Engineered E. coli Nissle 1917 for the reduction of vancomycin‐resistant Enterococcus in the intestinal tract

Abstract Vancomycin‐resistant Enterococcus (VRE) poses a serious threat in hospitals where they densely colonize the intestinal tracts of patients. In vulnerable hosts, these pathogens may translocate to the bloodstream and become lethal. The ability to selectively reduce VRE in the intestinal tracts of patients could potentially prevent many of these translocation events and reduce the spread of the pathogen. Herein, we have engineered Escherichia. coli Nissle 1917 to produce and secrete three antimicrobial peptides, Enterocin A, Enterocin B, and Hiracin JM79, to specifically target and kill Enterococcus. These peptides exhibited potent activity against both Enterococcus faecium and Enterococcus faecalis, the two most prominent species responsible for VRE infections. We first discuss the optimization of the system used to express and secrete the peptides. We then show that by simultaneously expressing these peptides, both E. faecium and E. faecalis were drastically inhibited. We then demonstrate a suppression of the development of resistance when supernatant from the E. coli producer strains was used to treat E. faecium. Finally, we tested the efficacy of the probiotic in a VRE colonization model in mice. These studies showed that administration of the engineered probiotic significantly reduced the levels of both E. faecium and E. faecalis in the feces of male Balb/cJ mice.

Enterococci have become ubiquitous infectious agents in nosocomial environments, where they act as opportunistic pathogens, due to their resistance to commonly used antibiotics. 2 The common occurrence of VRE in the native microbiota of patients plays a significant role in the spread of the pathogen. 3 It is estimated that approximately 10% of intensive care units patients are colonized with VRE upon admission and that another 10% will become colonized during their stay. 4 Low-level VRE colonization of the intestines is generally benign but becomes problematic when patients are treated with antibiotics. 5 Upon treatment, a patient's native gut microbiome is drastically altered and becomes an easily colonizable niche for the resistant enterococci. 6 In some patients, particularly those with compromised immune systems, the colonized VRE then translocate to other parts of the body, causing potentially lethal infections. 2 Frequently, even densely colonized patients exhibit no symptoms and act as long-lasting reservoirs of the pathogen, making it difficult to eradicate the bacteria from the hospital environment. 7,8 The ability to selectively eliminate VRE from the gastrointestinal (GI) tracts of densely colonized patients, or patients at risk of colonization, could help prevent both lethal translocation events and reduce hospital contamination. 3,6 Antimicrobial peptides (AMPs) are small peptides, typically less than 100 amino acids naturally produced by many organisms as a first line of defense against invading pathogens. 9 Over 70 AMPs have been tested and reported to have activity against enterococci. 10 Bacteriocins are a broad class of AMPs produced by bacterial species and often target other species phylogenetically similar to the producer strain. 11 A major benefit of the bacteriocins as an antimicrobial agent is their specificity compared to many traditional antibiotics. 12,13 Because of this specificity, they may offer a means of reducing unwanted pathogenic species, like VRE, from the gut microbiota while causing minimal changes to other potentially protective microbes. 12 However, a major challenge in using AMPs for internal infections is that they are often readily degraded in the body and thus cannot reach the site of infection in sufficient quantities when orally administered. 14,15 To overcome this delivery challenge, we are engineering probiotic bacteria that can survive passage through and temporarily reside in the GI tract. Once at the site of infection, they will produce and secrete the AMPs to eliminate the pathogen of interest. After the threat of infection has passed, probiotic administration will be ceased, and depending on the delivery organism selected, the probiotics will be naturally shed from the patient. [16][17][18] For this application, we have selected Escherichia coli Nissle 1917 (EcN) as our delivery organism. EcN is a well-established probiotic strain that has been used in humans for over a century. 16,19 EcN has been demonstrated to have numerous health benefits including antiinflammatory effects, induction of gut immune defenses, and treatment of both diarrhea and constipation. 19,20 Importantly, EcN has also been found to strengthen the integrity of the intestinal wall thereby preventing pathogen translocation. 21 To engineer EcN to deliver AMPs, one must identify both the appropriate peptides and a viable secretion system for those peptides.
Recently, an elegant study was published by Hwang et al. in which EcN was engineered to produce the bacteriocin pyocin S5 and release the protein via cell lysis in response to a quorum sensing signal produced by the pathogen Pseudomonas aeruginosa . 22 This mode of release via cell lysis is common for the production of large bacteriocins produced by and targeting Gram-negative species (ex. colicins and pyocins) but is rarely described for the smaller bacteriocins targeting Gram-positive species. 11,13 Several others have described heterologous production and secretion of bacteriocins from Gram-negative producers using dedicated secretion systems isolated from the original bacteriocin gene clusters. [23][24][25] Frequently, bacteriocin genetic clusters contain their own dedicated secretion system and heterologous production can be achieved by direct transfer of the cluster into a desired producer strain. 26,27 However, as mentioned above, most bacteriocins targeting Enterococcus, a Gram-positive genus, will be of Gram-positive origin.
Therefore, due to the vastly different membrane structure of Grampositive and Gram-negative species, it is unlikely that the secretion systems will be functional in the Gram-negative EcN. 28 One must then identify a secretion system that is both compatible with EcN and can secrete a variety of Gram-positive bacteriocins.
We previously developed a peptide expression vector for EcN that employs the Microcin V secretion system to secrete a variety of bacteriocins of Gram-positive and Gram-negative origin. 29 We showed that this modular peptide secretion system (pMPES) could secrete inhibitory concentrations of four bacteriocins with activity against VRE. However, it was expected that significant improvements were needed to make this a viable peptide delivery system for animal studies.
Herein, we describe improvements to the original modular secretion system and we examine EcN as a viable delivery vehicle for three anti-enteroccocal peptides; Enterocin A, Enterocin B, and Hiracin JM79. We begin by describing optimization of the peptide production system used for EcN. We then test the in vitro activity of these systems and demonstrate that this activity is indeed due to the secretion of the peptides. Finally, we test our engineered probiotic in a murine colonization model to evaluate its ability to reduce VRE in the GI tract.

| Plasmid design and construction
Previously we reported on a modular peptide expression system (pMPES) that used the Microcin V secretion system to express and secrete several AMPs targeting intestinal pathogens such as E. coli, Salmonella, and VRE. The original pMPES contained the entire 9.1 kb fragment originally isolated by Gilson et al. that encompassed the entire Microcin V production cluster. 30 This original Microcin V production cluster contains four genes required for Microcin V synthesis; cvaC, cvi, cvaA, and cvaB. 30 CvaC encodes for the Microcin V peptide, cvi encodes for the Microcin V immunity gene, and cvaA and cvaB encode for the secretion machinery. 30 The specific functions of these secretion genes and their proposed orientations in the membrane have been described by Hwang et al. 31 To create pMPES2, we first sought to reduce unnecessary genetic components to create a more well-defined, less cumbersome vector.
The two genes, cvaA and cvaB, are the only reported plasmid-derived components required for Microcin V secretion. 30,32 We thus anticipated that by isolating these genes and their promoter regions, we could attain peptide secretion. A 3.6 kb fragment containing only the cvaA and cvaB genes along with~150 basepairs up and downstream of the genes was isolated from the native Microcin V production cluster. The 3.6 kb fragment was then inserted into a backbone containing the ColEI origin of replication instead of the original p15A origin of replication. We note that the p15A origin of replication typically results in a lower copy number. 33 Figure 1 shows the genetic maps of pMPES and pMPES2.
The second primary modification incorporated in pMPES2 was the addition of an optimized, modular molecular cloning site. We aimed to make single and multiple peptide insertion rapid and standardized. This minimizes the time required for design and it makes comparisons between variables more controllable. In our modular system, we added a standard optimized, ribosomal binding site (RBS) to the 5 0 end of each peptide gene. On the 3 0 end, we added a standard primer binding site (PBS). We then designed a set of primers containing overlap regions that could be used to amplify any peptide flanked by the standard RBS and PBS. The overlaps incorporated by the primers enable individual or multiple peptides (up to five) to be simultaneously assembled into pMPES2 using Gibson assembly. This method is extremely beneficial when different combinations and orders of peptides are to be tested. Figure S1 (Supporting Information) depicts this modular assembly method.
The standard RBS used here was specifically optimized for this application using the RBS Calculator. 34,35 It was previously found that while the original pMPES was functional, its activity could be drastically improved by minor changes in the RBS. 29 It has been shown that the translation efficiency is significantly impacted by the genetic sequence immediately up and downstream of the RBS. 36,37 For the upstream sequence, we used the sequence produced by the the Salis Laboratory's RBS Calculator which is generated as part of the RBS optimization. 34 To reduce variability in the RBS function due to differences in the downstream genetic sequence, all peptide genes were expressed as a fusion of the Microcin V secretion tag (Vsp) and the mature bacteriocin. The sequence of Vsp was kept constant for all peptides and was used as the downstream sequence in the RBS optimization. We note that while further downstream regions will differ across peptides, it is the region proximal to the RBS that is thought to have the greatest impact on RBS function. 36 2.2 | Production of Enterocin A, Hiracin JM79, and Enterocin B As mentioned above, we opted to use EcN as the delivery organism for antienterococcal bacteriocins. We thus transformed pMPES2 and pMPES2:A, B, H, and BHA into EcN to generate respectively a control strain, a strain producing Enterocin A, Enterocin B, and Hiracin JM79 individually, and a strain producing all three peptides. Enterocin A and Hiracin JM79 were selected because they were the most potent bacteriocins targeting VRE in the original pMPES system. 29 Enterocin B was selected because of previous evidence that it may act synergistically or via a different mechanism of action with Enterocin A. 38 As mentioned above, all peptides were expressed as a fusion of the Vsp tag and the mature bacteriocin. The Vsp tag is thought to direct the secretion of the fused protein and is believed to be cleaved from the peptide upon secretion. 39 Figure 2 depicts the Vsp fusions for the three peptides tested; Enterocin A, Hiracin JM79, Enterocin B, and an operon of the three peptides referred to as BHA.
After generating the bacteriocin constructs, we then tested the activity of the modified EcN strains against two vancomycin-resistant clinical isolates, Enterococcus faecium 8E9 and Enterococcus faecalis V583R. E. faecium and E. faecalis were chosen because these two species are responsible for nearly all VRE infections. 40 We note that the addition of pMPES2 or pMPES2 expressing the enterocins did not impact the growth rates or morphologies of any of the E. coli strains used (EcN, EcN RN, and E. coli MC1061 F 0 ). In addition, plasmid stability was tested over 20 generations and no loss was observed.
Agar diffusion assays of the EcN strains against E. faecium 8E9 and E. faecalis V583R are shown in Figure 3. For these tests, Enterococcus was seeded in a solid growth medium plate then swabbed with the probiotic. EnterocinA Amino Acid Sequence FIGURE 2 Genetic configuration for expression of bacteriocin genes. All bacteriocins were expressed as a fusion of the Microcin V secretion tag (Vsp) and the mature bacteriocin sequence. The resulting Enterocin A amino acid sequence is shown as an example white dot is the probiotic, and the dark region is the zone of inhibition produced by EcN.
Based on these initial screens, Enterocin A appeared to be the most potent individual peptide system. We note that halo diameter not only depends on the inhibitory concentration of the peptide itself, but also on the production and growth rates of the producer strain and the growth rate of the pathogen. Thus, we find halo diameters can typically be used to compare probiotic activities against a given target strain but should not be used to compare between different targets. We also note that at higher activity levels, halo diameters become more similar in size making quantitative comparisons of producer activity difficult.
To better quantify the overall probiotic activities, we performed  Herein, bacteriocin activity from the producer strains is quantified in terms of bacteriocin units (BUs). 41 One BU is defined as the reciprocal of the highest dilution of the bacteriocin sample capable of inhibiting growth by 50%.
As expected from the agar diffusion assays, Enterocin A showed the highest level of activity among the individual peptides against E. faecalis V583R. Interestingly, EcN pMPES2:BHA showed a significantly higher activity than any of the three peptides independently.
This difference in activity was not reflected in the agar diffusion assays; however, this is not unexpected because as mentioned above, halo sizes become more similar at higher activity levels.
The activity observed from EcN pMPES2:BHA was greater than the sum of the activities of the individual peptides against both pathogens. This may be due to synergistic activity of the peptides or due to an increase in production efficiency. It is not uncommon for operon gene expression to be more efficient than individual gene expression systems, largely due to improvements in translation efficiency. 42 Several explanations have been posed for this increase in efficiency. It has been found that translation efficiency increases with the length of an RNA transcript, which is generally longer for polycistronic operons. 42 In addition, it has been proposed that as the ribosome progresses along the upstream gene, it can denature otherwise inhibitory secondary structure in the mRNA in the downstream RBS regions. 43,44

| Impact of operon organization
We next sought to test whether the order of the peptides in the operon would impact overall activity. This was done because it has been previously reported that operon efficiency can be drastically impacted by the order of the genes in an operon. 42 Figure S2 (Supporting Information) shows the supernatant activities of pMPES2 with the six different peptide operons. As used in the BHA naming convention, H indicates Hiracin JM79, A is Enterocin A, and B is Enterocin B. The order of the three letters indicates the peptide order in the operon.
As anticipated, we observed different levels of activity from the six configurations. In particular, ABH showed significantly higher activity against E. faecium 8E9 than all other constructs except BHA (P < 0.05). Interestingly, this uniquely high activity was not observed against E. faecalis V583R. These results may indicate that the ABH configuration produced more peptides exhibiting stronger activity against E. faecium 8E9 compared to E. faecalis V583R. Importantly, HAB and ABH consistently exhibited numerically lower activities against both E. faecium 8E9 and E. faecalis V583R compared to all other constructs (P < 0.1). For example, ABH supernatant exhibited less than one third the activity against E. faecium 8E9 compared to BHA, HBA, and BAH supernatant and only one sixth the activity of ABH supernatant. These observations support the importance of operon configuration that is often overlooked.
Note that in Figure S2  both individual and multi-peptide expression from this strain is significantly lower than from the wild-type EcN (P < 0.05). This is based on a one-tailed Student's t test assuming unequal variance (data not shown). This explains the inconsistent activity levels for BHA between Figure 4 and Figure S2 (Supporting Information). We selected this antibiotic-marked strain for use in mouse studies to enable enumeration in the feces; however, in the future, we may explore alternative spontaneous mutants that do not exhibit hindered production.

| Verification of peptide identity and activity
To verify that the anticipated peptides present in the supernatant were in fact produced and were the cause of activity, we performed sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) on the concentrated supernatants. Figure 5 shows the SDS- contain two arginine residues, four lysine residues, and one histidine residue. Enterocin A, however, contains five lysine residues but no histadine residues and most importantly, no arginine residues. Thus, one cannot readily compare the Enterocin A concentration to that of the other two peptides based on the SDS-PAGE band intensity.
To verify that these bands were responsible for the observed antienterococcal activity, we overlaid the protein gel on solid growth medium containing E. faecium 8E9, similar to the agar diffusion assays presented above. From Figure 5, one can see zones of inhibition on the indicator strain over the supposed peptide bands.
As a final verification of their identities, the peptide bands were extracted from the gel using in-gel trypsin digestion.

| Resistance prevention with the three-peptide system
In addition to drastically increasing overall activity, we have also observed that simultaneous production of the three enterocins reduces the development of resistance for E. faecium 8E9. Figure 6 shows the typical growth curves observed for E. faecium 8E9 and Though our endpoint supernatant activities provide insight into the total peptide production over 17 hr for our probiotics, a more realistic study requires that the pathogen and probiotic compete within the same culture. We thus performed a coculture inhibition   Figure SI3 shows the E. faecium From these results, it is clear that EcN RN pMPES2:BHA can compete with E. faecium 8E9 and that production of the three bacteriocins is able to eliminate E. faecium 8E9 without the emergence of resistance as seen in the supernatant inhibition assays. This lends promise to using this strain, despite its reduced activity compared EcN pMPES2:BHA.
Previously, we performed similar studies (both supernatant inhibition assays and coculture assays) involving Enterocin A, Hiracin JM79, and an additional peptide, Enterocin P. 48,49 In these studies, we consistently observed the same regrowth in E. faecium observed in the individual peptide cultures, regardless of the concentration of supernatant used. We found that the regrown population of E. faecium was stably resistant to the peptides. Enterocin A, Hiracin JM79, and Enterocin P are all class IIa bacteriocins and share the same extracellular protein target, a mannose phosphotransferase (ManPTS) (ManPTS). 49 The disruption of this transporter apparently renders E. faecium resistant to all three peptides.
We hypothesize that the lack of resistance development against pMPES2:BHA may be due to the simultaneous application of peptides with orthogonal mechanisms of action and mechanisms of resistance.
In this scenario, the pathogen would need to simultaneously develop two different resistant mutations. The probability of this occurring is orders of magnitude lower than for the development of a single resistant mutation.
It has been previously reported that Enterocin A-resistant E. faecalis mutants remain susceptible to Enterocin B and that Enterocin B-resistant E. faecalis mutants remain susceptible to Enterocin A. 38 To determine if Enterocin B shows the same orthogonal mechanism of resistance against E. faecium, we performed agar diffusion assays of EcN producing the individual peptides and BHA using two strains of E. faecium; E. faecium 6E6 and a previously identified E. faecium 6E6 ManPTS mutant. 49 Figure 7 shows the results of these activity assays.
One can see from these tests that while Enterocin A and Hiracin JM79 were inactive against the ManPTS mutant, Enterocin B and the three-peptide construct maintained activity. This supports the hypothesis that the addition of Enterocin B may eliminate the class IIa bacteriocin resistant subpopulation in the culture leading to an overall reduced resistance development.

| Studies in mice
We next sought to evaluate whether our probiotic could reduce VRE in the GI tract. To do this, we developed a VRE colonization model in mice, administered our engineered EcN in the water, and enumerated the VRE in the feces over time. In our colonization model, mice were administered~5 × 10 8 CFU/ml VRE (E. faecium 8E9 or E. faecalis V583R) in drinking water containing 250 μg/ml vancomycin for 8 days. Vancomycin was added because we had found it improved colonization stability.
After the colonization period, mice were provided with three different treatments in their water. The untreated group was given sterile water, the control group was given water containing~5 × 10 8 CFU/ml EcN RN pMPES2 (no bacteriocins), and the treated group was given water containing~5 × 10 8 CFU/ml EcN RN pMPES2:BHA.   The area under the CFU/g feces versus time curve was also calculated for each mouse and compared across groups. This area represents the total amount of VRE shed during the experiment.
Based on this parameter, treated mice exhibited significantly less VRE in their feces over the duration of the experiment compared to either the untreated or control groups (E. faecium: untreated, P = 0.02; control, P = 0.03, E. faecalis: untreated, P = 0.031, control = 0.003).
We must note that the experiments above were done in male Balb/cJ mice. We then repeated the same tests for female Balb/cJ mice. With the exception of one outlier in the treated group, we saw similar reduction of E. faecium 8E9 in female mice as observed in male mice ( Figure S4). However, we did not see this same reduction for the female E. faecalis V583R experiments ( Figure S5) 3 | METHODS

| Bacterial growth conditions
Bacteria and plasmids used in this study are reported in Table 1.
Unless otherwise noted, E. coli was grown in lysogeny broth (LB) broth (Fisher Scientific, Fair Lawn, NJ) with agitation at 37 C.
Spectinomycin sulfate was added at 100 μg/ml for vector selection when necessary. E. faecium and E. faecalis were grown statically in brain heart infusion (BHI) (Oxoid Ltd., Basingstoke, United Kingdom) medium at 37 C.

| Standard peptide insertion
All peptide DNA sequences, primers, and standardized parts men- 3.4 | In vitro activity assays 3.4.1 | Agar diffusion assays E. faecium and E. faecalis were grown overnight in BHI medium. The indicated producer strain was streaked from a freezer stock onto LB agar with spectinomycin. The following morning, molten BHI agar (3.7 g BHI, 1 g agar per 100 ml) was prepared and allowed to cool to

| SDS-PAGE and mass spectrometry verification 3.5.1 | Peptide concentration by AS precipitation
Fresh colonies of the producer strains were inoculated into 25 ml BHI medium and were incubated for~17 hr at 37 C. E. coli cultures were pelleted by centrifugation at 5,000g for 10 min and the supernatants were sterile filtered into to sterile tubes. Then, 11.25 g of ammonium sulfate salt was then added to each 25 ml sample to reach~70% saturation concentration. Samples were mixed by rotation at 4 C overnight. The following day, proteins were pelleted by centrifugation at 11,000g at 4 C for 10 min. Supernatant was removed, and the pelleted precipitate was resuspended in 250 μl sterile deinonized water then stored at −20 C.
The gel was run at 125V until complete (~90 min). It was then rinsed in deionized water and stained with SimplyBlue SafeStain for 1 hr.
The gel was then destained overnight and imaged the following day.

| Gel activity assay
The destained gel was imaged then placed using sterile tweezers on BHI agar seeded with E. faecium as described in the agar diffusion assay section. The plate was incubated overnight and was imaged the following day.

| Statistical analysis
The aim was to compare the mean levels of VRE in mouse feces in the treated group versus the untreated group and the control group. Two methods were used; day-by-day comparison and comparison of the total VRE shed during the experiment (area under the CFU VRE/g feces curve). For both analyses, CFU VRE/g feces were first log-transformed. To compare daily fecal counts across groups, a one-tailed Student's t test was performed between either the treated and untreated or untreated and control groups assuming unequal variance. The treated group was considered statistically reduced for p < 0.05.
Total VRE in the feces over time was compared to provide a single, cumulative analysis of the observed reduction over time. To do this, the integral of the CFU VRE/g feces was approximated using the trapezoidal rule from Day 0 to the final day of the experiment. The mean areas for the six mice in each group were then compared using a one-tailed Student's t test between either the treated and untreated or untreated and control groups assuming unequal variance. The treated group was considered statistically reduced for P < 0.05.

| CONCLUSIONS
In this study, we have significantly improved upon a modular peptide expression system for EcN, making a viable probiotic delivery system to target pathogens in the intestinal tract. We have successfully shown that the engineered strain produces and secretes the peptides at sufficient levels to drastically inhibit pathogens of interest in laboratory conditions. We have also provided evidence suggesting that the simultaneous production of multiple peptides may help prevent the regrowth of resistance to bacteriocins. In future studies, we must examine how engineered probiotics may impact the host microbiome and better establish their robustness in female mice and other strains.
In this study, we have focused on the production of peptides targeting vancomycin-resistant enterococci. However, because the secretion system used for pMPES2 is widely applicable for different peptides, we can readily modify our probiotic for use against a variety of Gram-positive and Gram-negative pathogens. Though further improvement and analysis will need to be done to achieve a clinical product, the in vitro and in vivo results presented here provide strong proof of concept evidence for pMPES2 as a probiotic-based bacteriocin delivery vector.

CONFLICT OF INTEREST
The authors declare that they have no conflicts of interest with the contents of this article.