Discovery and structure of the antimicrobial lasso peptide citrocin

We report the identification of citrocin, a 19-amino acid-long antimicrobial lasso peptide from the bacteria Citrobacter pasteurii and Citrobacter braakii. We refactored the citrocin gene cluster and heterologously expressed it in Escherichia coli. We determined citrocin's NMR structure in water and found that is reminiscent of that of microcin J25 (MccJ25), an RNA polymerase-inhibiting lasso peptide that hijacks the TonB-dependent transporter FhuA to gain entry into cells. Citrocin has moderate antimicrobial activity against E. coli and Citrobacter strains. We then performed an in vitro RNA polymerase (RNAP) inhibition assay using citrocin and microcin J25 against E. coli RNAP. Citrocin has a higher minimal inhibition concentration than microcin J25 does against E. coli but surprisingly is ∼100-fold more potent as an RNAP inhibitor. This suggests that citrocin uptake by E. coli is limited. We found that unlike MccJ25, citrocin's activity against E. coli relied on neither of the two proton motive force-linked systems, Ton and Tol–Pal, for transport across the outer membrane. The structure of citrocin contains a patch of positive charge consisting of Lys-5 and Arg-17. We performed mutagenesis on these residues and found that the R17Y construct was matured into a lasso peptide but no longer had activity, showing the importance of this side chain for antimicrobial activity. In summary, we heterologously expressed and structurally and biochemically characterized an antimicrobial lasso peptide, citrocin. Despite being similar to MccJ25 in sequence, citrocin has an altered activity profile and does not use the same outer-membrane transporter to enter susceptible cells.

We report the identification of citrocin, a 19-amino acid-long antimicrobial lasso peptide from the bacteria Citrobacter pasteurii and Citrobacter braakii. We refactored the citrocin gene cluster and heterologously expressed it in Escherichia coli. We determined citrocin's NMR structure in water and found that is reminiscent of that of microcin J25 (MccJ25), an RNA polymerase-inhibiting lasso peptide that hijacks the TonB-dependent transporter FhuA to gain entry into cells. Citrocin has moderate antimicrobial activity against E. coli and Citrobacter strains. We then performed an in vitro RNA polymerase (RNAP) inhibition assay using citrocin and microcin J25 against E. coli RNAP. Citrocin has a higher minimal inhibition concentration than microcin J25 does against E. coli but surprisingly is ϳ100-fold more potent as an RNAP inhibitor. This suggests that citrocin uptake by E. coli is limited. We found that unlike MccJ25, citrocin's activity against E. coli relied on neither of the two proton motive force-linked systems, Ton and Tol-Pal, for transport across the outer membrane. The structure of citrocin contains a patch of positive charge consisting of Lys-5 and Arg-17. We performed mutagenesis on these residues and found that the R17Y construct was matured into a lasso peptide but no longer had activity, showing the importance of this side chain for antimicrobial activity. In summary, we heterologously expressed and structurally and biochemically characterized an antimicrobial lasso peptide, citrocin. Despite being similar to MccJ25 in sequence, citrocin has an altered activity profile and does not use the same outer-membrane transporter to enter susceptible cells.
Microcins are small antimicrobial peptides with molecular masses below 10 kDa (1,2). They are produced by bacteria under conditions of stress as weapons against closely related bacteria, resulting in narrow-spectrum activity. Many microcins exhibit potent antimicrobial activity against susceptible strains (3). For example, microcin E492 has a minimum inhibitory concentration (MIC) 4 of 40 nM against Escherichia coli B (4), whereas microcin J25 has a MIC of 5 nM against S. enterica serotype Newport (5). The potent yet narrow-spectrum activity of microcins make them potential new antibiotics that have fewer unintended side effects on the microbiome than traditional broad-spectrum antibiotics.
Some microcins are unmodified peptides, but many undergo post-translational modification (1,2). Lasso peptides are a class of ribosomally synthesized and post-translationally modified peptides (2) that are characterized by their unique lasso structure (6,7). This lasso structure is formed by an isopeptide bond between the N terminus and an aspartate or glutamate side chain to form a 7-9-membered ring through which the C-terminal end of the peptide is threaded and locked in place. This constrained structure can confer high thermal stability and resistance against proteolytic degradation. For example, the well-studied antimicrobial lasso peptide MccJ25 remains threaded and functional after boiling in an aqueous solution at 100°C (8,9). In addition to antimicrobial activity, characterized lasso peptides have a wide range of biological activities, including receptor antagonism and antiviral activity (2).
Lasso peptide gene clusters contain a minimum of three genes, A-C. The A gene encodes the lasso peptide precursor with an N-terminal leader sequence, whereas the B and C genes encode enzymes required for maturation (Fig. 1A). Lasso peptides that exhibit antimicrobial activity have been found to also contain a D gene that encodes an ABC transporter. This transporter confers host immunity through active efflux of the toxic lasso peptide. The vast majority of characterized lasso peptide gene clusters have the ABC genes and, if a D gene is present, the ABCD genes, on a single putative operon. The notable exception to this is microcin J25's gene cluster, in which the mcjA gene is transcribed in the opposite direction of the mcjBCD genes (6). This study reveals a second example of a lasso peptide with this rare gene cluster architecture (Fig. 1C).
Studies on the antimicrobial lasso peptide microcin J25 offered insight into how a lasso peptide is taken up by susceptible strains, crossing the outer and inner membranes of Gramnegative bacteria. Microcin J25 hijacks the TonB-dependent transporter FhuA on the outer membrane by mimicking the natural substrate ferrichrome (10,11). It then requires the inner membrane protein SbmA to gain access to its intended target, RNA polymerase (10,(12)(13)(14). Most other studied microcins and colicins similarly rely on active transport across the outer membrane using one of the two proton motive force-linked systems, Ton and Tol-Pal, with the exceptions of microcin B17 and microcin C7, which rely on the outer membrane porin OmpF (1,15,16).
Here we report the discovery and characterization of a new lasso peptide, citrocin. Citrocin's gene cluster was initially identified using genome mining from enterobacterium Citrobacter pasteurii type strain CIP 55.13. C. pasteurii CIP 55.13 was isolated from a human diarrheal stool sample in Kentucky and deposited into the Collection de l'Institut Pasteur, France, in 1955 (17). We later also identified citrocin's gene cluster in the recently sequenced Citrobacter braakii type strain ATCC 51113, which was isolated from a snake in France (18). We were able to express citrocin using the native host C. braakii and heterologously in E. coli with a codon-optimized and refactored gene cluster. Purified peptide was used to test thermostability, obtain an aqueous NMR structure, and screen for antimicrobial activity against a panel of Gram-negative bacteria. We show that the Arg-17 side chain is critical for antimicrobial activity of citrocin. We also show that citrocin is a potent inhibitor of RNA polymerase in vitro. Finally, we investigated citrocin's uptake mechanism by generating E. coli variants with resistance and analyzing their genetic changes from the original sensitive strain, revealing the involvement of inner membrane protein SbmA. We confirmed that SbmA is required for uptake using an antimicrobial activity assay against a knockout strain. Surprisingly, the sequencing results and activity assays against outer membrane transporters and Ton/Tol-Pal knockouts indicate that citrocin crosses the outer membrane using a mechanism distinct from that of MccJ25.

Identification of citrocin biosynthetic gene cluster
The citrocin gene cluster was initially identified in C. pasteurii CIP 55.13 using an updated version of our genome mining method (19).We subsequently confirmed that it can also be identified using a BLAST search with McjB as query. The gene cluster is found on an 18,560-bp contig (CDHL01000044.1). A subsequent run of genome mining also identified the same gene cluster in C. braakii ATCC 51113. Notably, the gene cluster has a very low GC content of 27% compared with the 52% GC content of the C. pasteurii genome and with the 42% GC content of the contig. Although neither the C. pasteurii nor C. braakii genomes are to date fully assembled, the presence of common plasmid-associated genes on this contig suggest that the cluster may be located on a plasmid or a plasmid that has been integrated into the genome. This includes the presence of genes that encode a RepB family plasmid replication initiator protein and two type II toxin-antitoxin pairs (20) (Fig. S1). The predicted lasso peptide sequence has significant similarity to that of MccJ25. Fig. 1B shows that the two peptides have ten identical residues and four similar residues. However, there are two major differences: citrocin is shorter and has a net positive charge. Also like MccJ25 and unlike the gene clusters of all other lasso peptides characterized to date, citrocin's gene cluster has the citA precursor gene transcribed in the opposite direction of the rest of the genes in the cluster (Fig. 1C).

Heterologous expression and purification of citrocin
Because of the low GC content of the gene cluster, we chose to codon optimize and refactor it before attempting heterologous expression in E. coli. Previously, we had refactored the native MccJ25 gene cluster (without codon optimization) in a similar fashion and found a modest increase in production yield (21). The precursor gene citA was placed under an isopropyl ␤-D-1-thiogalactopyranoside-inducible T5 promoter, whereas the citBCD genes were placed under the constitutive promoter for the mcjBCD genes from the MccJ25 gene cluster (Fig. 1D). Citrocin was expressed in E. coli BL21 in M9 medium supplemented with amino acids and purified from the supernatant using HPLC with a yield of 2.7 mg/liter ( Fig. 2A). Mass spectrometry confirmed the expected lasso peptide's monoisotopic mass of 1880 Da (Fig. 2B).

Expression of citrocin from native host C. braakii
C. braakii was grown in M9 medium with and without amino acid supplementation at 30°C for 2 days. A C8 extract of the supernatant was analyzed by HPLC and LC-MS, confirming the production of citrocin with the same retention time and mass as the peptide heterologously produced in E. coli (Fig. S2). Production level was estimated using HPLC by integrating the peak

Citrocin is an antimicrobial lasso peptide
corresponding to citrocin and comparing it with a standard curve generated using purified citrocin (Fig. S3). C. braakii, when grown in M9 medium with amino acid supplementation, produced 0.7 mg/liter of citrocin.

Structural analysis of citrocin and comparison to MccJ25
Citrocin was prepared for 2D NMR experiments in 95% Fig. S4 shows the TOCSY and NOESY spectra. Simulated annealing of the structures was carried out using CYANA (22) followed by energy minimization using GROMACS with explicit water (23)(24)(25)(26). The top 20 structures were similar to each other (Fig. S5). The top structure is shown in Fig. 3

(A and B).
Citrocin has a right-handed lasso structure with the C-terminal end threaded through a macrocyclic ring formed by an isopeptide bond between Gly-1 and Glu-8. The C-terminal tail is sterically locked into place with Arg-17 above the ring and Tyr-18 below the ring. This results in a very short "tail" below the ring, consisting of just Tyr-18 and Gly-19.
The only other lasso peptide characterized with such a short tail structure is MccJ25. Three organic solution structures of MccJ25 in methanol or DMSO have been published (27)(28)(29), with the two methanol structures deposited into the PDB (codes 1PP5 and 1Q71). Additionally, a crystal structure of MccJ25 in complex with the outer membrane transporter FhuA is also available in the PDB (code 4CU4) (11). We have recently published a crystal structure of MccJ25 bound to E. coli RNAP (code 6N60) (14). The two solution structures in the PDB show similar ring and tail structures, although the structure of the loop region varies between them and is different from that of MccJ25 in complex with FhuA or RNAP (Fig. S6). This difference is likely due to the flexibility of the loop region while in solution.
A comparison of the 1PP5 structure with citrocin's aqueous structure shows that citrocin has a shorter loop region above the ring (Fig. 3, C and D). The loop region is more similar to MccJ25's 1PP5 structure than 1Q71 (Fig. S6). Additionally, citrocin has a net positive charge with positively charged Lys-5 and Arg-17 residues and a negatively charged C terminus. Interestingly, the positively charged residues are spatially close together (Fig. S7).

Thermostability
The thermostability of citrocin was tested by heating an aqueous solution to 95°C for 3 h. Part of the sample was then digested with carboxypeptidase. The heat-treated sample before and after digestion was analyzed by HPLC and MALDI. Typically unthreading of the tail, whether partial or complete, causes a retention time shift in the HPLC trace (30,31). Additionally, unthreading would expose the C-terminal tail allowing carboxypeptidase digestion to occur, which would result in both a retention time change in the HPLC trace, as well as a mass change. HPLC traces and MS data (Fig. S8) show that heat-treated citrocin before and after digestion has the same retention time and mass as unheated citrocin. Hence citrocin is a highly thermostable lasso peptide and able to resist unthreading for 3 h at 95°C.

Citrocin is an antimicrobial lasso peptide Antimicrobial activity screen
Purified citrocin dissolved in water was used for spot-onlawn activity assays. Citrocin was tested against strains of E. coli, Salmonella, Citrobacter, Pseudomonas, and Serratia (Table S1). Inhibition zones were observed against E. coli, Citrobacter, and Salmonella strains. The MICs are reported in Fig.  4. Citrocin had moderate activity against E. coli and Citrobacter strains with MIC values ranging from 16 to 125 M and some weak activity against Salmonella Newport with a MIC value of 1000 M. Its strongest activity was against enterohemorrhagic E. coli (EHEC) O157:H7 TUV93-0 with an MIC of 16 M. Interestingly, the EHEC strain was much more sensitive to citrocin than the E. coli lab expression strain BL21, which had a MIC of 100 M. Also of note, MccJ25 has a much stronger relative activity against Salmonella Newport (40 nM) than against E. coli (1 M) (32), suggesting that the two peptides differ either in uptake, mode of action, or both.

RNA polymerase inhibition
Next we tested for in vitro RNA polymerase inhibition using an abortive initiation assay against E. coli RNAP. Citrocin was tested at 1, 10, and 100 M concentration in triplicate, along with purified MccJ25 for comparison. At 1 M of citrocin, we observed only 15% of the transcription level compared with the level of the no-inhibitor control (Fig. 5 and Fig. S9). In contrast, 100 M of MccJ25 was required to achieve the same reduction of transcription (Fig. 5), which is in agreement with previous results (12). This 100-fold difference is surprising because citrocin has a higher MIC (16 -125 M) against E. coli than MccJ25 (ϳ1 M), yet citrocin is actually a much stronger RNAP inhibitor. This suggests that citrocin uptake is the limiting factor to its activity against E. coli.

Citrocin resistance
We generated citrocin-resistant variants of E. coli BW25113 by growing liquid cultures with varying concentrations of citrocin added in triplicate. We chose this strain because it is the parent strain of the Keio collection of single-gene knockouts. At 100 M of citrocin, none of the triplicate cultures grew after 18 h (Fig. 6A). At 12.5 M of citrocin, all three cultures showed some growth retardation but were able to grow in a linear fashion. At the intermediate concentrations of 25,50, and 75 M, however, the cultures either did not grow by 18 h or exhibited a long lag phase of ϳ8 h followed by exponential growth. The bacteria in these five cultures have spontaneously developed resistance to citrocin. This was confirmed by restreaking the cells in the absence of citrocin and growing new liquid cultures for confirmation of resistance using the spot-on-lawn assay (Fig. S10).
We then sequenced the resistant variants, as well as the WT E. coli BW25113 strain, and looked for genomic differences. One of the variants had a point mutation in the sbmA gene. Another variant had both a point mutation and deletion of part of the sbmA gene that resulted in a frameshift. A third variant had a large deletion that included the entire sbmA gene. SbmA is an inner membrane protein in E. coli of unknown physiological role that is required for the uptake of other antimicrobial peptides such as microcin J25 and microcin B17 (33). Fig. S11 shows the predicted transmembrane topology of SbmA using TMHMM (34) annotated with the locations of observed changes. The frequency of changes and deletions to sbmA suggested that it is also involved in transporting citrocin across the inner membrane. We confirmed that the Keio strain E. coli BW25113 ⌬sbmA (35,36) is resistant to citrocin via the spoton-lawn assay (Fig. 6B).
Surprisingly, we did not find a mutation in an outer membrane transporter nor in the Ton/Tol pathway proteins, which is what we originally anticipated because a similar experiment done to generate MccJ25-resistant variants found the most common mutations occurred in FhuA, in proteins in the Ton pathway, and in SbmA (10). However, our sample of citrocinresistant variants is relatively small, so we complemented this observation by testing citrocin against a panel of Keio knockouts, including all known TonB-dependent transporters, TonB, TolA, and porins (Table S2 and Fig. S12). Citrocin was active against the entire panel tested. We noted that the ⌬tolA Keio isolate 1 that we have is not correct (35) and knocked out the tolA gene from WT E. coli BW25113 using the method described by Datsenko and Wanner (37). Citrocin was active against this PCR-confirmed knockout ( Fig. 6B and Fig. S13). Citrocin had a noticeably lower MIC against the Tol system knockouts. This is likely due to the reduced fitness of cells because the Tol proteins contribute to membrane stability (38). Unlike MccJ25 and many other microcins and bacteriocins, citrocin does not rely on the Ton system, nor does it rely on the other PMF-linked Tol system for transport across the outer membrane.

Mutagenesis of Lys-5 and Arg-17
As mentioned above, the side chains of Lys-5 and Arg-17 are relatively close to each other in the 3D structure of citrocin, generating a patch of positive charge on one face of the molecule (Fig. S7). Given that citrocin does not appear to cross the outer membrane via any of the energy-coupled transport mechanisms (Table S2), we next considered the possibility that the positive charge on Lys-5 and Arg-17 is involved in transport across the membrane.
We generated K5A and K5E variants of citrocin, both of which expressed well and could be purified to near homogeneity (Fig. S14). Both of these variants still had activity against E. coli BW25113. The K5A variant was equivalent to the WT activity, whereas K5E was somewhat less efficacious (Fig. S15). Thus, neutralizing or even inverting the charge at K5 position does not strongly affect citrocin activity.
Next we generated substitutions at Arg-17. Given that Arg-17 is a steric lock residue and is likely involved in maintaining the threaded structure of citrocin, initially we generated R17L and R17E variants. Although we could detect some small amounts of the R17E variant by MS, it was not made in suffi-cient amounts for purification by HPLC. The R17L citrocin variant eluted as two peaks on HPLC (Fig. S16). When either of the two peaks were collected and reinjected to the LC-MS, two peaks again appeared. This indicated that the R17L variant may be interconverting between two distinct threaded topoisomers. We have observed similar behavior previously with the peptide benenodin-1 when it is heated (30). We saw no activity for the R17L variant against E. coli.
We were dissatisfied with the possibility that the R17L citrocin variant may have lost its activity simply because of a change in conformation, so we next generated an R17Y variant of citrocin. This peptide expressed well, eluted as a single peak on HPLC, and could be purified to homogeneity (Fig. S14). Furthermore, it was resistant to carboxypeptidase digestion, suggesting that it is in a threaded conformation similar to WT citrocin (Fig. S17). Despite this, the R17Y variant does not exhibit any antimicrobial activity against E. coli up to a concentration of 250 M (Fig. S18). Collectively these data indicate that the Lys-5 side chain is not important for citrocin's antimicrobial activity, whereas Arg-17 is important for both peptide stability and antimicrobial activity.

Discussion
In this study we heterologously expressed and characterized a new thermostable antimicrobial lasso peptide, citrocin. Although this peptide is similar to microcin J25 in sequence, it has an altered activity profile and does not rely on the same outer membrane transporter to enter susceptible E. coli. In fact, it appears to rely neither on the PMF-linked Tol-Pal or Ton systems, nor on porins. We have shown that the Arg-17 residue is critical for the antimicrobial activity of citrocin (Fig. S18). Arg residues are often implicated in cell-penetrating peptides (39). The electrostatic interaction between Arg-17 and the outer membrane may serve to increase the effective concentration of citrocin at the cell surface, which may allow for energy-independent uptake of the peptide into the periplasm.
One particularly compelling aspect of citrocin is its high potency as an RNA polymerase inhibitor in vitro (Fig. 5). Despite this high potency against its putative cytoplasmic target, citrocin has only modest activity against the enterobacteria tested here. It is possible that citrocin has a very narrow spectrum of activity and that it is actually highly potent against an unidentified enterobacterium. Further testing of citrocin against larger strain collections may reveal strains that are highly susceptible to citrocin. There is precedence for this type of narrow spectrum activity with microcin J25. Although it is very potent against Salmonella Newport, it is unable to kill several other strains of Salmonella (40). Another possibility is that the citrocin gene cluster is functioning like a toxin-antitoxin (TA) module with citrocin as the toxin and the CitD ABC transporter as the corresponding antitoxin. The citrocin gene cluster is found nearby two other TA modules (Fig. S1), and we hypothesize that the contig on which we found the citrocin gene cluster may correspond to a plasmid. As the toxin of a TA module, citrocin would only have to function intracellularly. This would remove the selective pressure on citrocin to maintain its ability to be taken up into susceptible cells. In this scenario, the ability for uptake may have evolved away from citrocin, explaining the modest antimicrobial activities we observe.
Although citrocin has modest antimicrobial activity against other enterobacteria, it was nonetheless noteworthy that citrocin is more potent against a pathogenic strain of EHEC than it is against laboratory strains of E. coli (Fig. 4). We were also surprised, given the similarity of citrocin to MccJ25, at the differ- Additional spot assays are provided in Fig. S12. The plate image for the ⌬fhuA knockout is reproduced in Fig. S12 for comparison purposes. Citrocin does not appear to rely on the Ton or Tol translocation systems.

Citrocin is an antimicrobial lasso peptide
ence in the spectrum of activity of these peptides. Most notably, Salmonella Newport, which is hypersensitive to MccJ25, is only killed by millimolar concentrations of citrocin. Antimicrobial activity of lasso peptides is multifaceted; the peptide must cross both the outer membrane and the inner membrane and must find a cytoplasmic target. Our observations here underscore the difficulty in predicting the antimicrobial activity of a lasso peptide from its sequence alone.

Materials
Cloning was done using XL-1 Blue E. coli cells. Expression was done using BL21 E. coli cells. All primers and gBlocks were purchased from Integrated DNA Technologies. All restriction enzymes were purchased from New England Biolabs. Picomaxx polymerase was purchased from Agilent Technologies. Strata C8 extraction columns were purchased from Phenomenex. Genomic preps were done using DNeasy blood and tissue kits from Qiagen. C. braakii ATCC 51113 was obtained from the American Type Culture Collection.
HPLC was done using an Agilent 1200 series instrument. Mass spectrometry analysis was done using a Bruker Ultra-Flextreme MALDI TOF/TOF (Princeton Proteomics and Mass Spectrometry Core Facility) and Agilent 6530 QTOF LC-MS. Plate reader experiments were done using a Biotek Synergy 4 instrument. Genomic sequencing was done using Illumina MiSeq (Princeton Genomics Core Facility).

Biosynthetic gene cluster identification
The gene cluster for citrocin was identified in the published genome of C. pasteurii strain CIP 55.13 using an updated version of our precursor-centric genome mining method (19) and BLAST searches.

Plasmid construction
The citABCD gene cluster was codon optimized using DNAWorks. The citA gene with an upstream RBS was assembled from the six oligonucleotides that were designed with DNAWorks (41). Assembly PCR consisted of an initial assembly PCR step with a mixture of the oligonucleotides followed by another PCR step to amplify the assembled gene. The gene was then cloned into the pQE-80 vector using EcoRI and HindIII restriction enzymes, creating pWC82. The codon optimized citBCD genes were purchased as two overlapping gBlocks with an upstream constitutive promoter from microcin J25's biosynthetic gene cluster. The two gBlocks were then assembled in an overlap PCR reaction and cloned into the NheI site of pWC82, creating pWC88, which was used to express WT citrocin. Gene fragments for citrocin variants K5A, K5E, R17L, R17E, and R17Y were generated by PCR and cloned into pWC88 digested with EcoRI and HindIII. The sequences of all primers and gBlocks used in this work can be found in Table S3.

Heterologous expression
E. coli BL21cells were transformed with pWC88 and used to express citrocin. Lasso peptide was expressed in M9 medium supplemented with 20 amino acids (0.05 g/liter of each amino acid) and 0.00005 wt% thiamine with 100 mg/liter ampicillin. The cultures were inoculated with an overnight LB culture to a starting A 600 of 0.02 and then grown at 37°C at 250 rpm up to an A 600 of 0.2-0.25. Cultures were then induced with 1 mM isopropyl ␤-D-1-thiogalactopyranoside and expressed for 20 h at 20°C, 250 rpm. Supernatant was harvested by spinning down the cultures at 4000 ϫ g at 4°C for 20 min.

Peptide purification
For a 500-ml expression, one Strata 6 ml/1-g column was activated with 6 ml of methanol and then washed with 12 ml of H 2 O. The supernatant was extracted through the column, which was then washed again with 12 ml of H 2 O before being eluted with 6 ml of methanol. The methanol extract was then rotavapped dry and resuspended with 500 l of 1:3 acetronitrile:water.
The peptide was purified from the resuspended extract using reverse-phase HPLC. Briefly, 50 l was injected onto a Zorbax 300SB-C18 semiprep column. An acetonitrile-water gradient flowing at 4 ml/min was used to separate the peptide from other compounds. The gradient started at 10% acetonitrile and increased linearly from 10 to 50% acetonitrile from 1 min to 20 min post-injection. Citrocin eluted with a retention time of 14.9 min, as confirmed by MS (electrospray ionization-QTOF and MALDI). Fractions containing citrocin were collected and lyophilized.
Citrocin variants were similarly purified with one modification. For variants K5A and K5E, the gradient started at 10% acetonitrile and increased linearly from 10 to 50% acetonitrile from 1 min to 30 min post-injection.

Citrocin expression in C. braakii
C. braakii was grown overnight in nutrient broth and then subcultured into 500 ml of M9 media supplemented with amino acids and thiamine (see heterologous expression section) or just thiamine to a starting A 600 of 0.006. Cultures were grown at 30°C, 250 rpm for 48 h. Supernatant was then harvested, extracted, and resuspended to 500 l as described above under "Peptide purification." The extract was analyzed by LC-MS, where citrocin was detected at the expected retention time. 10 l of the extract was then injected onto a Zorbax 300SB-C18 analytical column. The peak corresponding to citrocin was integrated, and concentration was determined using a standard curve created using purified citrocin.

Peptide thermal stability
Purified citrocin in water at 2.2 mg/ml was heated in a thermocycler at 95°C for 3 h. After heating, half the sample was treated with carboxypeptidase B and Y in 50 mM sodium acetate buffer, pH 6, for 3 h at 20°C. The heated peptide, before and after carboxypeptidase treatment, was analyzed by HPLC (Zorbax 300SB C18 analytical column) and by MALDI.

NMR studies
Purified citrocin sample was dissolved in 95:5 H 2 O/D 2 O at a final concentration of 3.8 mg/ml. TOCSY and NOESY spectra were acquired at 10°C with 60-and 100-ms mixing times, respectively. Chemical shifts for all protons were assigned Citrocin is an antimicrobial lasso peptide based on intra-and inter-residue connectivities seen in the TOCSY and NOESY spectra ( Fig. S4 and Table S4). Crosspeaks were manually picked from the NOESY spectrum and integrated. These cross-peak volumes were then used for calibration and as distance constraints in structural calculations performed using CYANA 2.1. Seven cycles of combined automated NOESY assignment and structural calculations were done, followed by a final structure calculation. Structural statistics are provided in Table S5. The top 20 structures were then energy-minimized in explicit solvent in GROMACS using a procedure described by Spronk et al. (23). Briefly, the peptide was placed in a simulation box and then solvated with tip3p water. The system was simulated for 4 ps, cooling from 300 to 50 K.

Accession codes
The atomic coordinates for the citrocin NMR structure reported in this paper have been deposited in the PDB (accession number 6MW6) and the Biological Magnetic Resonance Data Bank (accession number 30530).

Antimicrobial activity test
Antimicrobial activity was tested using a spot-on-lawn assay as previously described (21). Briefly, a 10-ml M63 agar plate was overlaid with 10 ml of M63 soft agar containing 10 8 CFUs of the target bacteria strain. After the soft agar solidified, 10-l dilutions of citrocin were spotted onto the plate and dried. The plates were then incubated at 37°C overnight and analyzed for inhibition zones the next day.

RNAP inhibition experiments
RNAP inhibition was tested in vitro using an abortive initiation assay. Each 10-l reaction contained 125 nM core RNAP, 625 nM 70 , and 50 nM T7A1 promoter DNA fragment in transcription buffer (100 mM KCl, 10 mM MgCl 2 , 50 mM Tris, pH 8.0, 10 mM DTT, 50 g/ml BSA). First, the core RNAP was incubated with 70 for 10 min at 37°C. Next, the T7A1 promoter DNA was added. After 10 min, heparin was added to a final concentration of 25 g/ml, and 100, 10, or 1 M of each peptide inhibitor was added to each reaction. After an additional 10 min at 37°C, RNA synthesis was initiated with the addition of an NTP mix consisting of 500 M CpA, 100 M UTP, and 0.1 Ci of [␣-32 P]UTP. The reactions were conducted for 10 min at 37°C and terminated with 2ϫ stop buffer (8 M urea and 1ϫ Tris-borate-EDTA). The reactions were heated at 95°C for 10 min and loaded on a 23% polyacrylamide gel (19:1 acrylamide:bis-acrylamide). Abortive products were visualized by exposing the gel on a GE storage phosphor screen overnight and digitized using a Typhoon phosphorimaging device. The data were quantitated using ImageJ.

Generation and sequencing of citrocin-resistant E. coli
A 5-ml LB culture of E. coli BW25113 was grown overnight. This overnight culture was diluted 300-fold to start 18 ϫ 135-l cultures in a 96-well plate. 15 l of citrocin was added to the culture at six concentrations with three replicates each. The final concentrations of citrocin were 0, 12.5, 25, 50, 75, and 100 M. The plate was incubated at 37°C with continuous shaking in a plate reader for 18 h, during which absorbance at 600 nm was measured every 10 min.
Five of the wells with citrocin in it had a long lag phase of 8 h before exhibiting exponential growth. The cells from these five wells and the WT E. coli BW25113 strain were streaked onto LB plates. An individual colony of each was then picked to confirm citrocin resistance via the spot-on-lawn assay and also used for genomic DNA extraction.
The six genomic DNA samples were sent to the Princeton University Genomics Core Facility where libraries were created using a Nextera DNA kit. The libraries were then sequenced with paired-end reads of 150 bp using an Illumina MiSeq system. Average genomic coverage for each sample was 20-fold. Data were processed and analyzed using a local instance of Galaxy. Sequences were mapped to the reference E. coli BW25113 genome using the BWA-MEM algorithm (42). FreeBayes was then used to find genetic variants (43). Additionally, genome interval coverage was analyzed using the genome coverage tool within the BEDTools package (44).