Stable neutralization of virulent bacteria using temperate phage in the mammalian gut

Elimination or alteration of select members of the gut microbiota is key to therapeutic efficacy. However, the complexity of these microbial inhabitants makes it challenging to precisely target bacteria without unexpected cascading effects. Here, we use bacteriophage to deliver exogenous genes to specific bacteria by genomic integration of temperate phage for long-lasting modification. As a real-world therapeutic test, we engineered λ phage to transcriptionally-repress shigatoxin by using genetic hybrids between λ and other lambdoid phages to overcome resistance encoded by the virulent prophage derived from enterohemorrhagic E. coli. We show that a single dose of engineered phage propagates throughout the bacterial community and reduces shigatoxin production in an enteric mouse model of infection without markedly affecting bacterial concentrations. We thus minimize the selection for resistance by relying on anti-virulence and not anti-bacterial action. Our work reveals a new framework for transferring functions to bacteria within their native environment.


1
The human gut microbiota is a collection of microbes colonizing the gastrointestinal 2 tract and has been associated with various aspects of human health. 1 While this community 3 typically works in concert with our bodies, substantial perturbations such as antibiotics or 4 infections can disrupt the microbial balance and lead to long lasting dysbiosis. 2 In some 5 instances pathogenic bacteria do so by transmitting virulence factors encoded by these 6 pathogens to commensal bacteria through plasmid-based 3 and phage-based horizontal gene 7 transfer (HGT). 4 Remediating diseases associated with these pathogens while minimizing 8 unintended and disruptive effects to the surrounding microbiota remains challenging 5 9 especially with the limited tools available for targeting particular species. 6 10 Our ability to manipulate the composition and function of the gut microbiota is presently 11 limited in terms of precision and durability. 6 Antibiotics non-specifically decimate swaths of gut 12 species, 7 dietary changes affect both the overall microbiota and mammalian host, probiotics 13 poorly engraft due to colonization resistance, 8 and even highly-specific lytic phages can cause 14 cascading effects in the bacterial community despite targeting specific species. 9 While in some 15 cases these strategies may show transient efficacy, the emergence of resistant mutants can 16 impact therapeutic effect. Broadly resetting the gut microbiota through fecal microbiota 17 transplants (FMTs) has been promising especially for treating Clostridium difficile infections, 10 18 but they are difficult to characterize and may transmit unintended traits such as obesity. 11

19
An alternative strategy is to modify bacterial function within its native environment. For   can lead to hemolytic uremic syndrome. 16 Of the two main Stx variants-Stx1 and Stx2-the 33 latter is ~1000-fold more toxic. 17 Similar to a number of other prophage-encoded virulence 34 factors, 18 Stx is not expressed while the phage is in a lysogenic state, i.e. stably integrated into 35 4 the bacterial genome. It is not until induction, whether occurring spontaneously or from stimuli 1 such as antibiotics, that the lytic life cycle is activated to produce Stx2 19 and progeny phage 2 that can spread virulence genes to commensal E. coli species. 20 3 Instead of an antimicrobial strategy for killing pathogens, a genetic-based anti-virulence 4 strategy could neutralize virulence before expression and minimize resistance until the bacteria 5 has been completely shed from the gastrointestinal tract. Temperate phages offer a solution as 6 they are genetically engineerable and can integrate into the bacterial chromosome as 7 prophages for long-lasting effect as they confer fitness advantages to the bacterial host. 21 8 Instead of relying on a non-native constituent of the gut that could face practical barriers for 9 efficacy, temperate phages are abundantly found in human gut bacteria 22-24 and can constitute 10 large portions of the bacterial chromosome. 25

11
To modify the function of specific species within the complex community in the 12 mammalian gut, in a manner that avoids the emergence of resistance, we report the use of a    repopulate over time ( Figure 1A, upper panel). In contrast, temperate phages can also integrate 31 their genetic material into the host chromosome as a prophage to co-replicate with the 32 bacterial genome during cell proliferation ( Figure 1A, iv-v). In a bacterial population, this leads 33 to the lysogenic conversion of phage-susceptible species that coexist with phage-resistant 34 species ( Figure 1A, lower panel). Because anti-bacterial approaches can enrich for resistance, 35 5 including phage therapy which typically utilizes lytic phages, we aimed to engineer temperate 1 phages to deliver an anti-virulence payload that neutralizes virulence in a manner that 2 minimizes the selection for resistance.

3
To illustrate the feasibility of using a temperate phage, we show that bacteriophage l 4 transduces a substantial fraction of targeted bacteria in the mammalian gut. As shown in 5 Figure 1B, we used a streptomycin-treated mouse model to quantitate temperate phage 6 lysogeny on E. coli colonizing the mammalian gastrointestinal tract. One day after colonization 7 with E. coli MG1655, we introduced lBH1 phage by oral gavage and collected daily stool 8 samples for analysis of bacterial and phage titers. We constructed lBH1 from l phage by 9 inserting an antibiotic resistance cassette for quantification of lysogens ( Figure 2D). After oral 10 administration of lBH1 phage, we found that fecal phage levels reached equilibrium 11 approximately two days later and persisted at substantial concentrations (> 10 6 pfu/g stool) for 12 the duration of the experiment ( Figure 1C). As phage in the absence of its cognate bacterial 13 host is undetectable in the stool of mice ~2 days after administration, 26 our results indicate that 14 lBH1 phage is capable of continuous replication in the gut, enabling its expansion throughout 15 the bacterial population from a single dose. Furthermore, introduction of lBH1 phage did not 16 significantly alter fecal E. coli concentrations ( Figure 1D), which is in sharp contrast to lytic 17 phages that can cause substantial reduction. 9 Using antibiotic selection, we quantified the 18 number of fecal E. coli harboring the lBH1 prophage and found a substantial fraction (~17 to 19 30%) remained lysogenized by days 7 to 10 ( Figure 1E). Overall, these results indicate that the 20 temperate phage l is capable of widespread modification of its cognate bacteria in the gut.    lambdoid phages to overcome the superinfection exclusion. 29 We found that the efficiency of 17 plating (EOP) of l phage against E. coli 933W was ~10 6 -fold lower than that of the non-lysogen 18 ( Figure 2B), confirming its superinfection exclusion. We verified that this effect is not due to a 19 cI-based immunity ( Figure S1). We then tested the EOP for genetic hybrids of l phage in which 20 the l immunity region is swapped with that of other lambdoid phages (e.g. 21, 434, and P22),

23
Moreover, hybridization with the Salmonella phage P22 resulted in near complete recovery of 24 EOP to 78% for limmP22dis phage, indicating that lambdoid phages from non-cognate 25 bacterial hosts could be a reservoir for genetic orthologs that maintain phage function while 26 circumventing superinfection exclusion mechanisms.

15
and 933W×cI indinserted into the non-essential b2 region of the phage genome. lBH2 phage is a 9 product of a phage cross between lBH1 and limmP22dis resulting in a phage containing KanR 1 and 933W×cI indgenes with a P22 immunity region in a l phage background. Their respective EOP 2 against E. coli 933W are shown to the right. Symbols represent biological replicates with bars 3 representing the geometric mean.

18
Efficient gene transduction enables the delivery of anti-virulence genes. We inserted 19 genes for 933W×cI indand a kanamycin resistant cassette (to quantitate lysogeny) into the non-20 essential b2 region of l, 30 producing lBH1 ( Figures 2D and S3). We confirmed that 933W×cI ind-21 expressed from lBH1 was functional ( Figure S4). To overcome superinfection exclusion from 22 the 933W prophage, we utilized a P22 immunity region instead of a l immunity region. A phage 23 cross between lBH1 and limmP22dis resulted in the replacement of ~6 kb of the immunity 24 region of lBH1 with a ~5 kb portion of that from limmP22dis while retaining 933W×cI indand 25 Kan R genes (Figures 2D and S3, and Table S3). This new phage, lBH2, showed improved EOP 26 to 90% ( Figure 2D) and demonstrated a functional loss of l immunity and gain of P22 27 immunity, as well as expression of functional 933W×cI ind-( Figure S4).

29
Anti-virulence phage inhibits Stx2 production in vitro 30 Transcriptional repression delivered by lBH2 phage neutralizes Stx2 production. As 31 outlined in Figure 3A, we tested the efficacy of lBH2 phage to inhibit Stx2 production from E. 32 coli 933W by mixing them at equal concentrations (MOI~1) and culturing for 8 h. We found 33 significantly less Stx2 produced in E. coli 933W cultures treated with lBH2 phage compared to 34 those untreated ("buffer") or treated with limmP22dis phage, the parental phage of lBH2 that 10 is capable of infecting E. coli 933W but lacks the 933W×cI indgene ( Figure 3B). Quantification of    including Stx-producing E. coli in mice, 33 mitomycin c injections can induce substantial 12 quantities of Stx that is otherwise too low to be detected in stool. Mice pre-colonized by E. coli 1 933W were orally treated with lBH2 phage and then received three doses of mitomycin c by 2 intraperitoneal injection to induce stx2 expression ( Figure 4A). Daily fecal samples were 3 collected for analysis of bacterial and Stx2 concentrations. After mitomycin c injections, we 4 quantified fecal Stx2 and found that lBH2 treatment reduced fecal Stx2 titers compared to 5 buffer on days 3 and 4 with the latter showing statistical significance ( Figure 4B). Furthermore, 6 mice receiving lBH2 phage showed a significant reduction in fecal Stx2 from day 3 to 4, 7 whereas buffer treated mice did not (p = 0.547; Wilcoxen test). Although lBH2 phage did not 8 completely repress Stx2 production, these are highly inducing non-physiological conditions 9 with fecal Stx2 concentrations (~10 2 to 10 3 ng Stx2/g mouse stool) in excess of those 10 encountered in human Stx-producing E. coli infections (~2-50 ng Stx2/mL human stool). 1 11 lBH2 phage lysogenizes E. coli 933W and does not affect its titer in the murine gut.

12
Quantification of total fecal E. coli 933W did not reveal markedly different concentrations between 13 buffer and lBH2 treated mice ( Figure 4C). Quantification of fecal E. coli 933W lysogenized by 14 lBH2 ( Figure 4D) showed that a substantial fraction of the population was transduced, with 15 geometric means between ~0.9% to 2.6% and individual samples reaching as high as 71%.     Here we demonstrate a genetic strategy for in situ anti-virulence treatment of bacteria 1 colonizing the gut. We genetically engineer temperate phage l to express a repressor that 2 neutralizes Stx production in E. coli and take advantage of the genetic mosaicism of lambdoid 3 phages to create a hybrid phage that is capable of overcoming phage resistance mechanisms.

4
We found that our anti-virulence phage not only efficiently infects, lysogenizes and inhibits Stx2 5 production from E. coli in vitro, but is also effective at propagating in the murine gut from a 6 single dose to significantly reduce Stx2 production in vivo.

25
To study the effect of the engineered temperate phage on Stx2-producing E. coli, mice 26 were treated with similar conditions as described above with the following modifications. On   Table S1. E. coli 933W was 5 generated by a previously described method, 37 in which 933W phage was produced from the 6 supernatant of a log phase culture of E. coli O157:H7 strain edl933 in a modified LB media (10 7 g/L tryptone, 5 g/L yeast extract, 5 mM sodium chloride, 10 mM calcium chloride, and 0.001%  (Table S3), with 100 µL of serially-diluted l 31 phage, incubated for 20 min at r.t. followed by addition with 3 mL of molten top agar (TNT 32 media with 0.3% top agar at 45°C) and poured onto TNT agar plates. After overnight 33 incubation at 37°C, top agar from the plate with the greatest plaque density was resuspended 34 into 5 mL of phage buffer (50 mM tris, 100 mM sodium chloride, 10 mM magnesium sulfate, 18 and 0.01% gelatin, pH 7.5), sterile filtered, and stored at 4°C. To isolate the recombinant 1 phage, 50 µL of crude phage lysate was mixed with 50 µL of E. coli C600 grown to log-phase 2 in LB and incubated for ~3 h at 37°C. After incubation, 100 µL was plated onto LB containing 3 50 µg/mL of kanamycin and grown overnight at 37°C with individual colonies re-streaked 4 twice. To additionally purify by plaque purification, colonies were grown overnight in LB and 5 their sterile filtered supernatants were spotted onto TNT top agar of E. coli C600. Individual 6 plaques were streaked onto LB containing 50 µg/mL of kanamycin and sequenced to confirm 7 insertion in the correct locus of l phage.   limm933W and limmP22dis ( Figure S4). Sequencing confirmed the presence of the P22 32 immunity region and 933WcI indgene ( Figure S3 and Table S3).

34
Quantifying phage and efficiency of plating. The infectivity of phage against E. coli was 1 quantified by the double overlay agar method in which E. coli MG1655 or E. coli 933W was 2 cultured overnight in TNT media, diluted 1:100 into fresh TNT media and cultured until mid-log 3 phase of which 50 µL was mixed with 700 µL of molten top agar (TNT media with 0.5% agar at 4 45°C) and poured into individual wells of a 6-well plate containing pre-poured TNT media with 5 1.5% agar. After hardening, 100 µL of phage serially-diluted in phage buffer was added and 6 incubated for 20 min at r.t. followed by aspiration. Plates were incubated at 37°C overnight and 7 then examined for titers of plaque forming units. Efficiency of plating was calculated as the titer 8 of phage on the E. coli 933W divided by its titer on the non-lysogenic E. coli.

10
In vitro assay of phage effect. E. coli 933W was cultured overnight in TNT media at 37°C, then 11 cells were washed once with fresh TNT media and diluted to OD600nm = 0.1 (~8 x 10 7 cfu/mL). At

31
The antibody-enzyme conjugate was previously prepared using an HRP conjugation kit (Abcam 32 #ab102890) with a rabbit anti-Stx2 antibody (List Biological Labs, #765L) according to the 33 manufacturer's protocol. After washing thrice with PBST, 100 µL/well of colorimetric reagent 1 of 2M H2SO4 to stop the reaction. Absorbance was measured at 450 nm.