Enhancing the antibacterial function of probiotic Escherichia coli Nissle: when less is more

ABSTRACT Probiotic bacteria confer multiple health benefits, including preventing the growth, colonization, or carriage of harmful bacteria in the gut. Bacteriocins are antibacterial peptides produced by diverse bacteria, and their production is tightly regulated and coordinated at the transcriptional level. A popular strategy for enhancing the antibacterial properties of probiotic bacteria is to retrofit them with the ability to overproduce heterologous bacteriocins. This is often achieved from non-native constitutive promoters or in response to host or pathogen signal from synthetic promoters. How the dysregulated overproduction of heterologous bacteriocins affects the fitness and antibacterial efficacy of the retrofitted probiotic bacteria is often overlooked. We have conferred the prototypical probiotic Escherichia coli strain Nissle (EcN) the ability to produce microcin C (McC) from the wild-type promoter and two mutant promoters that allow, relative to the wild-type promoter, high and low amounts of McC production. This was done by introducing specific changes to the sequence of the wild-type promoter driving transcription of the McC operon while ensuring that the modified promoters respond to native regulation. By studying the transcriptomic responses and antibacterial efficacy of the retrofitted EcN bacteria in a Galleria mellonella infection model of enterohemorrhagic E. coli, we show that EcN bacteria that produce the lowest amount of McC display the highest antibacterial efficacy with little-to-none undesired collateral impact on their fitness. The results highlight considerations researchers may take into account when retrofitting probiotic bacteria with heterogenous gene products for therapeutic, prophylactic, or diagnostic applications. Bacteria that resist killing by antibiotics are a major risk to modern medicine. The use of beneficial “probiotic” bacteria to make antibiotic-like compounds at the site of infection in the body is emerging as a popular alternative to the use of conventional antibiotics. A potential drawback of engineering probiotic bacteria in this way is that producing antibiotic-like compounds could impart undesired side effects on the performance of such bacteria, thereby compromising their intended use. This study highlights considerations researchers may take into account when engineering probiotic bacteria for therapeutic, prophylactic, or diagnostic applications.

the genes responsible for bacteriocin production are found alongside one or more cognate immunity determinants in chromosomally encoded or plasmid-borne gene clusters (6).Although the production of bacteriocins confers a competitive advantage to the producing strain, it is well recognized that bacteriocin production also imparts a fitness cost on the producing strain (7)(8)(9)(10)(11).Therefore, the bacteriocin synthesis genes are often under tight transcriptional control and only become activated in response to metabolic-or quorum-dependent signals, or as part of the SOS response (12).
In recent years, various efforts have been made to enhance the inherent antibacterial properties of probiotic bacteria by equipping them with the ability to produce heterol ogous bacteriocins against a pathogen of interest (13)(14)(15)(16)(17). Escherichia coli Nissle 1917 (EcN), a medically licensed probiotic, used for the treatment of gastrointestinal condi tions including ulcerative colitis and infant diarrhea, is often used as the model host to introduce heterologous bacteriocin genes (18)(19)(20).EcN is non-pathogenic, non-invasive, capable of colonizing the intestinal tract for long periods, and can provide protection against various enteropathogens by modulating host immunity, including by suppress ing Shiga toxin production from enterohemorrhagic bacteria and competing directly with pathogens for nutrients (21)(22)(23).EcN naturally produces two bacteriocins, microcin H47 and microcin M. The production of microcins H47 and M is dependent on the iron availability (24).This can limit the bacteriocin-dependent antibacterial properties of EcN as iron levels can substantially fluctuate in people depending on their diet, genetic factors, and intestinal health status (25).Therefore, conferring EcN (and other probiotic bacteria) with the ability to produce heterologous bacteriocins is an emerging and popular strategy to enhance the antibacterial prowess of probiotic bacteria.A survey of the current literature at the time of writing showed that, in all except one documented case, bacteriocin production was decoupled from its native regulation to achieve high amounts of the bacteriocin of interest (Fig. 1A).This was done either by introducing a synthetic promoter to constitutively drive transcription of bacteriocin genes or by allowing the transcription of bacteriocin genes in response to a host or pathogen signal (9,(13)(14)(15)(16)26).However, it is unclear whether the high amounts of bacteriocin produc tion, decoupled from native regulation, are warranted to obtain the desired antibacterial effect.This is important to know because sustained high amounts of production of heterologous bacteriocins could impose a metabolic burden and thereby compromise the inherent beneficial traits of EcN bacteria or introduce undesired collateral physiologi cal traits and/or fitness cost that might compromise the intended enhanced antibacterial efficacy of the engineered EcN bacteria.
Microcin C (McC) is a narrow spectrum yet potent translational inhibitor, which is produced naturally by some E. coli strains and predominantly affects closely related bacteria, including pathogenic Escherichia, Klebsiella, and Shigella species (27) (27)(28)(29) (Fig. 1B).An additional immunity determinant, MccF, is a separate transcription unit that is located downstream of mccE and is positioned in the opposite orientation to the mccA-E operon (30) (Fig. 1B).The transcription of the mccA-E operon is driven from an extended −10 promoter (Pmcc; Fig. 1B).Pmcc lacks an obvious −35 consensus element.Instead, a binding site for the catabolite repressor protein CRP is found centered ~60 nucleotides upstream of the transcription start site of mccA (Fig. 1B).Therefore, optimal transcription of mccA-E operon is subjected to stringent regulation and relies on the CRP and the RNA polymerase (RNAP) promoter specificity factor RpoS (30)(31)(32).Hence, McC is synthesized when the optimal carbon source becomes limited (33,34).Conversely, under nutrient replete conditions, the transcription of mcc genes is further repressed by the nucleoid-associated proteins, HNS and Lrp (30,32).By strategically altering the core Pmcc sequence, we made two mutant promoters which allowed higher (Pmcc HIGH ) and lower (Pmcc LOW ) transcription of mccA-E genes than the WT (WT) promoter (Pmcc WT ).Both mutant promoters were unaltered in their dependence on CRP and RpoS for their optimal activity.The properties of EcN strains with mccA-E genes under Pmcc HIGH and Pmcc LOW reveal that EcN bacteria that produce the lowest amount of McC display the highest antibacterial efficacy with little-to-none undesired collateral impact on their fitness and thereby highlight considerations that investigators might take into account when retrofitting EcN and other probiotic bacteria with bacteriocin genes or other gene products of therapeutic and prophylactic value.

The sensitivity of clinical E. coli isolates to McC
Our aim was to confer EcN bacteria the ability to produce different amounts of McC from Pmcc WT , Pmcc HIGH , and Pmcc LOW and to evaluate the consequences of doing so on the fitness and antibacterial efficacy of the EcN bacteria.Although it is well established that McC effectively stops the growth of E. coli and related pathogens, it is not known how sensitive clinical E. coli isolates are to McC and how prevalent McC resistance is.Therefore, we used a simple disc diffusion assay on Mueller-Hinton agar plates to evaluate the sensitivity of a small set of clinical E. coli isolates (obtained from the Imperial College National Institute for Health Research, Health Protection Research Unit in Healthcare Associated Infections and Antimicrobial Resistance) to McC.The enterohe morrhagic E. coli O157:H7 strain EDL933 (hereafter referred to as EDL933 strain) was used as the McC-sensitive reference strain.We also included broadly studied uropathogenic E. coli isolates CFT073 and UTI89 in our screen.None of the strains in the screen harbored any plasmids that contained the mccA-F genes.As McC is secreted into the culture media, McC for the disc diffusion assay was obtained by collecting and filtering the supernatant from an overnight culture of a standard laboratory E. coli bacteria containing a plasmid from which genes mccA-F were overexpressed.The size of the halo, indicating bacterial growth inhibition, surrounding the disc containing McC on plates containing the EDL933 strain, had a radius of ~4 mm (Fig. 2).Therefore, we deemed a halo size of <2 mm as having reduced sensitivity to McC (indicated by the dotted line in Fig. 2).Results revealed that of the 21 clinical E. coli isolates tested, 17 were as sensitive to McC as the reference strain, including CFT073 and UTI89, under aerobic growth conditions.We noted that two of the clinical isolates that displayed reduced sensitivity to McC under

Pmcc exists in different evolutionary clades but does not display differences in activity and regulation
Prior to constructing plasmids with mccA-E genes under Pmcc WT , Pmcc HIGH , and Pmcc LOW , we wanted to know how conserved the Pmcc core promoter and upstream regulatory sequences are.Therefore, we used the nucleotide sequence of the mccB gene as the query to identify all E. coli strains available in the NCBI RefSeq genomes database that potentially contained an McC-producing plasmid.We identified 185 unique E. coli strains from which we then compared the nucleotide sequence ~400 bp upstream of the first codon of MccA.Phylogenetic analysis revealed that Pmcc and the upstream regulatory sequences from the 185 E. coli strains fall into three distinct clades, with 69 and 108 of them represented in clades 1 and 2 and only 8 represented in clade 3 (Fig. 3A).The core Pmcc sequence and the CRP-binding site were conserved in all three clades (Fig. 3B).Notably, the sequences in clade 1 differed from those in clades 2 and 3 by the absence of a 21-bp-long sequence, located ~225 nucleotides upstream of the first codon of MccA.As the majority of Pmcc sequences either fell into clade 1 or 2, we focused our analysis on these two clades.We first wanted to determine whether the 21-bp-long insertion and other differences in the nucleotide sequence upstream of the core Pmcc sequence affected the activity and regulation of the promoter that belong to clades 1 and 2. To do so, we fused the ~400-bp-long sequence upstream of the first codon of MccA from a representative clade 1 and clade 2 promoter to green fluorescent protein (GFP) and placed this Pmcc-GFP transcriptional fusion in the low copy number (10-12 per cell) plasmid pACYC to generate plasmids pEB-C1-GFP and pEB-C2-GFP (Table 1).We then measured GFP production from pEB-C1-GFP and pEB-C2-GFP initially in E. coli strain MG1655 grown in lysogeny broth (LB).As shown in Fig. 3C, we did not observe any difference in GFP production from pEB-C1-GFP and pEB-C2-GFP in both strains grown as batch cultures.Although bacteria containing pEB-C1-GFP and pEB-C2-GFP did not differ in GFP production at the whole population level, we considered whether they differed at the single cell level.Hence, we took bacteria after 4 h of growth (when GFP production became detectable; Fig. 3C) in LB and enumerated GFP producing individual cells by fluorescence microscopy.As shown in Fig. 3D, we did not observe any differences in the number of individual GFP fluorescent bacteria in samples containing pEB-C1-GFP and pEB-C2-GFP.However, we noted that the activity of both clades of Pmcc is heterogenous under our conditions.As Pmcc is dependent on RpoS and CRP for optimal activation, we next determined whether clade 1 and clade 2 promoters differ in their dependency on RpoS and ability to respond to regulation by CRP.Hence, we compared GFP production from pEB-C1-GFP and pEB-C2-GFP in a ΔrpoS MG1655 strain and WT MG1655 strain grown in LB supplemented with extra glucose (note that the presence of excess glucose will abrogate CRP binding to DNA), respectively.As shown in Fig. 3E, we observed an ~ 2.7-fold reduction in GFP production from both pEB-C1-GFP and pEB-C2-GFP in the ΔrpoS MG1655 strain compared to WT MG1655 strain following 10 h of growth.This confirms previous work that showed that while the Pmcc is categorized as a RpoS-dependent promoter, it can be potentially used by RpoD, the major housekeeping promoter specificity factor in E. coli, when RpoS is unavailable (35).However, GFP production from pEB-C1-GFP and pEB-C2-GFP was fully attenuated when WT MG1655 bacteria containing pEB-C1-GFP and pEB-C2-GFP were grown in LB with extra glucose (Fig. 3F).This suggests that while Pmcc might not be fully reliant on RpoS, it has a strict requirement for CRP for activation of transcription.Overall, we conclude that, despite major evolutionary differences in the upstream regulatory sequences, clade 1 and clade 2 Pmcc promoters do not differ in their activity and regulation.

Construction and characterization of Pmcc HIGH and Pmcc LOW
Having established that clade 1 and clade 2 Pmcc promoters do not differ in their activity and regulation, we used a representative clade 1 Pmcc promoter to construct Pmcc WT , Pmcc HIGH , and Pmcc LOW .To make Pmcc HIGH , we introduced a suboptimal −35 sequence (5′-GTGACA-3′ instead of the optimal 5′-TTGACA-3′ sequence) in Pmcc WT (Fig. 4A).We expected that a suboptimal −35 consensus sequence together with the extended −10 element would subtly increase RNAP binding while retaining dependency on CRP, thereby leading to an overall increased transcriptional output from the modified promoter.Conversely, to make Pmcc LOW , we swapped the first 5′-AG-3′ of the AGTG extended −10 motif of Pmcc WT to a 5′-GT-3′ (Fig. 4A).We then fused Pmcc WT , Pmcc HIGH , and Pmcc LOW to GFP in pACYC to generate pEB-Pmcc WT -GFP, pEB-Pmcc HIGH -GFP, and pEB-Pmcc LOW -GFP ( As shown in Fig. 4D, and expected (see above), in the ΔrpoS MG1655 strain, the activity of all three promoters was reduced by ~1.5-2.4-foldcompared to their relative activity in WT MG1655 strain.However, GFP production was fully abolished when WT MG1655 bacteria containing Pmcc WT , Pmcc HIGH , and Pmcc LOW were grown in LB with extra glucose (Fig. 4E).Overall, we conclude that, Pmcc HIGH and Pmcc LOW display the expected difference in transcription activity compared to Pmcc WT and that the changes made to Pmcc WT sequence to generate Pmcc HIGH and Pmcc LOW do not affect their regulation.Next, we replaced the GFP in pEB-Pmcc WT -GFP, pEB-Pmcc HIGH -GFP, and pEB-Pmcc LOW -GFP with mccA-F gene cassette (as shown in Fig. 1B) to generate pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, and pEB-Pmcc LOW -mccA-F (Table 1) and determined whether the differences in GFP activity from the three promoters corresponded to McC production in EcN.To do so, EcN bacteria harboring pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, and pEB-Pmcc LOW -mccA-F were grown overnight in LB, and their supernatant harvested, filtered, and used as the source of McC.Different amounts [0%-16% (vol/vol)] of this supernatant were then used to supplement fresh LB media in 96-well plates, which were inoculated with enterohemorrhagic E. coli EDL933.As shown in Fig. 4F, ≥2% (vol/vol) of supernatant from EcN with pEB-Pmcc WT resulted in complete growth attenuation of E. coli EDL933; and moderate growth was only detected with ~1% (vol/vol) of supernatant from EcN with pEB-Pmcc WT -mccA-F.In contrast, in experiments with supernatant from EcN with pEB-Pmcc LOW -mccA-F, growth of E. coli EDL933 comparable to that in wells with 0% (vol/ vol) supernatant was detected with up to ~4% (vol/vol) supernatant (Fig. 4G).However, in experiments with supernatant from EcN with pEB-Pmcc HIGH -mccA-F, only ~0.5% (vol/ vol) of supernatant supported some growth of E. coli EDL933, and no growth was detected in well with ≥0.5% (vol/vol) supernatant (Fig. 4H).We estimate that approxi mately twofold more McC is produced by EcN containing pEB-Pmcc HIGH -mccA-F than in EcN containing pEB-Pmcc WT , and, conversely, greater than eightfold less McC is produced by EcN containing pEB-Pmcc LOW -mccA-F than in EcN containing pEB-Pmcc WT .Hence, the McC produced by EcN containing pEB-Pmcc HIGH -mccA-F and EcN containing pEB-Pmcc LOW -mccA-F differs by >16-fold.Overall, we conclude that Pmcc HIGH and Pmcc LOW confer EcN the ability to make high and low amounts, respectively, of McC compared to from Pmcc WT .

The impact of producing different amounts of McC on EcN fitness
To determine how producing high and low amounts of McC from pEB-Pmcc HIGH -mccA-F and pEB-Pmcc LOW -mccA-F affects the growth dynamics and long-term viability of EcN, we conducted growth and viability assays in LB.An additional control included EcN bacteria containing pEB-Pmcc MUT -mccA-F (in which the core Pmcc sequence was scrambled to inactivate the promoter; Fig. S1).As shown in Fig. 5A, the growth dynamics of EcN bacteria containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, and pEB-Pmcc LOW -mccA-F did not markedly differ during exponential phase of growth.Nonetheless, we detect a moderate adverse effect on growth when the bacteria entered the stationary phase of growth [which is when McC production starts from Pmcc (see Fig. 4B) and the adversity on growth correlated with the amount of McC produced (Fig. 5A)].Put simply, it appears that the amount of McC produced inversely correlates with the ability of the EcN to accumulate biomass in LB over a 24-h period of time.Thus, the production of lower amounts of McC in EcN bacteria containing pEB-Pmcc LOW -mccA-F did not seem to affect the ability of EcN to accumulate biomass (Fig. 5A, compare pEB-Pmcc LOW -mccA-F with pEB-Pmcc MUT -mccA-F).
To enumerate the proportion of viable cells in the population of EcN bacteria making different amounts of McC, we used overnight cultures of EcN to inoculate fresh LB media 2.5 × 10 7 colony-forming units (CFU) and measured CFU daily over a period of 6 days.As expected, on day 1, the adverse effect of McC production on growth (Fig. 5A) was reflected in the proportion of viable cells in the population of EcN bacteria producing different amounts of McC (Fig. 5B).From day 1 onward, the rate of decline in the proportion of viable cells in the population of EcN bacteria producing different amounts of McC was similar, and by day 6, the proportion of viable cells in the population of EcN bacteria with pEB-Pmcc HIGH -mccA-F was ~4.6-fold less than the proportion of viable cells in the population of EcN bacteria containing pEB-Pmcc LOW -mccA-F (Fig. 5B).Notably, the proportion of viable cells in the population of EcN bacteria containing pEB-Pmcc LOW -mccA-F and pEB-Pmcc MUT -mccA-F did not differ.Overall, EcN bacteria retrofitted with the ability to make McC from Pmcc LOW display better growth and fitness characteristics than bacteria that produce McC from Pmcc WT or Pmcc HIGH .
To determine how producing elevated and reduced amounts of McC affected the fitness of EcN in more depth, we compared the transcriptomes of EcN bacteria contain ing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, and pEB-Pmcc LOW -mccA-F to that of EcN bacteria containing pEB-Pmcc MUT -mccA-F.For the transcriptomics analysis, bacteria were grown in LB and harvested following 6 h when the cultures were in the early-stationary phase (recall that this is ~2 h after McC production has started under our conditions).As expected, the number of transcript reads as a percentage of total transcript reads that mapped to the mccA-F genes in EcN bacteria containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, pEB-Pmcc LOW -mccA-F, and pEB-Pmcc MUT -mccA-F correlated with the expected amount of mccA-F mRNA made from the respective promoters (Fig. 5C).We note that, in the case of EcN bacteria containing pEB-Pmcc HIGH -mccA-F, ~15% of total RNA in the cell corresponded to the mcc transcripts.For the comparative analyses of the transcriptomes, we defined differentially expressed genes as those with expression levels changed log 2 -fold change of >0.5 with a false discovery rate adjusted P < 0.05.As shown in Fig. 5D, ~6% of the genes in EcN containing pEB-Pmcc WT -mccA-F were differentially expressed compared to EcN containing pEB-Pmcc MUT -mccA-F.In contrast, ~32% of genes in EcN containing pEB-Pmcc HIGH -mccA-F were differentially expressed compared to EcN containing pEB-Pmcc MUT -mccA-F (Fig. 5F).However, only ~0.5% of genes in EcN con taining pEB-Pmcc LOW -mccA-F were differentially expressed compared to EcN containing pEB-Pmcc MUT -mccA-F (Fig. 5E).Notably, in EcN bacteria containing pEB-Pmcc WT -mccA-F and pEB-Pmcc HIGH -mccA-F, some of the genes that were upregulated by fourfold or higher were associated with asparagine (asnA and asnB) and methionine biosynthesis (metA, metE, metF, metK, and metR) and uptake (metN) (Fig. 5G).This could reflect the need for high levels of asparagine, methionine, and the methionine-derivative S-adeno sylmethionine (SAM) for McC production.Translation of the precursor heptapeptide MccA (MRTGNAN) requires two asparagine and one methionine molecules; methionine is further depleted as an indirect result of the McC maturation process, in which terminal asparagine of MccA is converted to a modified aspartate residue (38).The second stage of this process requires one molecule of SAM, which is synthesized from methionine by S-adenosylmethionine synthetase, the product of metK (38).Conversely, many genes that were downregulated in EcN bacteria containing pEB-Pmcc WT -mccA-F and pEB-Pmcc HIGH -mccA-F were associated with conferring acid resistance (hdeA, hdeD, gadA, gadB, gadC, gadE, slp, dctR, mdtE, and mdtF) (Fig. 5H).Hence, we considered whether the downregula tion of acid stress resistance genes was reflective of a decrease in the pH of the growth media that occurred as a consequence of McC production.However, as shown in Fig. 5I, the pH of the growth media following 6 h of growth of EcN bacteria containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, pEB-Pmcc LOW -mccA-F, and pEB-Pmcc MUT -mccA-F was similar and averaged around pH 7.75.We noted ivy, the product of which is an inhibitor of lysozyme, a constituent of the mammalian gut, is also downregulated in EcN bacteria containing pEB-Pmcc WT -mccA-F and pEB-Pmcc HIGH -mccA-F (Fig. 5H).EcN bacteria must be able to adapt to the predominantly acidic gut environment to effectively execute their intended antibacterial function.Therefore, we considered whether the downregulation of hdeA, hdeD, gadA, gadB, gadC, gadE, slp, dctR, ivy, mdtE, and mdtF imparts a fitness disadvantage on EcN bacteria producing elevated amounts of McC in the gut environment.To test this, we harvested EcN bacteria following 6 h of growth in LB and inoculated them into synthetic gastric fluid media (which had a pH of ~2.5 and contained lysozyme) and enumerated the number of viable bacteria following 20 min and 1 h of incubation.Results in Fig. 5J show that while ~10% of EcN bacteria containing Pmcc WT -mccA-F, pEB-Pmcc LOW -mccA-F, and pEB-Pmcc MUT -mccA-F survived 1 h of exposure to synthetic gastric fluid media, >99% of EcN bacteria containing pEB-Pmcc HIGH -mccA-F did not survive this challenge.Overall, we conclude that elevated production of McC results in metabolic perturbations that compromise the overall fitness of EcN bacteria.

The antibacterial efficacy of engineered EcN bacteria
Having established that producing high amounts of McC imparts a fitness cost on EcN bacteria, we investigated the antibacterial efficacy of EcN bacteria containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, and pEB-Pmcc LOW -mccA-F.For this, we incubated different number of cells of EcN bacteria containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, and pEB-Pmcc LOW -mccA-F and E. coli EDL933 bacteria in LB and enumerated the CFU of E. coli EDL933 bacteria following ~24 h of co-culturing.In these assays, we used E. coli EDL933 harboring the kanamycin resistance gene for specific selection of E. coli EDL933 cells.Initially, however, we used the EcN bacteria containing pEB-Pmcc MUT -mccA-F to evaluate the inherent, i.e., McC-independent, antibacterial activity of EcN.Results in Fig. 6A show that when ~ 2 × 10 7 EcN bacteria containing pEB-Pmcc MUT -mccA-F and E. coli EDL933 were co-cultured for 24 h in LB, there was an ~50%-80% reduction in the CFU of  antibacterial activity of EcN, we expressed the reduction in CFU of E. coli EDL933 when co-cultured with EcN bacteria containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, and pEB-Pmcc LOW -mccA-F as a percentage of CFU of E. coli EDL933 when co-cultured with EcN containing pEB-Pmcc MUT -mccA-F (referred to as McC-dependent antibacterial activity in Fig. 6B; see below).Notably, we did not detect any differences in the antibacterial activity of EcN containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, and pEB-Pmcc LOW -mccA-F co-cultured with 1:1 (Fig. 6B) or 1:1,000 (Fig. 6C) E. coli EDL933 cells or when the McC non-sensitive Staphylococcus aureus cells were co-cultured with EcN and E. coli EDL933 cells at 1:1:1 amounts (Fig. 6B).We conclude that, regardless of the amount of McC produced by EcN containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, and pEB-Pmcc LOW -mccA-F, the McC produced is sufficient to eliminate E. coli EDL933 under our co-culture conditions.Therefore, to better delineate the antibacterial efficacy of EcN containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, and pEB-Pmcc LOW -mccA-F, we evaluated it in the context of a Galleria mellonella larvae as an in vivo infection model for E. coli EDL933.Our plan was to prophylactically treat the larvae with a fixed dose of EcN bacteria containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, pEB-Pmcc LOW -mccA-F, Pmcc LOW -mccA-F or pEB-Pmcc MUT -mccA-F for 2 days, before challenging them with a lethal dose of E. coli EDL933 and enumerating the surviving larvae over a period of 8 days.Initially, we injected the larvae with different doses of EcN bacteria containing pEB-Pmcc MUT -mccA-F, pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, or pEB-Pmcc LOW -mccA-F or E. coli EDL933 to determine the appropriate dosages of the different bacteria to use in the planned experiment.Results revealed that a dosage 1 × 10 4 EcN cells is the maximum prophylactic dosage we could use to inject into the larvae without killing the larvae over a period of 8 days (Fig. 6C).Results also revealed that a dosage of 1 × 10 6 E. coli EDL933 cells killed ~66% of larvae just 24 h following injection (Fig. 6C).Therefore, for our planned experiment, we decided on prophylactically treating the larvae with 1 × 10 4 EcN cells containing pEB-Pmcc MUT -mccA-F, pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, or pEB-Pmcc LOW -mccA-F and, following 2 days of incubation, to challenge the treated larvae with 1 × 10 6 E. coli EDL933 cells.As shown in Fig. 6D, following 2 days after the challenge with E. coli EDL933, only ~22% of untreated larvae and ~30% larvae which were prophylactically treated with EcN bacteria containing pEB-Pmcc MUT -mccA-F survived.The survivability of the larvae was markedly improved when the larvae were prophylactically treated with EcN bacteria producing McC.We observed that EcN bacteria with pEB-Pmcc LOW -mccA-F is consistently better at protecting the larvae against killing by E. coli EDL933 compared to EcN bacteria with pEB-Pmcc WT -mccA-F or pEB-Pmcc HIGH -mccA-F.Overall, we conclude that, expectedly, the ability to produce McC increases the antibacterial effectiveness of EcN, but, consistent with the metabolic perturbations and associated fitness cost of producing McC, the EcN bacteria producing lower amounts of McC display markedly better antibacterial prowess than EcN bacteria producing relatively higher amounts of McC.

DISCUSSION
In the search for new antimicrobials, bacteriocins offer a promising alternative to conventional antibiotics.However, a major challenge with bacteriocins is their delivery in sufficient amounts to the site of infection.However, for infections of the gut, probiotic bacteria retrofitted with heterologous bacteriocin gene cassettes allow in situ production of sufficient amounts of bacteriocins to manage enteropathogens.This is becoming an emerging alternative strategy to promote the clearance of various enteric pathogens with, so far, successful application in different animal models (15)(16)(17).As EcN is a prolific probiotic colonizer of the human gut and exhibits beneficial effects in various intestinal diseases, EcN is often used as a host for introducing bacteriocins to the gut environment.Although bacteriocin gene expression is tightly controlled at the transcriptional level in native producers, EcN bacteria retrofitted to produce heterologous bacteriocins are often engineered to overproduce them from constitutive promoters or in response to host or pathogen signal from synthetic promoters (Fig. 1A).However, it is well accepted that bacteriocin production imparts a fitness cost on bacteria, by draining the primary metabolite pool or because of imperfect immunity of the producer to the bacteriocin (7,10).These costs can become more pronounced when bacteriocin production is decoupled from native regulation since bacteriocin production continues even when it is no longer needed or unlikely to confer a competitive advantage onto the producer (39).The impetus for this study was to investigate whether the dysregulated overproduction of heterologous bacteriocins confers any adverse physiological traits on EcN which might ultimately compromise the antibacterial efficacy of the retrofitted EcN bacteria.Although our bacteriocin of choice was McC, because it is non-toxic and has been demonstrated to be capable of preventing diarrheal disease in weaned piglets when used to supplement feed (40), we envisage that the concepts derived from this study are applicable to wide range of situations where EcN and other probiotic bacteria are engineered to produce heterologous gene products of interest for diverse applications.
We confirmed that McC is active against clinical E. coli isolates and that bacteria generally displayed good sensitivity to McC during anaerobic growth, which is reflective of the condition in the large intestine (41), thus qualifying McC as a good candidate for use in probiotic engineering.An RpoS-dependent promoter, Pmcc, drives the transcrip tion of McC genes and strictly relies on the transcription activator CRP for expression (30,32).The analysis of the core Pmcc sequences and upstream regulatory regions revealed the presence of two major evolutionary clades and one smaller one.Notably, despite differing by the presence of a 21-bp insertion in the regulatory region upstream of Pmcc, sequence divergences between the two larger clades did not result in any differences in transcriptional activity or regulation of representative Pmcc from the two large clades.We consider this as indicative of the strong evolutionary pressure for retention of tight transcriptional regulation of McC production, thus further underscoring the impetus for this study that dysregulated overproduction of bacteriocins in EcN and other probiotic bacteria can potentially have adverse effects on the fitness and desired antibacterial efficacy of such bacteria.
We constructed two variants of Pmcc, Pmcc HIGH and Pmcc LOW , that allowed high and low amounts of McC production compared to the WT Pmcc (Pmcc WT ), while full dependency on CRP, i.e., native regulation, was retained.Although Pmcc WT refers to the native sequence of Pmcc, we note that the production of McC from Pmcc WT in our experimental setup is not necessarily reflective of McC levels in native producers.This is because we introduced the McC operon into a plasmid with a higher copy number (10)(11)(12) than the copy number (1-5) of plasmids on which the McC genes are typically found in native producers (42,43).In fact, a previous study suggested that the production of McC from a non-native plasmid with a similar copy number to the plasmid used by us results in substantially higher amounts of McC produced than from any native plasmid (27).Nonetheless, under our conditions, the transcriptional activity and amount of McC produced from Pmcc HIGH and Pmcc LOW clearly followed the expected trend with respect to Pmcc WT .Although we are yet to investigate the kinetics of RNA polymerase occupancy and activity on Pmcc WT , Pmcc HIGH , and Pmcc LOW , the Pmcc promoter variants we have used in this study might benefit applications where inducer independent of accumulation of gene products of interest at different amounts is required.
The Overall, this study underscores the proverbial saying "less is more" should be an important consideration when retrofitting EcN and other probiotic bacteria with heterologous gene products for therapeutic, prophylactic, or diagnostic applications.Thus, the use of promoter variants that allow for production of the lowest effective amount of a gene(s) of interest in response to its native regulatory signals might be a robust strategy for future design of engineered probiotics.

Computational analysis of Pmcc
A BLAST search was conducted against the RefSeq genomes database using the mccB nucleotide sequence as a query.The sequence 400 bp upstream of the start codon of mccA was manually identified for each hit.Alignment of these sequences and construc tion of the phylogenetic trees were performed within Geneious.Clustal Omega with the mBed algorithm was used for the initial alignment, and the Geneious Tree Builder was then used to generate trees using the Tamura-Nei genetic distance model and the neighbor-joining tree build method.

Construction of plasmids and strains
Escherichia coli strains and plasmids used for this study are listed in Table S1.All plasmids except pEB-Pmcc LOW -GFP were constructed using Gibson assembly (44).The E. coli strains ECR22 and A25922R were identified as containing the mcc gene cluster (see above); the sequence ~400 bp upstream of the mccA genes of each of these strains was used as representative for the clade 1 and clade 2 transcriptional regulatory regions, respectively.These sequences were inserted directly upstream of a super-folder GFP gene in the plasmid pACYC184 (45), which was modified to remove the tetracycline resistance cassette.For promoter modification, pEB-C1-GFP was renamed pEB-Pmcc WT -GFP and used as a vector for constructing pEB-Pmcc MUT -GFP, pEB-Pmcc LOW -GFP, and pEB-Pmcc HIGH -GFP.Also, 120-bp complementary oligonucleotides encompassing the Pmcc HIGH and Pmcc LOW promoter regions were commercially synthesized and cloned into pEB-Pmcc WT -GFP, replacing the corresponding sequence in pEB-Pmcc WT -GFP.For construction of the pEB-Pmcc-mccA-F set of plasmids, the mcc gene cluster (from the translational start codon of mccA) was cloned in place of GFP on the pEB-Pmcc-GFP set of plasmids, such that mccA was directly downstream of the WT and modified Pmcc promoters.The mcc gene cluster was amplified from pp70, kindly provided by Konstantin Severinov (Rutgers University), and includes stand-alone immunity determinant mccF transcribed from an independent, native promoter (Fig. 1A).The E. coli EDL933 strain containing the kanamycin resistance gene downstream of hfq was constructed as previously described (46).

Bacterial growth
Bacteria were grown in LB at 37°C with shaking (~180 rpm) unless otherwise sta ted.Strains carrying a plasmid were grown in LB supplemented with 35 µg/mL chloramphenicol (to select for pACYC184 derivatives-see above).Overnight cultures of ΔrpoS MG1655 strain and EDL933 strain (if containing the kanamycin resistance gene) were supplemented with 50 µg/mL kanamycin.

McC susceptibility disc assays
For assays done under aerobic conditions, bacteria were grown to mid-exponential phase (OD 600nm of ~0.4).Two hundred microliters of culture were mixed with 3 mL of cooled molten soft Mueller-Hinton agar and poured onto a Mueller-Hinton hard agar plate with a depth of 4 mm.Discs infused with 3 µL of filtered supernatant from an McC producer strain (supplied by Syngulon) were applied.The plates were incubated at 37°C for 18 h, after which they were imaged, and the sizes of halos measured.The McC susceptibility assays under anaerobic conditions were done as above, but 20 µL of bacteria from the stationary phase of growth and deoxygenized molten/hard Mueller-Hinton media was used and grown in an anaerobic chamber for 48 h before imaging.All data shown in figures are from two biological replicates.

Plate reader assays
Bacterial cultures were set up in a 96-well plate and incubated in a BMG FlourStar Omega plate reader for 6-10 h.Optical density (OD 600nm ) readings were taken at 30-min intervals.For monitoring of GFP fluorescence over time, OD 600nm and OD 485nm readings were taken at 30-min intervals.OD 485nm values were divided by OD 600nm values at each time point to correct for cell density.All data shown in the figures are from three biological replicates.

McC susceptibility supernatant assays
Overnight cultures of EcN bacteria were centrifuged and washed in fresh LB to remove traces of any antibiotics and were then grown in LB without any antibiotics for 24 h.Subsequently, the supernatant of McC-producing bacteria was harvested and passed through a 0.22-µM sterile filter.Amounts of McC within the supernatant were estimated via growth inhibition assays against the McC-sensitive target EDL933 strain in a 96-well plate.Twofold serial dilutions of supernatant were applied, and growth monitored in a plate reader as described above.All data shown in the figures are from three biological replicates.

Viability assays
Overnight cultures of bacteria were grown to stationary phase in a 50-mL falcon tube and subsequently incubated for 6 days.At each time point, serial dilutions of each culture were plated on LB agar plates containing 35 µg/mL chloramphenicol, and the number of CFU/mL calculated.All data shown in figures are from three biological replicates.

Co-culture assays
Overnight cultures of EcN bacteria were centrifuged and washed in fresh LB to remove traces of any antibiotics.Co-cultures were set up in 50-mL falcon tubes.For co-cultures, where a ratio of 1:1 EcN:EDL933 or 1:1:1 EcN:EDL933:non-susceptible competitor S. aureus was used, a starting dilution of ~2 × 10 7 CFU/mL of each strain was used.For co-cultures with a ratio of 1:100 or 1:1,000 EcN:EDL933, ~2 × 10 5 CFU/mL and ~2 × 10 4 CFU/mL of the EcN strains were used, respectively, while the CFU of the EDL933 strain was maintained ~2 × 10 7 CFU/mL.Co-cultures were incubated for 24 h, then they were washed in PBS, serially diluted, and plated onto LB agar plates containing 50 µg/mL of kanamycin (to only select for the EDL933 strain).All data shown in the figures are from three biological replicates.

RNA-sequencing experiments
Bacteria were harvested during early stationary phase (OD 600nm of ~2.3-3).Three biological replicates of each strain were obtained from across two independent experiments.RNA was extracted using the PureLink RNA Mini Kit from Invitrogen.Extracted RNA was sent to the Core Unit Systems Medicine at the University of Würzburg for next-generation sequencing and alignment to the reference EcN genome assembly ASM354697v1, which includes the sequences of EcN plasmids pNissle1 and pMut2.Raw counts were analyzed with the DESeq2 Bioconductor package, using the Ashr package for LogFold2 shrinkage (47,48).All McC-producing strains were compared to EcN + pEB-Pmcc MUT -mccA-F.Statistical analyses for differential gene expression were performed using R version 4.2.2.

Galleria mellonella infection model
For initial work to the correct bacterial dose of the EcN and EDL933 strains, overnight cultures of the bacteria were harvested and washed twice in PBS.Bacteria were then diluted to a density of 10 6 , 10 7 , and 10 8 CFU/mL and mixed with filter sterilized 0.2% amaranth dye (wt/vol in PBS) at a ratio of 2:1 to aid with accurate injection.PBS and dye without bacteria were used as a negative control.G. mellonella larvae were randomly distributed into five experimental groups and injected with 15 µL of inoculum at the lower left hindleg.The larvae were incubated for 7 days in a 37°C, 5% CO 2 incubator, and survival monitored and recorded.Data shown in Fig. 5C is from five independent experiments with a cumulative total of n = 80 larvae per group.For the sequential injections, G. mellonella were distributed into 10 experimental groups (n = 20 per group).On day 1, the larvae were injected with PBS or 10 6 CFU/mL for one of the four EcN strains.The larvae were incubated in a 37°C, 5% CO 2 incubator for 2 days, after which time they were injected with either 10 8 CFU/mL EDL933 bacteria or PBS.The larvae were monitored for a further 5 days, and survival monitored and recorded.Data shown in figures are from two independent experiments with a cumulative total of n = 40 larvae.

pH measurements and gastric juice challenge
For pH measurements, supernatant was harvested during early stationary phase and sterile filtered.The pH was measured using a standard pH meter and recorded.The synthetic gastric juice medium was made as previously described by Booth and Frost, with the pH adjusted to 2.5 and 35 µg/mL chloramphenicol added (49).For the synthetic gastric juice challenge, bacteria were harvested during early stationary phase (OD 600nm of ~2.3-3), washed in PBS, and resuspended in synthetic gastric juice medium, and incubated for 20 min and 1 h.Serial dilutions of the bacteria were plated on LB agar plates containing 35 µg/mL chloramphenicol, and the number of CFU/mL calculated.Data shown in figures are from three biological replicates.
The genes responsible for the synthesis and secretion of, and immunity to, McC, mccA-F are carried on a low copy number (1-5 copies per cell) plasmid.The mccA-E genes are transcribed from a single operon and encode the structural peptide (MccA), enzymes (MccB, MccD, and MccE), which post-translationally modify MccA to produce mature McC, the McC export pump (MccC), and an immunity determinant (MccE)

FIG 1 (
FIG 1 (A) Table summarizing studies in which EcN has been engineered for heterologous bacteriocin production and the promoters used.(B) Schematic representation of the McC operon where the organization of the structural genes (mccA-E) and the immunity determinant (mccF) is shown with respect to the transcriptional regulatory elements.

FIG 2
FIG 2 The representative images of Mueller-Hinton agar plates show the "halo, " indicating bacterial growth inhibition, surrounding the disc containing McC.The graphs show the size of the halo produced (radius, in millimeters) on Mueller-Hinton agar plates containing the clinical E. coli isolates.The red bar indicates the McC-sensitive reference strain E. coli O157:H7 strain EDL933; the dark bars indicate the clinical isolates that displayed reduced sensitivity to McC.

FIG 4 (
FIG 4 (A) Sequences showing the nucleotide changes introduced (in bold) into Pmcc WT to generate Pmcc HIGH and Pmcc LOW .(B) Graphs showing growth (OD 600nm ) and GFP fluorescence (GFP/ OD 600nm ) from pEB-Pmcc WT -GFP, pEB-Pmcc HIGH -GFP, and pEB-Pmcc LOW -GFP in the E. coli MG1655 strain.(C) As in (B) but in the EcN strain.(D) Bar chart showing GFP fluorescence (GFP/ OD 600nm ) following 10 h of growth in E. coli ΔrpoS MG1655 strain.(E) As in (D) but in the E. coli MG1655 strain in the absence and presence of glucose.(F-H) Graphs showing inhibition of growth of E. coli EDL933 in the presence of different amounts of McC containing supernatant from EcN bacteria containing pEB-Pmcc WT -mccA-F (F), pEB-Pmcc HIGH -mccA-F (G), and pEB-Pmcc LOW -mccA-F (H).

FIG 5 (
FIG 5 (A) Graph showing the growth dynamics of EcN bacteria containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, pEB-Pmcc LOW -mccA-F, and pEB-Pmcc MUT -mccA-F in LB. (B) Graph showing the viability over time of EcN bacteria containing pEB-Pmcc WT -mccA-F, pEB-Pmcc HIGH -mccA-F, pEB-Pmcc LOW -mccA-F, and pEB-Pmcc MUT -mccA-F in LB. (C) Graph showing the number of reads as a percentage of total reads in the RNA-seq data that mapped to the mccA-F genes in (Continued on next page)
production of high amounts of McC has an adverse effect on the overall fitness of EcN bacteria.The transcriptomes of bacteria producing different amounts of McC indicate a strong correlation between the amount of McC produced and the extent of perturbation of metabolism.This could be due to the fact that substantial allocation of cellular resources is needed to support the production of high amounts of McC.Notably, the production of McC leads to the dysregulation of genes associated with acid resistance and degradation of lysozyme-both of which are hallmark features of the gastric environment.Consequently, EcN bacteria producing high amounts of McC do not survive stimulated gastric conditions as well as the EcN bacteria producing low amounts of McC.The compromised fitness of EcN bacteria producing McC is also reflected in the Galleria mellonella larvae infection model of enterohemorrhagic E. coli infection, in which EcN bacteria producing the lowest amount of McC show much higher antibacterial efficacy than EcN bacteria producing the highest amount of McC.It is possible that the fitness costs associated with high amounts of McC production potentially limit the ability of EcN bacteria to compete with the target pathogen.Alternatively, the metabolic perturbations associated with high amount of McC production could attenuate the other, McC-independent, probiotic functions of EcN.Furthermore, potential toxicity of McC on the G. mellonella larvae can also not be excluded.In sum, we anticipate the impaired antibacterial activity of EcN producing high amounts of McC (and conversely, the enhanced antibacterial activity of EcN producing low amounts of McC) would also be reflected in an active infection model, but this will be a subject of future experiments.

TABLE 1
Table showing bactrerial strains and plasmids used in this study.

Table 1
) and measured promoter activity in E. coli MG1655 and EcN bacteria.As shown in Fig.4B, as expected, in E. coli MG1655, the maximal rate of GFP accumulation from Pmcc HIGH was ~1.4-fold higher than the rate of GFP accumulation from Pmcc WT (10,170 GFP/OD 600nm /h for WT; 14,633 GFP/OD 600nm /h for high).This meant that after 10 h of growth in LB, the amount of GFP made from Pmcc HIGH exceeded that made from Pmcc WT by ~1.6-fold.Conversely, the rate of GFP accumulation from Pmcc LOW was ~2.9-fold lower than the rate of GFP accumulation from Pmcc WT (3,548 GFP/ OD 600nm /h for low).This meant that after 10 h of growth in LB, the amount of GFP made from Pmcc LOW was reduced by ~3.4-fold compared to that made from Pmcc WT .Similarly, in EcN bacteria, the rate of GFP accumulation from Pmcc HIGH was ~1.8-fold higher than the rate of GFP accumulation from Pmcc WT(2,679 GFP/OD 600nm /h for WT; 4,900 GFP/ OD 600nm /h for high).This meant that after 10 h of growth in LB, the amount of GFP made from Pmcc HIGH exceeded that made from Pmcc WT by ~1.9-fold.Conversely, the rate of GFP accumulation from Pmcc LOW was approximately twofold lower than the rate of GFP accumulation from Pmcc WT (1,357 GFP/OD 600nm /h for low).This meant that after 10 h of growth in LB, the amount of GFP made from Pmcc LOW was reduced by 2.3-fold compared to that made from Pmcc WT .In both, E. coli MG1655 and EcN bacteria, the amount of GFP produced from Pmcc HIGH and Pmcc LOW differed by ~5.4-and ~4.5-fold, respectively.