NMR structures and mutational analysis of the two peptides constituting the bacteriocin plantaricin S

The structure of the individual peptides of the two-peptide bacteriocin plantaricin S, an antimicrobial peptide produced by a Lactobacillus plantarum strain, has been determined in DPC micelles. The two peptides of plantaricin S, Pls-α and Pls-β, form an α-helix from and including residue 8 to 24 with a less structured region around residue 16-19 and an amphiphilic α-helix from and including residue 7 to 23, respectively. Activity assays on single amino acid-substituted GxxxG and GxxxG-like motifs show that substituting the Ser and Gly residues in the G9xxxG13 motif in Pls-α and the S17xxxG21 motif in Pls-β reduced or drastically reduced the antimicrobial activity. The two-peptide bacteriocin muricidin contains GxxxG-like motifs at similar positions and displays 40-50% amino acid identity with plantaricin S. Activity assays of combinations of the peptides that constitute the bacteriocins plantaricin S and muricidin show that some combinations are highly active. Furthermore, sequence alignments show that the motifs important for plantaricin S activity align with identical motifs in muricidin. Based on sequence comparison and activity assays, a membrane-inserted model of plantaricin S in which the two peptides are oriented antiparallel relative to each other and where the GxxxG and GxxxG-like motifs important for activity come close in space, is proposed.

. Sequences of the two peptides Pls-α (A) and Pls-β (B) investigated in this paper. Above the sequences, green bars illustrate sequence fragments that are active and red bars sequence fragments that are inactive according to Soliman et al. 12 . Below the sequences, the solid bars indicate the GxxxG and GxxxG-like motifs. The effect on activity by replacing the A, S or G residues in the black-barred A/S/GxxxA/S/G motifs with A, G or L and the bold W and Y residues with Y, W, L and R have been investigated in this paper. the peptides in the membrane, antimicrobial activity of variants where the N-and C-terminal aromatic residues shown in Fig. 1 were single substituted with Trp, Tyr, Leu and Arg, has been determined.

Results
Circular dichroism. The CD spectra of Pls-α and Pls-β are shown in Fig. 2A,B, respectively. Both peptides are unstructured in buffer solution but become structured in TFE and DPC. While there is little difference in the degree of structuring of Pls-α in TFE and DPC, there is a huge difference for Pls-β as observed in Fig. 2B. Furthermore, the CD spectrum of Pls-α in TFE has a curve that seems to have a negative minimum shifted towards lower nm values than that observed in DPC ( Fig. 2A). The CD results were best fitted, as evaluated based on RMSD and NMRSD values, using the routine Contin-ll in the CDpro program package 36 . The results are reported in supplementary section (Table S1). Due to the relatively drastic difference in the CD spectra in TFE and DPC, and the more bilayer-like properties of DPC, structure determination with NMR was done in DPC micelles. CD spectra of muricidin were recorded in similar environments ( Supplementary Fig. S1). The results indicate that the lengths of the helices of the two bacteriocins are similar in TFE (Table S1). A difference, that probably is due to the difference in C-terminal residues was, however, observed between Mur-α and Pls-α in DPC.
NMR structure determination. Chemical shifts were assigned using the 2D experiments described in methods and standard methods 37 . The assigned NOESY spectra are shown in supplementary section (Figs S2 and S3). The assigned chemical shifts have been uploaded to BMRB with the accession numbers 34278 and 34279. Chemical shift indexing 38 indicates possible helical regions from and including residue 5 to 25 in Pls-α and from and including residue 7 to 22 in Pls-β ( Supplementary Fig. S4A,B). Similar results were found by the application of TALOS+ 39 where the torsion angles indicated possible helical structuring from and including residue 6 to 26 in Pls-α and from and including residue 5 to 25 for Pls-β ( Supplementary Fig. S4C,D). The sequence plot illustrating the short and medium range NOE correlations observed from the two peptides in the NOESY spectrum ( Supplementary Fig. S4C,D) indicate a helical structure from residue 5 and out, with a potential helix break around residues 14 to 16, and from residue 7 to 21/22 for Pls-α and Pls-β, respectively. Structures calculated using restraints from NOESY and TALOS+ are shown in Fig. 3 while the structure restraints and structure ensemble characteristics are presented in supplementary section (Table S2). The calculated and refined structures show that Pls-α contains a straight α-helix from residue 8 to 24 with a break around residue 18, while Pls-β has an α-helix from residue 7 to 23. The structures and the α-helical break observed in some of the determined structures of Pls-α reflects the limited number of medium restraints observed in Supplementary Fig. S4C. The structure ensembles (Fig. 3A,B), cartoon representation of the structures (Fig. 3C,D) and structures illustrating the charged residues (Fig. 3E,F) are shown in Fig. 3. Both Pls-α and Pls-β contain hydrophobic C-terminal residues, whereas the N-terminal region consists of Lys and polar residues. The N-terminal region of the α-helix of Pls-α is partly amphiphilic containing the positively charged or polar residues His10, Tyr11 and Lys14; however, the polar residue Thr16 is on the opposite side of the helix (Fig. 3E). All residues from 17 and out are hydrophobic, as is observable from the representation of Pls-α with side chains (Fig. 3E). The three C-terminal residues are unstructured, while the seven-residue long N-terminal region folds back along the helix of Pls-α. The C-terminal positioning of the helix is in agreement with the observation that there was a higher degree of helicity in TFE of Pls-α fragments containing 21 and 16 C-terminal residues than in the full length peptide 12 .
The α-helix of Pls-β is amphiphilic containing the two polar, negatively charged residues Asp13 and Glu20, with an unstructured, short C-terminal tail and an N-terminal region that is partly folding back on the helix (Fig. 3B,D,F). To our knowledge, no CD-data exist for fragments of Pls-β. However, the N-terminal helical region of LcnG-β, one of the two peptides constituting the two-peptide bacteriocin lactococcin G 12,21 , shares 38% sequence similarity with Pls-β, and a similar helix to the one observed here. The secondary structures of both Pls-α and Pls-β peptides are comparable to those found for other two-peptide bacteriocins 22 and helicity is in Mutational effects on plantaricin S. Based on preliminary activity assays with plantaricin S against several bacterial strains, Lactococcus lactis IL1403 and Lactobacillus sakei NCFB 2714 were selected for further activity measurements of the mutated peptide variants (Supplementary Table S3). The Gly, Ala and Ser residues of the putative helix-helix stabilizing motifs ( Fig. 1) were single substituted with other small residues (Gly and Ala) or the large hydrophobic Leu residue. Furthermore, the terminal Trp and Tyr residues were substituted with Tyr, Trp, Leu and Arg to determine the membrane position of these residues. In total, 32 plantaricin S variants have been tested. Antimicrobial activity of peptide variants combined with the complementary wild type peptide was determined. Results of the GxxxG and GxxxG-like peptide variants in Table 1 are presented as minimum bacterial inhibitory concentration and relative minimum bacterial inhibitory concentration values with an 'x-fold' reduction (when x > 1) in activity of the peptide-combinations compared to the wild type minimum bacterial inhibitory concentration value. The antimicrobial activity measured was comparable between the two strains.
Replacing the Gly9 in Pls-α with an Ala residue reduced the activity 20-fold whereas this substitution was tolerated in position 13. However, replacing Gly residues (at position 9 and 13), with a Leu residue was highly detrimental, suggesting that the G 9 xxxG 13 motif in Pls-α is important for activity of the bacteriocin. On the contrary, the A 15 xxxG 19 motif is not important in helix-helix interactions as substituting with a Leu residue at these positions only slightly reduced the activity (1-to 5-fold). Substituting Ala15 with a Gly residue was somewhat detrimental, perhaps due to the more flexible Gly residue, which may destabilize the α-helical structure. For the Pls-β peptide, neither the S 6 xxxA 10 nor the A 10 xxxA 14 motifs had strong impact on the activity when substituted with Gly, Ala or Leu residues. Substituting Ser17 with a Leu residue in the S 17 xxxG 21 motif reduced the activity the most at this position compared to a Gly and an Ala residue, which reduced the activity 5-to 10-fold. Similar results were observed for Gly21 where replacing Gly with a Leu residue were detrimental on activity. These results indicate that the G 9 xxxG 13 motif in Pls-α and the S 17 xxxG 21 motif in Pls-β are important for bacteriocin activity.
The effect of single amino acid substitutions of the terminal Trp and Tyr residues of plantaricin S (Supplementary Table S4) showed that replacing Pls-α Y6 and Pls-β W26 with Arg were detrimental, while tolerated relatively well in Pls-α W23 and Pls-β W7. Replacement with Leu or Trp/Tyr residues was tolerated in all four positions. Activity of plantaricin S and muricidin peptide combinations. To investigate if the recently discovered bacteriocin muricidin 32 possesses antimicrobial activity, we performed activity assays of the individual peptides and combinations of muricidin and plantaricin S peptides ( Table 2). The minimum bacterial inhibitory concentration of muricidin was 2 to 3 times higher than that observed for plantaricin S against the two indicator strains tested. No or weak antimicrobial activity was observed for Mur-α and Mur-β individually. The fact that Mur-α has no individual activity within the concentrations tested, whereas the activity increases approximately 5000 times when Mur-α and Mur-β are combined as compared to Mur-β alone, clearly shows that muricidin is a two-peptide bacteriocin as predicted by Collins et al. 32 . Due to sequence similarities with plantaricin S, the activity of the combination of individual peptides of plantaricin S and muricidin was tested. Interestingly, the combinations Pls-α with Mur-β and Pls-β with Mur-α displayed high antimicrobial activity, only reduced by 3 to 5 times compared to plantaricin S.

Peptide combination b
Bacterial inhibitory concentration (nM)

Lab. sakei NCFB 2714
α + β 0.6 ± 0.1 0.17 ± 0.04 1 1 α(G9A) + β 11.4 ± 0.9 3.7 ± 1.0 20 20 Table 1. Bacterial inhibitory concentration and relative bacterial inhibitory concentration values from activity measurements of GxxxG and GxxxG-like peptide variants of Pls-α and Pls-β together with the wild type complementary peptide. a The relative bacterial inhibitory concentration is defined as a fold increase or decrease in activity compared to the wild type combination. b The peptides were added in equimolar amounts. α and β are Pls-α and Pls-β, respectively. Single amino acid substitutions are indicated in parentheses.

Peptide combination b
Bacterial inhibitory concentration (nM)

Discussion
In this study, we present the NMR structures of two peptides Pls-α and Pls-β, that constitute the bacteriocin plantaricin S, in DPC micelles and the CD structures of Pls-α, Pls-β, Mur-α and Pls-β in water, TFE and DPC micellar solutions. The peptides are unstructured in water and become α-helical upon interaction with membrane-mimicking environments 12,[19][20][21][22][23][24] . The α-helical structures determined by NMR of the two plantaricin S peptides are in agreement with the CD results, and the structures are similar to those observed for other two-peptide bacteriocins and proposed by Soliman et al. 12 . Previous predictions, however, indicated that the helices were from residues 15 to 25 and 4 to 15 12 . The structures determined here show that the peptides form linear helices from residue 8-24 with a break around residue 18 and 7-23 for Pls-α and Pls-β, respectively. Thus, the helices in the determined structures are both longer and positioned more centrally in the peptides than previously suggested.

GxxxG and GxxxG-like motifs important for plantaricin S activity. Several activity studies indicate
that the GxxxG and GxxxG-like motifs, identified in all characterized class IIb two-peptide bacteriocins so far, are important for the bacteriocin activity 19,25,31 in allowing close contact and stabilizing interactions between helices. A sequence alignment of Pls-α and Pls-β with two-peptide bacteriocins with, to our knowledge, most similar sequences are shown in Table 3. The sequence alignment of Pls-α, Mur-α and plantaricin E (PlnE) 40 illustrates that the G 9 xxxG 13 motif in Pls-α aligns perfectly with G 9 xxxG 13 in Mur-α and with the G 5 xxxG 9 motif found to be important for PlnE 25 . The alignment, furthermore, illustrates a higher degree of sequence similarity between Pls-α and Mur-α in the helix portion of Pls-α, residues 8 to 15, than in the more flexible N-and C-terminal parts. In agreement with the findings of Soliman et al. 12 , the sequence alignments and the activity measurements indicate that the G 9 xxxG 13 motif in Pls-α is highly important for activity.
Soliman et al. 12 showed that a 21-residue long N-terminal fragment that includes the S 17 xxxG 21 motif had activity similar to the wild type peptide when combined with Pls-β, however, a fragment that only included the A 10 xxxA 14 motif in Pls-β was inactive. This is in agreement with the finding that single amino acid changes were less tolerated in the S 17 xxxG 21 motif than in other investigated motifs in Pls-β. Table 3 shows that the S 17 xxxG 21 motif in Pls-β, the S 14 xxxG 18 motif in Mur-β and the G 18 xxxG 22 motif that previously has been shown to be important for lactococcin G-β (LcnG-β) activity 31 align, while neither the G 18 xxxG 22 motif in LcnG-β nor other N-terminal GxxxG and GxxxG-like motifs in Mur-β align with A 10 xxxA 14 in Pls-β. Furthermore, only minor changes in activity were observed when single amino acid replacements were performed in the previously predicted interaction motif A 10 xxxA 14 in Pls-β (Table 1). The observation that S 17 xxxG 21 motif in Pls-β is important for antimicrobial activity is not unique in that a SxxxG motif was also observed to be important in PlnF for its interactions with the complementary peptide PlnE in the bacteriocin plantaricin EF 25 .
Combinations of peptides from different bacteriocins are active. Activity assays on muricidin and the individual peptides that constitute this bacteriocin are reported here for the first time. Due to the large sequence similarities between the peptides of plantaricin S and muricidin, the activities of bacteriocin combinations were assayed to determine whether or not combinations of the peptides from different bacteriocins are active ( Table 2). The near plantaricin S activity of combinations of Pls-α and Mur-β and Pls-β and Mur-α (Table 2) indicate that motifs the two peptides have in common (Table 3) are important for their activity.
The nM minimum bacterial inhibitory concentration values observed for both plantaricin S, muricidin and peptide combinations further indicate that the bacteriocins act through a common receptor and not directly on the membrane, similarly to what has been observed for other one-and two-peptide bacteriocins 14,15,17,18 . No such receptor is, however, described for either of the bacteriocins in the literature to our knowledge. The highly similar  activities further indicate that the identical residues, which are mainly located in the α-helices of the peptides, are important for receptor interaction and that several residue changes may be accepted in the bacteriocin as long as these are not directly or indirectly affecting receptor interaction.

Dimer model of plantaricin s in bilayer.
In order for the two-peptide bacteriocin to exert an antimicrobial effect, the complementary peptides need to obtain the correct conformation to act on a receptor. The degree of structuring of most two-peptide bacteriocins, as was the case for Pls-α and Pls-β 12 , increases when combined in vitro compared to the sum of the parts 23,24 . We believe that this increased structuring is important for the receptor interaction, and that the structuring appear by the following mechanism; First the peptides of plantaricin S become structured in the membrane interface, then they interact and become one active bacteriocin dimer that inserts into the membrane, resulting in increased structuring of the bacteriocin. Similar mechanisms for the formation of bacteriocin dimers have previously been proposed for plantaricin S 12 and other two-peptide bacteriocins 19,25,31 . This membrane insertion is believed to be a prerequisite for the interaction of bacteriocins with a membrane-bound receptor 27,28 . It should be noted that at least in one two-peptide bacteriocin, carnobacteriocin XY, no in vitro interaction between the peptides has been observed 22 . Plantaricin S can form a parallel or anti-parallel transmembrane dimer complex when the G 9 xxxG 13 motif in Pls-α is in close contact with the S 17 xxxG 21 motif in Pls-β. If the G 9 xxxG 13 motif in Pls-α is close to the S 17 xxxG 21 motif in Pls-β in a parallel orientation, there will be little helix-helix overlap as the two motifs are in the N-and C-terminal ends of the respective helices. However, there is greater overlap between the helical regions of the two peptides in an antiparallel orientation. Activity of the variants where aromatic amino acids were replaced (Supplementary Table S4) are in overall agreement with the proposed model positioning the terminal aromatic residues Tyr and Trp in the membrane-water interface as illustrated in Fig. 4. The detrimental effect of replacing Pls-α Y6 and Pls-β W26 with Arg indicates that these residues are either positioned deeper into the bilayer than Pls-α W23 and Pls-β W7 or that the introduction of a positive charge affects the interaction with a receptor. Interestingly, neither of the active fragments observed by Soliman et al. 12 contain these aromatic residues, indicating that specific receptor interactions may be of importance. Replacing Ala15 and Gly19 in Pls-α and Ala10 and Ala14 in Pls-β with a Leu residue were well tolerated, thus these Ala residues may be positioned in the hydrophobic core of the membrane protected from water, in agreement with the proposed model (Fig. 4). Furthermore, the substitutions of Ser6 in Pls-β (Table 1) resulted in wild type activity, indicating that this polar residue is close to the membrane-water interface which also agrees with the proposed model. According to the positive inside rule 41 , the positively charged N-terminal of Pls-β is expected to be positioned on the cytosolic side of the membrane as illustrated in the model (Fig. 4). Previous MD simulations on plantaricin S, where the G 9 xxxG 13 motif in Pls-α was paired with the A 10 xxxA 14 motif in Pls-β, shows that this antiparallel model was able to partly insert into the upper membrane leaflet while the parallel model was not 12 . Unfortunately, no trajectories of the simulation were shown in the paper, making investigation into the model impossible. The relatively short length of the simulation (400 ns) compared to the several seconds full insertion may take 42,43 , may explain why the antiparallel dimer suggested here was not formed.

Materials and Methods
synthetic peptides. The wild type Pls-α, Pls-β, Mur-α and Mur-β peptides and single amino acid-mutated variants were purchased from GenScript. All peptides used in CD spectroscopy and NMR spectroscopy were ordered at a purity of >98%, while the peptides only used for activity measurement were ordered with a purity of >80%. Peptides used in activity measurements were dissolved in 40% 2-propanol upon arrival and stored at −20 °C, while the peptides used for structure investigation were dissolved on the day of use. The peptide concentrations were determined based on the molar extinction coefficients of the amino acids Tyr (ε 280 = 1200 M −1 cm −1 ) and Trp (ε 280 = 5560 M −1 cm −1 ) and absorption measured spectrophotometrically at 280 nm.
CD spectroscopy. CD spectra were recorded using a Jasco J-810 spectropolarimeter (Jasco International  47 NMR spectra were recorded. The data were acquired on a 600 MHz Bruker Avance II spectrometer with four channels and a 5 mm TCI cryoprobe (Bruker Biospin). NOESY and TOCSY spectra with mixing times of 120-300 ms and 60-80 ms, respectively, were used and the experiments were run at 298.15 K. Spectra were processed using the Topspin program (Bruker Biospin). DSS was used as a chemical shift standard, and 15 N and 13 C data were referenced using frequency ratios 48 .
Visualization and assignment of the spectra were performed using the computer program CARA 49 while SPARKY 50 was used for integration of NOESY peaks. The spectra were assigned with standard methods 37 . Dihedral angle restraints were obtained from the chemical shift values using the program TALOS+ 39 . Nuclear Overhauser effect (NOE) distance restraints were calculated from the peak volumes in the NOESY spectra with a NOESY mixing time of 120 ms. Cyana 2.1 51,52 was used for NOE assignment and restraint calculations. Structure calculations were done using CNS 53,54 and refinement in DMSO was done applying all restraints and RECOORD 55 . 100 structures were calculate and refined, the 20 structures with the lowest energy were analyzed further. The number of applied restraints and structure statistics are presented in Supplementary Table 1. Bacterial strains, growth conditions and activity measurements. Indicator strains were obtained from Laboratory of Microbial Gene Technology, Department of Chemistry Biotechnology and Food Science, Norwegian University of Life Science, Norway. All strains were grown at 30 °C without agitation in de Man-Rogosa-Sharpe (MRS) medium (Oxoid). The indicator strains Lac. lactis IL1403 and Lab. sakei NCFB 2714 were selected for mutational analysis based on sensitivity towards plantaricin S (Supplementary Table S3).
Antimicrobial activity of wild type and variants of Pls-α and Pls-β were analyzed in a microtiter plate assay system, essentially as described by Nissen-Meyer et al. 56 . In brief, each well of the microtiter plate contained MRS medium to a final volume of 200 μl, combinations of wild type and mutated variants of Pls-α and Pls-β (in 1:1 ratio), and indicator strain. The individual peptides were added at a concentration of 25 nM in the first well, with a two-fold dilution factor going from one well to the next. The muricidin peptides were added at a concentration of 25 nM in the first well, while the peptides of both muricidin and plantaricin S were added at 1000 nM concentrations for determining individual activity. Peptide combinations with decreased antimicrobial activity compared to the wild type combination were added at individual concentrations of 100 nM in the first well. Freshly made stationary phase cultures of indicator strains were diluted 1:50 and the microtiter plates were incubated for 5 hours at 30 °C. The growth of the indicator cells was measured spectrophotometrically at 600 nm by use of a Sunrise ™ Remote microplate reader (Tecan).
The bacterial inhibitory concentration was defined as the total amount of peptides added, at a 1:1 ratio, which inhibited the growth of the indicator strain by 50%. The relative bacterial inhibitory concentration value was quantitated in terms of fold increase or decrease in antimicrobial activity compared to the wild type combination.