Comparison of the Structure and Dynamics of the Antibiotic Peptide

Li , C.; Lewis , M. R.; Gilbert , A. B.; Noel , M. D.; Scoville , D. H.; Allman , G. W.; Savage , P. B. Antimicrob. Agents Chemother. 1999 43 1347 134...
0 downloads 0 Views 2MB Size
Download PDF
544

J. Phys. Chem. B 2009, 113, 544–551

Comparison of the Structure and Dynamics of the Antibiotic Peptide Polymyxin B and the Inactive Nonapeptide in Aqueous Trifluoroethanol by NMR Spectroscopy Jeffrey J. Meredith,† Antoine Dufour,‡ and Martha D. Bruch* Chemistry Department, Oswego State UniVersity, Oswego, New York 13126 ReceiVed: September 21, 2008; ReVised Manuscript ReceiVed: NoVember 7, 2008

The structure and dynamics of polymyxin B (PxB), an N-acylated cyclic decapeptide that displays antimicrobial activity against Gram-negative bacteria, is characterized by NMR and compared to results for the inactive nonapeptide, which is missing the N-terminal amino acid along with the attached acyl chain. Aqueous trifluoroethanol (TFE) was chosen as the solvent since the overall structure of PxB in TFE is similar to the structure when bound to vesicles. No differences were observed between the two peptides for 1H HR chemical shifts or patterns of cross peaks in NOESY spectra, indicating that the overall structures are quite similar. The sign and intensity of NOESY spectra obtained at different temperatures were used to assess the relative mobility of the peptides. For both peptides, differential mobility is observed in different parts of the molecule, with greater mobility observed for the linear portion than the ring and faster motion seen for the side chains than the peptide backbone. However, all motion is faster in the nonapeptide, indicating that the presence of the N-terminal acyl chain restricts the mobility of PxB compared to the nonapeptide, which lacks this structural feature. For both peptides, differential mobility is also observed within the cyclic portion of the peptide. This supports a proposed model whereby the more rigid residues serve as pivot points, allowing the ring conformation to change in response to different binding partners. However, conformational flexibility within the cyclic ring is not sufficient for antimicrobial activity since both the active and inactive peptides exhibit the same flexibility. The N-terminal acyl chain on PxB, which is essential for activity, exhibits rapid, independent motion, and this flexibility may facilitate penetration of the outer membrane. Introduction Biological activity of peptides and proteins has traditionally been associated with attainment of a fully folded, well-defined three-dimensional structure. Adoption of a specific, stable structure is often a requirement for biological activity, and structural changes can disrupt this activity. Furthermore, many biologically active peptides are largely disordered in water but fold into a specific structure upon binding to a membrane. These observations suggest that only well-defined, ordered structures are relevant for biological activity. However, recent studies suggest that disordered regions in partially folded proteins are also important for function.1-5 The inherent flexibility in these partially folded regions may enable the molecule to switch between multiple structures in order to interact with a variety of binding sites. Consequently, identification of flexible regions and determination of the range of structures available in partially folded molecules may aid in understanding the molecular basis for biological activity of a peptide or protein. Unfortunately, characterization of partially folded structures is more difficult than structure determination of a polypeptide chain that folds into a stable, well-defined structure, but insights can be gained from examination of molecular dynamics. Since structural flexibility results in greater mobility, characterization of the relative mobility of different regions of a molecule in combination with structural analysis can identify ordered and disordered regions in the polypeptide chain and determine the degree of flexibility within the molecule. * Towhomcorrespondenceshouldbeaddressed.E-mail:[email protected]. † Current address: Albany College of Pharmacy, Albany, NY 12208. ‡ Current address: Department of Chemistry and Hematology/Oncology, School of Medicine, Stony Brook University, Stony Brook, NY 11794.

Previous work suggests that conformational flexibility is important for antibiotic activity of the peptide polymyxin B,6 so characterization of this flexibility may enhance current knowledge of the mechanism of action of this peptide, which is not completely understood. Polymyxins are naturally occurring cyclic peptides produced by the Gram-positive bacteria, Bacillus polymyxa. Polymyxin B exhibits antimicrobial activity against Gram-negative bacteria7 and has been used clinically for over 50 years in antibiotic ointments for prevention of infection due to minor skin injuries. Despite years of study, the precise mechanism of this peptide is only speculatory in nature. PxB has been thought to act as a detergent, to disrupt the integrity of the cellular membrane, and to induce a hyperosmotic stress response which inevitably kills the organism.8-10 In order to accomplish this objective, PxB first binds to the lipopolysaccharide moiety (Lipid A) of the outer membrane. In doing so, divalent cations are released and the polysaccharide network of the outer membrane network can become slightly disarrayed.11 After this point, additional PxB molecules are permitted to penetrate into the periplasmic space where they are able to bind to the inner membrane and induce phospholipid transfer.12 This PxB-mediated phospholipid transfer has been shown in experiments involving fused vesicles.11 Since polymyxins are cationic molecules and are able to bind to the lipid A component of the outer membrane of a Gram-negative cell, the vast majority of bacteria are not likely to survive an encounter with this peptide without an uncommonly mutated lipid A site.13 It is for this reason that the study of antimicrobial peptides could be a door toward the development of new classes of compounds which would not allow bacteria to become resistant as easily.

10.1021/jp808379x CCC: $40.75  2009 American Chemical Society Published on Web 12/19/2008

Structure and Dynamics of Polymyxin B

Figure 1. Primary structure of polymyxin B. The N-terminal alkyl chain is heterogeneous, with different R groups possible; R ) H for the major isomer. The nonapeptide is formed by cleavage of the amide bond between Dab 1 and Thr 2.

The proposed mechanism for the antimicrobial activity of PxB requires the peptide to bind to both the inner and outer membranes of the cell, which contain divalent and monovalent phosphoester groups, respectively. PxB, which contains six 1,3diaminobutyric acid (Dab) residues and a D-phenylalanine, consists of a linear region, residues 1 to 3, and a ring caused by formation of an additional amide bond between the terminal amino group of Dab 4 and the C-terminal carboxylate of Thr 10 (Figure 1). The N-terminus is end-capped with a long alkyl chain, which is heterogeneous in commercially available PxB. Computer modeling studies indicate the existence of two separate and distinct binding sites in PxB, which can accommodate either monovalent or divalent ligands.6 However, computational results indicate that differences exist in the ring conformation of PxB depending on the nature of the ligands, and this is consistent with NMR results in aqueous solution which suggest that part of the peptide adopts a fixed structure while the rest is more flexible, allowing the ring pucker of the cyclic portion of the molecule to change.6 This flexibility may enable the peptide to engage in the complex set of molecular interactions required for antibiotic activity. We report further characterization of the model proposed previously for PxB by Bruch et al.6 through a thorough examination of the relative mobility of different structural regions of the molecule via two-dimensional NMR experiments. Aqueous trifluoroethanol (TFE) was chosen as the solvent because previously obtained results utilizing circular dichroism indicated that the peptide adopts a similar conformation in this solvent to that observed at phospholipid interfaces.12 Therefore, aqueous TFE is a simple membrane-mimetic solvent that is ideally suited for nuclear magnetic resonance (NMR) analysis. NMR is an ideal technique for structure determination in solution since NMR parameters such as chemical shift, coupling constant, and nuclear Overhauser effect (NOE) are sensitive to structural details. Furthermore, molecular motion occurring in the megahertz frequency range can be probed effectively by NMR since the sign and magnitude of the NOE is dependent on the frequency of molecular motion relative to the operating frequency of the spectrometer. Relative mobility was probed by examination of the sign and intensity of cross peaks in twodimensional NOE (NOESY) spectra as the temperature was varied, and 1H chemical shifts of HR protons were used to monitor changes in the overall structure of PxB. Since the cross peaks can be attributed to interactions between specific pairs of protons, mobility of different parts of the peptide can be probed separately, and differential mobility of different regions

J. Phys. Chem. B, Vol. 113, No. 2, 2009 545 is observed. Additional insights can be gained by comparison of the results on PxB with those obtained on the inactive analogue, polymyxin B nonapeptide (PxBn) The nonapeptide is obtained from PxB by cleavage of the Dab 1 residue and the attached alkyl tail. Like PxB, the nonapeptide interacts with lipid A and disrupts the outer membrane, but it does not display an equivalent ability to induce phospholipid transfer or hyperosmotic stress.8-10,12 The predominant difference in the two analogues of polymyxin B is that the active form, PxB, has a hydrophobic alkyl tail not present in the inactive form, PxBn. The alkyl tail has been implicated in the antimicrobial function of PxB since this tail is known to be essential for function, based on results for an eight amino acid polymyxin derivative, octapeptin.7 However, this tail has not been indicated to play a part in binding, as the inactive nonapeptide is known to bind to the outer membrane of a Gram-negative cell as efficiently as PxB.12 The nonapeptide exhibits the same differential mobility as PxB, but the rate of motion in all regions of the molecule is faster in PxBn, in contrast to previous work suggesting the nonapeptide is more rigid than polymyxin B.6 Experimental Methods Sample Preparation. PxB and PxBn were obtained from Sigma-Aldrich and used without further purification. Some heterogeneity in the structure of the alkyl chain on the N-terminus of PxB was detected by NMR, and a small amount of PxB was detected by mass spectrometry in the sample of PxBn. For NMR samples, approximately 5 mg of peptide was dissolved in a total of 700 µL of solvent, consisting of 50% TFE-d3, 10% D2O, and 40% HPLC-grade water (pH 4) by volume. The deuterated solvents were obtained from Cambridge Isotopes. The pH of the NMR sample was adjusted to between 3.5 and 4.0 using dilute HCl to minimize the exchange rate of the amide protons with the solvent. NMR Spectroscopy. Most NMR experiments were performed on a Varian Unity 300 MHz spectrometer, with some experiments performed on a Bruker DRX 400 MHz NMR or a Bruker DRX 500 MHz NMR. Temperature was controlled for all experiments, with temperatures below room temperature achieved with a FTS sample chiller, and the actual temperature was determined from the chemical shift difference in the proton spectrum of a methanol standard sample. Actual temperatures corresponding to 40, 25, 0, and -15 °C were determined to be 37.4, 23.6, 0.3, and -14.1 °C, respectively. 1H chemical shifts were referenced to the signal for TFE at 3.88 ppm. For all experiments, the water peak was suppressed using presaturation for 1 s prior to acquisition. All two-dimensional spectra were obtained in phase-sensitive mode using the method of States et al.14 and were processed using a phase-shifted Gaussian window function and were zero-filled to 1K in each dimension. Normal proton spectra were obtained with 16K data points covering a spectral width of 4000 Hz and a relaxation delay of 1 s. Line assignments for all protons in each amino acid were made with the assistance of total correlation spectroscopy (TOCSY), which were obtained with 256 increments of 1K data points covering a spectral width of 4000 Hz. Sixteen transients were acquired per increment using a mixing time of 75 ms and a spin-lock field of 5 kHz. NOESY spectra were acquired under identical conditions, except the mixing time was 300 ms. The phase was adjusted to produce positive diagonal peaks. As a result, negative NOESY cross peaks are obtained for fast motion and positive cross peaks are obtained for slow motion.

546 J. Phys. Chem. B, Vol. 113, No. 2, 2009

Meredith et al.

TABLE 1: Hr 1H NMR Chemical Shifts (ppm) for PxB and PxBn at Several Temperaturesa 25 °C

0 °C

-15 °C

residue

PxB

PxBn

PxB

PxBn

PxB

PxBn

Dab 1 Thr 2 Dab 3 Dab 4 Dab 5 Phe 6 Leu 7 Dab 8 Dab 9 Thr 10

4.56 4.34 4.52 4.24 4.52 4.40 4.15 4.35 4.25 4.09

4.02 4.54 4.28 4.56 4.44 4.12 4.38 4.26 4.11

4.58 4.30 4.50 4.21 4.52 4.34 4.11 4.31 4.24 4.07

3.97 4.50 4.24 4.52 4.40 4.13 4.28 4.25 4.11

4.59 4.37 4.50 4.20 4.49 4.41 4.14

3.95 4.49 4.23 4.51 4.38 4.13 4.26

b 4.25 4.12

b 4.12

a All shifts are measured in ppm. Estimated uncertainty is ( 0.02 ppm. b These shifts cannot be measured due to spectral overlap at -15 °C.

Results Overall Structure of PxB and PxBn. The structure of PxB was determined previously6 in 50% TFE at 25 °C, but the structure of PxBn is not known in this solvent. 1H chemical shifts of HR protons have been shown to be sensitive probes of secondary structure in polypeptides, with different ranges observed for a given amino acid in R-helix, β-sheet, or random coil conformation.15-17 Consequently, variation in HR chemical shifts is indicative of a conformational change. 1H chemical shifts for PxB and PxBn were measured from TOCSY spectra at several temperatures and are summarized in Table 1. For ease of comparison between the two peptides, the residues in the nonapeptide are numbered 2 to 10, not 1 to 9. There are no significant differences in HR chemical shifts between the two peptides at 25 °C except for Thr 2, which is shifted upfield in PxBn compared to PxB since this residue has an free amino group in the nonapeptide in place of an amide bond to Dab 1 in PxB. The similarities in chemical shifts between the two peptides indicate that there are no significant differences between the average structure of the PxB and PxBn. Furthermore, the chemical shifts of both PxB and PxBn are essentially invariant with temperature, which indicates the average structure does not change appreciably as the temperature is lowered. Another indicator of structure is the solvent accessibility of amide protons, which is reflected in the change in amide chemical shifts with temperature. Large variation in amide shift with temperature indicates that amide proton is exposed to the solvent, whereas small change is indicative of intramolecular hydrogen bonding to that amide. No differences are seen in solvent exposure between the two peptides, with the magnitude of most temperature coefficients greater than 4 ppb/°C, consistent with a lack of intramolecular hydrogen bonds in either peptide. Dynamics of PxB. The NOE is a through-space interaction between two spatially close protons, which manifests as a pair of cross peaks connecting the corresponding signals in a NOESY spectrum.18 The sign of these cross peaks depends on the frequency of motion affecting the interproton distance relative to the spectrometer frequency. Motion faster than the spectrometer frequency gives rise to negative cross peaks (opposite sign from diagonal), while slow motion results in positive cross peaks. For motion close to the spectrometer frequency, cross peaks are vanishingly small, even for very close proton pairs. The intensity of the cross peaks depends upon both the motional frequency and the interproton distance.18 For a given short distance, the intensity is near zero for motion near the spectrometer frequency and increases with slower molecular

motion. For a fixed motional frequency, the cross-peak intensity is proportional to the inverse sixth power of the interproton distance. The maximum distance for which cross peaks can be detected is typically 3-5 Å, with the upper limit increasing with slower motion. NOESY spectra of PxB at 25 °C obtained previously at 400 MHz revealed an array of positive cross peaks representing both sequential and intraresidue interactions between peptide protons.6 Since all peptide cross peaks are positive, motion is slow relative to 400 MHz, which corresponds to an effective correlation time greater than 0.4 ns. However, numerous negative cross peaks were observed between protons in the alkyl chain attached to the N-terminus. These negative signals indicate that this hydrophobic tail on the N-terminus is moving independently, exhibiting segmental motion that is considerably faster than that of the rest of the peptide. The tail cross peaks remain negative in NOESY spectra obtained at 500 MHz, indicative of a correlation time shorter than 0.32 ns at 25 °C for the tail. Somewhat different results are observed in the 300 MHz NOESY spectrum at 25 °C. The sign of the cross peaks is the same as at 400 MHz, with negative cross peaks observed within the tail and positive cross peaks in the rest of the peptide. However, only a few small sequential cross peaks are seen between amide protons at 300 MHz (Figure 2), corresponding to 2/3, 3/4, 9/10, and 10/4 Hδ interactions, whereas a network of these cross peaks are observed at 400 MHz. These interactions correspond to the shortest interproton distances between amide protons based on distance estimates obtained from 400 MHz NOESY spectra,6 so these cross peaks are expected to be the largest. All of these distances are estimated to be less than 3.5 Å, while the interactions not seen at 300 MHz have distances between 3.5 and 4 Å. Compared to 400 MHz spectra, all cross peaks are reduced in intensity at 300 MHz, with vanishingly small intensities for cross peaks between backbone protons corresponding to distances greater than 3.5 Å at 25 °C. If all motion in the peptide had the same effective correlation time, then positive cross peaks would be seen for all proton pairs separated by less than 3.5 Å. However, some interactions corresponding to short distances between sidechain protons are not visible at 300 MHz, suggesting that the side chains are moving faster than the backbone of the peptide. When the temperature is increased to 40 °C, all cross peaks are smaller as expected due to a decrease in molecular correlation time at the higher temperature. Since chemical shifts indicate that the average structure is invariant with temperature, interproton distances at 40 °C are expected to be similar to those measured previously at 25 °C.6 No NOEs are observed between amide protons, indicating that NOESY cross peaks will not be observed for interproton distances greater than 3.0 Å at 40 °C. Positive cross peaks are observed between some NH and HR protons, all corresponding to distances less than 2.5 Å. However, very few cross peaks are observed between amine and sidechain protons at 40 °C (Figure 3), with vanishingly small intensities observed in some cases where the estimated distance is quite short. Similarly, nearly all of the positive cross peaks observed between side-chain protons at 25 °C have disappeared at 40 °C (not shown). The absence of nearly all interactions to side-chain protons indicates that the side chains are moving faster than the backbone, with an effective correlation time for side-chain motion of approximately 0.53 ns at 40 °C. In addition to mobility differences between the backbone and the side chains, differential mobility is observed within different parts of the ring. There is evidence that residues 6, 7, 4, and 10

Structure and Dynamics of Polymyxin B

J. Phys. Chem. B, Vol. 113, No. 2, 2009 547

Figure 2. Amide region of 300 MHz NOESY spectra of PxB in 50% TFE at various temperatures. Top: 40 °C (left) and 25 °C (right). Bottom: 15 °C (left) and 0 °C (right).

are less mobile than the rest of the ring. Only two positive cross peaks are observed between side-chain protons, corresponding to interactions between the HR and methyl protons of Thr 10 along with large cross peaks connecting the nonequivalent Hγ methylene protons of Dab 4. In addition, positive cross peaks are observed for sequential and/or intraresidue interactions between HR and amide protons for residues 4, 6, 7, and 10, while cross peaks are not observed for these interactions corresponding to similar distances in residues 5, 8, or 9. Furthermore, the only positive cross peaks observed between amide and side-chain protons at 40 °C are due to NH/Hβ interactions in Dab 4, Phe 6, and Thr 10 (Figure 3). When the temperature is lowered at 15 °C, the intensity of all cross peaks is increased compared to 25 °C, reflecting decreased mobility at the lower temperature. More cross peaks between amide protons are present (Figure 2), indicating interproton distances less than 4 Å result in positive cross peaks at this temperature. Similarly, many additional cross peaks are seen to side-chain protons (Figure 3), and all peptide side chain cross peaks are now positive, indicating that the effective correlation time of side-chain protons is greater than 0.52 ns at 15 °C. By contrast, some negative cross peaks are observed between protons at the end of N-terminal tail, indicating that part of the tail is still moving fast relative to 300 MHz at 15 °C. Decreasing the temperature further to 0 °C results in more cross peaks, especially to side-chain protons. Specifically, positive cross peaks are observed between NH and Hγ protons of Dab residues, along with interactions between the aromatic ring of Phe 6 and HR protons of both Dab 5 and Leu 7. Cross

peaks due to an additional long-range interaction between amide protons of Dab 5 and Leu 7 are also observed. These cross peaks demonstrate the ability to see interactions between protons separated by longer distances due to the increase in the effective correlation time at lower temperature. When the temperature is further reduced to -15 °C, spectral overlap and line broadening makes definitive assignments difficult, but no additional cross peaks are apparent. Although all peptide cross peaks are large and positive at this temperature, cross peaks involving the end of the hydrophobic tail are still negative, even at -15 °C. However, the cross peaks to the methylene protons directly adjacent to Dab 1 are positive at both 0 and -15 °C, which implies that the mobility of the N-terminal tail increases down the chain with increasing distance from the peptide. The greater mobility of the tail is also reflected in the TOCSY spectrum at -15 °C (not shown). Many expected peptide cross peaks are small or absent in the TOCSY spectrum, especially cross peaks connecting amide and side-chain protons. These missing signals are attributed to short T1F relaxation times at low temperature due to reduced peptide mobility. However, all expected cross peaks connecting protons in the hydrophobic tail are clearly observed and are much larger than peptide cross peaks. This demonstrates the dramatic difference between mobility of the tail and the rest of the peptide, even at low temperature. Nonapeptide Dynamics. Unlike polymyxin B, the NOESY spectrum of the nonapeptide at 25 °C contains very few cross peaks, with a mix of positive and negative intensities observed (Figure 4).

548 J. Phys. Chem. B, Vol. 113, No. 2, 2009

Meredith et al.

Figure 3. Fingerprint region of 300 MHz NOESY spectra of PxB in 50% TFE at various temperatures showing interactions to amide protons. Temperatures from left to right: 40, 25, 15, and 0 °C.

This indicates that the nonapeptide has a shorter correlation time than PxB, with an effective correlation time near 0.52 ns at 25 °C. As in the case of PxB, however, differential mobility is observed within different regions of the nonapeptide. Most of the negative cross peaks observed at 25 °C correspond to interactions between side-chain protons of Thr 2 or Dab 3, which is consistent with greater mobility of the linear portion of the peptide compared to the ring. Differential mobility is also observed in different portions of the ring, similar to that observed for PxB. All cross peaks between amide protons and HR or sidechain protons are positive, indicative of slow motion, but the only cross peaks observed in this region involve amide protons of residues 4, 6, 7, or 10. Furthermore, the only positive cross peaks involving side-chain protons correspond to interactions involving Phe 6, Leu 7, or Dab 4, whereas small negative cross peaks are observed between side-chain protons within Dab 8 or Dab 9. No other cross peaks of either sign are observed for any protons in residues 5, 8, or 9. These results imply that residues 4, 6, 7, and 10 are more rigid than the rest of the ring, as was observed for PxB. When the temperature is lowered to 0 °C, many additional positive cross peaks are observed, and no negative signals remain. The NOESY spectra of PxBn and PxB are quite similar at 0 °C. Most of the same cross peaks are observed for the two peptides, but they are smaller in the spectrum of the nonapeptide, reflecting the greater mobility of PxBn. Some peaks are still missing in the spectrum of PxBn compared to PxB at 0 °C, but these peaks appear when the temperature is reduced further to -15 °C. The NOESY spectrum of PxBn at -15 °C is essentially the same as the NOESY spectrum of PxB at 0 °C.

Discussion In a prior effort to determine the structure of PxB from NMR data in aqueous TFE, simulated annealing was used to elicit several families of structures, which differ primarily in the ring pucker of the cyclic part of the molecule.6 Based on comparison of these structures along with analysis of chemical shifts, a model was proposed whereby part of the molecule, comprised of residues 6, 7, and 10, adopts a fixed structure, while the remainder of the ring is more flexible. Computer modeling of phosphoester binding to PxB indicated the existence of two separate and distinct binding sites, capable of binding to the headgroup of either one or two phosphotidylglycerols, but also capable of binding two phosphoester groups in the same ligand, as in lipid A. However, accommodation of these different substituents requires a change in the ring conformation, similar to the differences seen in the families of NMR structures generated from simulated annealing. These results led to a model in which residues 6 and 7 on one side of the ring and residue 10 on the other side serve as pivot points, while the rest of the ring changes conformation to accommodate different binding partners.6 This flexibility may be necessary to allow the peptide to engage in the diverse molecular interactions required for antibiotic activity. Our results are consistent with this model of fixed and flexible regions within the cyclic region of PxB. However, variabletemperature NOESY spectra demonstrate that residues 4, 6, 7, and 10 are less mobile than the rest of the ring, which suggests it is the 4-10 bridge, not just residue 10, that serves as the pivot point on one side of the ring. These results support the

Structure and Dynamics of Polymyxin B

J. Phys. Chem. B, Vol. 113, No. 2, 2009 549

Figure 4. 300 MHz NOESY spectra of the nonapeptide (PxBn) in 50% TFE at 25 °C (top) and 0 °C (bottom). Positive cross peaks are shown in black, negative in red.

model whereby the molecule flips between two conformations, which differ in ring pucker. Similar behavior to that proposed for PxB has been observed in proteins. In a study by Bemporad et al., enzymatic activity was observed for a nonnative, partially folded structure of an acyl phosphatase enzyme.4 Structural characterization of this partially folded state showed that the active site is largely unstructured and flexible, while the rest of the molecule acts as a scaffold that restricts the conformational space of the flexible region of the enzyme. Flexibility within a region of the protein also has been invoked to explain the structural heterogeneity observed in crystal structures of the protein ubiquitin when bound to different proteins. Analysis of residual dipolar couplings for ubiquitin in solution revealed a collective, pincer-like motion of those residues that are often involved in interfaces with other proteins upon formation of protein-protein complexes.5 Some residues of the protein are flexible, while others are relatively rigid, and the combination of rigidity and plasticity is proposed to be a crucial feature enabling ubiquitin to bind to a variety of partners with high affinity. In the case of PxB, the ability to change ring conformation may facilitate binding of PxB to different types of phosphoester ligands, which is essential for antimicrobial

activity. Indeed, synthetic cholic acid-derived mimics of PxB which are structurally rigid analogues do not exhibit antibiotic activity,19,20 which suggests that conformational flexibility is a key feature required for activity. However, the same differential mobility within the ring was also observed for the inactive nonapeptide, which suggests that the inactive peptide has the same ability to change conformations. Hence conformational flexibility appears to be a necessary, but not sufficient, condition for antimicrobial activity. In addition to mobility differences within the ring, mobility differences also are observed between the backbone and side chains for both peptides, with side chains exhibiting greater mobility, especially for Dab residues, which are the only residues in PxB with long, hydrophilic side chains (Figure 1). This agrees with differential mobility between backbone and side chains observed previously for linear peptides in aqueous TFE, particularly for lysine residues, which are structurally similar to Dab residues.21,22 The binding sites in PxB consist primarily of the terminal NH3+ groups of Dab residues, so the flexibility at the end of Dab side chains may make it easier for these cationic groups to bind to the anionic phosphoester groups on the cell membrane surface. Observation of flexible side chains

550 J. Phys. Chem. B, Vol. 113, No. 2, 2009 for both PxB and PxBn is consistent with similar binding propensities to the outer membrane of the cell observed for the two peptides. Differences in mobility are also observed between the cyclic and linear portions of both peptides. The linear portion is more mobile than the ring, as was observed previously for both PxB23 and PxBn24 in water. Our results provide a more detailed description of this differential mobility, demonstrating that the mobility increases with distance from the ring. For PxB, there is a gradual increase from Dab 3 to Dab 1, with the N-terminal alkyl chain moving much faster than the rest of the peptide. Differential mobility is also seen within the N-terminal tail, with mobility increasing down the chain as the distance from Dab 1 increases. The behavior of the N-terminal alkyl chain is of interest since this hydrophobic tail is critical for antibiotic activity of the peptide. Although specific conclusions cannot be drawn, the rapid, independent motion of the tail may provide clues to the molecular basis for the antimicrobial activity of PxB. The proposed mechanism of action is complex, involving a cluster of PxB molecules at the cell surface. Addition of PxB to vesicles causes formation of dimers, with approximately 5-6 PxB molecules required for each dimer-dimer contact.25 It is thought that PxB forms contacts between the inner and outer cell membranes and facilitates exchange of phospholipids between the two membranes, and this is the basis for antibiotic activity. This mode of action requires the peptide to traverse the outer membrane and enter into the periplasmic space between membranes. Observation of increased surface pressure of lipid A monolayers also suggests that the tail penetrates the outer membrane.26 The independent motion of the hydrophobic tail may provide the conformational flexibility necessary for penetration of the membrane. Furthermore, this flexibility, in combination with fixed structural features to restrict the conformational space of the molecule, may enable alignment of the N-terminal alkyl chains during cluster formation at the membrane surface, providing a hydrophobic driving force for penetration of the membrane. Despite moving independently from the peptide, the presence of the N-terminal tail restricts the mobility of the cyclic part of the peptide. It has been proposed that the nonapeptide is more rigid than PxB and lacks the conformational flexibility necessary for antibiotic activity.6 However, the nonapeptide exhibits faster motion than PxB, in direct contrast to this hypothesis. At first glance this is not surprising since the nonapeptide is smaller than PxB, resulting in a shorter overall correlation time. However, the decreased size of the nonapeptide is caused by removal of Dab 1 along with the N-terminal tail, and this section of the molecule is moving independently from the cyclic part of the peptide. Since the ring is the same size in both peptides and exhibits independent motion, it was anticipated that the effective correlation time for the ring motion would be the same for both peptides. By contrast, this correlation time is longer for PxB, indicating that the tail restricts the ring motion even though the tail engages in rapid, independent motion. The faster motion of the ring in PxBn suggests that a greater range of conformations is sampled by the nonapeptide compared to PxB. The presence of the tail in PxB restricts the ring motion, thus reducing conformational entropy, which may assist with formation of contacts among PxB molecules that are necessary for antibacterial activity. The presence of a flexible tail also may assist with stabilization of the multiple conformations required for binding to both monoanionic and dianionic phosphoester groups.

Meredith et al. No differences were observed between the average structures of PxB and PxBn in aqueous TFE based on chemical shifts and cross-peak patterns in NOESY spectra. This is consistent with equal binding propensities for lipid A observed for the two peptides. Consequently, differences in activity of the two peptides cannot be attributed to differences in their threedimensional structure. As discussed earlier, differential mobility within the peptide was observed for both PxB and PxBn, and thus it cannot explain the differences in antimicrobial activity between the two peptides. However, the existence of differential mobility does have important implications for structure determination based on NMR data. Methods for structure determination from NOESY cross-peak intensities typically assume the molecular motion is adequately described by a single rotational correlation time. Differential mobility demonstrates that multiple effective correlation times are needed to describe molecular motion for these peptides. The structure of PxB determined previously in aqueous TFE used NOESY cross-peak intensities to obtain distance restraints for a simulated annealing algorithm to determine compatible structures.6 Distance estimates were obtained by comparing the intensities of all NOESY cross peaks to the intensity observed for a known distance, the distance between the nonequivalent Hγ methylene protons of Dab 4. Since Dab 4 is relatively rigid compared to other parts of the molecule, the cross-peak intensity for this interaction will be larger than cross-peak intensities corresponding to similar distances in more mobile regions. Indeed this cross peak is large and positive even at 40 °C where most other cross peaks are negative or vanishingly small. Consequently, use of this intensity as an internal standard will underestimate distances for mobile regions if a single correlation time is assumed, as was done previously. This underscores the difficulties with using traditional approaches for structure determination to study partially folded polypeptide chains. Even a small, cyclic peptide with limited options for secondary structure cannot be adequately described by a single molecular correlation time. More accurate distance estimates can be obtained if intensity comparisons are limited to interactions in regions with similar mobility. Results of this study also suggest that the quality of structures determined from NOESY data can be improved in cases where the motional frequency is near the spectrometer frequency by reducing the temperature. For PxB and PxBn, chemical shifts of nonlabile protons are essentially invariant with temperature, as are the relative intensities of NOESY cross peaks, which indicates the overall structure does not change as the temperature is decreased. However, the decreased rate of motion increases the intensity of all NOESY cross peaks by the same factor. This has the effect of increasing the minimum interproton distance required for detection of cross peaks, providing distance estimates for longer distances. The increase in the number of restraints along with the ability to incorporate restraints for longer distances should result in better quality structures. In this case, the additional cross peaks observed at lower temperature could be used to refine the structure for PxB generated previously from 25 °C data.6 Conclusion The structure and dynamics of PxB and PxBn were characterized in aqueous TFE, a membrane-mimetic solvent, by NMR. Both peptides exhibit differential mobility, with greater mobility observed for the linear portion than the ring and faster motion seen for the side chains than the peptide backbone. All types of motion, including motion in the cyclic part of the peptide,

Structure and Dynamics of Polymyxin B are faster in the nonapeptide than for PxB, so the N-terminal tail, despite moving independently from the peptide, restricts the mobility of the ring. More importantly, certain residues in the ring are more rigid than others, which supports a previously proposed model whereby the conformation of the ring can change to facilitate binding to both divalent and monovalent phosphoester ligands.6 Although this ability to change conformation may be critically important for antibiotic activity, it is not sufficient for activity since both PxB and the inactive nonapeptide contain the same fixed and flexible regions within the ring. The hydrophobic tail on PxB, which is necessary for antibiotic activity, exhibits independent, rapid motion compared to the rest of the peptide. We propose this flexibility of the tail may help the peptide traverse the outer membrane and enter the periplasmic space. Acknowledgment. We are grateful to the University of Delaware and Montana State University for the use of their NMR spectrometers. Financial support for this work was obtained from the Scholarly and Creative Activities Committee at Oswego State University. References and Notes (1) Dunker, A.K.; Cortese, M. S.; Romero, P.; Iakoucheva, L. M.; Uversky, V. N. FEBS J. 2005, 272, 5129–5148. (2) Fink, A. L. Curr. Opin. Struct. Biol. 2005, 15, 35–41. (3) Dyson, H. J.; Wright, P. E. Nat. ReV. Mol. Cell Biol. 2005, 6, 197– 208. (4) Bemporad, F.; Gsponer, J.; Hopearuoho, H. I.; Plakoutsi, G.; Stati, G.; Stefani, M.; Taddei, N.; Vendruscolo, M.; Chiti, F EMBO J. 2008, 27, 1525–1535. (5) Lange, O. F.; Lakomek, N-A.; Fare`s, C.; Schro¨der, G. F.; Walter, K. F. A.; Becker, S.; Meiler, J.; Grubmu¨ller, H.; Griesinger, C.; de Groot, B. L. Science 2008, 320, 1471–1475.

J. Phys. Chem. B, Vol. 113, No. 2, 2009 551 (6) Bruch, M. D.; Cajal, Y.; Koh, J. T.; Jain, M. K. J. Am. Chem. Soc. 1999, 121, 11993–12004. (7) Storm, D. R.; Rosenthal, K. S.; Swanson, P. E. Annu. ReV. Biochem. 1977, 46, 723–763. (8) Oh, J.-T.; Van Dyk, T. K.; Cajal, Y.; Dhurjati, P. S.; Sasser, M.; Jain, M. K. Biochem. Biophys. Res. Commun. 1998, 246, 619–623. (9) Oh, J.-T.; Cajal, Y.; Dhurjati, P. S.; Van Dyk, T. K.; Jain, M. K. Biochim. Biophys. Acta 1998, 1415, 234–245. (10) Liechty, A.; Chen, J.; Jain, M. K. Biochim. Biophys. Acta 2000, 1463, 55–64. (11) Cajal, Y.; Jain, M. H. J. NIH Res. 1996, 8, 35–39. (12) Cajal, Y.; Ghanta, J.; Easwaran, K.; Surolia, A.; Jain, M. K. Biochemistry 1996, 35, 5684–5695. (13) Clausell, A.; Rabanal, F.; Garcia-Subirats, M.; Asuncion, Alsina, M.; Cajal, Y. J. Phys. Chem. 2006, 110, 4465–4471. (14) States, D. J.; Haberkorn, R. A.; Ruben, D. J. J. Magn. Reson. 1982, 48, 286–292. (15) Wishart, D. S.; Sykes, B. D.; Richards, F. M. J. Mol. Biol. 1991, 222, 311–333. (16) Wishart, D. S.; Sykes, B. D.; Richards, F. M. Biochemistry 1992, 31, 1647–1651. (17) Merutka, G.; Dyson, H. J.; Wright, P. E. J. Biomol. NMR 1995, 5, 14–24. (18) Keepers, J. W.; James, T. L. J. Magn. Reson. 1984, 57, 404–426. (19) Li, C.; Budge, L. P.; Driscoll, C. D.; Willardson, B. M.; Allman, G. W.; Savage, P. B. J. Am. Chem. Soc. 1999, 121, 931–940. (20) Li, C.; Lewis, M. R.; Gilbert, A. B.; Noel, M. D.; Scoville, D. H.; Allman, G. W.; Savage, P. B. Antimicrob. Agents Chemother. 1999, 43, 1347–1349. (21) Khandelwal, P.; Seth, S.; Hosur, R. Eur. J. Biochem. 1999, 264, 468–478. (22) Crandall, Y. M.; Bruch, M. D. Biopolymers 2008, 89, 197–209. (23) Pristovsek, P.; Kidric, J. J. Med. Chem. 1999, 42, 4604–4613. (24) Bhattacharjya, S.; David, S. A.; Mathan, V. I.; Balaram, P. Biopolymers 1997, 41, 251–265. (25) Cajal, Y.; Rogers, J.; Berg, O. G.; Jain, M. K. Biochemistry 1996, 35, 299–308. (26) Schro¨der, G.; Brandenburg, K.; Seydal, U. Biochemistry 1992, 31, 631–638.

JP808379X