Discovery of ubonodin, an antimicrobial lasso peptide active against members of the Burkholderia cepacia complex

Abstract

We report the heterologous expression, structure, and antimicrobial activity of a lasso peptide, ubonodin, encoded in the genome of Burkholderia ubonensis. The topology of ubonodin is unprecedented amongst lasso peptides with 18 of its 28 amino acids found in the mechanically-bonded loop segment. Ubonodin inhibits RNA polymerase in vitro and has potent antimicrobial activity against several pathogenic members of the Burkholderia genus, most notably B. cepacia and B. multivorans, causative agents of lung infections in cystic fibrosis patients.

Graphical Abstract

Ubonodin is a novel lasso peptide discovered by genome mining in a strain of Burkholderia. It has an unprecedented topology for a lasso peptide with an 18 aa mechanically bonded loop. Ubonodin exhibits potent antimicrobial activity against pathogenic Burkholderia species.

Introduction

Burkholderia is a genus of Gram-negative Proteobacteria comprised of resilient and ubiquitous bacteria that are mainly environmental saprophytes.^[1] Many of its members though, are opportunistic pathogens that can cause fatal diseases. Burkholderia mallei and Burkholderia pseudomallei, are classified as Tier 1 Select Agents by the US Federal Select Agent Program, causing glanders in animals and melioidosis in humans respectively.^[1–2] The Burkholderia cepacia complex (Bcc) consists of more than 20 closely related species of which many are opportunistic plant and human pathogens.^{[1, 3]} Bcc members are especially dangerous to patients with an underlying lung disease, such as those with cystic fibrosis (CF), causing deadly pneumonia. Bcc infections are difficult to treat due to their innate resistance to many antibiotics, their ability to persist even with aggressive antibiotic treatment, and their ability to acquire resistance to these antibiotics.^{[1, 3–4]} Members of the Burkholderia genus also encode a variety of natural products.^[5] Our group has focused on lasso peptides, a class of ribosomally synthesized and post-translationally modified (RiPP)^[6] natural products defined by their chiral rotaxane structure established via formation of an isopeptide bond between the peptide N-terminus and an acidic sidechain.^[7] To date two lasso peptides have been isolated from Burkholderia: capistruin^[8] from B. thailandensis and burhizin^[9] from B. rhizoxinica. While no activity has been reported for burhizin, capistruin has weak to modest antimicrobial activity against several Gram-negative bacteria including a strain of E. coli, a strain of Pseudomonas aeruginosa, and Bcc species. It was subsequently demonstrated that capistruin exerts its antimicrobial via inhibition of RNA polymerase.^[10] Here we present the discovery, structure, and antimicrobial activity of ubonodin, a novel lasso peptide encoded in the genome of B. ubonensis. Ubonodin is significantly larger than previously described lasso peptides, and has an unprecedented topology. More importantly, ubonodin has potent antimicrobial activity against multiple pathogenic Burkholderia, including Bcc species implicated in lung infections in CF.

Results and Discussion

We identified a lasso peptide gene cluster in the organism Burkholderia ubonensis MSMB2207 using our methodology for lasso peptide genome mining.^[11] This cluster also appeared in BLAST searches of the biosynthetic enzymes for citrocin, an antimicrobial lasso peptide produced by strains of Citrobacter.^[12] We were intrigued by the large size, 28 aa, of the core peptide of this putative lasso peptide (Figure 1), longer than any previously characterized example.^[7b] The lasso peptide gene cluster has 55% GC content, somewhat lower than the GC content of B. ubonensis genomes, which is ~67%. Currently, there are 306 B. ubonensis genomes in the RefSeq database, and 16 of them harbor this lasso peptide gene cluster (Table S1). We decided to refactor the gene cluster for heterologous expression in E. coli, a strategy that worked well for the production of citrocin.^[12] Briefly, the uboA gene encoding the lasso peptide precursor was placed under the control of a strong IPTG-inducible promoter while the uboBCD cassette containing the maturation enzymes and transporter were placed under a constitutive promoter (Figure 1, Figure S1). The refactored uboABCD gene cluster was introduced into E. coli BL21 which was able to produce 1.8 mg/L of a peptide with a monoisotopic mass of 3197.382 g/mol, which matches well to the predicted mass of the core peptide with one dehydration (3197.376 g/mol). In MS² experiments, this peptide fragmented minimally, similar to what was observed with the lasso peptide microcin J25 (MccJ25)^[13] (Figure S2). We named this peptide ubonodin after the organism that encodes it, B. ubonensis, and the Latin root for knot, nodum.

Figure 1.

Ubonodin sequence and biosynthetic gene cluster. A: Ubonodin is the largest lasso peptide discovered at 28 aa; the line shows the isopeptide bond between Gly-1 and Glu-8. B: Native ubonodin biosynthetic gene cluster C: Refactored ubonodin biosynthetic gene cluster.

We next set about determining the structure of ubonodin using 2D NMR experiments. A NOESY experiment was initially carried out with a long mixing time (500 ms) in order to assign all peaks along with COSY and TOCSY spectra. NOESY spectra were also acquired at shorter mixing times of 100 ms and 40 ms, with the 100 ms spectrum used for calculation of distance restraints (Figure S3, Table S2, Table S3). Structure calculations revealed an unprecedented topology for a lasso peptide with an 8 aa isopeptide-bonded ring, an 18 aa loop, and a short 2 aa tail (Figure 2, Figure S4). Previously, the largest loop region observed in a lasso peptide with 10 aa was in microcin J25 (MccJ25). Other large proteobacterial lasso peptides such as astexin-3 (24 aa) and sphingopyxin I (26 aa), are characterized by relatively short loop regions (5 aa for astexin-3 and 6 aa for sphingopyxin I) and longer C-terminal tails (Figure S5). Lasso peptides are often maintained in their [1]rotaxane structures by bulky steric lock residues that straddle the ring. In ubonodin, those residues are Tyr-26 and Tyr-27. This arrangement of steric lock residues is reminiscent of MccJ25 which uses Phe-19 and Tyr-20 as steric locks (Figure S6). The large 18 aa loop of ubonodin is its most compelling structural feature. The ubonodin NOESY spectrum includes strong amide-amide crosspeaks indicative of turns. The most prominent turn in the loop runs from His-15 to Trp-19, a mostly polar stretch of the peptide with sequence HIHDW. Strong crosspeaks between sidechain resonances for Ile-16 and Trp-19 support the presence of this turn. There is also a shorter turn present that runs from Met-22 to Ser-24.

Figure 2.

Comparison of solution structures of selected lasso peptides. The isopeptide-bonded rings of the peptides are colored yellow, the loops are colored magenta, and the C-terminal tails are colored green. Backbone nitrogen atoms are colored blue throughout. Sidechains of steric lock residues are shown. Ubonodin has a much larger loop than at 18 aa than any of the other lasso peptides. The long loop of ubonodin also include short turns not present in other lasso peptide loops. PDB/BMRB codes used to generate these figures are as follows: ubonodin: PDB 6POR (this work), microcin J25: PDB 1Q71, citrocin: PDB 6MW6, capistruin: BMRB 20014.

Though the loop structure of ubonodin is much larger and more complex than previously described lasso peptides, its ring and tail regions share some similarity to those found in microcin J25 (MccJ25) and citrocin (Figure 2, Figure S6). Both of these peptides are RNAP inhibitors,^{[12, 14]} and a recent crystal structure of MccJ25 bound to RNAP shows that the ring and tail regions of MccJ25 make multiple contacts with the secondary channel of RNAP.^[15] Thus we hypothesized that ubonodin would also function as an RNAP inhibitor. We carried out abortive transcription initiation assays with E. coli RNAP (Figure 3A). These assays confirmed that ubonodin inhibits transcription initiation, an activity and putative antimicrobial mode of action observed in several other lasso peptides.^{[10, 12–13, 16]} The potency of ubonodin in these assays was somewhat lower than that of MccJ25 (Figure S7), though this may be due to the fact that E. coli is not a natural antimicrobial target of ubonodin.

Figure 3.

Antimicrobial activity of ubonodin. A: Autoradiograph of abortive transcription initiation assays showing that ubonodin inhibits E. coli RNA polymerase. The heading in each gel lane is the concentration of ubonodin added to the assay in μM. CpApU* is the abortive transcript product. B: Spot-on-lawn assay showing the antimicrobial activity of ubonodin against Burkholderia multivorans. Concentration of ubonodin in each spot is given on the figure.

Encouraged by the RNAP-inhibiting activity of ubonodin, we tested ubonodin for antimicrobial activity against a panel of proteobacteria (Table 1, Table S4). Antimicrobial lasso peptides tend to have a narrow spectrum of activity, killing bacteria that are closely phylogenetically related. Ubonodin was unable to kill E. coli and Salmonella newport, strains that are susceptible to MccJ25 and citrocin. Given that ubonodin is encoded in the genome of a Burkholderia strain, we next tested ubonodin against other Burkholderia. We observed modest activity of ubonodin against the producing strain of the lasso peptide capistruin, B. thailandensis, and no activity against the plant pathogen B. gladiolii. The putative ubonodin producing strain, B. ubonenesis, belongs to the Burkholderia cepacia complex (Bcc), and we observed potent activity against two Bcc strains, B. multivorans and B. cepacia. These notorious species are frequently found in lung infections in CF patients.^[17] In spot-on-lawn assays, these organisms were inhibited by low micromolar concentrations of ubonodin (Figure 3B). The potency of ubonodin was affected by the media composition. For B. multivorans, the last active dilution in spot assays carried out in minimal M63 medium was 8 μM, whereas this increased to 20 μM on plates comprised of rich Mueller- Hinton medium. The spot-on-lawn assays were followed up by liquid growth assays in which the minimal inhibitory concentration of ubonodin was 4 μM against B. cepacia, and 31 μM against B. multivorans. The antimicrobial activity of ubonodin was also tested against the select agents B. pseudomallei and B. mallei. We also tested ubonodin against two attenuated (BSL-2) strains of B. pseudomallei, Bp82 and Bp576mn.^[18] While no activity was observed against any B. pseudomallei strains, we did observe growth inhibition of two B. mallei strains by ubonodin in spot assays. From this limited set of strains tested, we can conclude that ubonodin has the most potent activity against Bcc strains with some activity against strains in the pseudomallei/mallei group (Figure S8). Further testing must be carried out to understand the full spectrum of ubonodin’s activity; we expect ubonodin to have activity against other Bcc strains.

Table 1.

Minimum inhibitory concentration (MIC) of ubonodin against Burkholderia strains in Mueller-Hinton medium.

Strain	MIC via spot-on-lawn assay (μM)	MIC via liquid growth assay (μM)
B cepacia ATCC 25416	40	4
B. multivorans ATCC 17616	20	31
B. mallei Old ISU	40	>31
B. mallei NVSL 86-567-2	40	>31

We also studied the structure and activity of ubonodin upon heating to 50 °C and 95 °C. Upon heating, lasso peptides can unthread^[19] and partial backbone cleavage C-terminal to Asp residues has also been observed in lasso peptides.^[20] We heated a sample of ubonodin to either 50 °C or 95 °C for up to 6 h and observed that ubonodin retained activity against B. multivorans after heating to 50 °C, but not to 95 °C (Figure S9). Since a loss in activity can be due either to lasso peptide unthreading^[13a] or cleavage after Asp residues, we next examined a sample of ubonodin heated to 95 °C for 2h by LC-MS². The major species in this sample is still intact ubonodin (Figure S10). We also observed several new peaks and could assign many of them to cleavages of ubonodin C-terminal to each of the three Asp residues, two within the loop and one within the ring of ubonodin (Figure 4, Figure S11). Peptides cleaved after the two loop Asp residues (Asp-18, Asp-23, or both) remain threaded, generating a series of [2]rotaxane structures. Further cleavage of heat-treated ubonodin with carboxypeptidase, which hydrolyzes amino acids with a free C-terminus, confirmed the assignment of the [2]rotaxane peptides (Figure S11). We also observed species consistent with cleavage after Asp-3 in the ring of ubonodin plus an additional cleavage after either Asp-18 or Asp-23 residue in the loop (Figure 4, Figure S11). There is an additional peak with mass identical to intact ubonodin, but with a different retention time. We suspect that this peak corresponds to an unthreaded species of ubonodin as it is completely eliminated upon carboxypeptidase digestion (Figure S10, S12). The thermal degradation of ubonodin is unlike that of any other lasso peptide. Whereas most lasso peptides exhibit either thermostability or unthreading, ubonodin, due to the presence of multiple Asp residues, “self-destructs” into a variety of different peptide fragments.

Figure 4.

Schematic showing cleavage degradation products of heat-treated ubonodin. Intact ubonodin (center, blue box), can be cleaved at all the Asp residues (2 in loop, 1 in ring), generating a series of [2]rotaxane and branched peptides. Orange, green, and purple arrows/dotted lines correspond to cleavage after Asp-3, Asp-18, and Asp-23 respectively. Mass spectrometry evidence was seen for all species except the one boxed in red (Figure S10, Figure S11).

Next, we carried out mutagenesis on ubonodin to identify residues important for production and activity (Figure 5). We had hoped to utilize ubonodin as a starting material for peptide catenanes analogous to the MccJ25 catenanes we have described previously.^[21] Therefore we introduced Cys residues at the Pro-13, Met-14, and Trp-19 positions of ubonodin, as well as at the C-terminus of ubonodin, Gly-28. While all four of these variants were detected by LC-MS, only the G28C variant was produced at a quantity sufficient for purification (Figure S13). We also checked cell lysates for these variants, but found only minimal quantities of the peptides, detectable only by LC-MS. The G28C variant retained some antimicrobial activity against B. multivorans, but its activity was diminished relative to the wild-type peptide (Figure S14). We also carried out mutagenesis on the two steric lock residues, Tyr-26 and Tyr-27. While the Y26F variant of ubonodin was produced at roughly half of the wild-type level and retained antimicrobial activity, the Y27F variant was produced at levels only detectable by LC-MS. Substitution of either His residue (His-15 or His-17) with Ala surprisingly led to variants that expressed at near wild-type level and retained near wild-type antimicrobial activity against B. multivorans. This result suggests that these solvent-exposed His residues are not critical for antimicrobial activity. Finally, we generated a series of variants of ubonodin with conservative substitutions: I6L, D18N, I21L, D23N, and S24A. While all of these variants were detected by LC-MS in crude culture supernatant extracts, only the I6L and I21L variants were detected as a unique peak on HPLC (Figure S13). However, the peaks for the I6L and I21L variants of ubonodin were quite broad, suggesting that they may not exist as single defined structures. Our NMR structure suggests that the I6 sidechain packs against the Y27 sidechain, thus switching I6 to Leu may disrupt the fold of ubonodin. While other lasso peptides are tolerant to amino acid substitutions,^[22] ubonodin appears to be fairly recalcitrant to mutagenesis.

Figure 5.

Mutagenesis of ubonodin. A: Left: tolerance of ubonodin to amino acid substitutions. While all 13 single amino acid variants could be detected by LC-MS, only 6 were produced at a level sufficient for purification. The production level is coded as follows: green is at or near wild-type levels, yellow is less than 20% of wild-type, and red is only detectable by LC-MS. B: Antimicrobial activity of purified ubonodin variants. The H15A, H17A, and Y26F variants have near wild-type activity (green) while the G28C variant is less potent than wild-type. See also Figures S13–S14 for traces and spot assays on these variants.

Conclusion

Using genome mining and heterologous expression, we have produced a new antimicrobial lasso peptide, ubonodin. Ubonodin exhibits potent antimicrobial activity against several strains of Burkholderia, including B. cepacia and B. multivorans, two Burkholderia pathogens that commonly cause infections in CF patients.^[17b] We show that ubonodin is able to inhibit E. coli RNAP, suggesting that RNAP is the antimicrobial target of ubonodin. While ubonodin has activity against B. cepacia, B. mulitvorans, and B. mallei, it is poorly active against B. thailandensis and has no activity against B. gladiolii and B. pseudomallei. This narrow spectrum of activity may allow for therapeutic usage of ubonodin since it will only kill the target pathogens while leaving the healthy microbiome unscathed.

There is now an increasing body of work on antimicrobial lasso peptides,^[23] and the largest class of antimicrobial lasso peptides function as RNAP inhibitors. Since RNAP is a cytoplasmic target, the antimicrobial activity of these compounds is a product of both their uptake into susceptible cells and their engagement with RNAP. For example, MccJ25 and capistruin have similar K_i against RNAP^{[10, 16a]}, but MccJ25 is a much more potent antibiotic than capistruin. Similarly, the lasso peptides acinetodin, klebsidin, and citrocin all inhibit RNAP in vitro but exhibit only modest or no antimicrobial activity.^{[12, 16b]} In the case of citrocin, we observed only modest antimicrobial activity despite highly potent inhibition of RNAP in vitro, underscoring the importance of lasso peptide transport into susceptible cells. We hypothesize that the activity and the breadth of spectrum of ubonodin is dictated by its uptake into susceptible bacteria, a topic for future study. While MccJ25 crosses the outer membrane of susceptible cells via the ferrichrome transporter FhuA,^[24] there are no close homologs (>40 % identity) of FhuA in Burkholderia.

Though the sequence and structure of ubonodin differs greatly from those of MccJ25 and citrocin, each of these peptides includes Tyr residues at position 9 and at the penultimate position of the sequence. The C-terminal Gly residue is also conserved in each of these peptides. We suggest that this Tyr/Tyr/Gly motif is an excellent predictor of RNAP-inhibiting lasso peptides. This idea is supported by the recently-published crystal structure of MccJ25 bound to RNAP which shows multiple interactions between these residues (Tyr-9, Tyr-20, and Gly-21 in MccJ25) and the β and β’ subunits of RNAP.^[15] Acinetodin and klebsidin, also RNAP inhibitors, share the Tyr-Gly motif at their C-termini, though both of these peptides lack a Tyr residue at position 9.^[16b] The structure of ubonodin differs from any other characterized lasso peptide with an 18 aa-long loop region. While we observed turns in this loop region from our NMR structure calculations, structures of MccJ25 bound to RNAP and the outer membrane receptor FhuA^{[15, 25]} show significant remodeling of the MccJ25 loop region when bound to these proteins (Figure S6). We expect similar or even more drastic changes to the ubonodin loop when bound to its target(s) and transporters.

Experimental Section

All experimental methods are provided in more detail in the supporting information.

Plasmid construction-

The full ubonodin gene cluster was codon optimized for E. coli expression. The uboA gene was assembled from oligonucleotides and the uboBCD genes were assembled using synthetic dsDNA (gBlocks from Integrated DNA Technologies).

Peptide expression and purification-

Peptides were expressed using E. coli BL21 in M9 media supplemented with 20 amino acid solution (40 mg/L each). Cultures were induced at OD₆₀₀ absorbance of 0.2, and expressed at room temperature overnight. C8 solid phase extraction was then used to recover hydrophobic compounds, including ubonodin and its variants. Peptides were further purified from the extract using reversed-phase HPLC. Expected peptide masses were confirmed with LC-MS.

NMR-

¹H-¹H gCOSY, TOCSY, and NOESY experiments were conducted at 22 °C with an 800 MHz spectrometer. MestreNova was used for spectral processing and analysis. Cross peak volumes from the NOESY spectrum were used as distance constraints for structural calculations using CYANA 2.1. The top 20 structures were then energy-minimized using GROMACS.

RNA polymerase (RNAP) inhibition assay-

A previously described in vitro abortive initiation assay was used to test RNAP inhibition.^[12] RNA synthesis in the presence of varying concentrations of ubonodin was measured via incorporation of [α-³²P]UTP.

Antimicrobial activity assay-

Antimicrobial activity was tested using spot-on-lawn assays on agar plates and in liquid media. For spot-on-lawn assays, melted soft agar was inoculated with the target strain and then overlaid over an agar plate. After solidification, peptide dilutions were spotted onto the plate. Plates were then incubated overnight and evaluated for zones of inhibition. MIC for a spot-on-lawn assay was defined as the last concentration at which a zone of inhibition was observed. Liquid assay tests were done in 96 well plates. MIC for a liquid assay was defined as the lowest concentration at which there was no growth after 24 hours.

Thermostability assay-

Ubonodin was heated at 50 or 95 °C for 0, 2, 4 and 6 hours in a thermocycler. Heated peptide was analysed with LC-MS and activity was measured using the spot-on-lawn assay described above against Burkholderia multivorans ATCC 17616.

Data Deposition

The coordinates for the ubonodin structure have been deposited to the Protein Data Bank (PDB code 6POR).