineering for pathogen control.

1. Introduction

The global threat of multidrug-resistant (MDR) bacteria is becoming increasingly serious, resulting in significant growth in morbidity, mortality, and medical expenses. Recent analysis reports that approximately 8 million people worldwide died from 33 common bacterial infections in 2019 alone,¹ and bacterial infections have become the second leading cause of death in the world after ischemic heart disease. However, currently, the antibiotic pipeline is dwindling since many pharmaceutical companies stopped antibiotic discovery,²⁻⁴ making the development of new antibacterial drugs lag substantially behind the emergence of MDR bacteria.

Bacteriophages (phages) are among the most abundant biological entities on this planet.^5,6 These viruses specifically infect bacteria, and were discovered independently by Frederick Twort in 1915 and Felix dHerelle in 1917.⁷⁻⁹ Phage therapy, in which phages are used to treat bacterial infections in humans or animals, was applied extensively, especially in Georgia and Poland, in the 1930s. However, phage therapy fell out of favor in the West when penicillin was brought to the market in the early 1940s, although it continued in Eastern Europe and the former Soviet Union after World War II.^10,11 In the 21st century, the growing MDR crisis has revived phage therapy research. Clinical trials of phage therapy were launched in the European Union and the United States in 2013 and 2016, respectively, initially verifying the safety of phage treatment of drug-resistant bacterial infections.¹² In recent years, additional reports have indicated the value of phage therapy in the treatment of infections such as sepsis,¹³⁻¹⁵ urinary tract infection,¹⁶⁻¹⁸ osteomyelitis,¹⁹⁻²¹ and pneumonia.²²⁻²⁴ Clinical data on phage therapy have been recently reviewed elsewhere.²⁵ In general, the production of phages is potentially simple, rapid and cheap due to the self-replicating property of phages, although this possible advantage also depends on the bacterial host species. Phages are now being considered in various fields including the food industry,²⁶⁻²⁸ pathogen detection,²⁹⁻³¹ human bacterial infection treatment,^32,33 veterinary medicine,³⁴⁻³⁶ and plant disease control.³⁷⁻³⁹ In this review, we focus on developments in phage engineering for biomedical applications.

Unlike antibiotics, phages typically have very high specificity for a particular group of bacterial strains, often based on the interaction between the bacterial receptor and the phage receptor-binding protein, and therefore phages are envisioned as being capable of avoiding off-target effects in a microbiome context. However, for many important applications such as acute MDR bacterial infections, a narrow host range is problematic owing to the need to identify and develop a large number of phages for clinical use. Other present obstacles to phage therapy using naturally occurring, wild-type phages are potential undesired interactions, such as with the human immune system or with bacterial biofilms,^40,41 release of bacterial toxins (e.g., endotoxins⁴²), and genetic transduction. Much like antibiotics, it is also clear that phages have coevolved with their bacterial hosts, which continuously evolve defenses against phage infection (e.g., mutations in the receptor). While the persistence of this host–pathogen coevolution over billions of years might give one confidence that phages can be evolved to overcome bacterial resistance to them,^43,44 it cannot be taken for granted that the time scale of such evolution is compatible with human applications. In other words, generally speaking, phage therapy is fundamentally plagued by the fact that phages have been optimized through natural selection for their own survival and propagation, not for their biomedical utility. Therefore, to realize the promise of phage therapy and similar applications, phage engineering is required, encompassing genetic, chemical, and material techniques. The rapid development of next-generation biotechnologies and nanotechnology for engineering phages, combined with the rapidly expanding public databases of phage genomes and structures, may overcome the limitations of natural phages. This review covers recent progress, challenges, and opportunities in phage engineering techniques and the applications of engineered phages, with a focus on antibacterial strategies.

2. Engineering Phages by Chemical Modification

The abundant amino acids in phage capsids provide a variety of reactive functional groups such as amines,⁴⁵⁻⁵¹ thiols,⁵²⁻⁵⁴ carboxylic acids and phenols for bioconjugation (Figure 1).⁵⁵⁻⁵⁸ The local environments of these moieties are influenced by several factors, including steric hindrance from the surrounding protein structures and solution pH, which influence reactivity for chemical modifications. Chemical modification strategies, while powerful, often lack specificity as multiple residues, or even multiple types of residues, may react under the same conditions. This may lead to unpredictable side reactions or inconsistency among batches, and may affect the morphology, material properties, and possibly the biological activity of the phage.⁵⁹ However, phage capsids may also tolerate nonspecific modifications, as discussed in multiple examples below, and therefore the suitability of an approach may be best evaluated empirically. If higher specificity is required, low abundance amino acids, such as cysteine, might be utilized, possibly in conjunction with mutation of existing residues, or unnatural amino acids might be introduced. There is already a considerable literature on protein bioconjugation;⁶⁰ here we discuss the literature on phage bioconjugation.

Figure 1.

Representative chemical reactions for phage chemical modification.

2.1. Primary Amine Group Modification

Chemical modification of free amine groups (lysine residues and protein N-termini) on the phage capsid is a widely utilized strategy for bioconjugation to the phage surface. N-hydroxysuccinimide (NHS) esters,⁶¹⁻⁶³ isothiocyanates,^64,65 isocyanates and acyl azides are reactive electrophiles most commonly used for amine conjugations.^66,67 Among these, NHS ester chemistry is the most popular strategy due to the mild reaction conditions (e.g., pH 7–9, close to physiological conditions, and room temperature stirring for 2 h or 4 °C overnight), straightforward purification procedures and stable products.⁶⁸⁻⁷⁰ For example, Chen et al. reacted the capsid of M13 phage with N-succinimidyl-S-acetylthiopropionate (SATP) to produce acyl-protected thiols, which were deprotected with hydroxylamine to generate free sulfhydryl groups on the phage surface.⁷¹ The phage thiolation enabled subsequent gold nanorod attachment along the side wall of M13, a rod-shaped phage. Related reagents to NHS esters include sulfonic acid-modified NHS esters (Sulfo-NHS esters), which show better water solubility and are also widely used for modification of various phages, including M13, T4, and MS2.^48,62 NHS esters including a polyethylene glycol (PEG) linker have been used for phage modification to improve biocompatibility.^47,72 In cases where hydrolysis of the NHS esters is problematic, tetrafluorophenyl ester, which has better stability in alkaline conditions, can be considered.⁷³ Additionally, urea, thiourea and isothiocyanates are also commonly used for bioconjugation with amines to chemically modify phages.⁶⁵

Although NHS ester chemistry is widely reported as a reliable strategy for phage capsid protein modification, a nonspecific approach like this may have unintended consequences, such as reducing phage infectivity. Specificity can be increased somewhat by modulating the pH of the reaction solution, based on differences in the pK_a of desired reactive groups (e.g., the N-terminus α-amine has a pK_a around 8, while lysine ε-amino groups have a pK_a around 10).^74,75 Another parameter that can be adjusted is the reaction stoichiometry. An empirical kinetic model was developed by Jin et al. to predict the molar ratio of NHS ester reagent to target protein for achieving the desired degree of modification.⁴⁶ Interestingly, this research suggested that the maximum binding activity of the phage to streptavidin was achieved when the conjugation ratio of biotin to the phage major coat protein (pVIII) subunit was 0.03. In practice, the optimal pH and stoichiometry of reaction are often empirically determined for a specific conjugation scenario.

2.2. Carboxylic Acid Group Modification

Carbodiimide chemistry is widely used for bioconjugation of the solvent-accessible carboxylic acids (Glu, Asp, and protein C-termini) on the capsid proteins of phages. 1-ethyl-3-(-3-(dimethylamino)propyl) carbodiimide hydrochloride (EDC) and N′,N′-dicyclohexylcarbodiimide (DCC) are common carbodiimide reagents used to activate the carboxylic acid group as an electrophile. The activated O-acyl urea intermediates then react with primary amino groups to form stable amide bonds, leaving no chemical linker. For example, Hosseinidoust et al. used EDC chemistry to cross-link M13 phages to form antimicrobial phage microgels and hydrogels.^76,77 It is worth noting that the free amines and carboxylic acids on the capsid proteins may react to cause both intra- and interphage cross-linking when applying EDC chemistry. Therefore, if desired, phages can be cross-linked together by adding carbodiimide compounds into phage solutions at high concentrations. By the same token, it is important to adjust the concentration to reduce undesired cross-linking side reactions. Chen et al. utilized EDC chemistry together with cysteamine to modify the carboxylic acids to present thiols on the M13 capsid.^78,79 In M13, each major capsid protein (pVIII) contains at least 3 solvent-accessible carboxylic acid groups (Glu2, Asp4, and Asp5), with around 2700 copies of pVIII per virion. In this case, EDC chemistry modification led to around 1800 thiol groups per virion. The relatively low modification efficiency (∼22%) could be attributed to the dilute phage solution and EDC hydrolysis. To improve efficiency, NHS can be added to the coupling reaction to form esters that are more stable than the O-acyl urea intermediates. The NHS ester intermediates can react with free amines under mild conditions. This strategy is well documented for phage modification to label with fluorophores,⁸⁰⁻⁸³ attach nanomaterials,^47,84,85 immobilize on surfaces⁸⁶⁻⁹⁰ and load drugs.⁹¹⁻⁹³ Since EDC coupling is not a site-specific chemical modification strategy, any solvent-accessible carboxyl groups on the capsid protein can be modified. Thus, as with amine modification, this method may cause alterations of phage properties, depending on the specific reaction. Nevertheless, in one case, the binding affinities of the phages changed only slightly after bioconjugation,⁷⁸ probably influenced by a lack of solvent-accessible reactive sites in the receptor-binding protein.

2.3. Thiol Group Modification

The most commonly used strategy for thiol modification (amino acid Cys) is “click” chemistry with maleimides, which can react with thiols under mild conditions to generate stable thioether products.⁹⁴⁻⁹⁷ Although amines can also react maleimide, this reaction generally does not compete with the thiol, being roughly 1000 times slower at physiological pH.⁹⁸⁻¹⁰⁰ Compared to amine and carboxylic acid modifications, thiol modifications can be more specific. Free thiols are usually not abundant in phage capsid proteins, as Cys is a generally uncommon amino acid and tends to be oxidized into disulfide bonds. Although disulfide bonds can be broken into free thiols by reducing agents such as dithiothreitol or glutathione,^101−103 this procedure may cause the phages to lose some functions.^104,105 If desired, free thiols may be added by converting the amines or carboxylic acids into sulfhydryl groups (e.g., by SATP or EDC/cysteamine coupling,^71,78 respectively). Using maleimide–thiol chemistry, fluorescence probes,⁵⁷ drugs,¹⁰⁶ and nanoparticles^71,107 have been conjugated to a variety of phages. Furthermore, sulfhydryl groups are widely reported to form dative bonds with metals such as gold.¹⁰⁸ Therefore, hybrid phage-metal nanomaterials can be constructed by modifying the thiol groups, whether wild-type or chemically introduced, on the capsid proteins.

2.4. Chemical Modification of Phages Using Other Functional Groups

In addition to the widely used functional groups mentioned above, phage tyrosine residues can be utilized by reaction with diazoniums to generate diazo conjugates (Table 1).^66,100 Alternatively, tyrosine can be activated by laccase, an oxidoreductase enzyme, to generate a free radical species and react with acrylates.^109,110 Although the reaction conditions (e.g., pH) may be tuned to influence selection of the desired residues, these strategies may still cause nonspecific modification.¹¹¹

Table 1. Chemical Modification of Phage Capsids.

strategies	applications	ref
Chemical Modification Using Natural Amino Acid Residues
NHS and sulfo-NHS ester	labeling dyes	(62,142−144)
	conjugating folic acid	(145)
	immobilizing to gold nanorods	(71,146)
	conjugating biotin	(147−150)
TFP ester	labeling fluorophores	(151)
isothiocyanate	conjugating dyes	(65,91)
EDC coupling	introducing abundant thiol groups	(78,79)
	drug conjugation	(92,152)
	labeling photosensitizers	(80,82,153)
	conjugating AIEgents	(154)
	immobilization phages on surfaces	(86,155,156)
	conjugating DNA molecules	(157)
nucleophilic reaction between benzyl bromide or iodine and sulfhydryl groups	labeling AIEgents	(158,159)
diazonium coupling with tyrosine	interior surface modification of MS2 phage	(66)
laccase inducing free radical formation on the tyrosine	preparing M13-polymer hybrids	(110)
thiol–maleimide “click” chemistry	internal cross-linking of Qβ phage particles	(160)
PLP transamination	conjugation of PEG, dye	(45)
	conjugating contrast agent for molecular imaging	(51)
sodium periodate oxidation	conjugating of biotin and mannose	(161)
Chemical Modification Using Non-Natural Amino Acid Residues and Displayed Peptides
aortase-mediated capsid modification	incorporating proteins, peptides or reactive handles onto the pVIII protein of M13 phage	(112,114)
thiol–maleimide “click” chemistry	phage displaying thiol groups to conjugate maleimide labeled EGFR	(162)
displaying peptide containing selenocysteine	covalently attaching adenosine receptor ligands	(140)
displaying peptide with N-acryloyl-lysine, N-butyryl-lysine, N-crotonyl-lysine	constructing phage library for affinity selection, expansion of the chemical functionalities and diversities of phages	(54,117,139,163)
displaying peptide containing p-azidophenylalanine	labeling fluorescent dyes	(119)
displaying peptide containing (2,2′-bipyridin-5-yl)alanine	binding iron(III)	(164)

When desired, site-specific modifications can be achieved in combination with genetic engineering. In the AviTag system, a specific peptide sequence is inserted into the capsid protein, to be recognized and biotinylated by the enzyme BirA, allowing further site-specific functionalization via the biotin tag.¹¹² In the sortase system, a specific peptide tag sequence is inserted into the capsid protein and recognized by the sortase transpeptidase enzyme,^113,114 to be ligated to a provided peptide nucleophile. These approaches use naturally occurring amino acids in peptide tag sequences, and the specific recognition of the tag by the enzyme provides site-specificity.

Alternatively, another method to introduce unique chemical functionality through genetic engineering is encoding and introducing non-natural amino acids to the virions.^115−118 This approach extends the range of chemical modifications by incorporating functional groups such as ketones, azides and alkynes into the capsid proteins.^{115,117,119,120} Encoding of non-natural amino acids was originally pioneered by the Schultz lab through recoding of the amber codon to be identified by a tRNA charged with a non-natural amino acid.^121−124 There are now more than 200 non-natural amino acids that have been introduced into different proteins by this Genetic Code Expansion technology.^125,126 This technique has been used to introduce fluorophores,^127,128 drugs and create site-specific bioconjugation,^129−131 and has been applied to proteins expressed in mammalian cells, bacteria, and yeast as well as to phages.^132−139 To introduce a non-natural amino acid to a phage capsid protein, the bacterial host cells should contain a plasmid encoding the orthogonal tRNA, a cognate aminoacyl-tRNA synthetase to charge the non-natural amino acid, a phagemid encoding the mutant capsid protein, as well as the helper phage. The helper phage encodes proteins for genetic complementation of the phagemid. The non-natural amino acid is supplied in the cell culture. If amber suppression is successful, the progeny virions will contain a capsid protein displaying the non-natural amino acid. While technically challenging, insertion of non-natural amino acids into the phage capsid protein offers extensive opportunity for precise chemical modifications. Beech et al. chemically attached adenosine receptor ligands to phages via selenocysteine and showed that the modified virus can activate the adenosine A1 receptor.¹⁴⁰ Additionally, various non-natural amino acids have been inserted into phage capsid proteins for bioconjugation, capturing metal ions and switching functional ligands.^115,141 Although methods using genetic engineering are potentially powerful tools for phage modification, an important caveat common to these techniques is that the phage capsid protein must tolerate the genetic modifications, i.e., the mutations or insertions must not disrupt the structural integrity of the phage virion.

3. In Vivo Phage Engineering Techniques

Phage genetic engineering in vivo uses mutagenesis or homologous recombination to introduce mutations. Often, without an efficient selection strategy, a screening process is required to select the desired phage mutants.

3.1. Random Mutagenesis

Strategies like ultraviolet light^165,166 and chemical mutagens (e.g., alkylating agents^167,168 have been widely used for random mutagenesis). The desired phenotype is then selected from the obtained phage mutant pool to identify the corresponding gene and mutation, typically by sequencing. Heat-tolerant mutants of E. coli phages were successfully identified after random mutagenesis in this manner.¹⁶⁹ Although random mutagenesis is simple to apply, the obtained phage mutant pools are mixtures of unknown mutations, and phenotypic selection or screening is necessary. In addition, synergistic effects from multiple beneficial mutations may be difficult to discover and characterize.

3.2. Homologous Recombination

Homologous recombination is among the most well-documented approaches for phage genome engineering in bacterial hosts,^170−173 relying on a natural biological process in which sequences are exchanged between two DNA molecules with similar or identical sequences. This approach was utilized to cross phage genomes using related phages in the early history of phage research.^174,175 Two related phages with different phenotypic traits were used to coinfect a single host bacteria, enabling homologous recombination between the viral genomes, and leading to progeny phages with novel combinations of parental characteristics. This approach (phage cross) was employed to create chimeras of phage genes related to changes in host range, and to investigate the interplay between phages and newly discovered bacterial defense mechanisms.^176−178 However, the classic phage cross is restricted to recombination of pre-existing phage genomes. Facing this limitation, donor plasmids have been deployed to allow gene insertions, replacements, or deletions in the phage genome. In this approach, the desired gene is initially cloned into a replicative plasmid, flanked by regions of homology with the phage genome, which determine the specific location of gene integration. The donor plasmid is then introduced into a bacterial host, which is subsequently infected by the phage to be engineered (Figure 2). Homologous recombination of the plasmid and the phage genome leads to the integration of the foreign gene into the phage genome.

Figure 2.

Engineering phages with homologous recombination. Phage infects the host bacterium and injects the genome DNA into the host cell (A). The phage DNA recombines with the plasmid DNA which is partially (in green) homologous to the phage genome (B) and generates recombinant phage genomes (C). The genomes are packaged in recombinant phage particles (D) that are subsequently released (E).

As with random mutagenesis, low recombination efficiency usually necessitates an efficient screening method to identify the desired clone. To facilitate screening, phage-specific marker genes or reporter genes are often incorporated to allow specific selection for recombinant phage clones.^179−181 CRISPR-Cas systems have also been deployed to selectively eliminate the wild-type phage genomes,^182−185 thereby enriching the abundance of recombinants. Type I-E, II-A, and III-A CRISPR systems have been utilized to isolate recombinant phages that target both Gram-negative and Gram-positive hosts. Low recombination efficiency also limits the technique when multiple loci are being targeted. Since it is rare to achieve a clone carrying all the desired recombination events, such designs are usually created by sequential targeting of different loci.

3.3. Recombineering

Recombination efficiency can be very low (10^–4 to 10^–10),^186,187 especially for strictly lytic phages. Efficiency can be enhanced by utilizing the recombination systems of temperate phages, a process known as recombineering (homologous recombination-mediated genetic engineering). By expressing phage recombination proteins such as λ Red or the Rac prophage RecE/RecT proteins within the recombination host (e.g., E. coli), introduced DNA is processed into a single-stranded form that readily anneals with the phage genome (Figure 3). In particular, λ Red is highly effective in facilitating the recombination of linear DNA into phage λ, resulting in targeted modifications to the viral genome. When used in recombineering, these systems can reduce the homology arm length requirement in addition to boosting recombination frequency. In one case, Court et al.¹⁸⁸ engineered E. coli cells to carry a defective λ prophage as well as the pL operon, which mediates general and site-specific recombination and is controlled by a temperature-sensitive repressor. These cells were then infected with the phage to be modified. The functions of phage λ Red were activated by elevated temperature, leading to a relatively higher yield of recombinant phages compared with that through homologous recombination. However, the efficiency is still low and screening is important for this strategy.

Figure 3.

Engineering phages with recombineering. Phage infects the host bacterium, injects the genome into the host cell and foreign DNA is transformed into the cell (A). The phage genome recombines with the foreign DNA which can partially (in green) homology to the phage genome (B) and generates recombinant phage genomes (C). The genomes are packaged in recombinant phage particles (D) that are subsequently released (E).

3.4. Bacteriophage Recombineering of Electroporated DNA

While most recombineering is performed in E. coli or other Gram-negative strains, the discovery of RecE/RecT homologues in mycobacteriophage Che9c expanded the scope of these techniques. In the bacteriophage recombineering of electroporated DNA (BRED) technique, the viral genome and templates are electroporated into the recombineering-proficient host cells (Figure 4).¹⁸⁹ These cells carry a plasmid that encodes proteins including gp60 and gp61 of Che9c. The dsDNA substrate utilized in BRED consists of the DNA fragment for insertion, as well as homologous regions that flank the targeted area both upstream and downstream within the phage genome. Hatfull et al. initially utilized this technique to modify mycobacteriophages.¹⁹⁰ This technique resulted in a relatively high rate of recombinants, with >10% of plaques containing the desired mutant. The high rate enabled identification of mutants using a limited number of PCR reactions, without requiring an additional screening or selection step. BRED has been applied to phages that target bacterial hosts other than mycobacteria, such as E. coli and Salmonella enterica. While BRED offers a valuable tool for genetic engineering of phage genomes, the technique relies on high transformation efficiency due to the coelectroporation of the phage genome and editing template, and transformation efficiencies for large phage genomes are relatively low. One approach to overcome this limitation involves initial introduction of the editing template through electroporation, followed by infection of the bacteria with the phage. This method is known as bacteriophage recombineering with infectious particles (BRIP).¹⁹¹ In summary, compared to random mutagenesis, recombineering techniques allow specific phage genomes editing with relatively higher efficiency, and could be extended to a range of bacterial species.

Figure 4.

Engineering phages with BRED. Phage genome DNA is extracted (A) and coelectroporated into cells with recombineering DNA substrates (B). The recombination between the homologous parts (in black) yields recombinant genome DNA (C), which are packaged in recombinant phage particles (D) that are subsequently released (E).

3.5. Yeast-Based Phage Genome Engineering

Genetic engineering of phages in vivo in bacteria can be problematic due to detrimental effects of the phage and phage proteins on the host cells. Alternatively, brewer’s yeast (Saccharomyces cerevisiae) can be utilized as a host, using a yeast artificial chromosome (YAC) for recombination and gene cloning. In this approach, the phage genome is inserted into an S. cerevisiae-bacterial shuttle vector which carries homologous overhangs matching the ends of the phage genome, to enable recombination between the vector and the phage genome. The phage genomes are genetically modified and propagated in the yeast and then extracted and introduced into bacteria to produce engineered phage particles (Figure 5). Utilizing this method, Lu et al. manipulated phage host ranges through the modification of phage genomes within S. cerevisiae.¹⁸⁵E. coli phage structures were redirected to target pathogenic Yersinia and Klebsiella bacteria, while Klebsiella phage structures were redirected to capture E. coli, through the replacement of tail fiber proteins. These engineered phages effectively eliminated the new target cells and were utilized to specifically eradicate pathogens from diverse bacterial communities. Other E. coli phages, T7 and ΦX174, were also engineered through this technique.^185,192 A limitation of this strategy is the laborious process of extracting phage genomes from the yeast and introducing them to the bacteria host, which usually occurs with low efficiencies.

Figure 5.

Yeast-based assembly of phage genomes. Phage genome DNA is extracted (A) and coelectroporated into yeast cells with YAC molecules (B). The recombination between the homologous parts (in green) yields recombinant phage genome DNA (C), which is extracted (D) and in coelectroporated into bacterial cells (E). The resulting recombinant phage particles (F) are subsequently released (G).

4. Synthetic Phage Genomes In Vitro

Phage engineering strategies in vivo are generally limited to genetically tractable hosts and may have diminished efficiency for phages that negatively affect the cell. On the other hand, genetic engineering can be performed in vitro, with the DNA introduced into the host cell in the last stages. A classic method is restriction enzyme cloning, although assembly methods are increasingly used nowadays. For example, the phage genome is amplified in overlapping fragments by PCR, incorporating tailored mutations using synthetic oligonucleotides or genes as desired. These fragments are subsequently assembled in yeast cells or by homologous recombination. The synthetic genome is then introduced into the host environment, such as bacterial host species, L-form bacteria, or a cell-free transcription-translation (TXTL) platform, to initiate engineered phage production (Figure 6). The in vitro approach can be advantageous as it operates independently of the host bacteria, avoiding possible phage-host interactions and enabling a high level of customization and efficiency. For example, to maximize the potential space for genetic engineering, gene assembly allows phage genomes to be redesigned to avoid overlapping genes (“refactoring”). Endy et al. restructured the T7 phage genome to eliminate gene overlaps, ensuring that modifications made to the DNA within each gene did not impact the others.¹⁹³ Seventy-three genetic segments were identified and then organized into six distinct sections, bracketed by restriction sites. Molecular cloning was employed to construct the altered genomes in vitro in the individual sections, and three chimeric T7 phage genomes, each consisting of different engineered sections, were created. These refactored genomes were transformed into the bacterial host to generate viable phages. However, the recombinant phages exhibited significantly smaller plaques compared to the wild-type phage, indicating that refactoring of the phage genome can cause negative impact to phage fitness. Nevertheless, the ability to assemble phage genomes in vitro enables a wide range of genetic modifications.

Figure 6.

Synthetic phage genome in vitro. The phage genome is extracted (A), subcloned into different pieces (B) and further manipulated with desired mutations (C, D). The recombinant DNA fragments are inserted into the phage genome (E, F) and electroporated into E. coli (G) or L-form bacterium (H), yielding recombinant phages (I).

4.1. Gene Assembly in Vitro

In vitro assembly of synthetic oligonucleotides to construct complete phage genomes is an increasingly deployed strategy for phage engineering. In a landmark study for synthetic biology, the phage genome of φX174 was assembled from chemically synthesized oligonucleotides.¹⁹⁴ The process involved three critical parameters: (1) the pooled oligonucleotides were gel-purified to remove erroneous sequences; (2) the oligonucleotides were annealed at relatively high temperature (55 °C) to avoid mismatched assembly; and (3) the phage genome assembly was performed using polymerase cycling assembly, in which each terminal oligonucleotide can only be extended once. Then PCR amplification was employed to acquire larger quantities of complete genomic sequences. The infectivity of the synthetic DNA was found to be lower compared to natural DNA, indicating about 1 lethal error per 500 base pairs, but regardless, fully functional φX174 virions were successfully produced following electroporation into E. coli (Figure 7). The synthetic challenge increases with length due to technical difficulties in manipulating large DNA molecules in vitro (e.g., low transformation efficiency) and the rate of mutations. However, synthetic DNA technology has improved rapidly, with multiple commercial services available to create gene-length synthetic fragments. Wholly synthetic gene assembly enables almost any genetic modification, and in comparison to intracellular recombination, in vitro gene assembly is generally more efficient¹⁹⁵ and avoids the process of phage screening, thus making phage genetic engineering more straightforward and convenient.

Figure 7.

Gene assembly in vitro. The synthetic oligonucleotides (A) are annealed, assembled, and ligated to get double strand phage genome DNA (A), which is circularized when appropriate (B) and electroporated into host cells (C). The phage genome DNA is packaged into phage particles (D) that is subsequently released (E).

4.2. Synthetic Phage Assembly in Different Hosts

Several bacterial species, such as E. coli and P. aeruginosa, have well-established transformation protocols that are highly efficient. However, many, if not most, bacterial species present significant challenges in transformation, creating a bottleneck in effectiveness and speed. To broaden phage engineering to different host species, Xiao et al. established a stepping-stone host-assisted phage engineering (SHAPE) framework to express synthetic phage while minimizing use of the host organism.¹⁹⁶ In this framework, synthetic genomes are created either in vitro or in vivo using a modified E. coli strain carrying the plasmid pKD46, including the λ-Red recombination system. Expression of the phage in this “stepping-stone” E. coli host creates phage competent to infect the intended host species. The method was validated on 126 phage genomes from T7 and non-T7 families that infected multidrug resistant pathogens such as Klebsiella pneumoniae, Streptococcus enterica, Pseudomonas aeruginosa, and Acinetobacter baumannii. This work indicates that phage assembly and production can be performed in a model organism, with the host organism used only for steps requiring infection.

Phage genome transformation into Gram-positive bacteria is challenging due to the thick peptidoglycan layer. However, this obstacle can be overcome by utilizing L-form bacteria, which are wall-deficient cells that remain metabolically active and capable of cell division. L-form bacteria can be suitable for transforming large DNA molecules, including synthetic phage genomes of up to 154 kb. These cells are typically generated through extended culture in osmotically balanced media containing cell wall-activating compounds which disrupt the cell wall.¹⁹⁷ A variety of Gram-positive and Gram-negative bacteria can undergo a transition into the L-form state. Loessner et al. constructed custom synthetic phage genomes in vitro with smaller DNA fragments¹⁸⁷ and transformed the phage genomes into Listeria monocytogenes L-form cells. This “reboot” of the synthetic phage genomes through transfection resulted in production of phages beginning from synthetic DNA. Interestingly, L-form cells of Listeria facilitate the reactivation of both native and synthetic Listeria phage genomes. Furthermore, they also allow for cross-genus rebooting of Bacillus and Staphylococcus phages from DNA, expanding the scope to phages targeting other significant Gram-positive pathogens.

4.3. Rebooting Phage in a Cell-Free Environment

Another method to avoid transformation of nonmodel host organisms is to “reboot” the phage genomic DNA in a cell-free environment. In general, upon entering the host cell, phage DNA uses the bacterial transcriptional system to generate mRNA that is translated by the bacterial translation apparatus. In a cell-free system, given template DNA, cell extracts are usually used to provide the biomolecular components necessary for transcription and translation, and additional substrates (e.g., ATP, amino acids) are fed to the system (Figure 8).¹⁹⁸ For instance, Noireaux et al. successfully demonstrated the complete synthesis of the phage T7 by utilizing a cell-free TXTL system derived from E. coli BL21 Rosetta2.¹⁹⁹ After a few hours of incubation in a tube, billions of infectious T7 phages were generated in the reaction solution. Notably, genome DNA replication occurs simultaneously with the expression of viral genes, protein synthesis, and the assembly of the phage components. Several self-replicating phages have been produced by this approach including E. coli phage T4,²⁰⁰ T7,¹⁹⁹ ΦX174,²⁰¹ phiX174,²⁰² MS2,²⁰¹ and K1F.²⁰³ Cell-free systems from organisms other than E. coli have been reported,²⁰⁴ suggesting this technique could be applied in nonmodel hosts for phage engineering. TXTL systems can also be based on other bacterial species, including Bacillus megaterium, Pseudomonas putida, Clostridium autoethanogenum, and Vibrio natriegens, to produce corresponding phages or synthesize proteins.^205−210 While highly versatile, TXTL systems can be more expensive than bacterial production of phages and often requires optimization to ensure high yield.

Figure 8.

Rebooting phage in a cell-free environment. In a cell-free expression system, the phage genome DNA is transcribed into mRNA (A) and translated into protein (B), and DNA can replicate (C,D). Subsequently, the phage genomes are packaged into phage particles (E).

5. CRISPR-Cas Based Phage Engineering

CRISPR-Cas (Clustered Regularly Interspaced Short Palindromic Repeats, CRISPR-associated protein) systems are adaptive immunity systems ubiquitous in bacteria and archaea.^211−214 Small CRISPR RNAs (crRNA) guide Cas nucleases to bind to and create double-stranded breaks in DNA. These systems originally evolved to recognize and eliminate foreign nucleic acids of phage origin. Generally, the CRISPR-Cas immune mechanism consists of three distinct phases: spacer acquisition, crRNA biogenesis, and interference. In the spacer acquisition phase, a distinct, short, phage-derived sequence, known as a protospacer, is incorporated into the CRISPR loci, situated between partially palindromic DNA repeats with similar length. The cell thus “memorizes” fragments of phage sequences in a CRISPR array. During the crRNA biogenesis phase, the CRISPR array undergoes transcription to produce a long precursor crRNA, which is subsequently processed into mature crRNAs, each harboring a “memorized” sequence of the foreign DNA. In the interference stage, mature crRNAs and Cas proteins form effector complexes which specifically cleave newly invading nucleic acids that match the crRNA sequence. Thus, the bacterial cell retains information about prior phage infections, such that matching DNA is quickly degraded if encountered again. The CRISPR-Cas systems are categorized into two main classes based on phylogenetic and mechanistic differences.^215−217 Class 1 CRISPR-Cas systems (Types I, III and IV) encode multisubunit effector complexes. In contrast, Class 2 CRISPR-Cas systems (Types II, V and VI) use only a single subunit effector protein.

The CRISPR-Cas system can target the genomes in the wild-type phage and thus isolate the desired recombinant phages (Figure 9). Briefly, a bacterial cell harboring a donor DNA designed to edit a phage genome through homologous recombination is infected with the target phage. The resulting lysate contains both wild-type phages and a small fraction of recombinant phages. This lysate is then used to infect bacteria carrying a CRISPR-Cas system that specifically targets the unmodified phages, thereby selecting for the recombinant phages. This selection method overcomes the labor-intensive screening process for a recombinant phage in a mixed population and broadens CRISPR Cas-based editing to phages. At least five CRISPR-Cas systems (Types I, II, III, V and VI) have been utilized as tools for phage bioengineering (Figure 10).^182,218,219 The representative members of the tailed phage families such as Myoviridae, Siphoviridae, and Podoviridae were successfully engineered by applying the CRISPR-Cas systems. The specific defense mechanisms of each CRISPR–Cas Type, as well as the strategies used for phage genome editing will be discussed below.

Figure 9.

Utilizing CRISPR-Cas systems for phage genome editing. During the crRNA biogenesis (A), the repeat-spacer array is transcribed into crRNA (brown, blue, and green rectangles). During the interference (B), the crNRAs combine with the Cas proteins to form effector complexes, breaking down the phage genome DNA (C). Through homogeneous recombination, the DNA fragments acquire mutations in the desired genes from a donor DNA construct to form recombinant phage genome DNA (D,E), which is packaged and released subsequently (F).

Figure 10.

Representative CRISPR-Cas systems for phage genome editing, Type I-E from Escherichia coli,²²⁴ Type II-A from Streptococcus pyogenes,²²⁵ Type III-A from Staphylococcus epidermidis,²²⁶ Type V-A from Francisella novicida,²²⁷ Type VI-A from Leptotrichia shahii.²²⁸

5.1. Type I CRISPR-Cas Systems for Phage Genome Editing

The Type I-E system in E. coli is the best-illustrated model of the Type I CRISPR-Cas system.^220−222 In this system, Cas6 plays a crucial role in crRNA biogenesis by recognizing and cleaving the precursor crRNA of each repeat unit. Cas6 together with the crRNA, large type-specific subunit Cas8, two CasB subunits, six copies of Cas7 and Cas5, form the ribonucleoprotein complex Cascade. In this system, a protospacer adjacent motif (PAM) of 2 to 6 nucleotides is required for the interference. The helicase-nuclease Cas3 cleaves the DNA target when detected by the Cascade complex. Progeny phages with random mutations in the PAM may escape from the CRISPR immune system, thus escaping the selection and significantly increasing the background of the nonrecombinant phages. This system requires the participation and interaction of multiple genes and proteins, resulting in low efficiency and complicated operation procedures. Therefore, its application in phage engineering is yet limited. Nevertheless, Kiro et al. genetically engineered the E. coli phage T7 with the Type I-E CRISPR-Cas system, which led to moderate editing efficiency with 17 out of 44 plaques demonstrating the desired gene deletion.¹⁸² This editing efficiency is sufficient for most applications requiring the production of a specific recombinant plaque. Box et al. employed the Type I-E CRISPR-Cas system along with a homologous recombination template to engineer the Vibrio cholerae phage ICP1 which is a lytic phage of the Myoviridae family.²²³ In this study, a plasmid template carrying the targeting spacer as well as the donor DNA with desired mutations was introduced into a V. cholerae strain which contains the Type I-E CRISPR–Cas system in its genome. It turned out that multiple phage genome alterations occurred independently. A small deletion of 33 nucleotides containing exactly the protospacer region was identified in all the mutant phages examined. In addition, the larger gene deletion and gene replacement occurred at lower frequencies with efficiencies of 7/12 and 4/8, respectively, suggesting unintended complexities with the composition of the Type I-E CRISPR-Cas system.

5.2. Type II CRISPR-Cas Systems for Phage Genome Editing

Type II is the most utilized among the CRISPR-Cas systems. The endonuclease activities required for the interference with foreign DNA in this system rely on the single multidomain protein Cas9. The small enzymatic components and ability to easily program site-specific DNA targeting make Type II systems the most widely employed CRISPR system for phage genome editing. The small trans-activating crRNA or tracrRNA is critical for crRNA biogenesis. The tracrRNA encompasses a region complementary to the sequences derived from the repeats in the crRNA. The host-encoded nuclease RNase III cuts both the tracrRNA and precursor crRNA, which defines the 3′-end of the crRNA while the 5′-end is trimmed to form mature crRNAs. Cas9 can sense and cleave the target DNA as directed by the single guide RNA (sgRNA). A PAM of two to five nucleotides must lie at the 3′ end of the guide RNA to license interference and cleavage. The repair following the DNA cut may lead to random DNA insertion or deletion by nonhomologous end joining or a homologous DNA as a repair template by homology-directed repair.

Progeny phages in Type II CRISPR-Cas systems may acquire random point mutations in the PAM and increase results in decreased selection of the recombinant phages. The escaped phages with deletions of desired genes can be enriched by targeting the genes without a donor DNA. Compared with the Type I CRISPR-Cas systems, the Type II exhibits superior gene-targeting efficiency, affordability and ease of operation.^229,230 In the research of Martel et al.,²¹⁸ the researchers found that specific point mutations and significant deletions within the genome of the virulent phage 2972 can be engineered with the CRISPR-Cas Type II-A system of Streptococcus thermophilus, which serves as a selective pressure to enhance recombination efficiencies. This research employed four distinct isolates of S. thermophilus, each carrying a Type II-A CRISPR-Cas system that includes a natural spacer aiming at one of four phage genes with unidentified functions, to produce escape mutants. If the target gene is not essential for phage survival, then escaped phages with random deletions or nonsense mutations can be easily isolated. If, however, the gene is essential for phage survival, then it cannot be deleted. It turned out that every plaque examined harbored recombinant phages possessing the targeted mutation, indicating the CRISPR-Cas Type II system serves as an efficient mechanism for phage genomes engineering. Following this research, heterologous CRISPR-Cas Type II-A systems were introduced into diverse bacteria such as Escherichia coli,²³¹Klebsiella pneumoniae,²³²Bacillus subtilis,²³³Vibrio natriegens,²³⁴ and Listeria monocytogenes(235) for phage genome editing. Although widely used, a major obstacle of CRISPR-Cas9 systems is that phages can acquire resistance to it through different mechanisms, including escape mutations.

5.3. Type III CRISPR-Cas Systems for Phage Genome Editing

Type III systems are more complex compared with Type I and Type II systems. Different from Type I and II, the well-characterized Type III-A system found in Staphylococcus epidermidis eliminates the necessity for a PAM or seed sequence, thereby hindering the wild-type phages from escaping detection via single nucleotide substitutions.^236,237 This feature renders the Type III-A system advantageous in combating the development of escape mutants and enhancing the efficiency of phage genome editing. Furthermore, Type III systems demonstrate a dual functionality in DNA and RNA interference. Research indicated that the structure of Type III surveillance complexes bears remarkable resemblances to that of Type I surveillance complexes.^238−240 The Cas6 endonuclease which defines the 5′-ends of the crRNA is critical to the crRNA processing in the Type III-A system. In this system, the crRNA maturation on the 5′-ends is accomplished by further cleavage of non-Cas cellular nucleases. The resulting mature crRNAs are bound to the Cas10-Csm which is a multisubunit effector complex consisting of Cas10 (large type-specific subunit), Csm2 (in Staphylococcus epidermidis, small subtype-specific subunit), Cas5 (or Csm4) and multiple copies of Cas7 homologues (Csm3 and Csm5). Along with Csm6, this complex can recognize and cut the invading nucleic acids. In the Cas10-Csm system, at least three distinct nucleases are involved in the interference process triggered by the binding of the crRNA to the complementary RNA transcript. Among them, while Cas10 slices the nontemplate DNA strand, the Csm3 cuts the RNA in the protospacer region, and eventually the Csm6 steadily degrades the targeted transcripts.²⁴¹ Although a PAM is not required in the Type III system, the lack of complementarity between the tag, which is the eight nucleotides at the 5′ end of the crRNA, and the antitag, which is the opposing region next to the protospacer, is needed to allow the interference process.²⁴² Furthermore, in the Type III system, a few mismatches in the crRNA-protospacer pair do not cause much hindrance to the interference. It is worth noting that the prerequisite for the interference process in Type III systems is that the targeted locus should be actively transcribed. Therefore, the late genes in lysogenic phages are not engineerable by these systems, whereas genes in lytic phages are editable by these tools.²³⁶

The Type III-A system does not require a PAM or seed sequence, which helps enhance the phage engineering efficiency. Bari et al. employed the endogenous CRISPR-Cas10 system of Streptococcus pneumoniae for phage engineering.²⁴³ In this research, they introduced point mutations across various genetic loci in Podoviridae phage Andhra and Myoviridae phage ISP, conducting phage gene editing during the infection of the genetically modified bacteria. It demonstrated that the Type III-A system can serve as a specific counterselection mechanism for phage recombinants since the spacers targeting multiple gene loci in both lytic phages produced no phage escapers. After propagating the engineered strains with both the targeting spacer and donor DNA construct, all isolated phages showed the designed genetic modifications. Additionally, a plasmid with the complete Type III-A system was successfully incorporated into S. aureus, which was a CRISPR-deficient heterologous host, and worked efficiently.

5.4. Other CRISPR-Cas Systems for Phage Genome Editing

Recently, other types of CRISPR-Cas systems have been reported to be employed for genome editing of phages.^244−246 Among them, the Type V (CRISPR-Cas12a) system holds great potential in phage engineering.^247,248 The glucosyl hydroxymethyl cytosine (ghmC) genome of T4 phage shows strong resistance to most restriction endonucleases and different CRISPR-Cas systems. Dong et al. found that the type V CRISPR-Cas12a system can efficiently cut ghmC-modified genome to generate recombinant T4 phages,²⁴⁹ whereas the type II CRISPR-Cas9 system demonstrated no cleavage efficiency. The crRNA determines the efficiency of the Cas12a editing system. The double-stranded DNA breaks generated by the cleavage of the ghmC-modified genome facilitate the recombination between the phage DNA and the donor plasmid. The gene editing efficiency is as high as 100% in this research when the CRISPR-Cas12a system is employed.

Type VI (CRISPR-Cas13) system can be used for RNA-targeted modification and regulation.²⁵⁰ In the phage infection process, this system can significantly inactivate the host and thus effectively inhibit the escapers, which can be highly advantageous in phage engineering.^183,251 Adler et al. found the antiviral effect of LbuCas13a excellent and wide against various phages by testing it against E. coli phages of multiple phylogenetic groups.²⁴⁴ They deployed LbuCas13a for extensive phage editing by a two-step approach involving phage editing and sequence enrichment, and genome editing efficiency of 100% was achieved. It is worth noting that the targeting of Cas13a does not require the PAM, and in principle any position within or adjacent to the phage transcript is editable by Type VI system.

CRISPR-Cas systems in various organisms can be utilized to engineer phages systematically, allowing for the introduction of targeted and specific mutations. Among these systems, CRISPR-Cas9 is most widely utilized due to its usage duration over the years, high efficiency and versatility while other systems have been explored for phage engineering in different situations. The application of CRISPR-Cas for phage genome engineering depends on the natural or external defense mechanism of bacteria to recognize wild-type gene fragments. Nonetheless, phages can evolve resistance against immune responses triggered by specific CRISPR systems. Such resistance can be generated via mechanisms like escape mutations or obstruction of attachment,^252−255 resulting in lower genome engineering efficiency. Consequently, it is important to comprehend the features of various CRISPR systems and choose the suitable system to obtain the desired editing effect. The CRISPR-Cas systems utilized for phage genome editing are summarized in Table 2. Similar to other genome editing tools, one main issue of the CRISPR-Cas systems is the undesired mutations caused by off-target cleavage, the side effects of which may be ruled out through whole phage sequencing.

Table 2. Phage Genome Editing with CRISPR-Cas Systems^a.

CRISPR type	host organism and engineered phages	mutations introduced	editing efficiency	ref
I-E	E. coli	GD	38%	(182)
		T7	42%
	V. cholerae ICP1_2011_A	GD	100%	(223)
		GI	58%
		GE	50%
II-A	Streptococcus thermophilus	PM	100%	(218)
	2972	GD	100%
		GE	not determined
	Lactococcus lactisp2	GD	33% (1/3)	(256)
		PM	not determined
		GE	not determined
	E. coli	PM	100%	(231)
	T4	GD	100%
	E. coli	GD	T3:100%	(257)
	T3, T5, and T7	GI	T7:100%
		GE	T5:92%, T7
	Listeria monocytogenes	PM	not determined	(235)
	A511	GE	not determined
	Vibrio natriegens	GD	97%	(234)
	TT4, TT4P2	GI	100%
	Bacillus subtilis Goe1	GD	42%	(233)
		GE	5%
	Klebsiella pneumoniae phiKpS2	PM	100%	(232)
		GD	100%
		GI	88%
III-A	Staphylococcus epidermidis	SM	100%	(243)
	Staphylococcus aureus
	Andhra
	ISP
V	Escherichia coli	GD	100%	(249)
	T4	GI	100%
			100%
VI-A	Escherichia coli	GE	100%	(244)
	T4
	Pseudomonas aeruginosa	GD	12–53% (ΦKZ)	(183)
	ΦKZ, OMKO1, and PaMx41	GI	100% (PaMx41)
			7–25% (ΦKZ)
			50–71% (OMKO1)

^aNote: GD, gene deletion; GI, gene insertion; GE, gene exchange; PM, point mutation; SM, silent mutation.

6. Engineering Phages to Enhance Host Range

A single phage usually only infects a narrow range of bacterial strains belonging to a specific species. This high specificity may be desirable in some applications, but it also presents a significant obstacle to many applications. In addition, wild-type phages may not be available or yet discovered for specific hosts of interest. Thus, the ability to engineer phage specificity is an important goal. In general, the most specific interaction of the phage-host interaction is the initial binding between proteins of the phage capsid and the bacterial cell receptor, which enables subsequent steps of infection. Most phages are “tailed”, i.e., possess a tail whose fibers bind to the surface receptors of the host bacteria via receptor-binding proteins or peptides (RBPs). The structural variety of cell surface receptors is large,^258−260 including not only proteins but also lipopolysaccharide epitopes in Gram-negative pathogens recognized by T7-like phages,^261,262 as well as teichoic acids and pellicle polysaccharides in Gram-positive bacteria.^263−265 Several methods for genetic engineering of phage RBPs to alter or broaden the host range are being developed.

6.1. Engineering Phage RBPs through Selection

While the mutation rate of a gene is usually not related to its function under natural conditions,²⁶⁶ artificial mutations can be targeted to modify the phage RBPs.²⁶⁷ Artificial selection of phage RBPs involves production of a diverse phage population using its original working host, and selection of phages able to proliferate on an alternative host.^180,268,269 A typical procedure involves generating a phage library containing mutations in the RBP gene(s) through various methods such as error-prone PCR,^270−272 degenerate or random DNA synthesis,^273−275 and Golden Gate assembly in vitro.^276−278 Then the phage mutants of the library are screened on the alternative host. For example, Otsuka et al. investigated the outer membrane protein C (OmpC) of E. coli K12 and its interaction mechanism with the tail fiber protein of T4 phage.²⁷⁹ They identified critical amino acids in OmpC (P177 and F182 in loop 4) and T4 gp37 protein (position 937 and 942) for the adsorption process. Based on this, the researchers constructed a T4 phage mutant library with mutations in the distal tip region where the gp37 protein is located and selected multiple phage mutants that could adsorb to alternative receptors (OmpC of E. coli O157 or lipopolysaccharide of E. coli K12). In another case, the Lu lab demonstrated that the host range of E. coli phage T3 can be engineered through phage tail fiber mutagenesis,¹⁸⁶ producing viable phages that possessed subtle host-range alterations capable of targeting otherwise resistant bacterial mutants. Mutagenesis was specifically directed toward regions of the tail fiber known as the host-range-determining regions. Due to the relatively short length of these regions, the potential sequence space of these regions could be largely sampled. These approaches are informed by genetic and structural studies that identify important residues on the RBPs, allowing genetic diversity in the mutant library to be focused at the most productive regions. However, this is not strictly necessary. For example, Kilcher et al. identified the RBP of Listeria phage PSA (Gp15) and created a randomized synthetic gene library of the RBP using error-prone PCR.²⁸⁰ From this library, host range mutants were isolated and incorporated into a new polyvalent phage having an extended host range. To facilitate selection, the Qimron lab developed GOTraP (general optimization of transducing particles), a platform that links the phenotype with the desired genotype for phage engineering (Figure 11).¹⁷⁹ In this research, a T7 phage mutant with tail genes deleted infected E. coli hosts along with a plasmid containing tails from different sources and an antibiotic resistance marker. Then the resulting phages were screened to select the phages with tails that are compatible with the desired hosts. This approach facilitated the phage selection with improved transduction efficiency.

Figure 11.

Schematic illustration of GOTraP to extend phage host range.¹⁷⁹ (I) T7 phages without tail genes infects E. coli hosts carrying a plasmid with randomly mutated tail gene, antibiotic resistance gene, and packing signal gene. (II) The resulting phage library contains phage mutants with mutated tail protein. (III) Phage mutants are incubated with potential host strains and phages with compatible tail proteins recognizing the host bacteria can inject the plasmid with improved efficiency. (IV) Hosts requiring plasmid are selected on plates with antibiotics. The plasmids are extracted, mutated and transformed into the hosts. This process is repeated to optimize the tail gene and selected genes are sequenced to identify the desired mutations. Adapted with permission from ref (179). Copyright 2017 Elsevier.

With any directed evolution or selection project, success is subject to chance. Whether the DNA target of mutagenesis should encompass the RBP or focus on specific regions or residues depends on the frequency of discovering the desired phenotype. In other words, the phage “fitness landscape” for the new phenotype, over the sequence space being explored by the mutations, essentially governs whether the experiment can succeed in principle. If beneficial mutations are sufficiently numerous, then random sampling of RBP mutations may be sufficient to discover the phenotype; but if such mutations are rare, then greater focus on the critical regions is important to improve the chance of success given a finite library size. At this time, relatively little is understood about the molecular fitness landscapes of most phage RBPs, so library design in these cases is largely empirical.

6.2. Rational Design of Phage RBPs

In comparison to evolutionary approaches, phage RBPs may also be designed based on previously acquired knowledge. An important example is the design of phages with chimeric RBPs combining sequences from different phage types.^281−283 In a study complementing the evolutionary approach to engineering phage PSA, a high-resolution crystal structure of the RBP receptor-binding “head” domain was determined (Figure 12).²⁸⁰ Bioinformatic comparison to other Listeria serovar genomes identified related putative prophage-encoded RBPs. Swapping of the “head” domain resulted in the design and creation of chimeric phages with broader host ranges for various serovars. In another example, Tanji et al. used CRISPR/Cas9 to engineer the long and short tail fibers of E. coli T2 phage. Although T2 cannot infect the pathogenic E. coli O157:H7 strains,²⁸⁴ the E. coli phage PP01 does display broad infectivity to these strains. A chimeric T2 phage expressing the PP01 RBP was able to bind E. coli O157:H7, although the chimeric phages had substantially reduced infectivity. Tail fiber chimeras are a common approach for engineering tailed phages. Hu et al.²⁸² showed that swapping of tail fiber genes between different P. aeruginosa phages (JG004 and PaP1) through homologous recombination created chimeric phages capable of infecting the alternative host. Interestingly, in this system, a single mutation in the tail fiber gene of phage JG004 conferred extended host range compared to the original phage, emphasizing the potential versatility of phage tail fibers.

Figure 12.

Structure-guided rational design of chimeric RBPs.²⁸⁰ (A–C) Crystal structure of the PSA tail spike. (A) Ribbon diagram of the Gp15CTD homotrimeric. (B,C) Molecular surface of the RBP complex (B) and the PSA RBP with host-range mutants (C). (D) Analysis of the stem-neck conversation and crystal structure. (E–G) Head domains analysis revealed limited sequence identity of SV 4a, SV 5, SV 6b, and SV 1/2 (E), and they were used to construct the stem- and neck-chimeric phages (F). Host ranges and infection efficiencies of the synthetic phage chimeras (G). Adapted with permission from ref (280). Copyright 2019 Elsevier.

Engineering to alter the host bacterial genus is also possible. The Chen lab constructed five chimeric phages based on the E. coli phage M13, by swapping the receptor-binding domain of the RBP (g3p) with RBPs from other filamentous phages, to create phages that bind the human pathogens Pseudomonas aeruginosa and Vibrio cholerae, as well as two strains of the plant pathogen Xanthomonas campestris. While these phages can be produced in transformed E. coli, the virions do not infect or propagate via the phage life cycle on either E. coli or the alternative host.^78,79 On the other hand, Lu et al.¹⁸⁵ demonstrated that gene swapping between an E. coli phage and a Klebsiella phage resulted in chimeric phages that were able to replicate and lyse on the new host. Several of these studies highlight the difference between binding and infectivity and emphasize that binding is only the first step in infection. Depending on the application, binding may be sufficient (e.g., biosensor); in other cases, if infection is important (e.g., phage therapy), then additional engineering or evolution may be required if the chimeric phage proves to be unable to complete the infection cycle on the new host. The engineering of phage RBPs are summarized in Table 3.

Table 3. Summary of Engineering Phage RBPs.

strategies	phage vector	engineered proteins	host bacteria and receptor	ref
selection	ICP2	gp25	V. cholerae; OmpU	(267)
	T4	gp37	E. coli; OmpC	(279)
	PSA	gp15	Listeria; galactosylated wall teichoic acid	(280)
	T3	gp17	E. coli; LPS	(186)
rational design	M13	g3p	V. cholerae, P. aeruginosa, X. campestris (pv campestris), X. campestris (pv vesicatoria), E. coli (I+); CTXϕ, Pf1, ϕLf, ϕXv, Ιϕ1	(78)
	P2	gph	Salmonella; gp37	(285)
	T3	LPS	Klebsiella; gp11, gp12, and gp1	(185)
	P2	tail protein	S. flexneri M90T; tail protein G	(286)
	KP32, KP34, and KP36	tail protein	Klebsiella; N-terminal structural modules and C-terminal specificity modules of the RBP	(287)
	R2 pyocins	pyocin fiber protein	Pseudomonas aeruginosa; C terminus of the P2 tail fiber	(288)
	T2	long tail fiber protein (gene 37 and 38 encoding)	E. coli; tail fiber protein of IP008	(289)

7. Engineered Phages for the Detection of Bacterial Pathogens

Bacterial pathogens play a significant role in causing death and illness worldwide, with a particular prevalence in developing countries. Detection is important for a variety of applications, from identification of patient infections to ensuring food and water safety. Traditional methods for bacterial detection tend to be laborious and expensive, requiring trained personnel. For example, detection based on culturing is still widely regarded as the most reliable method for identifying bacterial pathogens. This process typically takes more than 48 h to allow for the selective growth of bacteria. On the other hand, enzyme-linked immunosorbent assays and lateral flow immunochromatographic assays offer rapid and straightforward biochemical immunoassays, but they can lack sensitivity. Polymerase chain reaction, biochips, and microarrays are among the nucleic-acid–based techniques used for pathogen detection. While these methods may be more rapid than culturing, most cannot differentiate between live and dead cells, limiting their applicability.^290,291

Phages can be rapidly and inexpensively produced in large quantities by infecting the host bacteria. If this can be done without cell lysis, purification steps can be minimized. As a result of the evolutionary arms race with bacteria covering billions of years, phages have developed strong binding affinity to the host cells. The virions are often robust and can retain activity in harsh environments (e.g., high temperatures, pH and organic solvents). The genetic diversity of phages is unmatched, and it is likely that phages exist to target all or nearly all bacterial species. Additionally, phages can potentially distinguish between live and dead cells since they only propagate in living hosts. Therefore, phages have attracted significant attention for applications in bacterial detection.

7.1. Phages Expressing Reporter Genes

One approach to detection relies on the ability of a nonlytic phage to infect and express genes in the host bacterial cell to be detected. The phage genome is engineered to contain a reporter gene. Upon infection, the reporter gene is expressed, leading to the production of a fluorescent compound or colorimetric marker. Widely used reporter genes include luciferase, fluorescent proteins, bacterial ice nucleoprotein, and β-D-galactosidase. The proteins produced by these genes within infected bacteria can be detected through colorimetric, fluorescent, or luminescent methods. A key feature of reporter phages is that they can differentiate between live and dead cells, as phages cannot express reporter genes in the latter. The representative engineered reporter phages for bacterial detection are summarized in Table 4.

Table 4. Summary of Representative Engineered Reporter Phages for Bacterial Detection.

strategy	phage	reporter	cloning method	pathogen	lod	detection time	ref
bioluminescence	A511::nluc; A006::nluc ΔLCR;A500::luc ΔLCR	Nluc, RLuc, GLuc, LuxAB	L-form rebooting	Listeria spp.	1–100 CFU/mL	3–24 h	(292)
	Wβ::luxAB-2	LuxA, LuxB, (spcR)	homologous recombination	B. anthracis	10–100 CFU/mL;104 CFU/g	6–12 h	(313,314)
	Y2	LuxAB	homologous recombination	Erwinia amylovora	4 × 103 CFU/mL	1 h	(315)
	E2, E4; K1, K4; EfS3, EfS7	NLuc	homologous recombination; CRISPR-Cas9 counter selection	E. coli, Enterococcus spp., and Klebsiella spp.	103 CFU/mL	5 h	(299)
	vB_KpP_TUN1	NLuc	in vitro phage genome assembly; phage rebooting	Klebsiella pneumoniae K64	10 CFU/well	a few hours	(316)
	HK620	LuxA, LuxB,	recombineering	E. coli	104 CFU/mL	1.5 h	(317)
	K1E	Nluc	in vitro phage genome assembly; phage rebooting	E. coli	5 CFU	not mentioned	(318)
colorimetry	T7LacZ	β-galactosidase (LacZ operon)	direct cloning using T7Select	E. coli	10–100 CFU/mL	7–8 h	(293,301)
	NRGp2 (T7)	ALP-Cex; ALP-CBM2a	direct cloning using T7Select	E. coli	8–10 h	1–1000 CFU/100 mL	(303,319)
	T7ALP	Alkaline phosphatase	direct cloning using T7Select;Homologous recombination; CRISPR-Cas counter selection	E. coli	9–16 h	1–1 × 105 CFU/mL	(320,321)
	T7	β-galactosidase (LacZ operon)	direct cloning using T7Select	E. coli	not determined	1 × 104 CFU/mL	(322)
	IP008BK, IP052BK	Cytochrome c peroxidase	homologous recombination	E. coli	16 h	4 CFU/g	(323)
electrochemistry	T7	ALP-GBP	direct cloning using T7Select	E. coli	12 h	1 CFU/100 mL	(305)
	T7	β-galactosidase	direct cloning using T7Select	E. coli	7 h	100 CFU/mL	(295)
fluorescence	TM4	GFP	phasmids	Mycobacterium spp.	2–3 days	1 CFU	(310,324,325)
	TM4	RFP	phasmids	Mycobacterium spp.	3–5 days	20 CFU	(294,309)
	PP01	GFP	phasmids	E. coli O157:H7	a few hours	Not determined	(326,327)
	T7	Maltose-binding protein	direct cloning using T7Select	E. coli	7 h	1000 CFU/mL	(320)
	λ	GFP	plasmid	E. coli	4–6 h	not determined	(308)

7.1.1. Bioluminescence Detection

Bioluminescence is a widely observed phenomenon among various life forms such as insects, fungi, bacteria and marine organisms.²⁹⁶ Luciferase reporter phages contain a gene for luciferase, an enzyme that oxidizes a substrate and causes light emission. An ideal luciferase should produce bright and long-lasting light emission while maintaining low background levels, and remain structurally stable under different environmental conditions. NLuc is an engineered luciferase widely used in phage reporters, that generates bioluminescence upon adding the substrate, with a signal half-life exceeding 2 h.²⁹⁷ The first reporter phage encoding NLuc was E. coli phage ΦV10, designed for the detection of E. coli O157:H7.²⁹⁸ Using reporter phage at a concentration of 1.76 × 10² pfu ml^–1, 5 CFU of target bacteria could be detected within 7 h. Recently, the Kilcher lab compared luciferases derived from different organisms, finding NLuc to be the most effective (Figure 13A,B).²⁹² NLuc was integrated into various Listeria phages for detection of various serovars. In particular, the broad-spectrum NLuc reporter phage A511::nluc_CPS was capable of detecting a single CFU of L. monocytogenes in 25 g of artificially contaminated food and drink in less than 1 day, with excellent specificity. The approach could be extended to detect common pathogens responsible for urinary tract infections, including E. coli, Enterococcus spp., and Klebsiella spp.²⁹⁹ Using patient urine samples, the bioluminescence induced by reporter phages was found to identify bacteriuria with high sensitivity, specificity, and accuracy, with detection of ≥10³ CFU/ml achieved within 5 h. Additionally, phage bioluminescence measured in urine correlated with antibacterial activity, highlighting the potential of reporter phages to aid screening of phages for personalized phage therapy (see below).

Figure 13.

Engineered phages expressing reporter genes for bacterial detection.^292−295 (A,C,F) Schemes illustrating bacterial detection with engineered phage for bioluminescence (A), colorimetric (C), and electrochemical (F) detection. (B) Bioluminescence time course assays of bacteria infected with the reporter phage.²⁹² Adapted with permission from ref (292). Copyright 2020 Meile et al. (D) Absorbance intensities vs detection time with T7_lacZ phage for bacterial detection.²⁹³ Adapted with permission from ref (293). Copyright 2017 American Chemical Society. (E) Fluorescence images of mCherrybombϕ infections in the presence of p-nitrobenzoic acid for detection of M. smegmatis.²⁹⁴ Adapted with permission from ref (294). Copyright 2018 Rondón et al. (G) Dependence of peak current obtained from differential pulse voltammetry curve on varying E. coli concentrations for different incubation times.²⁹⁵ Adapted with permission from ref (295). Copyright 2017 American Chemical Society.

An alternative system to Nluc is the bacterial lucificase operon (luxCDABE). This large operon (approximately 6 kb) consists of two genes encoding the luciferase (luxAB) and three genes encoding enzymes involved in production of the substrate, a fatty aldehyde (luxCDE). While luxAB can be used when the substrate is added externally, the complete luxCDABE operon does not require added substrate. Ryu et al. incorporated the luxCDABE operon into the ΦV10 genome to detect viable E. coli O157:H7 cells.³⁰⁰ Most phages, including tailed phages, have limited genetic capacity due to the defined geometry of the capsule (e.g., icosahedral). To accommodate the luxCDABE operon on the genome of ΦV10 phage, several nonessential phage genes were deleted. Regardless, the self-sufficient luxCDABE module eliminated the requirement to add luciferase substrate, allowing for a simplified workflow with repeated or continuous bioluminescence measurements from a single sample. Such a system could be advantageous in certain applications, such as in food manufacturing.

7.1.2. Colorimetric Detection

Colorimetric detection relies on enzymatic catalysis to convert a substrate into products whose color can be visually observed (chromogenic reaction). Beta-galactosidase (β-gal, encoded by the gene lacZ) is a widely used glycoside hydrolase derived from the lac operon of E. coli. The lacZ gene can be engineered into the phage genome, causing expression of β-gal after infection of the host cells. Cell lysis then releases the enzyme, which is mixed with the substrate. The Nugen lab demonstrated this method using E. coli phage T7^293,301 and the substrate chlorophenol red-β-d-galactopyranoside (CPRG), providing a visible red color, with high sensitivity (down to 10 CFU/mL) (Figure 13C,D).²⁹³

Another commonly used reporter gene is phoA encoding alkaline phosphatase (ALP), which catalyzes phosphate hydrolysis. Depending on the substrate, the reaction may be chromogenic, fluorogenic, or chemiluminescent. When ALP was engineered into phage T7, the reporter phage could detect 10³ to 10⁴ CFU/mL of E. coli in 6–8 h, depending on the substrate.³⁰² To improve sensitivity and remove contaminants from the detection solution, Talbert et al. engineered T7 to overexpress ALP fused to a carbohydrate-binding module (ALP-CBM). After cell lysis, the ALP-CBM was pulled down and concentrated using magnetic cellulose particles. With this enrichment step, the limit of detection (LOD) was lowered to 10 CFU/ml within 8 h. Thus, enzymes fused to an affinity tag can be used to enhance the sensitivity of reporter phage assays.³⁰³ While the readout is advantageous in that it does not require special equipment, maintaining sensitivity and accuracy can be challenging for colorimetric assays. Signal strengths can be low compared to alternative methods, and colored solutions can interfere with the visual signal, in addition to the effects of environmental factors (pH, salts, cell debris) that may inhibit substrate conversion.

7.1.3. Electrochemical Detection

The bacterial detection systems based on color or luminescence can be affected by sample color and turbidity. Electrochemical detection can avoid this pitfall by measuring the conductivity of the microbial growth medium. The reporter proteins catalyze redox reactions among the analytes, and the change in conductivity due to the conversion of small metabolites to larger charged metabolites by the bacterial cells can be measured. Rishpon et al. used lytic E. coli phage λ to infect E. coli K12.³⁰⁴ The released endogenous enzyme β-gal breaks down p-aminophenyl-β-D-galactopyranoside (β-PAPG) into p-aminophenol, which was oxidized at screen-printed carbon electrodes. Measurement of the current change yielded a LOD as low as 1 CFU/mL in 6 h. This strategy was further developed by Wang et al., using engineered T7 to overexpress β-gal for E. coli detection (Figure 13F,G).²⁹⁵ Immobilization of biomarkers on electrode surfaces can further enhance the sensitivity of electrochemical biosensors.³⁰⁵ A T7-based reporter phage was engineered to incorporate a gold-binding peptide fused to alkaline phosphatase (GBPs-ALP). The GBPs-ALP was released upon cell lysis and bound directly to a gold electrode surface. ALP activity was measured on the electrode through electrochemical linear sweep voltammetry, enabling detection of 1 CFU/mL of E. coli in drinking water in 9 h.

An alternative to lytic phages is temperate or chronic phages, which integrate into the host genome or establish an ongoing nonlytic infection, respectively. A system based on a derivative of the chronic E. coli phage M13 was engineered to deliver ALP.³⁰⁶ The phage could not lyse the host bacteria, and the reporter enzyme was instead transported and accumulated in the periplasmic space between the inner and outer cell membranes. However, both the substrate p-aminophenyl phosphate and the product p-aminophenol were able to permeate the periplasmic space, allowing conversion and product detection. This assay was highly sensitive (down to 1 CFU/mL of E. coli TG1) as well as rapid (<3 h). The method was later generalized to detect Bacillus cereus and Mycobacterium smegmatis with an LOD of 10 viable cells/mL in 8 h.³⁰⁷ Overall, phage reporters using electrochemical detection appear to be a promising avenue of development.

7.1.4. Fluorescence Detection

Fluorescent proteins such as green fluorescent protein (GFP) are well-known for generating highly sensitive and specific signals without added substrates, that are detectable by fluorimetry, fluorescence microscopy or flow cytometry. For instance, Namiki et al. modified the temperate E. coli phage lambda³⁰⁸ to express GFP to detect E. coli, with an LOD of a few bacteria under fluorescence microscopy after 4–6 h of incubation. Mycobacteria are particularly important targets for detection due to the long incubation times needed to culture cells (2 weeks). Mycobacteriophages expressing GFP or other fluorescent proteins in the host M. tuberculosis were able to differentiate between drug-resistant and drug-sensitive cells.^309−312 Introducing the antibiotic rifampicin or streptomycin simultaneously with the phage resulted in fluorescence only in resistant bacteria, enabling determination of drug sensitivity in under 24 h, even in mixed populations. An mCherry_bomb-delivering phage, with codon usage optimized for mycobacteria, was effective in a microscopy-based method for identifying Mycobacterium spp. and ascertaining rifampicin resistance directly from the sputum of tuberculosis patients in days (Figure 13E).²⁹⁴ An important limitation of fluorescence detection is the need for specialized equipment, as well as cellular autofluorescence, which limits point-of-care applications. Nevertheless, in a laboratory environment, fluorescence offers excellent sensitivity, versatility and convenience.

7.2. Engineered Phages Conjugated with Inorganic Nanomaterials

Nanoparticles are a fascinating group of materials with dimensions ranging from 1 to 100 nm and extremely high surface areas. These materials can possess exceptional magnetic, electrical, optical, mechanical, and catalytic properties that are significantly different from their bulk counterparts. The utilization of nanoparticles in conjunction with phages has the potential to enhance their effectiveness in bacterial diagnosis. A phage-bacteria-nanomaterial complex could be identified through various means, including color changes in the solution, particle aggregation, peaks in mass spectrometry, electrochemical signals, and shifts in surface plasmon resonance.

Quantum dots (QDs) are semiconductor nanocrystals that can convert a spectrum of light into a specific color. QDs have analogous optical properties to fluorescent reporter proteins like GFP, but higher sensitivity and better stability, as well as narrow spectral emission width enabling greater multiplexing. Edgar and colleagues engineered the E. coli phage T7 to include a small peptide substrate for in vivo biotinylation on the capsid protein,³²⁸ which was subsequently linked to streptavidin-coated QDs (Figure 14A–E). Replication of the recombinant phage in the bacterial host was detected via the streptavidin-coated QDs. This technique enabled 1-h detection of as few as 20 CFU/ml of E. coli in environmental samples using fluorescence microscopy. Fluorescence can be combined with magnetic particles as well. Holyst et al. developed bifunctional microparticles that possess both magnetic and fluorescent properties and were linked with T4 phages, which captured E. coli cells.³²⁹ The magnetic properties were used to separate the cells from a complex mixture, and flow cytometry was used to detect the fluorescent signal. The LOD for E. coli was around 10⁴ CFU/mL, with an incubation period of 15 min, giving a rapid but less sensitive assay. A drawback of fluorescence methods in general is the need for specialized equipment.

Figure 14.

Bacterial detection with engineered phages and conjugated inorganic nanoparticles.^78,328 (A) Scheme illustrating bacterial detection with engineered T7 phage and quantum dots (QD). (B) TEM images of T7 phage or phage bound to streptavidin-functionalized QDs targeted bacteria. (C–E) Using flow cytometry to detect phage–QD complexes. Plots of bacterial cells targeted by T7-myc (C) or T7-bio (D) phage after adding QDs. (E) Cell number vs fluorescence with the control and biotinylated phage. Adapted with permission from ref (328). Copyright 2006 the National Academy of Sciences of the USA. (F) Scheme illustrating bacterial detection with engineered M13 phage and gold nanoparticles (AuNPs). (G,H) TEM images of M13 phage before and after thiolation engineering. (I) Confocal microscopy image of engineered M13 phage capturing the host cell. Digital photos (J) and UV–vis spectra (K) exhibit the detection of bacteria with engineered M13 phage and AuNPs. Adapted with permission from ref (78). Copyright 2018 American Chemical Society.

The color of a gold nanoparticle suspension is extremely sensitive to changes affecting surface plasmon resonance (SPR) properties, such as particle aggregation. Chen et al. engineered E. coli phage M13 to display the RBPs targeting various Gram-negative pathogens (Figure 14F–K).^78,79 The phage capsid was nonspecifically thiolated, to enable formation of Au–S bonds. The phages were mixed with bacteria, and the phage-cell complexes isolated by centrifugation. Addition of gold nanoparticles led to aggregation with the complexes, leading to a visible color change from red to purple due to SPR changes, with an LOD of 100 CFU/mL cells in 30 min. Interestingly, the phage assay was robust to various environments including seawater, serum and milk, which can be a significant advantage compared with biomolecular affinity reagents such as antibodies and aptamers. This approach has been applied to a variety of phages in different environments.^330−332 The size of the gold nanoparticles can significantly influence the sensitivity of the assay.⁷⁹ Gold nanoparticles with larger dimensions (>20 nm in diameter) are usually not suitable for aggregation-based detection due to excessive colloidal stability. On the other hand, larger nanoparticles exhibit increased scattering properties. For example, the intensity of light scattering by 60 nm gold nanoparticles exceeds the light emitted by highly fluorescent dyes by 5 orders of magnitude.^333,334 Taking advantage of this feature, Watanabe et al. developed a plasmonic biosensor system that utilized S. aureus phage S13′ for binding to bacterial cells,³³⁵ leading to aggregation of gold nanoparticle-assembled silica nanospheres. Scattering intensity around the bacteria increased, with detection of S. aureus in 15–20 min and an LOD of 8 × 10⁴ CFL/mL.

Metal–organic frameworks (MOFs) are hybrid solids with a crystalline microporous structure, consisting of metal clusters connected by organic linkers. Phage-MOF hybrids combine the specificity of the phages and the optical properties of the MOF. For example, lytic phage were conjugated with IRMOF-3 (Zn₄O(NH₂–BDC)₃).³³⁶ When bound to the bacterial host cells, the MOF particles were sterically obscured by the bacteria, leading to a decrease in fluorescence intensity of the MOF, with an LOD for Staphylococcus arlettae of 10² CFU/mL. Sensitivity was improved with a different MOF, NH₂-MIL-53(Fe) based on iron instead of zinc,³³⁷ giving an LOD of 31 CFU/mL in both synthetic and real samples.

Single-walled carbon nanotubes (SWNTs) exhibit excellent properties for near-infrared imaging, owing to their photoluminescence within the 900–1400 nm range, significant Stokes’ shift, minimal autofluorescence background, and higher resistance to photobleaching in comparison to organic dyes. The Belcher lab functionalized the M13 phage with SWNTs for in vivo detection of pathogenic infections by fluorescence imaging.³³⁸ They attached an antibacterial antibody to the RBP of M13 via biotin–streptavidin interaction, demonstrating the detection of Staphylococcus aureus intramuscular infections and endocarditis. Overall, phage-nanomaterial conjugates present a promising approach to merging the biological specificity and affinity conferred by the phage with the special optical and material properties of inorganic nanomaterials.

7.3. Engineered Phages Conjugated with Organic Fluorescent Probes

Conjugation of fluorophores with phages could improve sensitivity and specificity compared to other methods (e.g., antibody conjugates). The natural multivalency of the phage surface enables the simultaneous presentation of targeting and signaling agents. In one case, a short peptide E3 was displayed on the g8p capsid protein of M13 phage,³³⁹ which was coassembled with the fluorescent dye Cy3 and silver nanoparticles, resulting in a 24-fold increase in fluorescence.

A different strategy is displaying a tag that can be used for fluorescent labeling. Wu et al. incorporated a TC-tag (an optimized tetracysteine sequence) into the RBP of M13 phage.³⁴⁰ The engineered phages entered the host cell through infection. A membrane-permeant biarsenical dye, fluorescein arsenical helix binder (FlAsH), was bound to the TC-tag, producing a strong fluorescent signal. With fluorescence microscopy, the tagged phage could detect as few as 1 CFU/mL viable target bacteria in a 40 mL sample in less than 3 h.³⁴¹

7.4. Phage-Based Surface Plasmon Resonance and Surface-Enhanced Raman Scattering Sensors

Analogous to bacterial detection by phage binding to electrochemical sensors, phages can also be combined with SPR and surface-enhanced Raman scattering (SERS) sensors, where excited plasmons in the substrate enhance the intensity of recorded spectra. These methods are real-time and label-free. In one study by Elliott et al.,³⁴²Salmonella-binding variants of M13 were identified by phage display selection, and attached to a dextran-modified gold chip. This SPR sensor exhibited high selectivity for Salmonella, as expected from the phage, and could detect as low as 3 CFU of Salmonella in 25 g of sample. Simonian et al. described label-free detection of Staphylococcus aureus using lytic phage 12600 to bind the cells on an SPR-based sensor, with LOD of 10⁴ CFU/mL.³⁴³ Indeed, detection of whole bacterial cells with SPR typically yields relatively low sensitivity, primarily due to challenges such as the restricted ability of the electromagnetic field to penetrate bacteria and the close resemblance in refractive index between the bacterial cytoplasm and the surrounding aqueous solution.

SERS sensors are modifications of SPR sensors that significantly amplify the Raman spectrum.^344,345 They have been effectively utilized in conjunction with other methods to identify bacterial cells, even in blood samples.³⁴⁶ Petti et al. developed a SERS sensor that can detect a single bacterial cell through multilayer structures incorporating phages, which led to significant enhancement of the SERS signal.³⁴⁷ However, most SPR and SERS-based sensors remain large and expensive, restricting their applications in point-of-care diagnostics.

7.5. Phage Components for Bacterial Detection

While phage virions have extensive utility and unique potential as described above, phage proteins themselves can also be integrated into biosensors, much like antibodies. Wang et al. linked purified tail fiber protein of E. coli phage T7 to magnetic beads to capture E. coli in water.³⁴⁸ Subsequently, the antimicrobial peptide polymyxin B was used to lyse the E. coli, and the released intracellular β-gal hydrolyzed the colorimetric substrate CPRG, resulting in a color change from yellow to purple. The LOD of this visible signal was 10² CFU/mL, and tests with four types of real water samples demonstrated recoveries of more than 80%. In another example. Bai et al. expressed recombinant tail fiber proteins derived from phages ΦAB2 and ΦAB6, which can recognize A. baumannii clinical isolates.³⁴⁹ The tail fiber proteins were also genetically tagged with a hexahistidine for linking to aluminum magnetic nanoparticles. Bacteria bound to the magnetic particles were isolated, and mass spectrometry validated the ability of the assay to distinguish between two A. baumannii strains according to design.

In some cases, phage properties are a detriment and phage proteins are therefore preferred. For example, filamentous phages are around 900 nm long, and may be oversized for some SPR and SERS-based biosensors in which binding events must occur within a specific distance from the transducer’s surface. Lytic phages destroy the bacterial cells in a short period, limiting the time available for analysis. Furthermore, novel phages may be difficult to produce and propagate, depending on the host strain and conditions for growth. Utilization of phage proteins may solve these challenges. Fu et al. created a fusion protein of GFP with the cell-binding domain (CBD) of a phage to detect methicillin-resistant Staphylococcus aureus (MRSA). The recombinant CBD demonstrated broad recognition of MRSA, but did not destroy the host cell, allowing assessment by flow cytometry.³⁵⁰ In a similar vein, Liu et al. characterized RBPs of three newly discovered phages (pO91, pO103, and pO111) obtained from hospital wastewater³⁵¹ and created genetic fusions to GFP, yielding a sensitive assay with LOD down to 33 CFU/mL. Fu et al. identified a virulent phage targeting P. aeruginosa PA1 in hospital sewage,³⁵² and predicted gene 069 to be the tail fiber protein. A recombinant version was produced in E. coli and shown to capture P. aeruginosa, with LOD of 6.7 × 10² CFU mL^–1 and 1.7 × 10² CFU mL^–1 through bioluminescent and fluorescent methods, respectively.

While most recognition strategies rely on the phage RBP, Liu et al. presented a different approach utilizing the major coat protein of M13 (pVIII). A peptide recognizing S. aureus was selected by phage display of pVIII,³⁵³ and bovine serum albumin-templated Co₃O₄ magnetic nanozymes (Co₃O₄ MNE) were linked to the pVIII fusion protein. The conjugates were used to target S. aureus from milk and the particles were separated magnetically. Subsequently, the bacteria were detected through the peroxidase-like activity of the Co₃O₄ MNE, achieving an LOD of approximately 8 CFU/mL. These examples demonstrate that phage coat proteins can have excellent properties as affinity reagents.

8. Engineering Phages for Bacterial Infection Treatment

The therapeutic use of natural phages dates back to their initial discovery in 1919, when d’Hérelle treated children with severe dysentery using phages.³⁵⁴ The idea of treating bacterial infections with a self-replicating agent is attractive, but the utilization of phages for antibacterial treatment was abandoned, particularly within Western medicine, due to the discovery and therapeutic success of antibiotics. However, more recently, the rise in antibiotic-resistant pathogens, coupled with the diminishing number of pharmaceutical companies engaged in the development and production of new antibiotics, poses a significant threat to human well-being.

Consequently, there has been a renewed interest in exploring the potential of phages as alternative antimicrobial agents. Phage therapy using wild-type phages is an active area of research and has been reviewed elsewhere.^10,11 Nevertheless, using natural phages for phage therapy may be constrained by factors such as the emergence of phage-resistant bacteria, immunogenicity, and the limited host ranges of natural phages. Fundamentally, naturally occurring phages have been optimized by natural selection for self-replication, and not for therapeutic properties. Recent research highlights engineered phages with improved features compared to their natural counterparts, including enhanced antibacterial activity, better biocompatibility, broader host ranges (discussed in section 5), and the ability to deliver therapeutic payloads to specific targets.

8.1. Phages to Reduce Bacterial Virulence or Increase Susceptibility to Antibiotics

Given the importance of antibiotics, one approach is to use phages as an adjunct therapy to augment antibiotic treatment. Collins et al. modified M13 to overexpress lexA3, a repressor of the SOS DNA repair system, to impair the ability of E. coli to respond to antibiotic treatment (Figure 15A,B).³⁵⁵ This technique led to orders of magnitude increased effectiveness of the antibiotic ofloxacin, and also reduced antibiotic-resistant mutants, leading to a substantial improvement in the survival rates of infected mice compared to ofloxacin alone or ofloxacin supplemented with wild-type phage. Phages can also be designed to deliver genes that reverse bacterial resistance to specific antibiotics.³⁵⁶ Qimron et al. utilized temperate phages to transport and integrate the genes rpsL and gyrA, which render genetically dominant sensitivity to the antibiotics streptomycin and nalidixic acid. Antibiotic-resistant strains of E. coli were lysogenized with the engineered phages. Indeed, bacterial susceptibility to both antibiotics was increased, with a 2–8-fold decrease in the concentration of antibiotic required to inhibit bacterial growth.

Figure 15.

Engineered phages to reduce bacterial virulence or increase susceptibility to antibiotics.^355,357 (A) Schematic illustration of engineered ϕlexA3 phage enhancing ablation of E. coli EMG2 with antibiotics.³⁵⁵ (B) Bacterial cell ablation ability of no phage, wild-type phage ϕ_unmod, and modified phage ϕ_lexA3 with 1 μg/mL ofloxacin. Adapted with permission from ref (355). Copyright 2009 the National Academy of Sciences of the USA. (C,D) Reducing Shiga toxin (Stx2) production by genetically engineered a temperate phage.³⁵⁷ (C) Schematic illustration of using for in situ repression of the virulence factor in a mouse model. (D) Schematic illustration of using phage to induce enforced lysogeny of endogenous prophage stx2 repression. Adapted with permission from ref (357). Copyright 2020 Hsu et al.

To reduce bacterial virulence, Silver et al. used temperate phage to insert genes into the bacterial genome (Figure 15C,D).³⁵⁷ The E. coli phage λ was utilized to deliver a transcriptional repressor gene to repress the yield of Shiga toxin. In the in vivo experiments, using the engineered phage can significantly reduce the production of Shiga toxin without affecting bacterial concentrations. Unlike traditional phage therapy, these strategies do not depend on the phage’s ability to eliminate bacteria directly, but rather on the capacity to transport genetic fragments to the host. While virulence reduction leaves the genetic potential for virulence expression intact, a potential advantage is that, because the cells are not killed, there is relatively low selection pressure for the bacteria to develop resistance to the phage. More study is needed to understand if this strategy is therapeutically appropriate.

8.2. Phages against Biofilms

Biofilms are clusters of microorganisms that adhere to surfaces through a matrix of polymers produced by the cells. Bacterial pathogens hidden in them are difficult to treat. Therefore, biofilms usually cause progressive and chronic infections. To address this challenge, Collins et al. engineered E. coli phage T7 to produce Dispersin B which can breaks down β-1,6-N-acetyl-d-glucosamine which is the main component of the biofilm (Figure 16A–G).³⁵⁸ While the wild-type phage was somewhat effective, reducing the bacterial count of the biofilm by 2 orders of magnitude, the engineered phage reduced the biofilm cells by approximately 4.5 orders of magnitude. Similarly, Loessner et al.³¹⁵ modified phage Y2 of the plant pathogen Erwinia amylovora to express a depolymerase gene, which degrades the exopolysaccharide capsule, resulting in a substantial enhancement of bacterial eradication. The engineered phage also hindered the ability of E. amylovora to colonize new surfaces (Figure 16H–J). These approaches are dependent on enzyme specificity. Since biofilms consist of a variety of bacteria that produce various matrix components, phage cocktails producing different enzymes to destroy the diverse extracellular makeup of natural biofilms may be required to enhance the efficacy of biofilm eradication. Alternatively, a more promiscuous enzyme can be used. Samanamud et al. engineered T7 phage to carry the gene for a lactonase enzyme having a wide spectrum of activity to disrupt quorum sensing,³⁵⁹ a mode of intercellular communication among bacteria that relies on acyl homoserine lactones (Figure 16K–P). Quorum sensing is crucial for biofilm development. While wild-type T7 phage decreased biofilm biomass modestly, treatment using the modified T7 phage on mixed-species biofilms containing E. coli and P. aeruginosa inhibited biofilm formation by 65–75%. Thus, both the biofilm matrix itself and biofilm formation mechanisms are significant targets for phage intervention.

Figure 16.

Engineered phages against bacterial biofilm. (A–G) Engineered DspB-T7 phage disperses biofilms.³⁵⁸ (A) Schematic illustration of removing biofilm with engineered T7 phages expressing DspB. (B,F) Time course of viable cell counts (B) and dose–response curves of mean cell densities (F) treated with T7control and T7DspB. (C,D) Scanning electronic microscopy images of T7DspB-treated biofilm (C) and untreated biofilm (D) after 20 h. (E,G) Time course of phage counts (E) and dose–response curves of phage number (G) after inoculating biofilm with T7DspB and the control. Adapted with permission from ref (358). Copyright 2007 the National Academy of Sciences of the USA. (H–J) Engineering of phages Y2::dpoL1-C inhibits biofilm formation.³¹⁵ (H) Plaque of Y2::dpoL1-C (right) gives them a clearer appearance compared with that of Y2 (left). (I) Time course of E. amylovora CFBP 1430 infected by different phages. (J) Engineered Y2::dpoL1-C inhibited E. amylovora to colonize on flowers. Adapted with permission from ref (315). Copyright 2017 American Society for Microbiology. (K–P) Inhibiting biofilm development by engineered T7 phage expressing quorum-quenching enzyme.³⁵⁹ (K–M) Degradation of AHLs in the condition of buffer (K) or wild-type T7 (L) or engineered T7aiiA (M). (N–P) Engineered phage degrading biofilms.³⁵⁹ (N) Different bacterial species are plated in the presence/absence of wild-type or engineered T7 phage. (O) Analysis of the dosage of phages on the development of bacterial biofilm. (P) Analysis of effect of wild-type and engineered T7 phages on the formation of dual-species bacterial biofilm. Adapted with permission from ref (359). Copyright 2014 American Society for Microbiology.

8.3. Phages Delivering Cas Nuclease

Delivery of CRISPR-Cas genes to pathogens is an emerging approach that converts an antiphage system into an antibacterial system, in a sense reversing the evolved biological function of CRISPR-Cas systems. Due to the ease of targeting specific sequences through design of the guide RNA, Cas nuclease can be directed against genes that mediate antibiotic resistance, pathogenicity, or virulence of the bacterial pathogen. Since not all bacteria possess a functional CRISPR-Cas system, implementation usually involves delivery of a Cas nuclease gene along with the CRISPR guide RNA gene. As with other uses of phages for gene delivery, this strategy requires the ability of the phage to infect and express genes in the host bacteria. Indeed, alternatives to phages for delivery include polymer-derivatized CRISPR particles and conjugative plasmids, although these present their own challenges. Nevertheless, due to the host specificity and evolved gene delivery mechanism, phages present an appealing strategy for transporting foreign nucleic acids, including the CRISPR-Cas systems, into the host. Phage gene delivery often relies on phagemids, which carry a limited set of genes but include a phage packaging signal. The phagemids are complemented by a helper phage that lacks a packaging signal, but supplies most of the phage genes in order to complete virion assembly. This method controls loss of the recombinant genes, although virion production may be inefficient. In one study, phagemid vectors containing CRISPR-Cas9 targeting a bacterial virulence gene were packaged in only 24% of the generated phage particles.³⁶⁰ Nevertheless, virulent strains of Staphylococcus aureus could be effectively eliminated while the avirulent ones were spared. The ease of reprogramming was also demonstrated by targeting the nuclease toward resistance-associated genes, which eradicated staphylococcal plasmids carrying drug-resistance genes, preventing their dissemination (Figure 17A–C). Using a similar approach, Lu et al. targeted both antibiotic resistance genes, including bla_NDM-1 and bla_SHV-18, which are responsible for broad-spectrum resistance and pan-resistance to β-lactam antibiotics, respectively, and virulence factors.³⁶¹ In this case, M13 phagemids carrying the CRISPR-Cas systems were delivered to E. coli. In vivo experiments using a Galleria mellonella (waxworm) infection model demonstrated the ability of the system to target a virulence factor (eae) of enterohemorrhagic E. coli O157:H7 (EHEC), conferring increased survival of infected larvae (Figure 17D–F).

Figure 17.

Engineered phages delivering Cas nuclease to kill bacteria specifically.^360−362 (A–C) Delivering CRISPR system to specifically ablate S. aureus.³⁶⁰ (A) Schematic illustrating ΦNM1 phage delivering a specific phagemid to S. aureus to kill the bacterial cells. (B) Effects of phagemid targeting drug resistance gene or the control. (C) Results of eliminating USA300^Φ in a bacterial community using the phagemid. Adapted with permission from ref (360). Copyright 2014 Springer Nature America, Inc. (D–F) Phages deliver RNA-guided nucleases (RGN) constructs as sequence-specific antimicrobials.³⁶¹ (D) Scheme illustrating phage delivering RGN to influence the physiology of the hosts. (E) Ablation of wild-type or drug-resistant EMG2 bacteria with engineered phage or control. (F) In vivo experimental results in G. mellonella larvae model treated with buffer, EHEC, ΦRGNeae, and ΦRGNndm-1. Adapted with permission from ref (361). Copyright 2014 Springer Nature America, Inc. (G–I) Utilizing lytic phages to enhance the population of bacteria sensitive to antibiotics.³⁶² (G) Scheme illustrating strategies to enhance bacterial population that are sensitive to antibiotics. (H) Enhancing the population E. coli resistant to phage. (I) Enhancing the population of E. coli sensitive to antibiotics. Surviving colonies from each culture were inoculated on plates having or lacking streptomycin or gentamicin. Adapted with permission from ref (362). Copyright 2015 the National Academy of Sciences of the USA.

An alternative to a phagemid system is using an infectious engineered phage delivering CRISPR-Cas in combination with a lytic phage. Often, one of the key obstacles in this approach is the limited space within the phage capsid due to its fixed geometry, particularly for tailed phages. In these cases, nonessential phage DNA must be eliminated to create room for new inserts. Beginning with the temperate E. coli phage λ, Qimron et al.³⁶² eliminated genes that were not required for phage replication and lysogenization, and integrated a CRISPR-Cas3 system that targeted both antibiotic resistance genes as well as the genome of a lytic phage. Treatment with the engineered temperate phage created two subpopulations of bacterial cells, being either transduced and thus resensitized to the antibiotic and immune to the lytic phage, or resistant to the antibiotic and sensitive to the lytic phage. Subsequently, the mixed population was treated with the lytic phage, which eliminated the latter subpopulation, leaving the first subpopulation that could be treated with antibiotics (Figure 17G–I). A caveat is that this approach relies on an understanding of the temperate phage’s wild-type genome, in order to engineer the deletions and insertions.

As with traditional phage therapy using wild-type phages, using phages for gene delivery requires a proper host range. Phages generally can only target a narrow range of bacteria due to their affinity for specific bacterial receptors. In addition to the approaches enhancing host range discussed above, another possible solution is the use of phage cocktails, i.e., combining different phages to enhance the delivery range of the antimicrobial system. Another important challenge, particularly when treatment duration is prolonged, is the evolution of bacterial resistance, through mutations, restriction enzymes, or even CRISPR-Cas systems. Indeed, restriction enzymes and CRISPR-Cas demonstrate that bacteria have evolved significant genetic programs to guard against phage infection. However, a guiding principle in this field is that phages have also developed various strategies to evade bacterial defenses, leading to an ongoing evolutionary arms race that has lasted billions of years. Combined with human ingenuity, there is cause for optimizm that phages can be evolved and engineered to overcome bacterial resistance that arises as phage-based therapies are applied.

8.4. Engineered Phages Delivering Antimicrobial Nanomaterials

Phages have the potential to transport therapeutic nanomaterials to bacterial pathogens with high specificity, thereby enhancing drug specificity and targeting. Chen et al. constructed chimeric phages that can specifically target five different bacterial species by modifying the RBP of M13 phages.⁷¹ Gold nanorods were immobilized along the side wall of the capsid through thiol-gold conjugation to yield a hybrid system named “phanorods”. The SPR absorption of the gold nanorods was tuned to the near-infrared spectrum, such that laser irradiation caused photothermal heating to kill attached cells. Phanorods targeting the pathogen P. aeruginosa killed more than 98% of target bacterial cells while largely sparing other bacterial species and mammalian cells (Figure 18A–C). The system was demonstrated to be effective in treating a mouse model of MDR P. aeruginosa wound infection.¹⁴⁶ Interestingly, the phanorods were significantly more effective than systemic antibiotic therapy in the animal model. In addition, since the phages were also photothermally inactivated after irradiation, they acted as delivery agents rather than self-replicating entities. Since this strategy allows drug-like control over dosing, in contrast to the biocontrol approach of standard phage therapy, the authors termed this “controlled phage therapy”.

Figure 18.

Engineered phages delivering antimicrobial nanomaterials. (A–C) Chimeric M13 phages deliver gold nanorods to specifically kill bacterial pathogens.^71,154,363 (A) Schematic illustration of constructing phanorod and killing bacterial pathogens specifically by irradiating the gold nanorods.⁷¹ (B,C) Cell viability assay of bacterial biofilm on mammalian cells, treated with phanorods, before (B) and after (row C) NIR irradiation. Adapted with permission from ref (71). Copyright 2020 the National Academy of Sciences of the USA. (D,E) Engineered phages deliver AIEgens for imaging, targeting and killing of bacterial pathogens.¹⁵⁴ (D) Scheme illustrating preparing phage-AIE conjugates TVP-PAP. (E) In vivo experimental results of using TVP–PAP to treat wounds infected with MDR P. aeruginosa in a mice model. Adapted with permission from ref (154). Copyright 2020 American Chemical Society. (F–I) Engineered phages with PEI for intracellular pathogen inhibition. (F) Scheme illustration of capping phage head with PEI enables endosomal escape in cell to eliminate intracellular bacteria.³⁶³ (G–I) Evaluation of bacterial treatment with PEI@P in vivo. Photos of the major organs from mice treated with PEI@P and the control groups (G). Distribution of bacteria in the major organs (H,I). Adapted with permission from ref (363). Copyright 2022 The American Association for the Advancement of Science.

Nanomaterial delivery can also be utilized for imaging in combination with therapy. The Tang lab modified the clinically isolated phage PAP targeting P. aeruginosa with an AIE molecule (luminogens with aggregation-induced emission property), TVP-S, to create a TVP–PAP bioconjugate.¹⁵⁴ This conferred fluorescence properties and the potential for photodynamic inactivation via ¹O₂ generation. Taking advantage of the phage’s targeting specificity, TVP–PAP enabled discriminative imaging as well as specifical ablation of the captured P. aeruginosa. An in vivo study confirmed that phage bioconjugates can effectively treat MDR bacteria infected wounds.

For bacteria that cause intracellular infections, MS2 labeled with AIE molecules were assembled with DNA sequences to create a spherical nucleic acid formulation,¹⁵⁷ MS2-DNA-AIEgen, enabling the phages to penetrate infected cells. Through in situ ROS generation when exposed to white light, MS2-DNA-AIEgen restored the activity of infected macrophage cells, and significantly enhanced healing of infected wounds in a diabetic mouse model (Figure 18D,E). In another approach, Liu et al. selectively modified the negatively charged phage head with cationic polymers polyethylenimine (PEI) through electrostatic interaction,³⁶³ resulting in phages with reversed charge while preserving activity. Due to the surface charge effect, the engineered phages are uptaken by intestinal epithelial cells and subsequently escape from endosomes to destroy the intracellular bacteria. This approach was validated in a murine model of intestinal infection, in which orally delivered PEI-capped phages (PEI@P) decreased the spread of pathogens to major organs (Figure 18F–I). These recent approaches highlight the potential for engineering phages to overcome delivery challenges. Conjugating phages with nanomaterials can amplify the synergistic effect of both components, potentially enhancing therapeutic efficiency of the nanomaterial and reducing toxicities. By using the phage for delivery but not self-replication, these approaches also can circumvent some risks of traditional phage therapy, such as undesirably rapid release of bacterial endotoxins, while enabling control over dosing. These mixed systems represent a large biochemical and nanomaterial space for exploration.

8.5. Phages and the Human Immune System

The immune system plays a crucial role in determining the fate of phages within the body. Phages are composed of numerous proteins, each capable of interacting with the immune system and being eliminated by the host defense system. Dual concerns are that the immune system may neutralize the phages, leading to lack of efficacy, or phages may be overly immunogenic, triggering an unwanted patient response. Chemical modifications (e.g., PEGylation) and microencapsulation (e.g., liposomal preparations) are used to decrease the virions’ interaction with the immune system. Compared to small molecules and even most biologics, phages are large, and this simple size disparity appears to play a crucial role in determining their ability to circulate in vivo, particularly due to clearance by the reticuloendothelial system (RES). Nevertheless, phage pharmacokinetics can be altered through modifications to the phage capsid proteins. In one early study, Merril et al. selected mutants of E. coli phage λ that could persist in the circulatory system by injecting phages into mice, collecting the surviving variants, amplifying the phages and repeating the selection cycle.³⁶⁴ Two λ variants showed significantly higher ability to avoid RES clearance 24 h post intraperitoneal administration compared to wild-type λ phage. Both variants shared a mutation in the major capsid protein E (Glu to Lys), and one variant also contained another mutation in the capsid D protein. Alternatively, Sokoloff and co-workers engineered phages to evade the complement system,^365−367 a humoral component of innate immunity that interacts with phagocytes, and can also be triggered by antibody binding. Starting with a library of E. coli phage T7 displaying random peptides on the phage capsid proteins, phages that were resistant to complement inactivation were identified. Interestingly, peptides with carboxy-terminal Lys or Arg residues demonstrated protection from complement-mediated inactivation through their interaction with C-reactive protein in rat blood. In human serum, several protective peptides featuring tyrosine residues were also identified. Thus, evasion of clearance mechanisms, including RES and the complement system, can be engineered into phage capsids to reduce neutralization of phages by the immune system. In addition to genetic engineering, phage formulations such as encapsulations can be attempted to enhance the ability of the viral particles to avoid immune inactivation and endure the hostile gastric environment, thus prolonging the circulation period.

On the other hand, an important risk of lytic phage therapy is the uncontrolled, rapid release of bacterial components, particularly endotoxins, due to extensive bacterial lysis, triggering a florid immune response and possibly septic shock. Indeed, endotoxins historically compromised some of the earliest phage therapy attempts. There are several approaches reported to address this challenge. As previously discussed, “controlled phage therapy” would allow tuning of the treatment over time to avoid sudden release of endotoxin. An alternative approach is engineering phages to be nonreplicative or lysis-deficient, including some of the genetic engineering approaches described above. In a report by Hagens and co-workers, the nonlytic phage M13 was genetically engineered to encode the restriction enzyme BglII or modified phage λS holin genes.³⁶⁸ Restriction enzyme delivery led to chromosomal DNA cleavage, while holin delivery caused bacterial membrane lesions. Both phages showed high ablation efficiency of E. coli while keeping the cell structure mostly intact, thus minimizing endotoxin release. Indeed, the levels of endotoxins released into the supernatants were significantly lower than a lytic phage control. Such systems could be further controlled by removing the replicative potential of the phage. In another research, the gene associated with the exporting protein in the nonlytic P. aeruginosa phage Pf3 was substituted with a restriction enzyme gene, resulting in a nonreplicative variant that delivered the restriction enzyme to directly infected cells but would not propagate beyond the original target cell.³⁶⁹ The engineered phage demonstrated effective killing of P. aeruginosa in vitro, with minimal endotoxin release. When a mouse model of P. aeruginosa infection, while both the engineered phage and a lytic phage had similar efficacy given a small bacterial inoculum, the engineered phage gave significantly higher survival compared to the lytic phage when the mice were challenged with a higher bacterial inoculum. The increased survival rate was associated with a reduced inflammatory response, consistent with the strategy to decrease endotoxin release.

Indeed, simply decreasing the lytic phenotype, without necessarily delivering additional genes, can improve phage therapy of lytic phages. Daly et al. compared the effect of phage therapy using lysis-deficient phage and wild-type E. coli T4 phages on mouse bacterial peritonitis.³⁷⁰ The lysis-deficient phage has a mutated holin gene, inhibiting cell lysis, but is still able to kill bacterial cells (nonlytic killing) due to phage nucleases and takeover of host machinery. The lysis-deficient phage, wild-type phage, or a β-lactam antibiotic were utilized to treat mice infected with B40sul E. coli. Treatment by the lysis-deficient phage showed significantly higher survival rates compared to the other groups, with notably lower endotoxin levels after treatment. Bacterial counts in peritoneal lavage fluid were initially higher when using the lysis-deficient phage (at 6 h), but ultimately dropped lower (at 12 h) compared to the other treatment groups. Concentrations of tumor necrosis factor and interleukin-6 were also lower when using the lysis-deficient phage, suggesting reduced inflammation, consistent with the lower endotoxin levels. Similarly, Ramachandran et al. constructed a recombinant temperate Staphylococcus aureus phage P954 which was lysis-deficient. The endolysin gene was inactivated by insertion of a marker gene for chloramphenicol resistance.³⁷¹ As expected, activation of the recombinant prophage did not lead to the lysis of host cells. Immunocompromised mice infected with pathogenic S. aureus were treated with the engineered phage and exhibited excellent therapeutic effects. These studies indicate that engineering phages to reduce lysis is a promising strategy to improve phage therapy.

These approaches highlight a common theme, that wild-type phages evolved for self-replication (e.g., optimizing release of virions by thorough cell lysis), and not for therapeutic use (e.g., avoiding sepsis or immune-mediated inactivation). Thus, engineered phages hold important potential for overcoming problems with phage therapy as currently practiced. Understanding and engineering the interaction between immune system and phages is an important area for future research.

9. Outlook

The growing interest in enhancing the safety and applicability of phages due to the global antibiotic resistance crisis has significantly advanced the field of engineering phages. However, there remain limitations in several aspects. First, most importantly, despite the age of the phage field, most phages are not well-characterized. This is a major issue that stems from the incredible number and diversity of the phage virome and affects both diagnostics and therapeutics. Most natural phages have not been cultured in a laboratory, while many known phages have not been thoroughly studied or are not easily manipulated genetically. As a result, only a few phages have been utilized as models for phage engineering. Despite multiple phage genome modification strategies that have been developed and proven effective for these phages, it is important to note that not all phages can be modified using these engineering techniques due to the specific limitations associated with each method.^372,373 The relative lack of knowledge about phage proteins also influences the current application scope. An important challenge with engineered phage-based sensors is their limited host range and thus limited target scope. Despite the significant advantages and impressive potential of using phages in biomedicine and biotechnology, the limited availability of receptor binding proteins targeting various pathogenic bacteria has hindered their applications. A critical area in which lack of knowledge about wild-type phages carries important risks is phage therapy applications. Phage therapy programs have been established in China, the United States, France, Belgium, as well as Eastern European countries such as the Republic of Georgia and Poland where phage therapy has remained as a tool to treat bacterial infection. However, a current lack of double-blind randomized controlled trials into their effectiveness and safety has significantly restricted the therapeutic applications of phages in humans, including genetically modified phages. Consequently, it is of critical significance to collect abundant rigorous experimental data to assess the efficacy and safety of phages. Further research characterizing phage biology and interactions with humans is needed to fill the existing critical knowledge gaps.

Second, the genome size of most phages except the filamentous virions is constrained by the capacity of the capsid. Most phages are “tailed”, having an icosahedral “head” containing the nucleic acids, and many nontailed phages also contain their DNA within a capsid having dimensions fixed by capsid protein geometry. This resulting space restriction significantly limits the modification of phage genomes as only small DNA fragments can be inserted or replaced. Depending on the intended application, this may or may not be an important limitation. For example, addition of a reporter cassette to an otherwise intact self-replicating phage genome could be problematic, but if the application does not require self-replication, the phage genome could tolerate large gene replacements. This issue does highlight the general theme that phages are optimized for self-replication, not for utility in a human world. While the icosahedral design may be widespread among viruses due to robustness of assembly from a small number of protein components, the fixed design is undesirable for genetic engineering. Alternatively, rod-shaped virions are more flexible in accommodating changes in genome size, since additional subunits can generally be incorporated to lengthen the virion.

Third, alterations in phage genome sequence may be pleiotropic, i.e., have multiple effects on phenotype. Alterations that are made for a specific reason (e.g., to alter the bacterial host) may have a negative effect on other aspects (e.g., phage infectivity). For instance, chemical or genetic modification of g3p protein of M13 phage can change the binding affinity to the host cell.^62,374,375 Many viral genomes contain overlapping genes due to the strong selection pressure to maintain a small genome size, and a particular mutation may therefore affect two genes. While this limitation can be overcome by refactoring, this process itself is a significant undertaking. Chemical and material modifications are relatively likely to exhibit multiple effects due to the nonspecific nature of the modifications (compared to genetic engineering). For example, chemical modification of surface proteins through PEGylation in two lytic myophages that infect Salmonella or Listeria species resulted in an extended blood circulation period in mice. However, this modification also led to a weaker ability to infect host cells.³⁷⁶ Thus, phage engineering often involves trade-offs in different properties, resulting in a multidimensional optimization problem.

A particular concern related to phage therapy applications is that bacteria can rapidly adapt in response to phage infection. Similar to how exposure to antibiotics can lead to the development of MDR bacteria, phage-resistant bacteria can also emerge in a short period following phage treatment. Bacteria have various mechanisms to avoid phage infection, such as mutation of cellular receptors, impeding the entry of phage DNA, cleaving phage nucleic acids (e.g., CRISPR-Cas systems and restriction systems), and inducing host cell death. However, in evolutionary time, phages can also counter these resistance mechanisms due to their genomic plasticity and fast replication. One approach, then, is to accelerate phage diversification and evolution in order to respond to bacterial resistance on a human time scale. In combination with informed designs (e.g., choosing slowly evolving receptors), the ability to design, mutate, and screen for new phenotypes in a high-throughput manner may allow phage engineering to keep pace with the evolution of bacterial resistance mechanisms.

A broader challenge in phage engineering pertains to the potential risks associated with the introduction of engineered phages into the environment, as they may lead to unforeseen impacts on the bacterial community. Such risks are inherent to biocontrol strategies including phage therapy. It is imperative to thoroughly assess these factors when designing phage genomes, and to take preventive steps whenever possible, such as limiting the number of replication cycles, employing complementation strategies, or using nonreplicative phages. Although phages cannot infect humans, plants, or animals directly, they have the potential to impact these organisms by modifying their microbiota. The impact of engineered phages on bacterial community dynamics, eukaryotic host organisms, and the broader ecosystem needs further investigation.

The development of artificial intelligence represents significant technological progress in biological and medical research, and phage engineering is not an exception. This may be an area of particular opportunity given the amount of uncharacterized genes (“dark matter”) in the virome and the relative lack of homology among phages, due to rapid evolutionary rates. Recently, several advanced AI-driven tools have been created to facilitate the prediction of the function and structure of phage proteins, and exploration of phage-host interactions. For example, Seeker is a deep learning tool that rapidly identifies various phages even with limited sequence information.³⁷⁷ On the other hand, VIBRANT utilizes a combination of machine learning (ML) and protein similarity techniques to automatically detect and evaluate the metabolic effects of the phages, outperforming conventional viral prediction software.³⁷⁸ Several other tools such as VirSorter2,³⁷⁹ PhageBoost,³⁸⁰ DEPhT³⁸¹ and Phanta³⁸² have been developed to detect diverse phages. Furthermore, computational tools have been developed to predict the phage-host interaction, which is of significant importance for phage therapy. For example, VirHostMatcher-Net utilizes a versatile network-oriented strategy that combines CRISPR sequences and alignment-free similarity measures,³⁸³ substantially improving the accuracy of predicting bacterial hosts. HostG can predict hosts based on a knowledge graph and graph convolutional network.³⁸⁴ These tools strengthen researchers’ ability to analyze the interaction between the phages and the host organisms, and design desired engineered phages for diagnostic and therapeutic applications. On the application side, in a recent report, Gally et al. developed ML models to select phage cocktails to address a bacterial infection.³⁸⁵ The ML models were trained on numerous phage-bacteria interactions. The authors developed random Forest models for each phage, resulting in characterization of a generalist phage that was active on over 20% of the bacterial strains.

This review highlights the twin challenges and potential of the current state of phage-based applications. A major challenge is expanding the knowledge base about phages, in which much depth of knowledge was previously developed on a handful of model phages in the formative years of molecular biology. It is now clear that this deep but narrow understanding of phages must be expanded many-fold in order to fulfill the engineering potential of phages in many applications. At the same time, information-rich approaches, including artificial intelligence, will be increasingly important for integrating newly acquired knowledge, such as novel phage genomes and microbiome interaction data, into a utilitarian framework to guide specific applications, such as suggesting mutations to engineer host specificity or selecting individual phages from a phage bank to design an optimal therapeutic cocktail. Nevertheless, the basis of these promising approaches is experimental data about phages, especially how various engineering strategies affect phenotypes. A multidisciplinary strategy integrating chemistry, synthetic biology, high-throughput techniques, nanomaterials and machine learning will be essential for translating an expanding knowledge base about phages and their engineering into real-world applications on nonmodel organisms.

Glossary

Abbreviations

MDR

multidrug-resistant

NHS

N-hydroxysuccinimide

Sulfo-NHS

sulfonic acid-modified NHS

SATP

N-succinimidyl-S-acetylthiopropionate

PEG

polyethylene glycol

EDC

1-ethyl-3-(-3-(dimethylamino)propyl) carbodiimide hydrochloride

DCC

N′,N′-dicyclohexylcarbodiimide

BRED

bacteriophage recombineering of electroporated DNA

BRIP

bacteriophage recombineering with infectious particles

YAC

yeast artificial chromosome

TXTL

transcription-translation

E. coli

Escherichia coli

S. cerevisiae

Saccharomyces cerevisiae

P. aeruginosa

Pseudomonas aeruginosa

SHAPE

stepping-stone host-assisted phage engineering

CRISPR

Clustered Regularly Interspaced Short Palindromic Repeats

CRISPR-Cas

CRISPR-associated

CRISPR RNAs

crRNA

PAM

protospacer adjacent motif

V. cholerae

Vibrio cholerae

tracrRNA

trans-activating crRNA

sgRNA

single guide RNA

S. thermophilus

Streptococcus thermophilus

ghmC

glucosyl hydroxymethyl cytosine

RBP

receptor-binding proteins or peptides

OmpC

outer membrane protein C

DT

distal tip

GOTraP

general optimization of transducing particles

CPRG

chlorophenol red-β-d-galactopyranoside

ALP

alkaline phosphatase

LOD

limit of detection

β-PAPG

p-aminophenyl-β-d-galactopyranoside

GFP

green fluorescent protein

QDs

quantum dots

SPR

surface plasmon resonance

MOF

metal–organic frameworks

SWNTs

single-walled carbon nanotubes

SERS

surface-enhanced Raman scattering

MRSA

methicillin-resistant Staphylococcus aureus

E. amylovora

Erwinia amylovora

AIE

aggregation-induced emission

RES

reticuloendothelial system

ML

machine learning

Biographies

Huan Peng is currently an associate professor at the College of Life Science and Technology, Huazhong University of Science and Technology. He received the Dr. rer. nat. in RWTH-Aachen University in Germany, and completed postdoctoral training at Maastricht University in The Netherlands, University of California, Santa Barbara and Los Angeles in the United States. His lab studies chimeric phages functionalized with nanomaterials for bacterial pathogen control.

Irene A. Chen is a Professor of Chemical and Biomolecular Engineering and Chemistry and Biochemistry at the University of California, Los Angeles. She attended Harvard University, earning degrees in Chemistry (A.B.), Biophysics (Ph.D., advised by Jack Szostak), and Medicine (M.D.), and completing a postdoctoral Bauer Fellowship in systems biology. Her laboratory studies synthetic protocells and phage engineering for both fundamental understanding and biomedical applications.

Udi Qimron is a Professor of Microbiology at Tel Aviv University. He earned his Ph.D. from Ben-Gurion University, where his research focused on Salmonella evasion from the mammalian immune system. He then pursued postdoctoral training at Harvard University under Prof. Charles C. Richardson, studying genes involved in T7 phage growth. His laboratory is dedicated to advancing phage therapy, CRISPR-Cas technologies, mammalian sex determination, and bacterial immune systems, with a particular focus on innovative strategies to combat antibiotic resistance and enhance phage engineering for therapeutic purposes.

Author Contributions

CRediT: Huan Peng, Irene A. Chen, Udi Qimron conceptualization, investigation, visualization, funding acquisition, supervision, writing-original draft, writing-review and editing.

The authors declare no competing financial interest.

Special Issue

Published as part of Chemical Reviewsspecial issue “Synthetic Biology”.