Pandonodin: a proteobacterial lasso peptide with an exceptionally long C-terminal tail

Abstract

Lasso peptides are a family of ribosomally synthesized and post-translationally modified peptide (RiPP) defined by their threaded-ring topology. The N-terminus of the peptide forms an isopeptide bond with an aspartate or glutamate sidechain to create a 7- to 9-aa macrocyclic ring, through which the rest of the peptide is threaded. The result is a highly constrained 3D structure. Even though they share a threaded-ring feature, characterized lasso peptides vary greatly in sequence and size, ranging from 14 to 26 aa. Using genome mining, we identified a new lasso peptide gene cluster with a predicted lasso peptide that is 33 aa long. Here we report the heterologous expression of this new peptide, pandonodin, its NMR structure, and its unusual biophysical properties. Pandonodin has a long, proteolytically resistant 18-residue tail of low sequence complexity, which limits its water solubility. Within this tail is a 6 aa disulfide-bonded macrocycle that serves as a steric lock to maintain the lasso structure. This disulfide bond is unusually stable, requiring both heat and high concentrations of reductants for cleavage. Finally, we also show that segments of the C-terminal tail of pandonodin can be replaced with arbitrary sequences, allowing for the construction of pandonodin-protein fusions.

Lasso peptides are a family of RiPP (ribosomally synthesized and post-translationally modified peptide)¹ natural products, the structure of which resembles a threaded lariat or lasso.^2–3 The lasso structure, formally a [1]rotaxane, is generated via a backbone-sidechain isopeptide bond.^4–6 The isopeptide bond generates an N-terminal macrocyclic “ring” through which the C-terminal thread passes. While all lasso peptides contain an isopeptide bond, these structures can be further diversified by post-translational modifications,⁷ including disulfide bonds,⁸ phosphorylation,⁹ and acetylation.¹⁰ Another hallmark of the lasso peptide family is its diversity in sequence; both highly polar^11–13 and highly hydrophobic¹⁴ members of the family have been studied. Another source of diversity among the lasso peptides is their size (FIGURE 1). While the prototypical lasso peptide, microcin J25,¹⁵ is 21 aa long, other examples have been found that are as short as 14 aa.¹⁶ At the other end of the spectrum are a series of large, polar lasso peptides that are encoded in the genomes of α-proteobacteria.^11–13 The current record holder for the largest lasso peptide is sphingopyxin I at 26 aa.^{13, 17} Leveraging our genome mining software, we have had a long-standing interest in finding the limits on size for lasso peptides. Here we report the discovery of the largest lasso peptide reported to date, pandonodin, at 33 aa.

Figure 1.

Diversity in lasso peptide size and topology. Crystal structure of sphingopyxin I (left), the largest lasso peptide characterized to date at 26 aa long. NMR structure of canucin (center), the smallest lasso peptide characterized to date at 14 aa long. NMR structure of 21-aa long microcin J25 (right), shows the variability in loop versus tail length. The sidechains of the steric lock residues are shown in each structure. The crystal structure of sphingopyxin I (gray backbone) was drawn from the coordinates file PDB: 5JQF. Sphingopyxin I, the main product from heterologous expression, was a 5 aa truncation of the predicted full-length lasso peptide. The last 5 aa, ITADD, (yellow backbone), were modeled in using PyMOL. Canucin was drawn from a coordinates file provided by reference 16. Microcin J25 was drawn from PDB file 1Q71.

The biosynthetic steps for lasso peptide biosynthesis are well understood^{14, 18–21} and follow the general biosynthetic logic of other RiPPs.²² Precursors to lasso peptides, referred to as A proteins, are ribosomally synthesized and include an N-terminal leader peptide and C-terminal core peptide.²³ The precursor must first be cleaved by a dedicated cysteine protease (B protein) at a scissile bond between the leader and core peptides. The cleaved core peptide is then matured into the lasso peptide via the action of the C protein, homologous to asparagine synthetases and β-lactam synthetases. The C protein adenylates a Glu or Asp sidechain and also presumably preorganizes the core peptide into a near lasso structure. The lasso peptide is formed upon attack of the N-terminus on the activated Glu or Asp. The genes for the A, B, C proteins are found in clusters within bacterial genomes, enabling their discovery by genome mining methods.^{7, 11, 24} Beyond these core A, B, and C proteins required for lasso peptide biosynthesis, additional genes are found within these clusters. The most common lasso peptide accessory protein found in gene clusters is an ABC transporter (D protein) which functions to pump the mature lasso peptide out of the cell.^{18, 25–26}

Results

Unusual lasso peptide precursors in Pandoraea genomes

We originally found a hit for a lasso peptide gene cluster in the organism Pandoraea norimbergensis27 using our precursor-directed genome mining strategy.¹¹ The Pandoraea genus²⁸ belongs to β-proteobacteria, and there are currently only 10 species with accepted names. As such, members of the Pandoraea genus have not been studied extensively as natural product producers. No lasso peptides from Pandoraea have been reported, though two examples have been discovered in the genomes of phylogenetically related Burkholderia.^{13, 29} The P. norimbergensis lasso peptide gene cluster includes a precursor protein PanA of 60 aa, 27 aa of which comprise the leader peptide along with a predicted core peptide of 33 aa (FIGURE 2A). In the genome of P. norimbergensis, the last 8 bases of the panA gene overlap with the 5’ end of the annotated panB gene. The panB gene also includes a GTG codon at position 4 of the protein which may be the bona fide start codon. Lasso peptide gene clusters tend to organize the B, C, and D genes into an operon with only small gaps between the ORFs for these genes. Therefore, it was surprising to note that there is a large gap (73 bp) between the end of the B ORF and the beginning of the C ORF. Overall the P. norimbergensis gene cluster has GC content of 65%, comparable to the 63% GC content for the genome. No other genes associated with lasso peptide gene clusters (isopeptidases,¹² kinases,⁹ disulfide oxidoreductases,⁸ methyltransferases,³⁰ or acetyltransferases¹⁰) are near this gene cluster. Indeed, the nearest ORFs to the gene cluster are 200–300 bp away from the ends of the cluster. While we originally identified the lasso peptide gene cluster in the P. norimbergensis genome, further BLAST searching revealed that similar clusters are found in the genomes of 6 other members of the Pandoraea genus. For four of these examples, the gene cluster is intact, though some point mutations occur in the A gene. The other two clusters are disrupted, and likely are non-functional (FIGURE S1).

Figure 2.

Pandonodin gene cluster. (A) Native pandonodin gene cluster architecture and the predicted lasso peptide product. Note the unusual overlap of the predicted panA and panB genes. The PanA core peptide is also shown with putative locations of the isopeptide bond and disulfide bond. (B) Refactored pandonodin gene clusters. In one version, the entire gene cluster was placed under the inducible tet promoter. In the other version, panA was placed under an inducible T5 promoter while panBCD were placed under the control of a constitutive promoter.

Heterologous expression of the P. norimbergensis lasso peptide gene cluster

P. norimbergensis was cultured and genomic DNA was isolated, allowing for cloning of the putative lasso peptide gene cluster. Since lasso peptides are often produced only at low quantities or not at all in their native producers under standard culture conditions, we did not attempt to isolate the peptide directly from P. norimbergensis. We took two approaches toward heterologous expression of the gene cluster. In the first, the entire panABCD gene cluster was placed under the control of an inducible tet promoter. We also carried out a “light refactoring” of the gene cluster in which the putative operon comprising the panBCD genes was left intact and placed under the control of a constitutive promoter (FIGURE 2B).³¹ In this refactored construct, the panA gene was placed under the control of an IPTG-inducible T5 promoter. Both of these strategies have been used previously to heterologously express lasso peptides.^{11, 31–32}

These two different plasmids were introduced into E. coli BL21, and lasso peptide production was induced at 20 °C overnight. Analysis of the culture supernatants by HPLC revealed a prominent, late-eluting peak in both cultures (FIGURE 3). This peak was collected and injected onto LC-MS, revealing a species with a monoisotopic mass of 3306.78. This mass is ~20 amu less than that of PanA core peptide, which we rationalized as the formation of one isopeptide bond and one disulfide bond (predicted monoisotopic mass of 3306.75). We named this putative lasso peptide pandonodin, based on the genus in which it is encoded, Pandoraea, and the Latin root for knot, nodum. In MS² analysis of pandonodin, we observed fragmentation from the C-terminus back to Cys-21, consistent with the presence of a disulfide bond between Cys-16 and Cys-21. Additional fragmentation was observed N-terminal to Cys-16, back to the putative isopeptide bonded ring at Glu-8 (FIGURE S2A). This analysis suggests a topology that includes the isopeptide bonded ring between Gly-1 and Glu-8 as well as an additional linkage via the disulfide bond. Three topologies for this set of linkages are possible (FIGURE S2B). In the first, the disulfide-bonded macrocycle is found fully within the C-terminal tail of the lasso peptide. In the second, the disulfide bond is fully present within the loop. Finally, intermediate to these possibilities is a catenated structure in which the disulfide bonded macrocycle is formed around the isopeptide bonded ring. The MS² data are consistent with the first two topologies, but not the catenane structure, which would be expected to be more recalcitrant to fragmentation.

Figure 3.

Pandonodin identification and purification. (A) HPLC chromatogram of pandonodin supernatant extract. Wild-type pandonodin was expressed from both expression vectors as mainly full-length pandonodin, with a minor C-terminal truncation product (ΔC4). (B) Mass spectrum of peak collected from HPLC corresponding to the expected mass of full length pandonodin lasso peptide with a disulfide bond. The masses for the doubly, triply, and quadruply-charged ions are noted. The inset shows the isotopic distribution of the triply-charged species.

We scaled up the production of pandonodin to 4 L and purified it to homogeneity using HPLC in order to determine its yield and generate material for NMR. Unfortunately, pandonodin has essentially no extinction coefficient at 280 nm, precluding the use of absorbance for concentration determination. Instead, we measured the mass of the purified, lyophilized peptide on a balance. Approximately 8 mg of pandonodin was isolated from 4 L, giving a yield of ~2 mg/L culture. Upon attempting to resuspend the lyophilized pandonodin in water, a cloudy suspension was obtained (FIGURE S3), suggesting that pandonodin was only sparingly soluble in water. Analysis of a dilution series of this sample by dynamic light scattering (DLS) showed that pandonodin remains insoluble at concentrations as low as 0.125 mg/mL (FIGURE S4). This suspension of pandonodin was relyophilized and resuspended in methanol, which readily dissolved the peptide.

NMR analysis of pandonodin reveals a new topology for lasso peptides

Given the poor solubility of pandonodin in water, we prepared an NMR sample (6 mg/mL, 1.8 mM) in CD₃OH. A set of 2D spectra (gCOSY, TOCSY, NOESY) were initially acquired, allowing for the assignment of many resonances in the peptide (FIGURE S5, TABLES S1, S2). The pandonodin sequence is relatively low complexity with regards to amino acid usage, especially in the putative loop and tail segments. From residues 9–33, only 10 different aas are present, and there is also a palindromic sequence, TLPPLT, which complicates the NMR assignment. We later acquired a ¹H/¹³C HSQC spectrum to clear up any ambiguities in the assignments. Structural models were built using CYANA specifying one isopeptide bond between Gly-1 and Glu-8 and one disulfide bond between Cys-16 and Cys-21. These calculations revealed a right-handed lasso structure, consistent with all other published lasso structures (FIGURE 4). Following the 8 aa isopeptide bonded ring, the loop of the lasso peptide is comprised of residues 9–15, for a 7 aa loop segment. The remainder of the peptide, 18 residues, is found in the C-terminal tail portion. Only minimal NOE cross peaks between non-adjacent amino acids were observed in the tail region, suggesting that the tail is unstructured. In contrast, extensive NOE cross peaks between the Leu-15, Cys-16, Phe-17, and Lys-18 sidechains and the ring were observed (TABLE S3), suggesting that this segment of the peptide is threaded through the ring.

Figure 4.

Pandonodin NMR structure in methanol-d3. (A) Top structure of pandonodin. The backbone of all residues and the sidechains of Glu-8, Leu-15, Cys-16, Phe-17, and Cys-21 are shown as sticks. The disulfide bond between Cys-16 and Cys-21 is in yellow. Leu-15, Phe-17, and the disulfide-bonded macrocycle serve as steric locks, maintaining the lasso structure (B) Overlay of the top 18 structures showing the variability in the structure of the C-terminal tail.

Such a long C-terminal tail in lasso peptides is unprecedented. As mentioned above, sphingopyxin I, at 26 aa, is the current record holder for the largest lasso peptide.¹³ Its tail segment is 11 aa. Other examples of lasso peptides encoded in α-proteobacterial genomes also have relatively long tails, like astexin-3, with its 10 aa tail. Besides its size, the 18 aa pandonodin tail is noteworthy for its stability. It is known that the C-terminal tails of lasso peptides are susceptible to proteolysis both in their native producers³³ and in heterologous expression systems. For example, though sphingopyxin I presumably has an 11 aa tail after maturation, the major product observed in heterologous expression is a variant in which the 5 C-terminal aas are truncated (FIGURE 1). Similar C-terminal truncations have been observed in lasso peptides such as astexin-2¹² and benenodin-1.³⁴ Full-length pandonodin is the major product upon its heterologous expression, with only small amounts of a ΔC4 truncation product (ie missing the C-terminal 4 aa) observed (FIGURE 3).

Stability of pandonodin

In order to probe the stability of pandonodin, we treated a solution of it with heat (95 °C for 2 h) or with a mixture of carboxypeptidases B and Y. Upon heat treatment, greater than 95% of the peptide remained as intact pandonodin as judged by the retention time on the LC-MS (FIGURE S6). Two minor products were observed; one with the same mass as pandonodin but a drastically different retention time, and one with an increase of 18 amu relative to pandonodin. We hypothesized that the +18 amu species corresponds to hydrolysis after Asp in the ring of pandonodin, which has been observed in other lasso peptides.¹² It is possible that the minor product with the same mass as pandonodin is an unthreaded variant of the peptide.³⁵ Carboxypeptidase treatment of lasso peptides can report on the topology of the lasso peptide (i.e. whether it is threaded or not).³⁶ Carboxypeptidase treatment of unheated pandonodin resulted in only small amounts of digestion products corresponding to the removal of 2 or 6 aa from the C-terminus of the peptide. Carboxypeptidase treatment of heated pandonodin led to the degradation of the putative unthreaded species as well as C-terminal truncations of 2 or 6 aa from threaded pandonodin. Minor products suggesting promiscuous activity of carboxypeptidase cleaving within the loop of pandonodin were also observed. In summary, our data show that pandonodin is remarkably stable upon thermal challenge. Its resistance to carboxypeptidase cleavage beyond the C-terminal 6 aa of the tail is likely due to the presence of multiple proline residues in the tail.

We were intrigued by the lack of disulfide bond scission upon heating of pandonodin, which is reminiscent of the thermal stability of other entangled disulfide-bonded peptides such as kalata B1.³⁷ Therefore, we tested the susceptibility of pandonodin to reduction with either 20 mM DTT or 5% (~641 mM) β-mercaptoethanol (β-ME). At room temperature, only a tiny fraction of the peptide was reduced in the 5% β-ME condition; no reduction products were observed under DTT reduction (FIGURE S7). Next we carried out similar experiments but heated the samples to 95 °C for two hours. Under these conditions, partial reduction of pandonodin was observed with two distinct peaks exhibiting a mass corresponding reduced pandonodin. Under the DTT reduction conditions, less than 10% of the peptide was reduced. However, under the more forcing β-ME conditions, a majority of pandonodin was reduced. In both conditions, a 3^rd peak was observed corresponding to reduced pandonodin +18 amu. Under the β-ME conditions, we also observed a small peak that may correspond to β-ME adducts (+78 amu). We also subjected these heated, reduced samples to carboxypeptidase digestion. Similar to oxidized pandonodin, we observed elimination of the putatively unthreaded species upon carboxypeptidase digestion. However, threaded, reduced pandonodin was not cleaved. As noted above, carboxypeptidase has promiscuous activity against pandonodin, which was observed again these experiments. The reduced, unthreaded species of pandonodin appears to be cleaved after Ile and Leu residues at three positions.

Given the recalcitrance of pandonodin to reducing agents, we generated a variant of pandonodin in which both Cys residues (Cys-16 and Cys-21) were changed to Ala, thus abolishing the disulfide linked macrocycle. This variant, pandonodin C16A C21A (pandonodin AA for brevity’s sake) was produced robustly, both as the full-length peptide and its ΔC4 variant, consistent with what was observed for wild-type pandonodin. Compared to wild-type pandonodin, however, more of the peptide was truncated to the ΔC4 variant (FIGURE S8A). Full-length pandonodin AA elutes at 21.4 min on our HPLC gradient as compared to 25.5 min for full-length wild-type pandonodin (FIGURE 3A). Full-length pandonodin AA is produced at ~1.2 mg/L, ~60% of the expression level of full-length pandonodin. Upon analysis of the peak corresponding primarily to full-length pandonodin AA by LC-MS, we noticed two minor products corresponding to the ΔC12 and ΔC13 variants of pandonodin AA (FIGURE S8B). This observation, coupled with the increased amount of ΔC4 observed for pandonodin AA suggest that its tail region is more susceptible to proteolysis than that of wild-type pandonodin. Suspecting that pandonodin AA would be more susceptible to thermal unthreading, we heated a sample of pandonodin AA at 95 °C for 2 h. Pandonodin AA still did not unthread fully over this period of time, but significantly more pandonodin AA unthreaded than did wild-type pandonodin (FIGURE S6). The major product following heating of pandonodin AA, however, has a mass 18 amu larger than pandonodin AA (FIGURE S8B). We also observed a +18 amu species upon pandonodin heating as well (FIGURE S6), but this species is much prominent for pandonodin AA. As mentioned above, we expect this +18 amu species to correspond to cleavage C-terminal to the Asp residue in the ring. We provide experimental evidence for this below in the context of pandonodin truncation variants.

Truncations of the pandonodin tail

Given that the long C-terminal tail of pandonodin is its most unusual feature, we next asked whether shortening the tail would have any effects on the maturation of pandonodin. We have previously studied the truncation of the C-terminal tail of the lasso peptides astexin-2 and astexin-3.³⁸ Removal of up to 6 aa from the C-terminus of astexins-2 and −3 was possible. However, removal of more than 3 aa from the C-terminus led to a precipitous drop in the yield of peptide produced. Since pandonodin has a much longer tail than astexin-3, we were interested in how much of the tail could be removed while still retaining the ability of the peptide to be matured correctly. Another consideration with pandonodin is that the lasso peptide is exported from the cell through an ABC transporter. We were interested in whether tail truncations could still be correctly exported.

We generated a series of truncation variants of pandonodin in which 4, 8, 12, 14, 16, or 17 aa were removed from the C-terminal tail (FIGURE 5). We refer to these variants as pandonodin ΔCX where X is the number of aa removed from the C-terminus. All of these truncation variants were produced and exported in our heterologous expression system at levels detectable by HPLC except for ΔC16 and ΔC17. The ΔC16 variant was observable by LC-MS, but the ΔC17 variant was not detected at all. These results are supported by our NMR structure since the ΔC16 variant truncates the tail back to Phe-17, the steric lock residue. The ΔC4 and ΔC8 variants are produced well, and only the oxidized forms of these peptides were observed. The ΔC12 variant, the last truncation capable of forming a disulfide bond, expresses reasonably well, but both oxidized and reduced forms of the peptide were observed. Interestingly, the ΔC14 variant, which eliminates the disulfide-bonded macrocycle, is produced robustly at nearly half of the production level of the full-length wild-type peptide. The fact that so much of the tail of pandonodin can be removed without disrupting the maturation of the lasso is in stark contrast to what we observed previously for astexins-2 and −3.³⁸

Figure 5.

Pandonodin truncation variants. (A) Sequence of wild-type pandonodin and the truncation variants. (B) HPLC chromatograms of the variants. Arrows indicate where each peptide eluted, which was confirmed by mass spectrometry. For pandonodin ΔC12, there were two peaks that corresponded to the ΔC12 peptide, with the earlier retention time corresponding to free cysteine residues and the later retention time corresponding to the disulfide-bonded peptide. Pandonodin ΔC16 was not produced at a level detectable by HPLC.

Given the robust expression of the ΔC14 variant and our suspicion that the C-terminal tail of pandonodin limits its water solubility, we tested the ΔC14 variant for water solubility and characterized it further. Gratifyingly, pandonodin ΔC14 was soluble in water to a concentration of ~2 mg/mL (1.0 mM). We generated an NMR sample of the ΔC14 variant and acquired 2D spectra that share some features with the full-length pandonodin spectra in methanol (FIGURE S9). Unfortunately, the quality of these spectra were too poor to allow for full assignment and structure calculations. Besides NMR, we subjected pandonodin ΔC14 to carboxypeptidase treatment, heating to 95 °C, and further carboxypeptidase treatment following heating (Figure 6). Carboxypeptidase did not cleave pandonodin ΔC14, strongly suggesting that it is maintained in the lasso structure. Upon heating pandonodin ΔC14, we observed two peaks in the LC-MS with identical m/z to pandonodin ΔC14, but with different retention times. Both of these peaks disappeared upon carboxypeptidase treatment (Figure 6A), suggesting that they are unthreaded species. We compared the MS² spectra of the threaded and unthreaded species (FIGURE S10). The parent ion is among the most abundant peak in the MS² spectrum of threaded pandonodin ΔC14, suggesting that it is difficult to fragment, an observation noted previously for lasso peptides.⁴ In contrast, the unthreaded species fragmented readily back to the isopeptide bonded ring. We also observed a species with an additional hydrolysis relative to pandonodin ΔC14 (FIGURE 6). A similar species was observed for full-length pandonodin and pandonodin AA (FIGURES S6 and S8). The abundance of this species in the pandonodin ΔC14 heating experiments allowed us to analyze it by MS/MS, confirming that this species is a branched peptide, cleaved in the ring C-terminal to aspartate (FIGURE S11). Finally, upon heating pandonodin ΔC14, we observed minor products corresponding to disulfide bonding of two molecules of pandonodin ΔC14. Given that these species are also eliminated upon carboxypeptidase treatment, it is likely that these dimeric species are comprised of unthreaded peptides. In summary, from these experiments we can conclude that pandonodin ΔC14 is threaded, but it is more susceptible to thermally-induced unthreading and ring cleavage than full-length pandonodin.

Figure 6:

Stability of pandonodin ΔC14. A: Treatment of pandonodin ΔC14 with carboxypeptidase (orange trace) does not result in cleavage of the peptide, demonstrating that it is threaded. Heating of pandonodin ΔC14 for 2 h at 95 °C (gray trace) results in peptide unthreading (peaks 2 and 3), cleavage of the isopeptide bonded ring after Asp (peak 1) and formation of disulfide bonded dimers of these species (peaks 4 and 6). Most of these peaks disappear upon carboxypeptidase treatment (yellow trace), confirming that they correspond to unthreaded species. B: Table of masses for peaks 1–6. See also MS² data on these peaks in Figures S10 and S11.

C-terminal protein fusions to pandonodin

We have previously demonstrated the fusion of globular proteins to the C-terminal of another lasso peptide, astexin-1, which has a 9 aa tail.³⁹ The direct fusion of proteins to the C-terminus of astexin-1 did not result in the formation of any lasso peptide at the N-terminus. Instead, a flexible linker comprised of Gly and Ser needed to be introduced between the C-terminus of astexin-1 and the protein fusion partner. We reasoned that the longer tail length of pandonodin would allow for the fusion of protein partners to its C-terminus. To this end, we designed genes encoding for a pandonodin-GFP fusion and a pandonodin-A1 fusion (FIGURE 7A). The A1 protein is a highly-expressed artificial leucine zipper protein capable of self-dimerization.⁴⁰ In both the pandonodin-GFP and pandonodin-A1 fusions, a thrombin cleavage site (LVPRGS) was introduced between the C-terminal alanine of pandonodin and the beginning of the fusion partner. These genes were introduced to the refactored pandonodin gene cluster which includes the panBCD operon to mature the peptide (FIGURE 7A). Although the ABC transporter gene panD was also present in the cluster, we did not expect the fusion proteins to be exported. Both of the fusion proteins were expressed well, though several other bands putatively corresponding to these proteins were also observed (FIGURE S12). For the GFP fusion, the green color of the protein was retained. We used mass spectrometry on this protein mixture to identify which species were present. For both pandonodin-A1 and pandonodin-GFP, the most prominent products were a fully processed pandonodin-protein fusion and a completely unprocessed PanA-protein fusion (FIGURE 7B). We did not observe any species corresponding to unlassoed pandonodin fused to protein. These results are in contrast to what we observed in previous experiments with astexin-1 fusions. In that work, the leader peptide of the lasso peptide was always cleaved.³⁹ In the pandonodin-A1 fusion, the masses observed for PanA-A1 and pandonodin-A1 both correspond to the pandonodin disulfide bond being formed. However, we observed both oxidized and reduced forms of the pandonodin-GFP fusion. For the unprocessed PanA-GFP fusion, only the reduced form was observed (FIGURE S12).

Figure 7:

Fusion of proteins to the pandonodin tail. A: Gene cluster for production of pandonodin fusion proteins. The protein of interest, either GFP or A1, is fused to the C-terminus of the pandonodin with an intervening thrombin cleavage site (LVPRGS). B: Major products observed in pandonodin-GFP and pandonodin-A1 fusions are either unprocessed pandonodin precursor (left) or fully processed pandonodin (right). Cleaved but not cyclized precursor is not observed. Gels and mass spectrometry data for these experiments are in Figure S13. C: Pandonodin tail swapping experiments. In these fusions, the final 12 aa of pandonodin are replaced with either full-length A1 protein or the first 12 aa of A1. D: For fusion of full-length A1 to PanA ΔC12, no conversion to the lasso is observed; the major product is unprocessed precursor and the minor product is precursor cleaved by PanB but not cyclized. For fusion of the first 12 aa of A1 (A1₁₂) to PanA ΔC12, conversion to the lasso and formation of the disulfide bond is observed in peptide exported to the supernatant, but the tails of these hybrid lasso peptides are highly truncated.

Given that the tail portion of pandonodin was dispensable with regard to its maturation, we also asked whether non-native sequences could be appended to a pandonodin variant with a truncated tail. Specifically, we generated two new constructs with C-terminal fusions to the pandonodin ΔC12 variant (FIGURE 7C). As discussed above, this is the shortest tail truncation that retains the disulfide bond. First, we directly fused A1 to pandonodin ΔC12 without any intervening thrombin cleavage site. This protein was expressed at high levels. According to SDS-PAGE analysis, the major product corresponds to unprocessed PanA ΔC12-A1, though there are also lower molecular weight bands suggesting that some protein was processed to a lasso or at least cleaved by PanB (FIGURE S13). Mass spectrometry analysis of this protein mixture showed both unprocessed PanA ΔC12-A1 as well as cleaved but not lassoed pandonodin ΔC12-A1, corresponding to linear pandonodin core peptide fused to A1 (FIGURE 7D). Thus fusion of full-length A1 to the pandonodin ΔC12 precursor still allows for the PanB protease to function. PanC, the pandonodin cyclase, however, cannot process this substrate further.

Finally, we generated a construct in which the first 12 aa of A1 were appended to the C-terminus of pandonodin ΔC12 (FIGURE 7C). We called this construct pandonodin ΔC12-A1₁₂. Given that A1 is a highly soluble protein, we expected this hybrid peptide to be more soluble than the parent pandonodin. This construct allowed us to probe whether a short, unrelated sequence could be appended to the C-terminus of pandonodin ΔC12. Expression of the gene cluster encoding pandonodin ΔC12-A1₁₂ led to only small amounts of products retained in the cells (FIGURE S14). In contrast, the supernatant of these cultures contained species that eluted near pandonodin (FIGURE S14). We carried out LC-MS on the cell extract and an extract of the culture supernatant. In addition, two distinct peaks from the culture supernatant at retention times of 22.5 min and 24 min were collected from the HPLC and analyzed. The cell extract contained some full-length, lassoed, and disulfide bonded pandonodin ΔC12-A1₁₂ as well as a range of its different truncations (FIGURE S14). The culture supernatant included three major species corresponding to pandonodin ΔC12, pandonodin ΔC13, and pandonodin ΔC12-Ser (FIGURE 7D). The pandonodin ΔC12-Ser species likely results from proteolytic cleavage of pandonodin ΔC12-A1₁₂ by E. coli proteases back to the first residue of A1, Ser. Both pandonodin ΔC12 and pandonodin ΔC12-Ser retained the disulfide bond, while pandonodin ΔC13 is missing the C-terminal cysteine residue (FIGURE 7D, FIGURE S14). Small amounts of other truncations of pandonodin ΔC12-A1₁₂ were observed in the supernatant with the longest species observed being pandonodin ΔC12-A1₁₀. These studies suggest that the tail residues of pandonodin can be swapped for arbitrary sequences while still retaining the ability to correctly form the lasso. However, the native pandonodin tail is especially protease resistant; when the 12 C-terminal aa are replaced with a different sequence, the tail is readily proteolyzed. Since pandonodin ΔC12-A1₁₂ is mostly proteolyzed prior to export, unfortunately we cannot conclude whether the PanD ABC transporter is as efficient at exporting this “tail-swapped” pandonodin variant as it is at exporting native pandonodin.

Discussion

Here we present the heterologous expression and structure of the lasso peptide pandonodin, encoded in several genomes within the Pandoraea genus. Our initial interest in this peptide stemmed from its unusually large core peptide sequence. At 33 aa, pandonodin represents the largest lasso peptide studied to date. Of the 33 aa in pandonodin, 18 aa comprise the C-terminal tail. This tail segment includes a 6 aa disulfide-bonded macrocycle which serves as a steric lock stabilizing the lasso structure. Our NMR analysis does not indicate the presence of any secondary structure (such as helicity) within the tail portion of pandonodin beyond the disulfide macrocycle. Despite this, the tail of pandonodin is largely resistant to proteolysis by carboxypeptidase. The tail portion of pandonodin could be truncated by as many as 14 aa while still retaining a high level of production. Full-length pandonodin is poorly soluble in water, instead forming a stable suspension. In contrast, the tail-truncated form of pandonodin, pandonodin ΔC14, was highly soluble in water. This suggests that the tail portion of pandonodin dictates its solubility in water. The tail segment of pandonodin is a low complexity sequence including multiple proline residues, so one explanation for the poor solubility of the full-length peptide is that multiple low complexity tails of the peptide aggregate together.

In addition to this solubility behavior, pandonodin is also noteworthy because of its thermostability and the proteolytic resistance of its tail. Upon heating to 95 °C, full-length pandonodin exhibits only minimal unthreading. More surprising is that the disulfide bond remains intact upon either heating to 95° C or treatment with reducing agents at room temperature; only a combination of heating and a high concentration of reducing agent was able to break the disulfide bond. Treatment of pandonodin with carboxypeptidase led to only minimal cleavage, which is likely due a combination of the poor solubility of the pandonodin and the presence of multiple proline residues within the tail. When we swapped the native pandonodin tail for a different sequence (FIGURE 7), the tail was readily proteolyzed by E. coli, showing that the tail sequence is responsible for its proteolytic stability. In summary, while our original interest in pandonodin stemmed solely from its large size, our experiments here show multiple surprising biophysical properties of pandonodin.

Methods

Cloning

All constructs were cloned by ligating restriction enzyme-digested insert and vector. All inserts were generated using 1-step or overlap PCR. The PCR template for A1 and GFP came from vectors previously reported.³⁹ A more detailed description of the constructs cloned is provided in the Supporting Information (SI), along with all primers and constructs used in this study (TABLE S4).

Expression and purification of pandonodin and its truncation variants

Expression was performed using Escherichia coli BL21 in M9 media supplemented with 0.00005 wt% thiamine, 20 amino acids (0.05 g/L of each amino acid), and 100 mg/L ampicillin. Cultures were induced at OD₆₀₀ of 0.2, with 200 μg/L anhydrotetracycline (aTc) if they contained a pASK75-based plasmid, or 1 mM of isopropyl-ß-D-thiogalactopyranoside (IPTG) if they contained a pQE-80-based plasmid. Cultures were then grown at 20 °C for 20 hours. Supernatant was harvested and extracted with C8 solid phase extraction columns. The extract was then eluted with methanol, rotavapped dry, and resuspended in 1:1 acetonitrile water. Peptide was then purified using reverse-phase HPLC with a C18 column. A more detailed description is provided in the SI.

Expression and purification of pandonodin fusion constructs

Expression was performed using E. coli BL21 in LB media with 100 mg/L ampicillin. Cultures were induced at OD₆₀₀ of 0.5 with 1 mM IPTG, and expressed at 20 °C for 20 hours. Cells were then harvested and lysed. Protein was purified using standard Ni-NTA affinity chromatography. A more detailed description is provided in the SI.

NMR

Purified pandonodin was quantified by an analytical lab balance, and then dissolved to a final concentration of 6 mg/mL in methanol-d₃ (CD3OH). The following 2D experiments were acquired with the 800 MHz Bruker Ascend NMR spectrometer at the Princeton University NMR Facility: ¹H-¹H gCOSY, ¹H-¹H TOCSY (80 msec mixing time), ¹H-¹H NOESY (500 msec mixing time), and ¹H-¹³C HSQC. All acquisitions were done at 295 K. Chemical shifts for all protons were assigned based on intra residue connectivities (gCOSY, TOCSY, HSQC) and inter residue connectivities (NOESY). Cross peaks were manually picked and integrated from the NOESY spectra in MestReNova version 11.0. These peak volumes were then used for distance calibration and as upper distance restraints for structural calculations using CYANA. Seven cycles of automated NOE peak assignment and structure calculations were performed starting with 100 initial structures, followed by a final structural calculation. The top 20 structures were analyzed. Of these 20 structures, 2 had major clashes involving the disulfide bond, and were removed. The remaining 18 structures with the lowest target function values were deposited into the PDB database (PDB code 6Q1X).

Stability assay

Pandonodin thermostability was tested with and without reducing reagents. Forty microliter samples of aqueous pandonodin at 5.8 mg/mL was prepared with the peptide alone, with 20 mM dithiothreitol (DTT), or with 5% β-mercaptoethanol (BME). These were then each split into 4 × 10 μL samples, where one sample was not further processed, one sample was digested with carboxypeptidase, one sample was heat-treated, and the last sample was heat-treated and then digested with carboxypeptidase. All samples after addition of reducing reagent were allowed to incubate at room temperature for 1 hour before any further processing. For samples that required no further processing steps, the samples were placed at 4 °C until all samples were ready for analysis by LC-MS. Heat-treatment was conducted at 95 °C for 2 hours in a thermocycler with a heated lid. Carboxypeptidase digestion was carried out in 100 μL reactions with 50 mM sodium acetate buffer, pH 6, 1 unit of carboxypeptidase B (Sigma-Aldrich) and 1 unit of carboxypeptidase Y (Affymetrix) for 3 hours at room temperature. All other samples were diluted to the same concentration with sodium acetate buffer before analysis. Two microliters of each sample was analyzed by LC-MS, where the first two minutes were diverted to waste.

Mass spectrometry analysis of pandonodin variants and fusion proteins

For LC-MS and LC-MS/MS analysis, 1–10 μL of either HPLC-purified pandonodin variant, solid-phase extract of culture supernatants including pandonodin variants, or partially purified pandonodin fusion proteins (see SI methods for more details) were injected onto the Zorbax 300SB-C18 (2.1 mm × 50 mm, 3.5 μm) column with a 1260 Infinity II system for separation using gradient elution. The mobile phase A consisted of water with 0.1% formic acid, and phase B consisted of acetonitrile with 0.1% formic acid. The gradient program used was as follows: 10% B from 0–1 minute, linear gradient from 10–50% B from 1–20 minutes, linear gradient from 50–90% B from 20–25 minutes, and isocratic elution at 90% B from 25–30 minutes. The separated species were directly sprayed to an Agilent 6530 q-TOF instrument for mass detection and fragmentation.