Sustained in situ protein production and release in the mammalian gut by an engineered bacteriophage

Abstract

Oral administration of biologic drugs is challenging because of the degradative activity of the upper gastrointestinal tract. Strategies that use engineered microbes to produce biologics in the lower gastrointestinal tract are limited by competition with resident commensal bacteria. Here, we demonstrate engineering of bacteriophage (phage) that infect resident commensals to express heterologous proteins released during cell lysis. Working with the virulent T4 phage, which targets resident, non-pathogenic E. coli, we first identify T4-specific promoters with maximal protein expression and minimal impact on T4 phage titers. We engineer T4 phage to express a serine protease inhibitor of a pro-inflammatory enzyme with increased activity in ulcerative colitis and observe reduced enzyme activity in a mouse model of colitis. We also apply the approach to reduce weight gain and inflammation in mouse models of diet-induced obesity. This work highlights an application of virulent phages in the mammalian gut as engineerable vectors to release therapeutics from resident gut bacteria.

Ed summary:

Biologics are delivered to the gut using phage that infect resident bacteria.

Introduction

Oral administration of drugs is an ideal delivery route because of its non-invasive nature,^1,2 but it remains infeasible for many types of compounds. If the target is in the intestine, the drug must be able to survive the acidic and proteolytic conditions of the upper gastrointestinal tract (GIT)^2,3 as well as potential degradation by microbial proteases. For biological therapeutics that are sensitive to denaturation and digestion, the upper GIT is an especially challenging obstacle. If the drug’s destination is systemic circulation, then it must also be absorbed at biologically significant quantities, which requires specific physicochemical properties allowing it to traverse the intestinal epithelia.^1,4 Even if a drug and/or its formulation is capable of overcoming these challenges, frequent dosing (e.g., daily) is associated with noncompliance and has been shown with half of chronically ill patients not adhering to the prescribed dosing regimen.⁵

Bacteriophages (phages) are prokaryotic viruses that are well-known for their ability to redirect host machinery towards the expression of phage-encoded genes during the infection process. In addition to active research focused on using virulent phages as anti-bacterials (“phage therapy”),⁶ emergent work has shown that gut viruses, ~90% of which are phages, are numerically dominant elements of the gut microbiome, can outnumber their bacterial counterparts especially within the mucosal lining of the GIT,^7,8 and can persist for long periods of time.⁹ Longitudinal analysis has shown that phage sequences can persist in humans for more than a year,¹⁰ with more recent work demonstrating a multi-year persistence of the highly prevalent crAssphages.⁹ One such phage, ΦCrAss001, is a virulent phage for the gut commensal Bacteroiodes intestinalis.¹¹ More generally, co-existence between virulent phages and their bacterial hosts have been observed in monocolonized,¹² gnotobiotic,¹³ and conventionally colonized mice.¹⁴ With implications for the efficacy of phage therapy,¹⁵ this coexistence offers a possible opportunity for alternative utilizations of virulent phages. We hypothesized that a therapeutic protein could be encoded into a virulent phage so that it is co-expressed with phage genes during the infection of a resident gut bacterium and then released into the extracellular space by phage-induced cell lysis. Furthermore, we hypothesized that phage-bacterial coexistence would lead to a sustained production of phage-encoded genes.

Herein, we demonstrate that genetically engineered virulent phages can be used to reprogram bacteria to express and release therapeutic proteins. First, we demonstrate that the virulent phage, T4, releases substantial quantities of the intracellularly-produced fluorescent marker protein, superfolder green fluorescent protein (sfGFP), into the culture lysate under two conditions: when the sfGFP gene is encoded in the genome of the host bacterium and when it is encoded in the genome of the phage. Then, we use a mouse model to show that a single dose of phage encoding the sfGFP gene leads to significant in situ production of sfGFP in the murine mucosa. Finally, we demonstrate the feasibility of our proposed approach in two murine models. In the first model, we use T4 phage to express the serine protease inhibitor B1a (Serpin B1a) protein to inhibit the activity of the pro-inflammatory enzyme, neutrophil elastase, which is upregulated during colitis. In the second model, we use T4 phage to express the protein ClpB, which has a discontinuous 5 amino acid motif that is homologous to the anorexigenic neuropeptide α-melanocyte-stimulating hormone, and has shown to induce satiety to reduce weight gain during obesity.^16,17 We show that T4 phage engineered to express ClpB reduces weight gain in a mouse model of diet-induced obesity (DIO). Collectively, our data shows that virulent phages can be leveraged to express and release heterologous proteins with physiological effect on the mammalian host.

Results

Phage propagation releases intracellular protein

As depicted in Figure 1A, we hypothesized that bacterial infection by a virulent phage would enable the release of intracellular protein to the extracellular environment. While we ultimately aimed to encode heterologous genes into the phage genome, we first determined whether phage-induced lysis of bacteria would release significant quantities of intracellular protein. We incubated the virulent T4 phage with an E. coli K-12 strain that constitutively-expresses the sfGFP gene from the bacterial chromosome. Hereafter, we use “sfgfp” to refer to the gene and “sfGFP” to refer to the protein product of the sfgfp gene. We found that this fluorescent E. coli is readily lysed by T4 phage (Supplementary Figure S1) and that the phage-induced lysis of bacteria led to substantially more sfGFP in the culture supernatant. As shown in Figure 1B, fluorescent E. coli cultures with T4 phage had greater fluorescence per cell in the supernatant (~0.5-1.0 x 10⁶ AU/OD) compared to bacteria cultured without phage (~7 x 10³ AU/OD). In contrast, fluorescence of the bacterial cells (“cell pellet”) was comparatively similar (~10⁵ AU/OD). Together, this data indicates that T4 phage facilitates the release of intracellular proteins into the extracellular environment.

Figure 1.

Phage-induced lysis of bacteria releases intracellular proteins. (A) Scheme of the potential utility of this strategy in the gut, by introducing a recombinant phage to target an endogenous commensal gut bacterium. (B) Biological replicates of batch culture E. coli and wildtype T4 phage demonstrates the presence of phage increases the mean extracellular quantity of sfGFP (n=3).

Engineered phage produces and releases recombinant protein

Next, we aimed to quantitate whether significant amounts of phage-encoded, heterologous protein could be produced and released during the infection of bacterial cells. Phage infection generally leads to a disruption of host cell processes in favor of those that are phage-specific. For T4 phage, the preferential transcription of early T4 promoters over host promoters is due in part to a greater promoter strength of the former and the modification of host RNA polymerases to preferentially transcribe T4 promoters.¹⁸ This ultimately leads to the transcription of phage promoters in sequential stages (Figure 2A) to optimize the assembly of viable viral particles.^18,19 Because this process is carefully orchestrated, we aimed to determine the promoter and stage of activation (i.e., early, middle, or late) for maximal expression with minimal impact on phage propagation. We generated a small library of recombinant T4 phages containing select promoters driving sfgfp gene expression (Figure 2B), and inserted this construct into a non-essential hypothetical membrane protein (ac gene) that has been previously used for T4 recombination.²⁰ Past studies have shown that the selected promoters are highly expressed during T4 phage infection.^18,21,22 After 12 hours of their culture with E. coli K-12, we measured the sfGFP protein released into the supernatant by fluorescence. As shown in Figure 2C, the absence of a promoter (“No promoter”) or use of a strong constitutive E. coli promoter (“J23119”) did not lead to significantly more sfGFP fluorescence than a non-engineered T4 phage that lacks the sfgfp gene (“Wildtype”). However, one early promoter (motB, p=0.024) and two late promoters (gp18, p=0.0035; gp22, p=0.01) led to significantly greater sfGFP levels in the supernatant. Titers of phages encoding these promoters showed ~2-fold reductions compared to wildtype T4 phage, though none achieved statistical significance (p=0.0807, 0.0513, and 0.1556, respectively) (Figure 2D). All promoters tested produced viable phage. Based on our criteria of maximal sfGFP levels and minimal impact on phage titer, our subsequent experiments used the gp22 promoter, which regulates transcription of the major prohead scaffolding core protein.^23,24

Figure 2.

Survey of T4 phage promoters for in vitro sfGFP production. (A) During T4 infection, promoters are expressed in a series of stages. Representative promoters were selected from early, middle, and late stages of expression. (B) To survey the protein production from these promoters, they were encoded upstream of sfgfp and inserted into the ac region of the T4 phage genome to produce a small library of recombinant T4 phages. (C) After 12 h of coculture of the T4 library with E. coli, the fluorescence and (D) phage concentration was measured in the supernatant (n=3 biological replicates). Bars represent mean values. Statistical analyses were performed by one-way ANOVA compared to wildtype T4 phage with Dunnett’s multiple comparison test.

Engineered phage expresses proteins in the mucosa

To determine the feasiblity of this approach for the in vivo production of a protein in the gut microbiota, we used a streptomycin-treated mouse model of E. coli colonization. Human commensal E. coli strains, including the laboratory-adapted strain E. coli K-12 originally isolated from human feces, do not reliably colonize the murine gastrointestinal tract. Previous studies have shown that streptomycin administration is effective at maintaining streptomycin-resistant E. coli colonization, allowing for the study of E. coli function in mouse models of the mammalian gut.^25,26 As schematically outlined in Figure 3A, we provided streptomycin in the drinking water, colonized with a single dose of E. coli and then administered wildtype T4 phage or T4::sfgfp phage (n=6 per group). After 4 days of co-colonization between bacteria and phage, we found that both are largely present (Figure 3B), though fecal E. coli levels for some mice drop below our level of detection. Quantitation of fluorescence in the mucosa of unfixed intestinal sections by fluorescence microscopy showed significant increases in GFP fluorescence (p<0.0001) in mice colonized with T4::sfgfp phage over those colonized with wildtype T4 phage (Figure 3C). Representative images show that the wildtype T4 phage condition had low levels of fluorescence due to autofluorescence whereas the T4::sfgfp phage condition had substantial mucosal fluorescence (Figures 3D and 3E, respectively).

Figure 3.

In vivo sfGFP production by engineered phage. (A) Schematic representation of the mouse model. (B) Fecal E. coli and phage concentrations immediately prior to phage administration and at experimental conclusion for wildtype T4 (control) and T4::sfgfp phage conditions (n=6 mice per condition). (C) Mean fluorescence intensity of the murine mucosa was significantly higher for mice colonized with T4::sfgfp phage compared to wildtype T4 (n=6 mice per condition). (D) Fluorescence was also visually evident within fluorescence microscopy images of unfixed colonic sections, with minimal autofluorescence in mice receiving wildtype T4 phage, while (E) mice colonized with T4::sfgfp had substantially greater mucosal fluorescence. Scale bars = 50 μm. Dashed white lines represent the intestinal mucosa. (F-O) Colonic sections were fixed and fluorescently stained with nuclear (blue) and mucin (red) stains (n=6 mice per condition). Images shown are representative results from n=3 mice per condition. Scale bars = 10 μm. (F,K) Macroscopic and (G,L) higher magnification views show overlays of DAPI, FITC, and TRITC channels whereas individual high magnification channels for (H,M) FITC, (I,N) TRITC, and (J,O) DAPI are shown separately. All bars represent mean values. Statistical analysis was performed by a two-tailed unpaired t-test.

For additional validation that GFP fluorescence is localized in the mucosa, we fixed intestinal samples and stained nuclei with DAPI and mucins with rhodamine labeled wheat germ agglutinin and Ulex Europaeus agglutinin I. Although GFP fluorescence is greatly diminished, it remained sufficiently fluorescent for localization. As shown in Figure 3F, the FITC-channel fluorescence in the mucosa of mice colonized with wildtype T4 phage is insubstantial when compared to the autofluorescence of the tissue. This is clearly shown at higher magnification (Figure 3G) with the channels separated (Figures 3H, 3I, and 3J). In contrast, mice colonized with T4::sfgfp phage had substantial mucosal GFP fluorescence as evidenced by the orange color due to overlapping green and red (Figure 3K). Higher magnification (Figure 3L) confirms an increase in GFP fluorescence (Figure 3M) that colocalizes to the mucosa (Figure 3N) on the luminal side of the tissue, away from the nuclei (Figure 3O). Collectively, these results indicate that heterologous genes, driven by phage-specific promoters, can be expressed to significant levels during phage propagation. Furthermore, this production in the mucosa is in close proximity to the intestinal epithelium where these gene products would exert effect.

T4::serpin reduces neutrophil elastase activity

To determine whether proteins produced by our phage-based method were active in the mammalian gut, we engineered T4 phage to produce a serine protease inhibitor (serpin). During ulcerative colitis (UC), a type of inflammatory bowel disease, neutrophil infiltration and granule release leads to an increase in neutrophil elastase (NE) activity.²⁷ Not only does NE lead to sustained pro-inflammatory effect,^28,29 it can also impair the function of therapeutic antibodies.²⁷ Past studies have shown that the NE activity in mice with chemically-induced colitis have disease trajectories that parallel UC patients, and that the treatment with a small molecule NE inhibitor by intraperitoneal injection, twice daily, suppressed NE activity.³⁰ Serpins are potent protease inhibitors produced by host cells to regulate NE activity, which if left unregulated increases tissue damage.²⁸ Therefore, to attenuate NE activity we engineered T4 phage to encode a serpinb1a gene to produce a Serpin B1a protein to specifically inhibit murine NE. Hereafter we refer to the gene and protein as serpin and Serpin, respectively.

Using the same strategy described for the expression of the sfgfp gene, we inserted the serpin gene containing a N-terminal His₆-tag into the T4 genome, driven by a gp22 promoter (T4::serpin phage). This protein is the murine ortholog to the human protein serpin B1.³¹ To determine if Serpin produced by T4::serpin phage was active, we purified Serpin from in vitro phage cultures and incubated it with NE. Remaining NE activity was then measured using a fluorogenic peptide substrate. We found that Serpin produced by T4::serpin is active with a comparable inhibitory effect against NE (IC₅₀ = 1.7 ± 0.9 nM) to that of the small molecule NE inhibitor, sivelestat (IC₅₀ = 1.2 ± 0.2 nM).

To test the efficacy of Serpin produced by T4::serpin phage in the mammalian gut, we used a chemically-induced colitis mouse model as shown in Figure 4A. Mice were colonized with E. coli followed by a single oral dose of wildtype T4 phage or T4::serpin phage (n=5 per group). Intestinal colitis was established with seven days of dextran sodium sulfate (DSS), a method previously shown to induce NE activity that parallels UC.³⁰ As shown in Figure 4B and 4C, fecal E. coli and phage levels persisted for the duration of the experiment without substantial differences between wildtype T4 and T4::serpin conditions. On Day 9, we found that T4::serpin treated mice gained significantly more weight than wildtype T4 treated mice (Figure 4D), and had significantly lower fecal NE activity (Figure 4E). Analyses of colonic neutrophils revealed similar neutrophil densities between phage conditions (Figure 4F, Supplemental Figure S2), but that these neutrophils from T4::serpin treated mice were enriched with CD24 markers (Figure 4G). CD24 is a key anti-inflammatory mediator expressed on resolving neutrophils,^32,33 an effect largely due to the formation of the CD24-Siglec axis. This structure suppresses DAMP-related inflammation³⁴ and has a predicted anti-inflammatory role in colitis.³⁵ Anti-inflammatory leukocytes can potentially propagate their effects to neighboring resident cells through CD24-mediated inter-cellular communications^36,37 leading to improved tissue homeostasis and regeneration.³⁸

Figure 4.

Reduction of DSS-induced colitis via Serpin production. (A) Schematic of 1.5% DSS-induced colitis mouse model, assessing Serpin reduction of inflammation and NE activity. (B) Fecal bacteria and (C) phage concentrations for the wildtype T4 (control) and T4::serpin phage conditions was consistent throughout the experiment, including during DSS administration (n=5 mice per condition). (D) Weight change and (E) fecal neutrophil elastase concentration based on protease activity was determined after 7 d of DSS, on Day 9 (n=5 mice per condition). (F) Analyses of colonic cells by flow cytometry measured neutrophil fractions, and the density of (G) CD24 markers on these neutrophils (n=4 mice for the wildtype T4 phage condition and n=5 mice for the T4::serpin condition). Statistical analyses were accomplished via two-tailed unpaired t-test.

T4::clpB reduces host weight in a diet-induced obesity model

We next examined whether this phage-based system could address a disease on a longer time scale, such as obesity. Previous studies have shown that daily oral administration of bacteria expressing an endogenous chaperone protein ClpB reduces weight gain in mouse models of obesity¹⁷ and in human trials.³⁹ Analyses of the ClpB sequence revealed a discontinuous five amino acid motif with homology to the mammalian peptide α-melanocyte-stimulating hormone (α-MSH) that is recognized by anti-α-MSH antibodies in western blot and by anti-ClpB/anti-α-MSH cross-reactive antibodies from ClpB-immunized mice.¹⁶ ClpB also induced the secretion of satiety-inducing hormones (peptide YY and glucagon-like peptide-1) from enteroendocrine cells⁴⁰ in a manner similarly observed with α-MSH binding to the melanocortin 4 receptors (MC4R) in enteroendocrine cells in vitro and in mice.⁴¹ This strategy for introducing MC4R agonists to treat obesity is the subject of pharmaceutical development that includes the FDA-approved cyclic peptide setmelanotide.⁴² With the possibility that intestinal ClpB could lead to reduced weight gain in the host, we engineered T4 phage to encode the clpb gene to produce the ClpB protein. Quantitation by ELISA of N-terminal His₆-tagged ClpB in cocultures of T4::clpB with E. coli K-12 showed significant levels in the supernatant (Supplemental Figure S3) corresponding to 0.88 ± 0.24 μg per g of supernatant protein or 37 ± 21 μg per L of culture (mean ± s.d., n=3). To determine if engineering T4 phage significantly altered its infection parameters, we generated one-step growth curves (Supplementary Figure S4A and S4B) and found that the burst size of T4::clpB phage (54 ± 33, n=3) was significantly reduced compared to wildtype T4 phage (195 ± 73, n=3) (Supplementary Figure S4C). The latency was unchanged (29.7 ± 1.7 min, n=3; and 30.2 ± 1.2, n=3; respectively) (Supplementary Figure S4D). This indicates that the diversion of cellular resources to produce ClpB leads to fewer progeny phage per infection event.

To determine if in situ ClpB production by our phage-based approach can have a physiologically-relevant impact on the host, we used a mouse model of DIO as outlined in Figure 5A. After colonization with E. coli K-12, three experimental groups of mice (n=8 or 9 per group) received a single oral dose of either buffer, wildtype T4 phage, or T4::clpB phage (Day 0). On Day 3, mice were transitioned from a standard mouse chow to a high fat diet (45%) to induce obesity. Fecal phage titers show that both wildtype T4 and T4::clpB phages maintained colonization of the murine gut, though there was large inter-individual variation, especially for wildtype T4 phage (Figure 5B). Fecal E. coli concentrations also showed sustained colonization despite the introduction of wildtype T4 or T4::clpB phages (Figure 5C), highlighting a coexistence between phage and bacteria that is consistent with previous observations.¹³ It is notable that E. coli concentrations appear to drop on Day 11 for all groups and recover thereafter, which may be due to transient effects of the high-fat diet introduced on Day 3 or a mutational selection leading to a resurgence of mutant E. coli that are better adapted for the murine gut, as has been observed in the past.^25,26 At sacrifice, analysis of the intestinal mucosa showed substantial phage and E. coli concentrations, confirming that the mucosa, in addition to the lumen, is colonized (Figures 5B and 5C).

Figure 5.

The efficacy of T4::clpB to reduce food consumption and weight gain. (A) Schematic representation of the in vivo experiment using a murine model of diet-induced obesity with C57BL/6 mice. Mice were colonized with E. coli alone “no phage” (control, n=8 mice), wildtype T4 phage (control, n=9 mice), or T4::clpB phage (n=9 mice). (B) Phage measured by plaque assay and (C) bacteria measured by selective culture show that wildtype T4 and T4::clpB are able to coexist with their E. coli host in vivo and that the presence of phage does not markedly alter E. coli colonization (n=8 mice for the E. coli alone condition and n=9 mice for wildtype T4 and T4::clpB conditions). (D) Fluorescence microscopy of E. coli mixed with wildtype T4 (control) or T4::clpB phage after fixation and labeling of liquid cultures for DNA (DAPI), and ClpB (anti-α-MSH antibodies) and (E) quantification of ClpB fluorescence from individual bacteria (n=27 cells for wildtype T4 and n=40 cells for T4::clpB). (F) Fluorescence microscopy of murine intestinal samples labeled for DNA, mucin (WGA/UEI) and ClpB, and (G) its quantification of ClpB fluorescence in mice colonized with wildtype T4 (control) or T4::clpB phage. Scale bars = 10 μm. Of wildtype T4 mice (n=3), 20, 15 and 20 cells were counted from each mouse. Of T4::clpB mice (n=3), 20, 22 and 20 cells were counted from each mouse. Different shading of symbols represent different mice. (H) Longitudinal weight measurements were collected for E. coli (n=8), E. coli + wildtype T4 phage (n=9) and E. coli + T4::clpB phage mice (n=9), and for a subset of mice, (I) food consumed on a per day basis (n= 5 mice per condition) and (J) overall activity was measured over a 24 h period before sacrifice for E. coli (n=4), E. coli + wildtype T4 phage (n=5) and E. coli + T4::clpB phage mice (n=5). (K) Cytokine panel of serum at sacrifice. Samples with undetectable concentrations were set to the limit of detection. (B,C,E,K) Lines represent median values and (H,I,J) bars represent mean values. Statistical analyses were performed by (E) two-tailed unpaired t-test , (G) mixed model analysis , (H,J,K) two-way ANOVA or (I) one-way ANOVA with (H,I,J,K) Tukey’s multiple comparison test.

For insight into the potential mechanism of observed phage and bacterial coexistence, we cross-streaked fecal E. coli colonies against phage to determine the frequency of phage resistance. Of the 1,920 colonies tested across phage-colonized mice and timepoints before and after phage administration (Days 0, 3, and 18), all colonies were phage susceptible. This suggests that resistance development may either be below our detection limit, E. coli persists in spatial refuges inaccessible to phage⁴³ and/or that phage resistance leads to poor bacterial fitness.⁴⁴ Additionally, we confirmed that T4::clpB phage increased the expression of ClpB protein with E. coli fluorescence microscopy. As shown in Figure 5D and 5E, E. coli (n=40) cultured with T4::clpB phage had significantly greater labeling with an anti-α-MSH antibody compared to E. coli (n=27) cultured with wildtype T4, in vitro. This anti-α-MSH antibody cross-reacts to bind ClpB. Similar examination of intestinal samples from mice at Day 3 shows significantly greater ClpB production from bacteria in mice (n=3 per group) colonized with T4::clpB compared to wildtype T4 phage (Figures F,G).

During the course of this experiment, murine weights showed that mice colonized with T4::clpB phage had significantly reduced weight compared to mice colonized with no phage or wildtype T4 phage (Figure 5H). Past work has shown that mice gavaged daily with bacteria encoding the clpB gene (e.g., E. coli or Hafnia alvei) had reduced weight gain that was associated with reduced daily food intake.¹⁷ We similarly observed reduced food consumption in mice colonized with T4::clpB phage (Figure 5I), which was associated with increased night activity (Figure 5J).

DIO leads to inflammation in murine models,⁴⁵ which mimics the low-grade chronic inflammation in humans.^46,47 As further support for a physiological effect of in situ ClpB production, we found that several pro-inflammatory cytokines were reduced in mice colonized with T4::clpB phage compared to wildtype T4 phage (e.g. IL-1α and IL-1β), or both wildtype T4 and bacteria alone (IL-23). Other cytokines were also reduced, including MCP-1, IL-17A, and GM-CSF, but did not reach statistical significance (Figure 5K). IL-23 is linked to an array of autoimmune inflammatory diseases,⁴⁸ however its role in obesity is less clear.^49,50 An upregulation in IL-23 stimulates T-cell differentiation, leading to an increase in Th17 cells,^48,51 which leads to over production of IL-17 and several other proinflammatory cytokines including IL-1.^48,51 Reductions in IL-23, IL-1α/β, and IL-17A likely points to a decrease in IL-23/IL-17 axis formation^48,51 and an overall decrease in systemic inflammation. MCP-1 levels further support this claim, with high concentrations of this cytokine being found in obese individuals^52-54 and linked to both macrophage invasion of adipose tissue⁵² and heightened insulin resistance.^52,54 Histopathological analysis of liver and colonic tissue revealed no significant differences in lymphoid aggregates in the colon or the number and diameter of inflammatory foci in the liver (Supplementary Figure S5). However, mice colonized with T4::clpB phages had significantly thicker colonic mucosa compared to mice colonized with wildtype T4 phage or bacteria alone. Increased mucosal thickness is correlated with gut barrier function and studies have shown that the stimulation of enteroendocrine cells to release peptide hormones like GLP-1 can improve barrier function during DIO.⁵⁵

A practical concern for any therapeutic intervention mediated by the gut microbiota is whether such approaches lead to unintended adverse effects on the mammalian host. In a follow up murine study, we used mice similarly colonized with E. coli alone (n=5), E. coli with wildtype T4 (n=5) or E. coli with T4::clpB phages (n=5) as described in Figure 5A, but sacrificed on Day 3 for analyses prior to the high-fat diet. As shown in Supplementary Table S1, serum chemistry analysis revealed no significant differences in glucose, urea nitrogen, creatinine, or total protein between mouse groups. Serum alanine aminotransferase values were significantly lower in wildtype T4 and T4::clpB phage conditions compared to the E. coli alone condition, which may indicate an amelioration of hepatocellular injury mediated by phages, though all were within normal expected ranges. Whole-body histopathology of mice showed no significant aberrations among organs between groups (Supplemental Figure S6).

Finally, we examined whether our engineered phage induced significant alterations to the gut microbiota that may contribute to the observed effects in the DIO mouse model. We had previously confirmed that no E. coli was present in the murine gut after streptomycin until we introduced E. coli K-12. Also, T4 phage is specific to E. coli,⁵⁶ and so it would not directly infect other bacteria present. However, virulent phages can induce indirect effects to the gut microbiota as mediated by interbacterial interactions.¹³ Examination of the fecal microbiota of mice on Day 3 revealed no significant differences in observed OTUs or Shannon diversity between experimental groups (Supplemental Figures S7A and S7B). Qualitative comparison of bacterial composition at the Order level revealed a loss of Erysipelotrichaceae in the wildtype T4 group (Supplemental Figures S7C) and distinctive clustering of mice by experimental groups as shown by principal coordinate analysis of Bray-Curtis dissimilarity (Supplemental Figures S7D). When comparing taxa (Supplemental Figure S7E), we found several bacterial Families that had significant differences between groups but only two Families, Erysipelotrichaceae and Paraprevotellaceae, were significantly increased for the T4::clpB group compared to the other two groups. These Families are enriched in obese⁵⁷ and overweight individuals,⁵⁸ respectively, and do not plausibly explain the reduced weight gain in the T4::clpB condition.

Discussion

Our work demonstrates that an engineered virulent phage, T4, can orchestrate the production and release of heterologous proteins in the mammalian gut during the normal phage infection process, leveraging its coexistence with its bacterial host for a sustained in situ effect. By engineering T4 phage to express sfgfp under a late T4 phage promoter, gp22, we show that this reporter protein is significantly expressed in vitro and in vivo. We then show that T4 producing the protease inhibitor, Serpin B1a, can reduce NE activity in the murine gut during colitis. Finally, show that expression of a protein, ClpB, can significantly reduce weight gain in a DIO murine model.

Virulent phages are abundant and prevalent components of the human gut that can exhibit a long-term stability.⁹ Exemplar is the prevalence of crAssphage, a Bacteroides phage found in healthy humans across diets and geography.⁵⁹ By using this co-existence we show it is possible to direct a resident bacteria to express proteins. This is an alternative to engineering a bacterium to both produce and secrete the biologic, which can compromise the bacterium’s ability to colonize the gut.^17,60-62 While we generally detect both phage and bacteria in the gut, on occasion we observe levels below detection in one sample that rebound in the next sample.It is possible that high concentrations of T4 phage may make it difficult to accurately measure E. coli colonies due to phages killing the bacteria on the culture plates.⁶³ Also, in the DIO mouse model it appears that wildtype T4 phage levels trend downward for some mice despite a continued persistence of T4::clpB phages which have a lower burst size. Although we remain able to observe phenotypic effects due to our engineered phage, it is important to account for such potential dynamics in the mammalian gut.

We confirmed that heterologous gene expression during T4 phage infection required phage-specific promoters, which ensures that gene expression is contingent on phage infection. Expression from the T4 phage genome progresses through early, middle, and late stages with the transcription of late T4 promoters requiring T4-encoded RNAP-binding proteins Gp33 and Gp55 among other factors.⁶⁴ This limits the possible scenarios of expression from late phage promoters to the latter phases of successful phage infection. A major concern with any genetically-engineered microbe is potential dissemination to the environment. For bacteria with targeted effect in the gut, a number of biocontainment strategies have been investigated including nutrient auxotrophy,⁶¹ temperature-sensitive toxin-antitoxin systems,⁶⁵ and chemically-dependent CRISPR systems.⁶⁶ Similar strategies that conditionally allow gene expression or phage propagation could be integrated into future strategies. Additionally, phages are unable to independently proliferate and require a bacterial host, which provides an additional layer of biocontainment.

Our work is a proof-of-concept demonstration that an engineered virulent phage can be used for the in situ production of heterologous proteins. There remain several considerations for the future application of this work. First, we used E. coli as the bacterial host because it is prevalent in the human gut with generally high colonization densities (10⁷ −10⁹ cfu/g stool).^67,68 Although we had comparable colonization levels in our mouse model, our experiment was conducted under controlled environmental and dietary conditions. The impact of several factors including diet, medication, and inter-individual variation would likely be influential and warrants future study. Second, some phages have narrow strain-level specificity for bacterial hosts and the success of this approach may depend on whether a susceptible strain is present. Strategies used in phage therapy can be used for guidance, including using broad host range phages, phage cocktails, or testing a phage bank against bacterial isolates.⁶⁹ Such approaches would likely also be useful if phage resistance develops. While we do not observe substantial alterations in the murine microbiome, the possibility of phages disrupting the gut microbiome should be monitored. Third, our understanding of the significance of gut phages to the mammalian host remains limited. Studies have shown that the phage particle itself can have several host interactions such as binding to the intestinal mucosa to increase local bacterial lysis,⁷ interacting with the immune system,⁷⁰ and translocating across intestinal epithelial cells.⁷¹ Phages may have a yet unknown beneficial role in the gut. There is a core of virulent phages in the gut microbiome of healthy individuals that is depleted in individuals with inflammatory bowel disease.⁷² Finally, the choice of protein to express is important. Using Serpin B1a, we disrupted neutrophil inflammation by reducing NE activity and reprogramming neutrophils as shown with increased anti-inflammatory surface mediator CD24. CD24 is a potent immune-suppressor that can propagate tissue homeostasis and reduce surrounding cellular inflammation.^36,37 Using ClpB, we reduced weight gain associated with diet-induced obesity over multiple weeks. This resulted in a reduction of pro-inflammatory cytokines and an increase in mucosal thickness. Considering that these diseases are complex and multifaceted, the in situ production of other protein therapeutics to target other aspects of these diseases, whether together, at different stages, or alone, would be exciting future lines of investigation.

Methods

Molecular cloning

Genetic constructs were designed in Benchling and assembled using Golden Gate Assembly⁷³ (NEB, Catalog # E1601L) with assistance from the NEB Golden Gate Assembly Tool. Plasmids were transformed into DH5α cells (NEB, Catalog # C2987H) and verified for accuracy via Sanger sequencing (Virginia Tech Genomics Sequencing Center). For challenging plasmids, NEB Stable cells (NEB, Catalog # C3040H) were used. T4 phage promoters were identified from previous screens,²¹ and the constitutive E. coli promoter J23119 from past work.⁷⁴ The fluorescent E. coli used to study the release of intracellular fluorescent protein is a sfgfp⁺, mrfp⁺ E. coli K-12 MG1655 strain, a gift from Stanley Qi.⁷⁵ While both GFP and RFP are expressed, the former is used as the intracellular indicator protein. The sfgfp gene used in T4 phage was cloned from this fluorescent strain while clpB was cloned from E. coli Nissle 1917 using PCR to include an N-terminal His₆-tag. The serpinb1a gene sequence was synthesized via gBlock (GeneArt, Thermo Scientific DNA) prior to N-terminal His₆-tag addition and further cloning. DNA oligos were custom synthesized by Thermo Fisher and have been provided in the supplementary materials.

In general, genetic constructs in a pGGASelect vector contained a promoter, a ribosome-binding site, and a gene (e.g., sfgfp, serpinb1a, or clpB) with flanking regions of ~200 bp of homology to the ac gene in the T4 genome. DNA sequences are provided in Supplemental Table S2. Recombinant T4 phages were generated by spontaneous recombination.^20,76,77 Plasmid-containing E. coli were grown to log phase and cultured with wildtype T4 phage (ATCC, Catalog # 11303-B4) at a multiplicity of infection (MOI) = 0.01 overnight at 37°C. Lysates were treated with chloroform, centrifuged (4600 x g/10 min/4°C) to remove bacterial debris and sterile-filtered with a 0.45 μm syringe filter. To select for recombinant phages from the crude phage lysate, 100 μL phage lysate and 300 μL log phase E. coli DH5α were mixed with 3 mL of molten LB top agar (containing 0.3% agar) at 50°C and 15 μL of 17 mg/mL acriflavine hydrochloride, then overlaid onto LB agar plates. After overnight culture at 37°C, plaques were purified by suspension into phage buffer and plaque assay against E. coli K-12 (ATCC, Catalog # 700926). Recombinant phages were identified by Sanger sequencing from purified plaques.

Fluorescence assays

To measure intracellular protein release by phage infection, fluorescent E. coli K-12 was grown at 37°C to log phase (OD_600nm = ~0.3-0.5) in TNT medium (10 g tryptone, 5 g sodium chloride, and 1 mg thiamine in 1 L water) and diluted to an OD_600nm of 0.05 before infection with wildtype T4 phage (MOI = 1). To separate supernatant from bacterial cells, samples were centrifuged at 2000 x g for 5 min. After the supernatant was removed, bacterial pellets were resuspended in an equivalent volume of TNT medium. OD_600nm and fluorescence (ex. 485 nm/em. 528 nm) were measured from supernatant and resuspended bacterial pellets. Additional fluorescent E. coli and wildtype T4 phage growth experiments were used to supplement those above, growing E. coli in TNT medium overnight at 37°C, before back diluting to log phase. Cultures were diluted to an OD_600nm of 0.05 or 0.35 before adding either phage buffer or wildtype T4 phage at a MOI of 1 or 0.001. All conditions were incubated at 37°C in a BioTek Cytation 3 Imaging Reader overnight, measuring OD_600nm every 5 min for 12 h.

To measure sfGFP production when sfgfp gene is encoded in the T4 phage genome, E. coli K-12 was grown to log phase in M9 minimal media supplemented with 0.4% glucose and 1 μg/mL thiamine, and diluted to an OD_600nm of 0.05. The bacterial culture was then mixed with wildtype T4 phage, engineered T4 phages, or the buffer vehicle (MOI = 0.001) and shaken at 37°C for 12 h. Bacterial cells were removed by centrifugation at 18000 x g for 30 sec, before measuring the supernatant fluorescence (ex. 485 nm/em. 528 nm). Phage concentrations in the supernatant were quantified by spot assay.

In vitro inhibition of neutrophil elastase via phage-generated serpin B1a

Wildtype and engineered phage lysates were prepared for the purification and concentration of Serpin protein. 50 mL of E. coli K-12 cells were grown to log phase (OD_600nm = ~0.3-0.5) in LB broth and infected with phage (MOI = 0.001) at 37°C overnight. The resulting lysate was centrifuged (16,000 x g/1 h/4°C) and filtered with a 0.2 μm filter to remove bacterial debris prior to His₆-tag purification. A small aliquot of culture was taken prior to centrifugation and plated for cfu and pfu measurements.

Ni-NTA resin (G Biosciences, Catalog # 786-939) was prepared according to manufacturer’s instructions and incubated with purified phage lysate (1 h/4°C/shaking 200 rpm). Poly-Prep chromatography columns (Bio-Rad, Catalog # 7311550) were used to isolate the resin, discarding the resulting flow through. The resin was washed three times with 5 column volumes (CV) of Wash buffer (1x PBS), before eluting the purified Serpin protein five times with 2 CV of Elution buffer (1x PBS, 150 mM imidazole). Concentration of protein and buffer exchange were accomplished utilizing a 3 kDa centrifugal filter (Pall Corporation, Catalog # MCP003C41). The resulting elution was centrifuged (3000 x g/30 min/4°C) and concentrated before performing four subsequent washes with HEPES buffer (20 mM/pH 7.4), increasing the speed and duration of centrifugation for the final spin (4000 x g/45 min/4°C). This would result in an approximate 100-fold increase of Serpin protein. Predicted total protein concentration was assessed via Bradford assay.

Assessment of NE inhibition via phage-based Serpin or the small molecule inhibitor sivelestat (Caymen Chemical, Catalog # 17779) followed previous protocols.²⁸ In short, 100 ng/μL purified NE enzyme (R&D Systems, Catalog # 4517-SE-010) was incubated with 100 ng/μL cathepsin C in equal volumes for 2 h at 37°C to induce enzyme activity, before diluting to 0.2 ng/μL in NE assay buffer (50 mM Tris, 1M NaCl, 0.05% w/v Brij-35, pH 7.5). His₆-tag purified Serpin protein – and sivelestat resuspended to 0.1 mM in DMSO – were serially diluted in NE assay buffer. 20 μL of 0.2 ng/μL NE and 20 μL of diluted sivelestat or Serpin were incubated for 30 min at 37°C prior to the addition of 10 μL of fluorogenic elastase substrate MeOSuc-AAPV-AMC (EMD Millipore, Catalog # 324740-25MG) at 500 μM. Kinetic cleavage was measured at room temperature for 12 h (ex. 380 nm/em. 460 nm) to determine enzyme activity and inhibition.

In vitro production of the ClpB protein via T4 phage infection

E. coli K-12 cells were grown to log phase (OD_600nm = ~0.3-0.5) in a 50 mL LB broth culture, before infecting with wildtype T4 phage or T4::clpB overnight. Post incubation, a small aliquot of culture was taken and plated for cfu and pfu measurements. The remaining culture was centrifuged at high speed (16,000 x g/1 h/4°C) to pellet residual bacteria and cellular debris. The resulting supernatant was filtered with a 0.2 μm filter to ensure sterility, storing aliquots at −20°C.

In vitro ClpB protein concentration was determined via a His₆-tag ELISA (GenScript, Catalog # GSCRPT-L00436), adhering to manufacturer’s instructions. Interpolated values were normalized to total protein concentration in the lysate, obtained using a Bradford assay (Fisher Scientific, Catalog # AAJ61522AP).

Additional measures of ClpB production were accomplished via fluorescent imaging.⁷⁸ In short, E. coli K-12 cells were grown overnight in LB at 37°C and back diluted to mid-log (OD_600nm = ~0.4). 10 mL of culture was infected with either wildtype T4 phage or T4::clpB (MOI = 0.1) and allowed to grow for 20 min at 37°C. 5 mL of co-culture was centrifuged (4000 x g/5 min) and resuspended in 1 mL of 4% paraformaldehyde (PFA), fixing for 15 min at room temperature. Fixed cells were centrifuged (2000 x g/5 min) and washed twice with 1X PBS prior to another centrifugation and resuspension in 70% EtOH (10 μL per 1E8 cfu). The culture was incubated (1 h) in a nutator before adding 5 μL of cells to 5 μL of DI water, centrifuging a final time (2000 x g/2 min) prior to resuspension with 130 μL of 1X PBS. The solution was added to a poly-L-lysine slide and allowed to adhere for 1 h at 4°C, before incubating with blocking buffer (2% fetal bovine serum, 1% bovine serum albumin, 0.2% Triton X-100, 0.05% Tween 20) for 1 h at room temperature. Slides were incubated (2 h) with a primary rabbit anti-alpha-MSH antibody (Origene, Catalog # TA364148)¹⁶ diluted 1:100 in blocking buffer, prior to washing with 1X PBS for 3 min. A 1:1000 diluted goat anti-rabbit secondary antibody (Abcam, Catalog # AB150081) conjugated with Alexa Fluor 488 was then incubated for an additional 2 h, washing the resulting slide twice with 1X PBS for 3 min each. Slides were counter stain with DAPI (1 μg/mL) for 15 min at room temperature, washed briefly with DI water and dried completely before mounting with Prolong Glass Antifade Mountant (ThermoFisher Scientific, Catalog # P36980). Slides were cured for 24 h in the dark.

Fluorescent imaging was accomplished via the Nikon AXR point scanning confocal built around an inverted Nikon Eclipse Ti2 microscope, equipped with 8 solid state lasers, 3 GaAsP PMTs and 1 NIR GaAsP PMT. Images were captured with a PLAN APO λD 60x/1.42 Oil Immersion objective (MRD71670) under fast scanning speed. Mean fluorescent intensity measurements were compared via FIJI ImageJ software.

Phage burst size and latency period

Assessment of wildtype and engineered T4 phage burst size and latency period followed a previously established protocol with minor alterations.⁷⁹ In short, E. coli K-12 was grown overnight in LB at 37°C before back diluting into 10 mL of LB and growing to mid-log (OD_600nm = ~0.4). Cultures were incubated on ice for 10 min and centrifuged (3700 x g/5 min), removing the supernatant and resuspending in M9 minimal media (0.4% glucose). Cultures were divided into 5 mL conical tubes and infected with either wildtype T4 or T4::clpB (MOI = 0.01). After incubating at 37°C for 7 min, both conditions were diluted 100-fold in LB. The resulting dilutions were incubated at 37°C for 40 min, collecting 100 μL aliquots every 5 min prior to centrifugation (18000 x g/30 sec) and assessment of free phage via spot assay. Burst size was calculated by identifying the change between initial and final phage concentration, and dividing by the number of infected cells. Latency period was determined as previously described, calculating the period at which 50% of the virions had been released.

Animal Experiments

Housing and husbandry

Mice were obtained from Jackson Laboratories and allowed to acclimate for one week. Mice were housed in rooms with controlled air, temperature (68-79°C), and humidity (30-70% relative humidity) with a 12h:12h light:dark cycle. Food and water were provided ad libitum, unless otherwise noted. All animal studies were performed in compliance with Institutional Animal Care and Use Committee guidelines at Virginia Tech under protocols #20-097 and #23-260.

Mucosal sfGFP via engineered phage

C57Bl/6 mice (Stock 000664, female, 7-8 wks old, n=12) were randomized, before receiving 5 g/L of streptomycin provided in the drinking water (Day −4), which was replaced every 3 days. After 1 day of streptomycin treatment (Day −3), mice were fasted for 4 h and then received by oral gavage 100 μL of bacteria suspended in 1X PBS. Food was returned immediately after oral gavage. Bacterial inoculum was prepared from streptomycin-resistant E. coli K-12 cultures grown overnight in LB broth with 100 μg/mL of streptomycin, washing twice and diluting 10-fold in 1X PBS. After 3 days, on Day 0, mice received 100 μL wildtype T4 phage or T4::sfgfp by oral gavage. Phage solutions were prepared from 10⁸ pfu/mL phage stocks by diluting 1:10 into 0.1 M sodium bicarbonate immediately before gavage. Stool was regularly collected for bacteria and phage quantification.

To quantify fecal bacteria and phage, fecal pellets from each mouse were suspended in phage buffer at 50 mg/mL and homogenized manually with sterile wooden sticks. Large debris were allowed to settle for 1 min prior to selective plating on MacConkey agar plates containing 100 μg/mL of streptomycin to quantify fecal E. coli. The remaining fecal sample was treated with 10 μL of chloroform and centrifuged at 4600 x g for 10 min at 4°C. Phage was quantified by spot assay from this supernatant. Measurements were collected longitudinally for each mouse and experimental condition.

After euthanasia, colonic tissue was preserved in an OCT frozen block (Tissue-Tek, Sakura Finetek Catalog # 4583; Tissue-Tek Cryomold, Sakura Finetek Catalog # 4566) and maintained at −80°C prior to tissue processing.

Engineered phage reduction of DSS-induced colitis via protease inhibition

Murine models of DSS-induced colitis mirrored those above with slight alterations. On Day 0, C57Bl/6 mice (Stock 000664, female, 6 wks old) received 100 μL of wildtype T4 or engineered T4::serpin phage by oral gavage. On Day 2, dextran sulfate sodium salt (MP Biomedicals, Catalog # 160110, Lot # S8634) was supplemented in the drinking water at 1.5% to induce colitis, being replaced every 3 days to ensure potency. Stool was also collected intermittently to assess phage and bacterial concentrations, as well as fecal neutrophil elastase activity.

Immediately following euthanasia, intracardiac blood samples were collected and allowed to coagulate at room temperature for 30 min, before centrifuging (18000 x g/10 min/4°C) to obtain the resulting serum. Segments of colon tissue were isolated and preserved in 10% neutral buffered formalin for histology.

Lamina propria cell isolation and Flow cytometry

Colon tissue was collected at the end of the DSS-water regimen. Lamina propria cells were prepared as previously described.³³ Briefly, after cleaning with ice-cold 1X PBS, colon tissue was cut into ~0.5 cm small pieces, and then incubated in HBSS containing 5 mM EDTA, 5% fetal bovine serum, and 1 mM DTT to dissociate epithelial cells. The remaining pieces were minced and incubated in the digestion solution provided by Lamina Propria Dissociation Kit (Miltenyi Biotec, Catalog # 130-097-410) according to the manufacturer’s instruction. The lamina propria cells were resuspended in the FACS buffer (1X HBSS supplemented with 2% fetal bovine serum).

For flow cytometry analyses, the cells were pre-incubated with Fc Blocker (BD Biosciences, Catalog # 553142), then stained with fluorescent-conjugated antibodies against mouse CD11b (Biolegend, Catalog # 101226), Ly6G (Biolegend, Catalog # 127605), CD45 (Biolegend, Catalog # 103113), and CD24 (Biolegend, Catalog # 101813). After washed, the cells were resuspended in FACS buffer containing Propidium Iodide (Thermo Fisher, Catalog # P3566) and detected by FACS Canto II (BD Biosciences). Data were analyzed with FlowJo (Ashland, OR).

Fecal neutrophil elastase activity assay

Following previous protocols,²⁸ neutrophil elastase was extracted from fecal samples resuspended at 100 mg/mL in Fecal Protein Extraction Buffer (FPEB; 50 mM Tris, 150 mM NaCl, pH 7.5). Resuspended material was incubated on ice for 30 min, vortexed briefly every 5 min, before centrifuging (2000 x g/10 min/4°C) and isolating the resulting supernatant. Samples were then filtered through a 0.22 μm centrifugal filter (15,800 x g/30 min/4°C) (Corning Costar, Catalog # CLS8160) and appropriately diluted in FPEB.

Determination of NE concentration in stool followed previously established protocols with minor alterations.²⁸ After the purified NE enzyme was activated, it was diluted in NE assay buffer to generate a standard curve. Then, 25 μL of sample or standard were incubated with 25 μL of a fluorogenic neutrophil elastase substrate (200 μM) at room temperature, measuring the kinetic cleavage for 12 h (ex. 380 nm/em. 460 nm).

Engineered phage treatment of diet-induced obesity

Murine models of DIO mirrored those above with minor alterations. For the DIO model, C57Bl/6 mice (Stock 000664, male, 6 wks old) weighing 19 to 22 grams⁸⁰ were randomly assigned to three experimental groups. On Day 0, mice received 100 μL of a phage buffer vehicle, 100 μL of wildtype T4 phage or 100 μL of T4::clpB phage by oral gavage. On Day 3, food was switched from a standard rodent chow (Teklad 2918) to a high-fat diet (Research Diets D12451i). Food was replaced weekly during which stool was collected and body weights measured. Fecal E. coli colonies isolated from mice receiving wildtype T4 or T4::clpB phages were tested for phage resistance by cross-streaking against 10⁸ pfu/mL of the same purified phage. At least 20 colonies per mouse per timepoint were examined. Measurements were collected longitudinally for each mouse and experimental condition. Prior to sacrifice, a subset of mice was individually housed – cage dimensions mirrored those of the home cage (5.5”W, 6.5”H, 9”L) – and monitored via the Oxymax Comprehensive Lab Animal Monitoring System (CLAMS) (Columbus Instruments). Food (Research Diets D12451i) and water, supplemented with streptomycin at 5 g/L, were provided ad libitum. All mice were allowed to acclimate overnight, before collecting CO₂ production, oxygen consumption, animal activity, and energy expenditure every 2 min for a complete day-night cycle (7 AM – 7AM).

Immediately after euthanasia, blood was collected by intracardiac bleeding, followed by colonic and liver tissue preservation in an OCT frozen block (Tissue-Tek, Sakura Finetek Catalog # 4583; Tissue-Tek Cryomold, Sakura Finetek Catalog # 4566) or in 10% formalin (Fisher, Catalog # 316-156). Intestinal mucosa was collected from select colon sections by gentle scraping with a metal spatula and suspension at 50 mg/mL in phage buffer. Serum samples (n=5) were extracted as described above and assessed for cytokine concentrations with the LEGENDplex (BioLegend, Catalog # 740446) Mouse Inflammation Panel (13-plex). All samples were analyzed using a FACSCanto II (BD Biosciences, Reference # 338960) in technical duplicate and processed according to the manufacturer’s instructions.

Engineered phage impact on murine health

Murine health assessments of T4::clpB mirrored those previously described with minor alterations. On Day −3, C57Bl/6 mice (Stock 000664, male, 7 wks old) were gavaged with E. coli K-12, allowing colonization levels to stabilize for three days. On Day 0, mice were gavaged with phage buffer, wildtype, or T4::clpB phages, permitting co-colonization to endure for three additional days before sacrifice. A full body necropsy was performed to identify murine health status, sacrificing each mouse via CO2 asphyxiation prior to heart puncture. Collected blood samples were processed as previously described to obtain the serum and submitted to ViTALS for a chemistry panel. All associated tests were performed in a Beckman Coulter AU480 within 3 h of sample submission, with the resulting levels being evaluated by a board-certified veterinary pathologist, Dr. Priscila da Silva Serpa, using reference intervals.⁸¹ Organs were collected and fixed in 10% neutrally buffered formalin. Tissue samples were processed and stained by the Virginia Tech Animal Laboratory Services at the Virginia-Maryland College of Veterinary Medicine before assessment of inflammation. Tissue samples were evaluated histologically by a board-certified veterinary pathologist, Dr. Teresa Southard. Colons were scored for lamina propria inflammation (0-1 = normal complement of resident inflammatory cells; 2 = multifocal areas with increased cellular infiltrates; 3 = diffuse and/or severe cellular infiltrates); crypt dilation/necrosis (0= no crypt changes detected; 1 = rare crypts [fewer than 4] dilated and/or lined by necrotic epithelial cells; 2 = moderate numbers of crypts affected [4-10]; 3 = more than 10 affected crypts). Mucosal thickness was measured in an area where crypts were perpendicular to the epithelial surface, and the diameter of the largest submucosal lymphoid aggregate was measured. Other changes were also noted, such as vascular mineralization in the colon of one mouse.

Assessment of microbial communities via 16S sequencing was accomplished via SeqCoast Genomics. Cecum samples were lysed with MaxMAX microbiome bead beating tubes and extracted via the DNeasy 96 PowerSoil Pro QIAcube HT kit prior to sequencing. Samples were prepped for 16S V3/V4 amplicon sequencing using the Zymo Quick-16S Plus NGS Library Prep Kit and unique dual indexes. Post amplicon generation, sequencing was performed on the Illumina NextSeq2000 platform using a 600-cycle flow cell kit to produce 2x300bp paired reads. 30-40% PhiX control (unindexed) was spiked into the library pool to support optimal base calling of low diversity libraries on patterned flow cells. Read demultiplexing, read trimming, and run analytics were performed using DRAGEN v4.2.7, an on-board analysis software on the NextSeq2000. All associated bioinformatics was performed using open-source software Qiime2 (version 2019.10.0) and DADA2 (version 2019.10.0) within the SeqCoast established pipeline.^82,83

Histology and fluorescence imaging of sfGFP

All tissue preparation, slicing, and H&E staining was performed by Virginia Tech Animal Laboratory Services at the Virginia-Maryland College of Veterinary Medicine. Mucin and gastric epithelial cell nuclei staining was accomplished via sectioning of OCT preserved murine colonic tissue. Select tissue slices were fixed and dehydrated.⁸⁴ Post fixation, all slides were incubated with blocking buffer (2% fetal bovine serum, 1% bovine serum albumin, 0.2% Triton X-100, 0.05% Tween 20) for 1 h at room temperature and washed twice with 1X PBS. Slides were incubated with 40 μg/mL WGA (Vector Labs, Catalog # RL-1022), 10 μg/mL UEA I (Vector Labs, Catalog # RL-1062-2), and 1 μg/mL DAPI (Thermo Scientific, Catalog # 62248), diluted in 1X PBS, for 15 min at room temperature. All slides were washed briefly with DI water and allowed to air dry before mounting.

Fluorescence microscopy for unfixed slides was performed with a Nikon Eclipse Ti confocal microscope, utilizing a 20x/0.5 CFI Plan Fluor objective (MRH00201). Mean fluorescence intensity of unfixed slides was determined with ImageJ by measuring the intensity of ten mucosal regions per image across two regions per mouse and then normalized to the background tissue intensity. Imaging for WGA, UEA I, and DAPI stained slides was performed with a Nikon CSU-W1 SoRa spinning disk confocal built around an inverted Nikon Eclipse Ti2 microscope, utilizing a 20x/0.8 CFI Plan Apochromat Lambda D objective (MRD70270) and a 60x/1.49 CFI Apochromat TIRF Oil Immersion objective (MRD01691) under 2.8x magnification. Images were captured with a Hamamatsu Photonics ORCA-Fusion BT sCMOS camera under fast scanning speed, 50 micron spinning disk pinhole size, and 7 total laser lines. All processing was performed on Nikon NIS-Elements software.

Phage-derived ClpB production histology and fluorescence imaging

Colon staining for ClpB production in vivo mirrored that described above. OTC frozen colonic tissue sections were dried at room temperature for 30 min prior to fixation with 4% PFA (10 min), washing briefly with 1X PBS and DI water. Tissue sections were then incubated with increasing concentrations of EtOH (50%, 80%, 95%) for 3 min each before washing briefly with 1X PBS. Blocking buffer (2% Fetal Bovine Serum, 1% Bovine Serum Albumin, 0.2% Triton X-100, 0.05% Tween 20) was added for 1 h at room temperature, followed by incubation with the primary anti-alpha-MSH antibody (Origene, Catalog # TA364148) diluted 1:100 for 2 h. Slides were washed for 3 min with 1X PBS before secondary antibody (Abcam, Catalog# AB150081) incubation for an additional 2 h, diluting 1:1000 prior to slide application. Stained tissue was washed twice with 1X PBS for 3 min each, before counter staining (DAPI, WGA, UEA I) and mounting as described above.

Assessment of fluorescence intensity of bacterial cells within colonic tissue was performed on a Nikon CSU-W1 SoRa spinning disk confocal built around an inverted Nikon Eclipse Ti2 microscope, using a 60x/1.49 CFI Apochromat TIRF Oil Immersion objective (MRD01691) under 2.8x magnification. Images were captured with a Hamamatsu Photonics ORCA-Fusion BT sCMOS camera under fast scanning speed, 50 micron spinning disk pinhole size and 7 total laser lines. Mean fluorescent intensity measurements were taken from E. coli cells within the colon of 6 mice colonized with bacteria and either wildtype or engineered T4 phage (n=3 per condition), subtracting the associated background fluorescence and comparing the resulting values via FIJI ImageJ and JMP Pro statistical software.

Availability of materials

The recombinant strain T4::sfgfp phage is available from the Félix d’Hérelle Reference Center for Bacterial Viruses at the Université Laval. Other strains are available from the corresponding author upon request.

Supplementary Material

Supplementary Information

Supplementary Figure S1: Fluorescent E. coli K-12 (n=3 biological replicates) growth alone and infected with wildtype T4 phage at different multiplicities of infection (MOI) in TNT medium, as measured in a plate reader, with bacteria starting at an (A) OD_600nm=0.05 or (B) OD_600nm=0.35 in a 1-cm pathlength.

Supplementary Figure S2: Representative gating strategy for flow cytometry.

Supplemental Figure S3: Quantitation of His₆-tagged ClpB and supernatant protein of phage lysates (n=3 biological replicates). Total lysate concentration analyzed via Bradford reagent (A) was used to normalize His₆-tag ELISA results (B), indicating significant production of His₆-tagged protein via phage infection (C). Statistical analyses were performed by a two-tailed unpaired t-test.

Supplemental Figure S4: Phage parameters of wildtype T4 and T4::clpB phages. (A) One-step growth curve for wildtype T4 phage with E. coli K-12 (control) and (B) T4::clpB phage with E. coli K-12. Dashed lines represent individual replicates and solid lines represent means. (C) Burst size and (D) latency calculated from the one-step growth curves (n=3 biological replicates). Statistical analyses were performed by a two-tailed unpaired t-test.

Supplemental Figure S5: Histological assessment of colon and liver tissue inflammation after diet-induced obesity (n= 8 mice per condition). (A) Colon lymphoid aggregates showed no significant difference between conditions, but (B) mucosal lining thickness was significantly increased among mice colonized with bacteria and engineered phage compared to other conditions. (C) Quantification of liver inflammatory foci and (D) foci diameters showed no between conditions. Statistical analyses were performed by one way ANOVA with Tukey’s multiple comparison test.

Supplemental Figure S6: Representative histological images of all major organs from a full body necropsy post colonization with bacteria E. coli alone, E. coli + wildtype T4 phage, and E. coli + T4::clpB (n=5 mice per condition). Images shown are representative results from n=5 mice per condition. There was no observable inflammatory response from phage colonization, or production and release of protein via engineered phage infection.

Supplemental Figure S7: Cecal microbiome profiling of mice by 16S amplicon sequencing on Day 3, including E. coli alone (control) and co-colonized with wildtype T4 or T4::clpB phage (n=5 mice per condition). (A) There were no statistical differences in OTUs or (B) Shannon diversity. (C) Relative bacterial abundances of individual mice at the Order level, and (D) principal coordinate analysis of Bray-Curtis dissimilarities. (E) Comparison of bacterial abundance at the Family level shows significant changes for some taxa between experimental groups. Statistical analysis by two-way ANOVA with Tukey’s multiple comparison test (****,p<0.0001).

Supplemental Table S1. Serum chemistry of mice colonized by either E. coli alone, E. coli and wildtype T4 phage, or E. coli and T4::clpB phage.

Data availability statement

Source data for figures in the main text and supplementary information are provided with this manuscript. 16S amplicon reads are available on Mendeley Data (Baker, Zach. 16S Illumina sequencing reads. Mendeley. https://data.mendeley.com/datasets/pt9m6b9s4z/1/doi: 10.17632/pt9m6b9s4z.1 (2019)).

Code availability statement

No code was generated in this study.