2H-NMR structure analysis of membrane-bound peptides

ARTICLE IN PRESS

Flow charts for the systematic solid-state 19F/2H-NMR structure analysis of membrane-bound peptides Erik Strandberga,*, Anne S. Ulricha,b,* a

Karlsruhe Institute of Technology (KIT), Institute of Biological Interfaces (IBG-2), Karlsruhe, Germany KIT, Institute of Organic Chemistry, Karlsruhe, Germany *Corresponding authors: e-mail address: [email protected]; [email protected] b

Contents 1. Introduction 2. Flow chart steps 2.1 Preliminary work (Chart 1) 2.2 NMR experiments (Chart 2) 2.3 Further analysis (Chart 3) 2.4 Rotational diffusion analysis (Chart 4) 3. Choice of lipids 4. Concluding remarks Acknowledgements References

2 3 3 10 22 28 30 34 34 34

Abstract Solid-state NMR (SSNMR) is one of the most useful methods to investigate membranebound peptides and proteins, in order to elucidate the structural basis of their diverse biological functions. In the case of short peptides, SSNMR is now performed on a routine basis in several labs. By employing side-chain 19F- or 2H-isotope-labelled peptides in static oriented samples, SSNMR can provide information about conformation, alignment, dynamics, oligomerization, and aggregation behaviour. Changes in and transitions between these structural properties tend to be directly related to the functional mechanism of the peptides. Here, we present an easy-to-follow description of the methodological SSNMR approach in the form of flow charts. The successive steps needed, from peptide synthesis and functional tests, on to NMR sample preparation and NMR experiments, and further on to data analysis, are described. At each stage, questions to be answered and tasks to be performed are explained, and some representative results are illustrated. Relevant challenges and pitfalls are discussed, and possible reasons for problems and their possible solutions are considered.

Annual Reports on NMR Spectroscopy ISSN 0066-4103 https://doi.org/10.1016/bs.arnmr.2019.08.002

#

2019 Elsevier Ltd All rights reserved.

1

ARTICLE IN PRESS 2

Erik Strandberg and Anne S. Ulrich

Keywords: Membrane-bound peptides, Peptide structure and dynamics, Side-chain 19 F- or 2H-labelling, Solid-state NMR data analysis, Oriented membrane samples, Antimicrobial peptides, Flow charts

1. Introduction Many important biological functions are performed by membraneactive peptides, hence it is of great scientific and practical interest to understand their functional mechanisms in the context of their membrane-bound structures. Solid-state NMR (SSNMR) can provide detailed information about the 3D conformation, the membrane alignment, segmental and overall dynamics, oligomerization and/or aggregation characteristics of a peptide in a model membrane under quasi-native conditions, which can be composed of virtually any desired lipid composition. We have previously used SSNMR on a range of different antimicrobial peptides (AMPs) [1–12], cell-penetrating sequences (CPPs) [13–16] and fusogenic segments [17,18], to determine their conformation and orientation within macroscopically oriented membranes. All information was gathered from isotope labels that were selectively incorporated into the peptide side-chains. In particular, we have introduced a set of 19F-labelled amino acids as highly sensitive NMR reporter groups [7,8,10,19–24], and these results were verified on the basis of completely non-perturbing 2H-labels in the same systems. In either labelling approach, we have found that it is essential to include dynamics in the data analysis, in order to obtain the correct orientation of the peptide backbone; hence this SSNMR method also provides information about the mobility of the peptide in the lipid bilayer, which gives important clues on oligomerization and aggregation [25–30]. For illustrative purposes, we will focus here largely on the comparison of simple α-helical and regular β-stranded amphiphilic peptides, though the same principles and observations have been obtained for a wide range of other membrane-bound structures, such as kinked helices, cyclic and knotted peptides, and even highly flexible segments that are intrinsically disordered in the 2D plane of the lipid bilayer. Once the peptide orientation and dynamics have been determined in one system, the next step is to vary the experimental conditions, in order to identify important functional factors such as peptide concentration, membrane composition, lipid phase state, pH, hydration, etc. In several cases, like PGLa [4,7] and MSI-103 [2,3], we observed a concentration-dependent reorientation of these amphiphilic α-helices, suggesting an oligomerization step

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

3

and/or membrane insertion as part of their functional mechanism. It turned out that the ability of surface-bound amphipathic peptides to flip into an inserted state depends critically on the spontaneous curvature of the lipids, as found for PGLa, MAG2, and MSI-103 in a wide range of lipid systems [2,31]. We have also noted a temperature-dependent reorientation of peptides, where PGLa was found to stay on the membrane surface at higher temperatures but inserted into the membranes in the gel phase [32]. We and other groups have systematically studied the tilt angle of designated transmembrane helices (TMHs) as a function of hydrophobic mismatch [26,33,34], and examined the effect of flanking residues at each end of the hydrophobic stretch [35–38]. We also monitored the aggregation of flexible monomeric peptides into large immobilized assemblies, either well oriented on the membrane surface, such as KIGAKI [39], or as unordered powders [14], and how such aggregations can be modified by the introduction of D-amino acids into the peptides [14,39]. From these diverse examples, it should be clear that there is a plethora of information that can be gained from this kind of solid-state NMR experiment on peptides in membranes. In this review, we explain the different steps involved in a typical solid-state NMR investigation of side-chain isotope-labelled peptides in lipid bilayers, by presenting easy-to-use flow charts in which the different steps can be followed. The flow charts in Charts 1–4 give a general overview, and details are explained in the text, according to the following legend: (T1…): Task, to be performed experimentally; (Q1...): Question, to be answered as a basis for further progress; (A1…): Answer, as obtained from the structural behaviour of the peptide; and (E1…): Example, as highlighted by our (published) experiences. The flow charts are equally valid both for CF3 and CD3 labels, but to keep the text in the boxes consistent and to cover any special precautions of 19F-NMR, only the CF3 scenario is shown in the flow chart. The explaining text also covers the CD3 case.

2. Flow chart steps 2.1 Preliminary work (Chart 1) (Q1) Are there at least n
4 positions in which the peptide can be non-perturbingly CF3-labelled? In the present approach, the behaviour of the peptide backbone is characterized from the anisotropic dipolar (19F) or quadrupolar (2H) couplings of

ARTICLE IN PRESS 4

Erik Strandberg and Anne S. Ulrich

START

More labels needed

(Q1) Are there at least n≥4 positions in which the peptide can be non-perturbingly CF3-labelled?

no

yes 19

(T1) Synthesize n F-labelled analogues (T2) Purify peptides in the absence of TFA/TFE (T3) Record CD spectra in detergents/vesicles

(Q2) Are the CD lineshapes of at least 4 analogues comparable with that of the wild type?

no

yes (T4) Perform functional assays, e.g. MIC. cell uptake, toxicity, or vesicle fusion

(Q3) Are the functional activities of at least 4 analogues comparable with that of the wild type?

no

yes

Go to NMR measurements (Chart 2)

LEGEND Question

Go to next box

Task

Answer (result)

Successful end of project

Go to next chart

Unsuccessful end of project

Chart 1 Work that needs to be done before NMR experiments can start. (TFA— trifluoroacetic acid; TFE, trifluoroethanol; CD, circular dichroism; MIC, minimal inhibitory concentration).

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

5

START

no

Maybe 2 populations. Go to T5 and try different P/L

Try other solvents, lower P/L, other lipids

(T5) Reconstitute peptides in oriented samples (T6) Measure 31P-NMR in horizontal alignment

(Q4) Do the 31P-NMR spectra show a narrow downfield signal that confirms well-oriented samples? yes

(T7) Measure 19F-NMR in horizontal alignment

no

(Q4) Do the 19F-NMR spectra show a narrow triplet that confirm a well-aligned peptide? yes (Q6) Is there only one triplet?

no

no yes

Go to T1/T2 in Chart 1 Check purity of peptides (may have racemized) Check TFA/TFE contamination

(T8) Perform RMSD analysis to determine τ, ρ, and dynamics (T9) Display 19F-splittings on a dipolar wave plot

(Q7) Is there a unique solution with a low RMSD for any regular secondary structure element?

no

Go to further analysis (Chart 3)

yes (A1) Peptide conformation (A2) Alignment in bilayer (τ and ρ)

Go to dynamical analysis (Chart 4)

Chart 2 NMR experiments and data analysis.

CF3 or CD3 groups that are selectively incorporated into specific structurally constrained amino acids, one by one. To provide meaningful information, any such label may be placed only at a position where it does not disturb the structure or function of the membrane-bound peptide. Ala-d3 can replace

ARTICLE IN PRESS 6

Erik Strandberg and Anne S. Ulrich

START

(Q8) Is there more than one solution?

no

(Q9) Is there no good solution (with sufficiently low RMSD)? yes

yes More labels needed. Go to Q1 in Diagram 1.

(Q10) Do all labels give the same splitting ?

no yes

no

(Q11) Does a window over fewer consecutive labels (n≥4) give a solution with low RMSD? yes

(A3) Frayed N/C terminus (A4) Peptide helix with kink (E3) KIA14, TP10 (E4) B18, alamethicin

(Q12A) Splittings ≈ 0 kHz?

yes

(A5) Unbound peptides (E5) SSL-25

no

(Q12B) Maximum splittings ≈ +15 kHz?

yes

(E6) KIGAKI at high conc.

no

(Q12C) Splittings ≈ +8 kHz?

(A6) Oriented β-sheet assemblies

yes

(A7) Disordered peptides (E7) KIGAKI at low conc., TP10

Go to dynamical analysis (Chart 4)

no

(Q12D) Splittings ≈ -7 kHz?

no

Go to dynamical analysis (Chart 4) yes (A8) Immobilized aggregates

Chart 3 Further data analysis.

Ala without any disturbance, while other amino acids replacements would need to be considered as Ala-mutants. CF3-labelled amino acids are intrinsically non-natural, therefore great care has to be taken that they do not induce any significant local or global perturbation. Note that for each labelled position a separate peptide has to be synthesized and characterized.

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

7

START (T10) measure NMR in vertical alignment

(Q13) Do the CF3 splittings scale by a factor of -0.5?

yes

(A9) Rotational diffusion of peptides, high dynamics (E9) PGLa, BP100, GrS

no (A10) Immobilized peptides; no rotational diffusion (E10) MSI-103

(Q14) Are the splittings around -7 kHz, the same as for the horizontal alignment?

yes

(A11) Powder; the peptide is aggregated (E11) MAP, TP10 (under some conditions)

no

END Chart 4 Rotational diffusion analysis.

Every peptide with a single mutation compared to the wild-type sequence must of course be tested to check for disturbances. When deciding on the positions to be labelled, however, an educated guess can be made about which positions can be substituted—one by one—without significant disturbance, based on our available arsenal of designated 19F-labelled amino acids. For bulky hydrophobic amino acids such as Val, Leu and Ile, we found that CF3-Phg [3,8] and CF3-Bpg [13,23,39,40] usually do not disturb the peptide. The (R)- or (S)-stereoisomers of CF3-Ala have also successfully been used to replace the non-canonical amino acid α-aminoisobutyric acid (Aib) in the peptaibols alamethicin [1] and harzianin KH-VI [12], and in these cases, both stereoisomers can give useful constraints [1,12]. Several CF3-Pro analogues have been developed that can be used in place of Pro [41–43]. Additionally, for aromatic Phe, a suitable CF3 analogue is available [44,45]. To replace the polar H-bonding residues Ser and Thr, other

ARTICLE IN PRESS 8

Erik Strandberg and Anne S. Ulrich

CF3-containing amino acids have been developed [46]. Charged amino acids are usually essential to peptide function and should not be replaced by uncharged labels. With the recent development of CF3-containing amino acids that are suitable for replacing cationic Lys [47] and Arg residues [48], it is now possible to label almost all classes of amino acids with CF3 groups. Efforts are currently underway to design 19F-labelled amino acids with anionic side chains for structure analysis. To characterize any properly folded secondary structure element, a minimum of n ¼ 4 labels are needed to obtain a meaningful fit of the NMR data, and more labels are helpful to obtain a more reliable fit of the data. If possible, six or more positions should be used, but good results have also been obtained with as few as four labels for an α-helical segment [8,28,34]. It is possible to combine the structural information from different types of 19F-labelled amino acids on the same peptide, provided that the respective orientations of the CF3 groups are taken into account. For example, harzianin KH-VI was labelled at four positions with CF3-Bpg (replacing Leu and Ile) and at three positions with CF3-Ala (replacing Aib), and the data analysis included all data points [12]. We have often used the antimicrobial peptide PGLa as a test case for new amino acids [sequence: GMASKAGAIAGKIAKVALKAL-NH2], so it has been specifically labelled with Ala-d3 in 11 separate positions [4–6,49], with CF3-Phg in five positions [7,8], with CF3-Bpg in four positions [23,24], and with a CF3-containing Lys analogue in three positions [47]. By combining such labels, it was possible to observe almost every position along the sequence of PGLa and verify its membrane-bound structure with high accuracy [47]. (T1) Synthesize n 19F-labelled analogues. The set of n specifically labelled peptides is usually synthesized using standard solid-phase peptide synthesis (SPPS) methods [50]. Most of the sequence can be coupled on an automated peptide synthesizer, but it is recommended to couple the 19F- or 2H-labelled amino acids manually, using only a threefold excess of amino acid to save material. Only a single label should be placed onto each peptide, hence multiple syntheses are needed to obtain a sufficient number (n
4) of data points. In some reports, two or three Ala-d3 labels have been placed on the same peptide with different degrees of labelling (for example, 100% at one position, and 70% at another), so that splittings could be assigned according to intensity [37]. This only works well, however, if the splittings are significantly different, i.e., if the labels are placed essentially orthogonally onto the helix. (T2) Purify peptides in the absence of TFA/TFE.

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

9

Trifluoroacetic acid (TFA) and trifluoroethanol (TFE) are often used additives or solvents to purify peptides with HPLC, or they are used in sample preparation. However, for 19F-NMR, these and all other 19F-containing compounds must be strictly avoided. TFA and TFE bind avidly to peptides and give strong 19F-NMR signals, which can be more prominent than, and often overlap with, the peptide signals that should be measured. We tend to obtain good 19F-NMR results with HCl in place of TFA in the HPLC purification step [22]. (T3) Record CD spectra in detergents/vesicles. Circular dichroism spectroscopy (CD) is a sensitive method for determining the secondary structure of peptides. To ascertain that the isotope-labelled amino acid does not lead to a change in structure, CD experiments should be performed in the presence of vesicles made from the same lipids as will be used in the NMR experiments, to be compared with the CD results of the wild type (WT; unlabelled) peptide. (Q2) Are the CD line shapes of at least four analogues comparable with the wild type? If some of the labelled peptides give a different CD spectrum than the WT, i.e., when the labelling has affected the structure, this analogue should not be used in the NMR analysis. In case there remain less than n ¼ 4 labelled peptides with essentially the same CD spectrum as the WT, further labelled peptides must be prepared and tested. If this is not possible, the project has to be terminated— though we have not encountered such situation yet, given the wide range of 19 F-labels and/or non-perturbed Ala-d3-mutants that tend to be available. (T4) Perform functional assays, e.g., MIC, cell uptake, toxicity, or vesicle fusion. Even if the structure (as seen by CD) of a labelled analogue is the same as that of the WT, the analogue should also be tested in order to ascertain that the labelling has not disrupted its membrane interactions and/or biological function. Depending on the type of peptide, different assays can be applied, for example on antimicrobial function, cell uptake, toxicity, vesicle leakage, fusogenic activity, etc. (Q3) Are the functional activities of at least four analogues comparable with the wild type? Only labelled peptides with similar activity as the WT should be used in the NMR analysis, hence at least n ¼ 4 such analogues are needed; otherwise, further labelled peptides should be prepared and tested. If the answer to Q3 is yes, the project can proceed to the NMR experiments described in Chart 2.

ARTICLE IN PRESS 10

Erik Strandberg and Anne S. Ulrich

2.2 NMR experiments (Chart 2) (T5) Reconstitute peptides in oriented samples. To calculate the orientation of a peptide in a lipid bilayer from anisotropic 19 F- or 2H-NMR parameters, the measurements are best performed using macroscopically oriented membranes samples, because especially the CF3 triplets tend to be sharp and show a unique sign. We prepare the peptide-lipid bilayers on thin glass plates [19,51,52], as virtually any desired lipid composition can be used this way. It is also possible to obtain oriented peptide samples in bicelles [53–56], but there are two caveats: (a) the bicelle composition cannot usually be varied freely due to limited stability, and (b) amphiphilic peptides do not find a characteristic flat bilayer surface but will sense the strong curvature at the rim, which may lead to uncharacteristic behaviour. To prepare oriented samples on glass plates, peptides and lipids are codissolved in organic solvents and spread onto the plates. The plates are then dried under vacuum to remove the solvents and subsequently hydrated in a chamber with a controlled relative humidity. We have often obtained good results with mixtures of CHCl3 and MeOH, sometimes with the addition of small amounts of H2O, but the solvents for any specific peptide-lipid system may need to be optimized by trial and error. To avoid contamination of the NMR spectrum, TFE and other 19F-containing solvents should be strictly avoided when studying 19 F-labelled peptides [22]. The choice of lipids is an important topic in this kind of study and will be further discussed in Section 3. (T6) Measure 31P-NMR in horizontal alignment. The degree of orientation in the NMR samples should be as high as possible. The easiest way to monitor the orientation of the phospholipid bilayers is by performing 31P-NMR experiments. The phosphate in the lipid head groups gives a strong NMR signal, so it is usually possible to obtain a high-quality spectrum within a few minutes, provided that a flat-coil 31P-NMR probe head is available to accommodate the oriented sample. In the usual “horizontal” alignment of the glass plates (i.e. where the membrane normal is parallel to the static field direction of the NMR magnet), a narrow 31P-NMR signal at approximately 25–30 ppm is a sign of well-oriented lipids. Poorly oriented lipids give a broad signal, or even a powder line shape with an upfield peak at around 15 ppm [57]. (Q4) Do the 31P-NMR spectra show a narrow downfield signal that confirms well-oriented samples? It is important not only to find a narrow 31P-NMR signal from oriented lipids, but also to have as few non-oriented membrane regions in the sample as possible. The corresponding disordered powder component can usually

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

11

be estimated by integrating the two NMR signals. If more than 80% of the lipids are well oriented, the sample is usually good enough. To obtain good orientation, as little sample as needed should be deposited on each glass plate. We have found that a rule of thumb is to have no more than 7 μg/mm2 on the glass plates (i.e. 1 mg dry material on each glass plates with a size of 18 mm 7.5 mm) to obtain well-oriented bilayers. The quality of alignment, however, also depends strongly on the lipids and peptides used. If the orientation as judged from the 31P-NMR signal is bad, several steps can be tried to improve it. One possibility is to use different solvents in the sample preparation. A lower peptide-to-lipid ratio (P/L) usually leads to better orientation, but on the other hand several P/L values are often needed to investigate the concentration-dependent peptide self-assembly. Another lipid system can also yield improved orientation, though it is hard to predict which lipid-peptide combination will work best; furthermore, lipid properties can also have dramatic effects on peptide orientation and self-assembly (see Section 3). Notably, 31P-NMR gives information only about the lipids, and often the peptides will still misbehave. Good lipid orientation is not a guarantee of good peptide orientation. In other cases, however, good peptide signals can be obtained even when only 50% of the lipids are well-oriented (especially in cases where the P/L is high). Therefore, even if 31P-NMR shows a bad orientation, it can still be worth running some exploratory 19 F- or 2H-NMR experiments on the sample. Another type of information available from the 31P-NMR spectrum concerns the phase of the lipids. Usually, a lamellar liquid crystalline phase is desired, as it is most relevant for biological membranes. 31P-NMR is well suited to indicate which phase or phases the lipids assume [57,58]. If there are other lipid phases present (for example, isotropic or hexagonal phases), this indicates that the peptide has disturbed the lamellar phase very significantly. Such observation may be interesting per se, but these samples are unsuitable for further structural analysis. One reason for non-lamellar phase formation can be an excessively high P/L (>1:20). If non-lamellar phases are found under such conditions, it may well be worth going to lower peptide concentrations, given that 19F-NMR is capable of picking up signals for P/L ratios down to 1:3000 [59]. (T7) Measure 19F-NMR in horizontal alignment Once well-oriented samples have been obtained, 19F- or 2H-NMR measurements should be performed in a horizontal alignment of the sample. To get undistorted 19F-NMR spectra of CF3 groups, a simple one-pulse

ARTICLE IN PRESS 12

Erik Strandberg and Anne S. Ulrich

experiment works best [52], and we often implement an anti-ringing pulse sequence [60]. For 2H-NMR, a solid echo sequence should be used [61]. (Q5) Do the 19F-NMR spectra show narrow triplets that confirm a well-aligned peptide? A CF3 group rotates fast around the C–CF3 bond, averaging all 19F-NMR interactions so that all three fluorine nuclei give the same signal. However, this signal is split into a triplet because of 19F–19F dipolar coupling. Each coupling splits the signal in two, and since each 19F nucleus senses two other, identical nuclei, the two couplings give rise to a 1:2:1 triplet in the spectrum. Both the dipolar coupling ΔdFF, seen in the spectrum as the distance between two adjacent peaks in the triplet, as well as the chemical shift CSF of the central peak, depend in the same way on the angle θ between the C–CF3 bond vector and the magnetic field direction, i.e., as a function of the 2nd Legendre polynomial (3cos2θ–1). This simple relationship makes it possible to determine not just the magnitude but also the sign of the dipolar coupling (which is not usually available in NMR), given that the triplet is shifted either downfield (positive sign) or upfield (negative sign) from the isotropic position [8]. The local orientational angle of each 19F-label is thus derived from the accurate CF3 dipolar splitting, using the equation ΔdFF ¼ 17 kHz  (3cos2θ-1) [8]. By combining the individual angles θ from several measured 19F labels, it is possible to calculate the orientation of the peptide unit in the bilayer [8], provided that the backbone is folded in a known conformation. The same information of peptide alignment can in principle be obtained also from the 19F chemical shift, but in practice, it is much easier to read out the dipolar coupling with high accuracy, given that exact spectral referencing of oriented samples is difficult [8,19,59]. To obtain the needed orientational constraints, the triplets must be well resolved. If the lines are too broad, another sample preparation method may have to be tested. In the case of Ala-d3 labels, the oriented 2H-NMR spectrum contains a doublet from the CD3-group due to the quadrupolar interaction. The quadrupole splitting Δνq reveals the orientational information according to Δνq ¼ 84 kHz  (3cos2θ – 1), though two indiscriminate solutions are possible due to the unknown sign of Δνq. A well-resolved splitting is needed, which again requires a well-aligned sample with little inhomogeneity. (Q6) Is there only one triplet? If there is more than one CF3-triplet of CD3-doublet observed in the spectrum, this leads to problems with the data interpretation and analysis. Likely reasons for multiple CF3-triplets can be (i) TFA or TFE contamination in

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

13

the sample; (ii) badly purified peptides; or (iii) racemization of peptides. Case (i) can be recognized if all samples have a triplet with the same splitting, which is usually very intense with narrow lines. Peptides should be purified without TFA/TFE present, and ideally never be in contact with these compounds at any stage (e.g. a lyophilizer should be used where no other TFA/TFE-containing samples are dried, and a designated set of pipettes should be used for fluorine-free organic solvents). To avoid problem (ii), peptides should be purified to give a single HPLC peak and a single peak in a mass spectrum. Regarding problem (iii), specifically CF3-Phg is well known to racemize avidly under basic conditions [8]—for this reason, CF3-Bpg is a better choice of label [23,24], which racemizes only rarely in certain cases [62]. If there appears one dominant triplet and another less intense triplet consistently throughout all samples (and if their splittings vary between the labelled positions), it may be possible to analyse the two sets of peaks independently in terms of two different populations of peptide. However, it will be more interesting to vary the sample composition (peptide concentration, lipid type, pH, etc.) or experimental conditions (temperature), until single sets of triplets are obtained under distinctly different conditions, in order to describe the structural response of the overall peptide-lipid system. For Ala-d3 labels, problems (i) and (iii) do not occur. However, the natural abundance of 2H can result in background signals. In particular, a signal at the isotropic position tends to arise from deuterium in the water, and at low P/L there can be a signal from deuterium in the lipids with a characteristic splitting of approximately 60 kHz in oriented samples [18]. It is thus recommended to hydrate the samples with deuterium-depleted water to minimize background signals. (T8) Perform RMSD analysis to determine τ, ρ, and dynamics When at least four well-defined 19F-NMR triplets or 2H-NMR splittings have been obtained, they can be analysed in terms of the peptide orientation in the membrane, as it has been described in detail elsewhere [8,34]. In short, to determine the molecular orientation relative to the membrane normal, the peptide structure must be known, i.e., whether it forms a regular α-helix or any other well-defined 3D fold. Additionally, the exact orientation of the 19F- or 2Hlabel relative to the peptide backbone must be known, which in Ala-d3 and some of the 19F-labelled amino acids (CF3-Phg, CF3-Bpg) is aligned with the Cα-Cβ bond vector. Then, the peptide orientation can be calculated from at least 4 data points (ΔdFF or Δνq) in terms of a tilt angle τ and an azimuthal angle ρ, as defined and illustrated in Fig. 1 for the case of an α-helix.

ARTICLE IN PRESS 14

Erik Strandberg and Anne S. Ulrich

A

B τ = 93o

ρ = 135o

Fig. 1 Illustration of tilt angle (τ) and azimuthal rotation angle (ρ). As an example, we show the orientation found from 2H-NMR data for the α-helical peptide MSI-103 [KIAGKIAKIAGKIAKIAGKIA] in a POPE/POPG (3/1) membrane [2]. The amphiphilic peptide is shown in dark blue (6 Lys residues are seen to reach upwards), and the yellow box represents the lipid membrane. (A) The tilt angle is defined as the angle between the membrane normal and the helix axis (going from the N- to the C-terminal). (B) The azimuthal angle describes the rotation of the peptide around the helix axis, where we define ρ ¼ 0° as the starting point when a vector from the helix axis to the α carbon of Lys-12 (marked with a white circle) lies parallel to the membrane surface.

The whole-body motions of the peptide backbone in the fluid lipid bilayer must be taken into account in the NMR data analysis. We have examined the influence of the type of dynamic model applied to the data analysis and found that excessively simplified models can lead to false structural results [27–30,63]. We therefore recommend to use a dynamical model in which the range of fluctuations in τ and ρ are explicitly included as standard deviations στ and σρ [28,29]. In this case, at least five data points are needed in order to obtain a unique solution to the data fit. Only for peptides that are moderately mobile, a simpler model is appropriate, where the dynamics are described by an order parameter Smol [2,6,8,28]. In that case Smol works as a scaling factor with a value between 0 and 1, where 0 means complete isotropic motion, which averages all splittings to zero, while 1 means that the peptides have a fixed orientation in the membrane [8]. (T9) Display 19F-splittings on a dipolar wave plot. The RMSD analysis will always give a nominal best-fit solution. However, a reliable result should not have a larger root mean square deviation (RMSD) between experimental and calculated splittings than approximately 1 kHz for CF3 data, or 4 kHz for CD3 data. It is often possible to obtain an RMSD of 0.5 kHz for CF3 data, or 2 kHz for CD3 data. To assess the quality of the fit, the theoretical best-fit splittings should be compared with the experimental splittings, both in a dipolar/quadrupolar wave plot, as well as in a plot of RMSD as a function of τ and ρ. This way, single outliers can be identified

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

15

and remeasured, or possibly discarded if there is some indication that the label in this particular position has affected the structure and/or function of the peptide. Examples of such plots for the case of 19F-NMR data are illustrated in Figs. 2 and 3, yielding the orientation of 19F-labelled MSI-103 peptides in oriented bilayers of DMPC at P/L ¼ 1/200 [2]. Fig. 2A shows an RMSD plot, where the RMSD was calculated for each combination of τ and ρ and colour coded according to its value. A darker colour indicates a better fit, while white means that the RMSD value is larger than 3.8 kHz. This plot shows that there is only one minimum, i.e., only a single combination of τ/ρ values gives a good fit. It is also clear that only a narrow range of τ and ρ values around this best-fit pair can give reasonable RMSD values. Fig. 2B shows the experimental data points and the calculated helical wave curve corresponding to the best-fit solution for τ and ρ. In this figure, the dipolar couplings are plotted as a function of the amino acid number along the peptide sequence, and the data points are shown at the positions corresponding to the labelled residues assuming an ideal α-helix. The fit is seen to be good, because all points are close to the curve. Fig. 3 shows the same experimental data and helical wave analysis as in Fig. 2, but in this case, the positions of labelled residues are projected onto the 360° circumference of the helix, similar to a helical wheel projection. Because of the expanded horizontal scale, it is easier to check the distance from the data points to the curve. Fig. 3A shows the same best-fit curve as Fig. 2B, illustrating the different type of projection. The numbers next to the data points indicate the residue number. Fig. 3B shows the same best-fit curve (solid line) and some theoretical curves for different values of the tilt angle τ, in order to illustrate how the curve depends on this parameter. Dipolar waves are shown for tilt angles of 84°, 94° and 104°, keeping the azimuthal angle fixed at the best-fit value of 126°. The shape of the curves changes with τ, and it can be noted that the curves are so sensitive to the value of τ, that a change in τ by a few degrees is clearly visible. For example, close to the data point from position 13 (approximately 100° at the horizontal axis), the theoretical splitting changes approximately 3 kHz for a change in the tilt angle of 10°. In Fig. 3C, the same curve is illustrated for different values of ρ, using a fixed tilt angle of 94°. Here, curves are shown for ρ values of 116°, 126° and 136°. The shape of the curves does not change with ρ, but the curve is shifted along the horizontal axis. At locations where the curve has a steep

ARTICLE IN PRESS 16

Erik Strandberg and Anne S. Ulrich

180 RMSD / kHz 0.00 - 1.80 1.80 - 2.00 2.00 - 2.30 2.30 - 2.80 2.80 - 3.80 3.80 -

A 160 140

100

τ/

o

120

80 60 40 20 0 0

20

40

60

80

100 120 140 160 180

ρ/

o

15 B

Dipolar coupling / kHz

10

5

0

-5

-10

-15 6

7

8

9

10

11

12

13

14

Residue number Fig. 2 RMSD plot and dipolar wave plot, using data from four 19F-labelled MSI-103 peptides in oriented DMPC bilayers at P/L ¼ 1/200 [2] (fitting was done using the simplified Smol model to account for dynamics). (A) RMSD plot. The RMSD value is calculated for all combinations of τ and ρ, and the value at each point is colour coded. The darkest point is the best fit, in this case τ ¼ 94° and ρ ¼ 126°. (B) Dipolar wave plot where the 19F-19F dipolar coupling is shown as a function of the amino acid number along the peptide sequence. The four data points were collected for CF3-Bpg placed one-by-one at positions Ala-7, Ile-9, Ala-10 and Ile-13. The solid line indicates the best-fit curve for these data.

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

A

17

B

15

15 10

Dipolar coupling / kHz

Dipolar coupling / kHz

10 13

5

7 9

0

10

–5 –10 –15

5 0 –5

Tilt angle 84o 94o 104o

–10 –15

0

100

200

300

0

Position around helix / degrees

C

D

15

200

300

15 10

5 0 Azimuthal angle 116o 126o 136o

–5 –10 –15

Dipolar coupling / kHz

10 Dipolar coupling / kHz

100

Position around helix / degrees

5 0 Order parameter 0.6 0.7 0.8

–5 –10 –15

0

100

200

Position around helix / degrees

300

0

100

200

300

Position around helix / degrees

Fig. 3 (A) Dipolar wave plots, where the dipolar coupling is now shown in a different manner than in Fig. 2B, namely as a function of the angular position around the helical wheel, where the data points are labelled according to the residue number. Compared with Fig. 2, all data points are projected onto a single turn of the helix. (B) Dipolar wave plots for different values of the tilt angle, with the azimuthal angle constant at ρ ¼ 126°, to illustrate the accuracy of the curve fit. The shape of the curves changes with τ, and it can be observed that the two peaks (around 100° and 280°) respond in an opposite manner to modulation of τ. The curves respond very sensitively, hence a change of τ by a few degrees is clearly visible from the data. (C) Similar to (B), but with curves shown for different values of the azimuthal angle ρ, with a fixed tilt value of τ ¼ 94°. Upon variation of ρ, the shape of the curve remains the same, but it is shifted along the horizontal axis. A shift in ρ of a few degrees can thus be reflected by large changes in the splittings. (D) Similar to (B), but curves are shown for different values of the order parameter Smol. The shape and position of the curve are the same in all cases, but the amplitude scales with the value of the order parameter.

slope, such as positions 7 or 9, the theoretical dipolar couplings can change by approximately 5 kHz for a change in ρ of 10°, whereas at some other positions, such as position 13 in this example, the resulting change is smaller.

ARTICLE IN PRESS 18

Erik Strandberg and Anne S. Ulrich

Fig. 3D shows the effect of the order parameter Smol, which is used to account for peptide mobility in the simplified dynamical data analysis. The shape and position of the curve are the same in all cases, but the amplitude scales with the order parameter. An order parameter of 1.0 gives the maximum splittings and corresponds to a completely immobile peptide. An order parameter of Smol ¼ 0 would correspond to a peptide undergoing fast isotropic motion, in which case all splittings would be scaled to zero. This is what happens with free peptides in solution, where the fast isotropic tumbling averages out the dipolar couplings and would make this orientational analysis meaningless. For peptides in membranes, however, typical order parameter values between 0.6 and 1.0 have been reported [2–4,6,7,40,49,64]. Small peptides are more mobile; for example, the 11-mer harzianin HK-VI in DDPC lipids was found to have an order parameter of approximately 0.35 [12]. In this example of the 21-mer MSI-103, the best-fit value of the order parameter is 0.7, which is used in all the helical curves of Figs. 2B and 3A–C. The same peptide MSI-103 is used in Fig. 4 as an example to illustrate the analysis of 2H-NMR data. In this case, however, the orientation derived from seven 2H-labelled analogues is calculated for POPE/POPG (3/1) membranes at a P/L ¼ 1/50 [2]. Fig. 4A shows an RMSD plot, which is analogous to that in Fig. 2A. The RMSD is calculated for each combination of τ and ρ, and colour coded according to its value. This plot shows that there is only one minimum, i.e., only a single set of τ/ρ values gives a good fit. It is also clear that only a narrow range of τ and ρ values around this best-fit solution can give a reasonable RMSD. (This range is smaller than in Fig. 2A, partly because there are more data points, hence the orientation is better defined, and partly because of the different RMSD scales resulting from the 19F dipolar and 2 H quadrupolar splittings). Fig. 4B shows the experimental data points and the calculated helical wave curve corresponding to the best-fit values of τ and ρ. In this figure, the quadrupolar splittings are plotted as a function of the residue number of the label, and the data points correspond to the labelled residues in an ideal α-helix. Since the sign of the splitting is not known, absolute values are shown for both, the data points and the theoretical curves. The fit is good, as all data points are close to the curve. Fig. 4C and D also show the experimental data and the helical wave in a helical wheel projection. It can be seen that the curve is not a perfect fit to the data points, which is consistent with the best-fit RMSD obtained here of at least 3.3 kHz.

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

B

180

RMSD / kHz 0.00 - 3.30 3.30 - 3.50 3.50 - 3.80 3.80 - 4.30 4.30 - 5.30 5.30 -

160 140

τ /o

120

x x x x x

100 80 στ /o

40

60

20

40

80 Quadrupolar splitting / kHz

A

19

60

40

20

0 0

20

20

σρ /

40 o

0

0 0

20

40

60

C

7

8

9

10

11 12 13 14 Residue number

15

16

17

18

D Tilt angle 90o 100o o

110

60

o

93

11 40 13

10

9

20

Azimuthal angle

80 Quadrupolar splitting / kHz

80 Quadrupolar splitting / kHz

6

80 100 120 140 160 180 ρ /o

120o 150o 60

135o 11

40 13 20

7 17

10

7

9

14

17

0

14

0 0

100 200 Position around helix / degrees

300

0

100 200 Position around helix / degrees

300

Fig. 4 RMSD plot and dipolar wave plots, using data from seven 2H-labelled MSI-103 peptides in POPE/POPG (3/1) oriented bilayers [2]. Compared to the 19F-NMR dipolar splittings in Figs. 2 and 3, the quadrupolar data can only be analysed in terms of absolute values, since positive and negative splittings cannot be discriminated in the 2H-NMR spectrum. (A) RMSD plot, calculated for all combinations of τ and ρ, with a colour code to indicate the deviation of the fit relative to the experimental data. The darkest point corresponds to the best fit parameters, in this case τ ¼ 93° and ρ ¼ 135°. The crosses correspond to the τ/ρ pairs shown in panels (C) and (D). In this example, the explicit dynamical model was applied to the data analysis, hence στ and σρ were included in the fit (inset in panel A), to obtain information on the fluctuations of the peptide backbone around the axes defining τ and ρ. (B) Dipolar wave plot, in which the quadrupolar splitting is shown as a function of the amino acid number along the peptide sequence as in Fig. 2B. (C) Dipolar wave plot showing the quadrupolar splitting as a function of position around the helical wheel as in Fig. 3. The numbers adjacent to each data point correspond to the residue numbers that were labelled with Ala-d3. Curves are shown for different values of the tilt angle, with a constant azimuthal angle of ρ ¼ 135°. The shape of the curves changes with τ, and it can be observed that the two peaks (around 100° and 280°) respond in an opposite manner to modulation of τ. The curves are sensitive enough that a change in τ by a few degrees is clearly visible. (D) Similar to (C), but with curves shown for different values of the azimuthal angle ρ, with a fixed tilt angle value of τ ¼ 93°. The curve retains the same shape, but it moves along the horizontal axis. A shift in ρ of a few degrees can correspond to large changes in the splittings.

ARTICLE IN PRESS 20

Erik Strandberg and Anne S. Ulrich

Fig. 4C shows the best-fit curve (solid line) and some theoretical curves for different values of the tilt angle τ, in order to illustrate how the curve depends on this parameter. Quadrupolar waves are shown for tilt angles of 90°, 100° and 110°, keeping the azimuthal angle fixed at the best-fit value of 135°. The shape of the curves changes with τ, and it can be noted that the curves are so sensitive to the value of τ, that a change in τ by a few degrees is clearly visible. For example, at approximately 100° on the horizontal axis, the theoretical splitting changes by approximately 20kHz for a change in tilt of 10°. In Fig. 4D, the same curve is illustrated for different ρ values of 120°, 135° and 150°, using a fixed tilt angle of 93°. The shape of the curves does not change with ρ, but the curve gets shifted along the horizontal axis. At locations where the curve has a steep slope, such as position 9, the theoretical quadrupolar splittings can change by approximately 1 kHz for a change in ρ of 1°, whereas at some other positions, such as position 11 in the present example, the resulting change is much smaller. From Figs. 3 and 4, it is clear that the solid-state NMR approach is quite sensitive to peptide orientation, and that it is able to resolve the nominal τ and ρ values to within 2–5°. However, the calculated curves always have to be based on an underlying 3D structure of the peptide. When dealing, e.g., with α-helical peptides, certain conformational parameters must be used to model the helix structure (usually “ideal” helices are assumed), which may not reflect the true structure of the peptide. This issue has been discussed elsewhere [7,33,34] and leads to a somewhat larger uncertainty in the actual peptide orientation. However, for any particular peptide like MSI-103, the use of a consistent structural model will reveal with very high accuracy the relative changes in helix orientation, which occur when experimental conditions are varied such as membrane thickness, temperature or lipid composition. For the analysis of the 2H-NMR data, 7 labelled positions were used, hence there were enough data points to implement the more advanced dynamical model. This model contains two dynamical parameters, στ and σρ, which describe the Gaussian width of the fluctuations of the τ and ρ angles, respectively. More details about this model are available elsewhere [28,29]. The inset in Fig. 4A is an RMSD plot for these dynamical parameters στ and σρ, with the same colour coding as in the main figure. In fact, in this case the full RMSD analysis is a function of four variables and therefore covers a four-dimensional space, but for illustrative purposes is shown as two 2D plots. In the τ-ρ plot, the RMSD values are given as a function of only τ and ρ, with the (overall optimized) values of στ and σρ

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

21

fixed at their best-fit values. Similarly, in the inset, the RMSD values are shown as a function of only στ and σρ, with τ and ρ fixed at their overall best-fit values. In this example, the best-fit values are στ ¼ 0 and σρ ¼ 28°. As seen in the inset, σρ is better defined than στ. How the helical waves depend on the dynamics, and in particular on στ and σρ, has been described in detail elsewhere [28,29]. Most of the amphiphilic antimicrobial peptides that have been studied in depth, were found to have a relatively low mobility, with values of στ and σρ in the 0–30° range [2,65,66]. Long-axial rotation around the membrane normal is obviously dependent primarily on peptide concentration and surface coverage. Yet, in terms of rigid-body fluctuations it was consistently found that shorter peptides tend to wobble more than longer ones. The 11-mer peptide BP100 [67] was found to be more mobile than the 14-mer KIA14 peptide [66], which in turn is more dynamic than the 21-mer MSI-103 peptide [2]. Some rather interesting results regarding peptide dynamics were observed for transmembrane model helices of the WALP-type. These peptides are completely hydrophobic with a central Ala-Leu stretch, carrying one or more Trp or other aromatic residues as “anchors” at each terminus [38,68–70]. It had been initially expected that these peptides would be oriented completely upright in the membrane, and that they would only rotate rapidly around their helix axis, which would then coincide with the membrane normal. In such a scenario, the azimuthal angle ρ has no meaning, hence the 2H- or 19F-splittings from different labelled positions should all give the same value. However, this expectation was found to not be correct, as the WALP-type peptides were found to have a small but non-zero tilt angle within the membrane and a preferential mean value of ρ [33,34]. Depending on the exact nature, position and number of the aromatic anchoring residues, the fluctuations in ρ could be dramatically different. Some peptides have σρ values >100°, which means they are able to almost, but not quite, rotate freely around the helix axis, whereas other related peptides have a moderate σρ of only approximately 30° [71]. (Q7) Is there a unique solution with low RMSD for any regular secondary structure element? The 3D structure of a peptide can often be determined from CD if it possesses a regular conformation such as an α-helix or a hydrogen-bonded β-sheet. We have successfully used the SSNMR approach on several α-helical peptides and a few β-sheet–forming sequences [39]. Other types of helices can also be analysed, such as regular polyproline helix type II peptides [16,72]. In the best-case scenario, any regular structure should give a

ARTICLE IN PRESS 22

Erik Strandberg and Anne S. Ulrich

single solution in the τ-ρ plot with a low RMSD. A successful NMR analysis should thus provide the following results (answers A, with examples E): (A1) The peptide conformation is confirmed by a good fit of the NMR data to a specific structural model. (E1) For several membrane-active peptides, an α-helical structure has been confirmed, e.g., for PGLa [6,8,49], magainin 2 [65], MSI-103 [2,3], and KIA14 [66]. For KIGAKI, a good fit was found for an untwisted β-sheet [64]. (A2) The alignment of this peptide structure in the bilayer is by the same analysis uniquely determined in terms of τ and ρ (under the particular conditions studied). (E2) We have obtained and compared the molecular alignment for various membrane-active peptides in many different lipid systems. The most thoroughly studied systems include PGLa [6,8,49], magainin 2 [65], MSI-103 [2,3], KIA14 [66], BP100 [13,67,73] and KIGAKI [64]. A few representative results have been selected for Fig. 5. Namely, for PGLa (and many other antimicrobial peptides), we found that the molecular orientation changes dramatically with the concentration of the peptide. At a low peptide-to-lipid ratio (P/L ¼ 1/200), the amphiphilic PGLa helix lies flat on the surface of the membrane, but at a P/L of 1/100 or higher, an obliquely tilted alignment was found (Fig. 5A and B) [6]. The same behaviour has been seen also for, e.g., MAP [75], MSI-103 [2,3], and KIA14 [66]. On its own, MAG2 lies flat on the membrane surface at P/L ¼ 1/50 (Fig. 5D) [65], but when it is mixed with its synergistic partner peptide PGLa, there is a change in orientation of both peptides (Fig. 5C). PGLa undergoes a dramatic realignment into a transmembrane orientation [49], whereas MAG2 becomes obliquely tilted [65]. These two peptides are naturally present in the same frog skin and are known to act synergistically, which can now be explained by the 3D structure of the hetero-oligomeric PGLa/MAG2 complex with an enhanced pore-forming ability [65,76]. If the answer to Q7 is “no,” then it is necessary to perform further analysis (Chart 3).

2.3 Further analysis (Chart 3) (Q8) Is there more than one solution? Sometimes more than one minimum is seen in the RMDS plot, meaning that several τ/ρ combinations are nominally compatible with

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

A

B PGLa/DMPC/DMPG 1/150/50, τ ≈ 98°

23

C

PGLa/DMPC/DMPG 1/37.5/12.5, τ ≈ 125°

E

D

PGLa/MAG2DMPC/DMPG 1/1/75/25, τ PGLA ≈ 160°, τMAG2 ≈ 120°

F MSI-103/POPC 1/50, τ ≈ 95°

H

G MSI-103/DMPC 1/50, τ ≈ 125°

I WALP19/DMPC 1/20, τ ≈ 4°

MAG2/DMPC/DMPG 1/37.5/12.5, τ ≈ 90°

MSI-103/DMPC/lyso-MPC 1/33/17, τ ≈ 150°

J WALP23/DMPC 1/100, τ ≈ 22°

K WALP23/DMPC 1/100, τ ≈ 30°

WALP23/DMPC 1/100, τ ≈ 36°

Fig. 5 Structure and orientation of peptides in membranes (yellow boxes) determined from solid-state NMR data. (A) PGLa (red) at low concentration lies flat on the surface (τ  98°) of DMPC/DMPG (3/1) membranes [6]. (B) At higher concentration (P/L
1/100) it is obliquely tilted in the membrane, possibly due to dimerization (τ  125°) [6]. (C) PGLa and MAG2 (green) in a 1/1 mixture in a DMPC/DMPG (3:1) membrane. PGLa is inserted almost upright in the lipid bilayer (τ  160°), lining a transmembrane pore [49], while MAG2 remains obliquely tilted on the surface (τ  120°) [65]. (D) MAG2 in DMPC/DMPC (3/1) at P/L ¼ 1/50 lies flat on the membrane surface (τ  90°) [65]. (E) MSI-103 (blue) lies flat on the surface in POPC at P/L ¼ 1/50 (τ  95°) [31]. (F) In DMPC, however, at the same concentration MSI-103 is obliquely tilted, possibly as a dimer, (τ  125°) [31]. (G) MSI-103 is fully inserted with a transmembrane pore-lining orientation in DMPC/lyso-MPC (2/1) at P/L¼ 1/50 (τ  150°) [74]. (H–K) The model transmembrane peptides of the WALP-type (orange) are tilted in the membrane according to hydrophobic mismatch, so that the hydrophobic stretch of the peptide can be optimally accommodated within the hydrophobic part of the membrane [25,33,34].

the NMR data, and it may not be clear which one is the correct solution. This situation is more likely to occur when there are only few data points, and in that case several curves corresponding to different τ/ρ combinations can be fitted well to the same data points in the dipolar or quadrupolar wave

ARTICLE IN PRESS 24

Erik Strandberg and Anne S. Ulrich

plots. Sometimes, especially if the 19F or 2H labelled amino acids could only be placed within a narrow section of the helical wheel (i.e. only on one face of the helix), there may also be several different solutions corresponding to different dynamical parameters [13]. In that case, more labelled analogues should obviously be prepared and included in the analysis, if at all possible, provided that they pass all criteria mentioned above (no change in the CD spectrum nor in their functional activity compared with the wild type, and giving well-oriented samples with well-resolved triplets). These additional labelling positions should be chosen such that they give distinct splittings that can be used to differentiate between the different possible curves, in order to pick out the right solution. If this is not possible, it might still be feasible to find out which of the multiple solutions is correct by complementing the 19F- and/or 2H-NMR data with some other measurements, for example, 15N-NMR or oriented CD [13]. If this is also not possible, or these additional experiments are also inconclusive, then the project must be abandoned with ambiguous results. Such outcome, however, has not yet occurred in our hands. (Q9) Is there no good solution (with sufficiently low RMDS)? In some cases, even though there are enough (more than four) labelled analogues yielding well-resolved splittings, no solution with a sufficiently low RMSD is found. In such cases, a critical analysis of the observed data and/or the underlying structural assumptions should be made, as described in the following. Obviously, if some of the data points stem from peptide analogues where CD or functional assays gave questionable results, those points should have been excluded. (Q10) Do all labels give the same splitting? If all splittings are essentially the same, this is a clear sign that the peptide falls into a special category of regular structures described under Q12. Even if the splittings differ from one another, it might still be possible to fit a subset of these splittings (from a consecutive stretch within the sequence) to a meaningful secondary structure with a meaningful membrane orientation, as described under Q11. (Q11) Does a window over fewer consecutive labels (n
4) give a solution with low RMSD? If the analysis of the full set of data does not give a good fit, it is possible that the full length of the peptide does not belong to the same uninterrupted secondary structure element. In that case, it may well be possible to fit a shorter stretch of consecutive amino acids. If this works, a meaningful result can still be obtained.

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

25

(A3) Frayed N/C terminus. It has been observed that residues close to the peptide termini sometimes do not fold into a well-defined structure, but instead remain flexible and/or unstructured [13,66,67]. Two or three residues at each end of the peptide should therefore be avoided as labelling positions (unless such fraying effects are of interest). (E3) For the short α-helical peptide KIA14 (with a length of 14 amino acids), all positions except for the four lysines have been replaced with Ala-d3 in order to specifically investigate terminal fraying. It was found that positions 1–3 were not helical, but that the helix started at position 4 and continued all the way to the amidated C-terminus [66]. Likewise, in GWALP-type peptides, an unwinding of the termini has been observed using 2H-NMR in several lipid systems [77]. For the peptide BP100, on the other hand, every residue from the N- to the C-terminus were found to be part of the contiguous α-helix [67]. We have also come across membrane-bound peptides, in which a long stretch is unstructured and the remaining part folds into a well-defined helix, hence a significant number of consecutive data points did not fit to the assumed secondary structure. Examples are TP10 [15] and SSL-25 [40]. (A4) Peptide helix with kink. Even if CD analysis shows the peptide to be largely α-helical, there might still be a kink in the helix, especially if it contains a Pro in the sequence. In this case, helical segments on either side of the kink can still be analysed, provided that there are enough data points (at least four) in each of these two segments. (E4) We found this behaviour in the antimicrobial peptide alamethicin, which carries a central Pro in the helix [1]. Additionally, the fusion peptide B18 was found to have a helix-kink-helix structure within a membrane [17]. Also the hydrophobic transmembrane model helix GWALP23-P12, which has a Pro in the middle of the sequence, was found via 2H-NMR to assume a helix-kink-helix structure in lipid bilayers [78]. The next case concern the scenario where all splittings are the same (or almost the same), which will yield direct information on special structural properties, without the need to fit any data. (Q12A) Splittings  0 kHz? If all labelled positions give small splittings very close to 0 kHz, this is a sign that the peptide is not actually bound to the membrane, but is instead swimming around in the bulk water surrounding the membrane. Such situation is usually not observed in oriented samples that are prepared on glass plates, since these membranes have a comparatively low degree of hydration, with no excess water. Additionally, CD spectra should in such

ARTICLE IN PRESS 26

Erik Strandberg and Anne S. Ulrich

cases show unstructured peptides, which already indicates at the beginning of the study that they are not membrane-bound. (A5) Unbound peptides. This may be an important result, rejecting the original hypothesis that the peptide in question is membrane-bound or membrane-active. This observation may thus be the start of a different line of analysis for the peptide, but it marks the end of the membrane-based study. (E5) This was observed for the cell penetrating peptide SSL-25 in membranes containing 25 mol-% cholesterol [40]—though not in the presence of any other lipid compositions. (Q12B) Maximum splittings  +15 kHz? The maximum possible splitting of a (freely rotating) CF3 group is around +17 kHz [8,79,80]. To produce this splitting, the CF3 group must be oriented with the C–CF3-bond vector almost parallel with the magnetic field (which is along the membrane normal). At the same time, the peptide cannot be very mobile, since any deviation from this orientation leads to a partial averaging of this maximum splitting. If such a large splitting is thus observed for all of the labelled positions, this means that all positions have the CF3 group pointing in the same direction, namely perpendicular to the bilayer surface. The only regular and immobilized secondary structure that is compatible with such a set of NMR data corresponds to an untwisted β-sheet assembly lying face-down on the membrane. Due to the intrinsic geometry of β-sheets, all the Cα-Cβ vectors of the labelled side chains will be either inserted into the bilayer or point out of the membrane at a 90° angle. For a CD3 group the equivalent maximum splitting is around 75 kHz [3,81]. (A6) Oriented β-sheet assemblies. (E6) Such a regular and immobilized kind of structure was found for the amphiphilic model peptide KIGAKI at high concentrations [39]. In this designer-made sequence, every second amino acid is a hydrophobic Ile or Ala (alternating with polar Lys or Gly). As the Ile positions were labelled one-by-one with CF3-Bpg, all these labelled positions resulted in splittings of approximately +15 kHz. (Q12C) Splittings  +8 kHz? If all 19F splittings are close to +8 kHz, this can be an indication that the peptide is intrinsically unstructured within the two-dimensional plane of the lipid bilayer. In this situation, there are no hydrogen bonds to restrict the peptide backbone, which is therefore flexible and highly mobile, as also reflected by a β-sheet type of CD spectrum. When CF3-Bpg (or any other large hydrophobic CF3-containing amino acid) is introduced in such an unfolded sequence, the 19F-label cannot provide information about any secondary

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

27

structure, but it will instead reflect the tendency of the CF3 group to insert as deeply as possible into the hydrophobic core of the membrane. If the peptide was not flexible but immobilized (e.g. as β-sheets in Q12B), this would give a large splitting of around +15 kHz. However, in unstructured and highly dynamic peptides, which are anchored to the bilayer surface but swimming around in the fluid membrane, the local orientation of the CF3 group will be strongly fluctuating, with a notionally averaged orientation that points straight down into the membrane. This typically leads to a splitting of approximately +8 kHz. The range of observed splittings might be larger in this case than in the other “constant splitting” cases, since the flexibility and mobility might be different along the sequence of the peptide. If all values are between +5 and +9 kHz, this may thus be taken as an indication of an unstructured peptide, provided that the CD analysis shows a β-sheet type of line shape. (A7) The peptide (or some segment of it) is unstructured. (E7) Disordered peptides or segments of peptides have been observed for TP10 [15], SSL-25 [40] and KIGAKI at low concentration [39]. (Q12D) Splittings  -7 kHz? If any of the 19F-NMR line shapes are comparatively broad with a splitting of approximately 7 kHz, this is a strong indication that the peptide has aggregated and lost its orientational order. That is because disordered CF3 groups produce a broad powder spectrum with a characteristic peak-to-peak splitting of 7 kHz (i.e. upfield of the isotropic chemical shift, but with significant intensity extending downfield by twice this range). To check whether aggregation has occurred, a dynamical analysis should be performed (see Chart 4), and in many cases also the 31P-NMR spectra of the phospholipids will show a significant powder contribution. In case only one or a few labelled positions give a powder spectrum, this indicates that these particular labels have perturbed the peptide folding, so these data points must be discarded. If all labelled positions of the peptide show a powder spectrum, however, this sequence has an intrinsic tendency to aggregate in membranes without any preferential conformation or ordered structure. This behaviour is not uncommon, but may often be relieved by reducing the peptide concentration or choosing a different lipid composition. (A8) The peptide forms an immobilized aggregate. In 2H-NMR the sign of the splitting is not known, hence it may not be clear whether 37 kHz (which corresponds to the characteristic peak-to-peak splitting of a disordered CD3 group) reflects an aggregated peptide or rather a particular well-ordered side chain alignment in the membrane. A dynamical

ARTICLE IN PRESS 28

Erik Strandberg and Anne S. Ulrich

analysis can distinguish these situations, but a clear sign of copious aggregation is the case where all labelled positions give a broad powder spectrum. In oriented 2H-NMR spectra, splittings of approximately 60 kHz are sometimes observed for all labelled positions. This spectral component, however, is a signal from the ordered lipid acyl chains and is due to the natural abundance of deuterium in the hydrocarbon chains [18]. Such background signals are a major problem when working with very low peptide concentrations, making it hard to perform 2H-NMR on peptides in membranes at P/L ratios lower than approximately 1:400. (This is also close to the limit due to the low sensitivity or 2H-NMR.) In that case, 19 F-labelling is clearly preferential. If all splittings are the same, but do not fit any of the cases described above, this might be due to some contamination of samples or artefacts in the spectra (due to NMR spectrometer problems). Such cases fall outside the scope of this review. To obtain more information about peptide mobility and its ability to diffuse freely in the membrane, additional experiments followed by a simple inspection of the data should be performed, as described in Chart 4.

2.4 Rotational diffusion analysis (Chart 4) Before a final conclusion about the peptide orientation and dynamics can be drawn, an important aspect of peptide mobility should be examined, namely aggregation versus free lateral diffusion in the lipid bilayer (which is usually in the liquid crystalline state under biologically relevant conditions). This analysis is especially important if one (or more) of the 19F-NMR splittings is approximately 7 kHz (37 kHz in the case of 2H-NMR), since this splitting is characteristic of disordered powder spectra and often reflects aggregation of the peptide. Any data point corresponding to an aggregated peptide should obviously be excluded from the 3D structure analysis. (T10) Measure 19F-NMR in a vertical alignment. The macroscopically oriented peptide-lipid sample should be measured with the membrane normal perpendicular to the magnetic field (“vertical” sample alignment), i.e., flipped manually by 90° relative to the standard “horizontal” alignment of the original experiment. (Q13) Do the CF3 splittings scale by a factor of 0.5? If peptides are mobile and undergo free lateral diffusion within the oriented membrane sample (faster than the kHz time scale of the NMR experiment), this leads to rotational averaging around the bilayer normal. This averaging

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

29

show up in the spectra, as the splittings are scaled by a factor calculated according to (3cosφ2-1)/2, where φ is the angle between the membrane normal and the magnetic field. For φ ¼ 0° (horizontal sample alignment) this factor is 1, meaning that the rotational diffusion dynamics have no impact on the spectrum and thus cannot be detected. However, when φ ¼ 90° (vertical sample alignment), the factor is 0.5, hence the size of the splitting will change in the case of free rotational diffusion. For CF3 groups, the chemical shift (the position of the triplet in the spectrum) will also change and indicate the sign change. (A9) Rotational diffusion. If the splittings are scaled by a factor of 0.5 (or for 2H-NMR by an absolute factor of 0.5), this indicates that the peptide is undergoing fast rotational diffusion in the membrane, so it is not immobilized or aggregated. (E9) Lateral diffusion of peptides, indicating rotational freedom in the plane of the membrane, has been found for virtually all peptides when examined at low P/L [2,3,6,7,14,18,33,34,38], and in some cases, e.g., for the short peptide BP100, also at very high peptide concentrations up to P/L ¼ 1:10 [13,67]. If peptides do not undergo rotational diffusion around the bilayer normal, then the spectra in the vertical sample alignment will be fundamentally different from the sharp triplets found at φ ¼ 0°. Some kind of powder-like line shape will appear, where the magnitude of the φ ¼ 90° splitting depends also on other types of dynamics present (in many cases a CF3 splitting of 7 kHz, or a CD3 splitting of 37 kHz will be present). In this case, the lack of dynamics can be readily detected, so these peptides are considered immobilized—regardless whether they are assembled in a well-ordered array or whether they are disordered and aggregated. (A10) Immobilized peptides; no rotational diffusion. (E10) Immobilization of a well-folded MSI-103 α-helix was observed at and above P/L ¼ 1:50 in lipid systems where the peptide was in an obliquely inserted state. In other lipid systems where the peptide lay flat on the surface, rotational diffusion was always observed at this P/L [2]. Such difference in lateral rotation, while maintaining an ordered structure, is indicative of peptide oligomerization and/or self-assembly (see below). (Q14) Are the splittings about 7 kHz, the same as for the horizontal alignment? If peptides are aggregated, i.e., immobilized and disordered in the membrane, a powder line shape will be observed, with a splitting of approximately 7 kHz (37 kHz in the case of 2H-NMR), which does not change

ARTICLE IN PRESS 30

Erik Strandberg and Anne S. Ulrich

when the orientation of the sample is changed. In such cases, both φ ¼ 0° and φ ¼ 90° give the same powder spectra with 19F-NMR splittings of approximately 7 kHz. (A11) Powder; the peptide is aggregated. (E11) Peptide aggregation (accompanied by a loss of structure and/or order) has been observed under some conditions for TP10 [15] and MAP [14], and it is also seen in many other cases when the P/L ratio is too high. Interestingly, MSI-103 yields powder spectra in cholesterol-containing membranes [2], but not in other lipid environments tested. When all labels show an aggregated state, the orientation or structure of the peptide cannot be determined because there is no order in the sample. If only one or a few positions yield a powder spectrum, then the other labels can still be used to calculate the peptide structure and orientation.

3. Choice of lipids It is possible to prepare oriented NMR samples on glass plates using virtually any lipid composition desired. The SSNMR method is therefore very useful to study peptide-lipid interactions over a wide range of conditions, in order to find out how these interactions influence peptide conformation, orientation, dynamics, self-assembly, aggregation behaviour, etc. An interesting aspect concerns, e.g., the role of hydrophobic mismatch between membrane thickness and peptide length, which has been studied by several groups using solid-state NMR. It is well known that hydrophobic transmembrane peptides which are longer than the thickness of the lipid bilayer, will increase their tilt angle to become optimally embedded in the membrane. Some examples are WALP and KALP peptides, which have been characterized with 2 H-NMR in membranes of different thicknesses. It was found that their orientations fit almost perfectly to the tilt angle expected from simple mismatch theory, as illustrated in Figs. 5H–K [25,33,34]. GWALP and related model peptides have also been studied systematically in membranes of different thicknesses [82]. Recently, the concept of hydrophobic mismatch has been extended also to amphipathic peptides, as it was found that these peptides can change their orientation from surface-bound to transmembrane depending on the spontaneous curvature of the lipids. In short, in lipids with a negative spontaneous curvature (with a small head group volume compared to the acyl chains, giving them a cone-like shape, such as with PE lipids),

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

31

peptides are preferentially oriented flat on the membrane surface. On the other hand, in lipids with a positive spontaneous curvature (with a large head group volume compared to the acyl chains, giving them an inverted conelike shape, such as with lyso-lipids), a more-inserted, or even transmembrane orientation is favourable [2,31,83,84]. In Fig. 5, we show as an example the orientation of MSI-103 in different lipids at a constant peptide-to-lipid molar ratio of 1:50. In POPC (which possesses a negative spontaneous curvature), the peptide lies essentially flat on the membrane surface (Fig. 5E) [31]. In DMPC (which possesses a small positive spontaneous curvature), the peptide is obliquely tilted in the membrane, possibly as a dimer (Fig. 5F), while in DMPC/lyso-MPC (with a large positive spontaneous curvature), the peptide assumes a transmembrane orientation (Fig. 5G) [85]. Synthetic lipids are available from several companies, with different acyl chains and head groups, giving them distinctly different physicochemical properties. Some of these lipids used in solid-state NMR studies, and some of their properties are listed in Table 1. When choosing a lipid for an NMR sample, the phase transition temperature is very important. Since biological membranes exist in a lamellar liquid crystalline phase, the samples should be measured above Tm, the gel-to-liquid crystalline phase transition temperature (unless there is some reason to study lipids in the gel phase). It is advisable to measure at temperatures at least 10 °C above Tm, because peptides may broaden the phase transition, and in some preparations some gel phase may still be present above the nominal Tm. PE lipids in particular have a tendency to form hexagonal phases above TH, hence this lamellar-to-hexagonal phase transition temperature should also be considered. PC lipids with a wide range of acyl chains, from 6 to 24 carbon atoms lengths, are most suitable for investigating the effects of membrane thickness. Due to the high phase transition temperature Tm of long-chain saturated chains, lipids with unsaturated chains are often preferred in such series of experiments [68,95]. To examine the effects of spontaneous lipid curvature, PE and lyso-lipids can be added to a lamellar lipid matrix in order to produce negative or positive spontaneous curvature, respectively [2]. Lipids with branched diphytanoyl chains (with a negative spontaneous curvature) have also been used in NMR studies [96]. For DMPC samples containing an admixture of lyso-MPC, we have obtained well-oriented samples with up to 33 mol-% lyso-MPC [74], but non-lamellar phases can start to form at higher concentrations of lyso-lipids.

ARTICLE IN PRESS 32

Erik Strandberg and Anne S. Ulrich

Table 1 Common synthetic lipids used in SSNMR studies and their phase transition temperatures. Lipid Full name Acyl chains Tm (°C)a TH (°C)j

N/Ab

DHPC 1,2-dihexanoyl-snglycero-3-phosphatidylcholine

di-6:0-PC

DDPC 1,2-didecanoyl-snglycero-3-phosphocholine

di-10:0-PC

DLPC

1,2-dilauroyl-snglycero-3-phosphatidylcholine

di-12:0-PC

2

DTPC

1,2-ditridecanoyl-snglycero-3-phosphocholine

di-13:0-PC

14

DMPC 1,2-dimyristoyl-snglycero-3-phosphatidylcholine

di-14:0-PC

24

DPPC

1,2-dipalmitoyl-snglycero-3-phosphatidylcholine

di-16:0-PC

41

DSPC

1,2-distearinoyl-snglycero-3-phosphatidylcholine

di-18:0-PC

55

DMoPC 1,2-dimyristoleoyl-snglycero-3-phosphatidylcholine

di-14:1cis9-PC

DPoPC 1,2-dipalmitoleoyl-snglycero-3-phosphatidylcholine

di-16:1cis9-PC

36c

DOPC 1,2-dioleoyl-snglycero-3-phosphatidylcholine

di-18:1cis9-PC

18

DElPC 1,2- dielaidoyl-snglycero-3-phosphatidylcholine

di-18:1trans9-PC 12

DEiPC 1,2-dieicosenoyl-snglycero-3-phosphocholine

di-20:1cis11-PC 4

DErPC 1,2- dierucoyl-snglycero-3-phosphatidylcholine

di-22:1cis13-PC 12

DNPC 1,2-dinervonoyl-snglycero-3-phosphatidylcholine

di-24:1cis15-PC 26

DPhPC 1,2-phytanoyl-snglycero-3-phosphatidylcholine

branched chains < 120d

POPC

16:0-18:1cis9-PC 2

1-palmitoyl-2-oleoyl-snglycero-3-phosphatidylcholine

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

33

Table 1 Common synthetic lipids used in SSNMR studies and their phase transition temperatures.—cont’d Lipid Full name Acyl chains Tm (°C)a TH (°C)j

DMPA 1,2-dimyristoyl-snglycero-3-phosphatidic acid

di-14:0-PA

50e

DMPE 1,2-dimyristoyl-sn-glycero-3phosphatidylethanolamine

di-14:0-PE

49f

DPoPE 1,2-dipalmitoleoyl-sn-glycero-3phosphatidylethanolamine

di-16:1cis9-PE

DOPE

1,2-dioleoyl-sn-glycero3-phosphatidylethanolamine

di-18:1cis9-PE

DElPE

1,2-dielaidoyl-sn-glycero-3phosphatidylethanolamine

di-18:1trans9-PE 37g

POPE

1-palmitoyl-2-oleoyl-sn-glycero-3- 16:0-18:1cis9-PE 25 phosphatidylethanolamine

44k 16

DMPG 1,2-dimyristoyl-snglycero-3-phosphatidylglycerol

di-14:0-PG

23h

DOPG 1,2-dioleoyl-snglycero-3-phosphatidylglycerol

di-18:1cis9-PG

18

POPG

1-palmitoyl-2-oleoyl-snglycero-3-phosphatidylglycerol

16:0-18:1cis9-PG 2

DMPS

1,2-dimyristoyl-snglycero-3-phosphatidylserine

di-14:0-PS

POPS

1-palmitoyl-2-oleoyl-snglycero-3-phosphatidylserine

16:0-18:1cis9-PS 14

LysoMPC

1-myristoyl-2-hydroxyl-snglycero-3-phosphatidylcholine

14:0-PC

10 64 71

35i

N/Ab

a Reported temperature for transition between gel phase and liquid crystalline lamellar phase in fully hydrated lipids, from Ref. [86]. b No lamellar phase, forms micelles. c From Ref. [87]. d From Ref. [88]. No phase transition was found in the interval  120 °C to +120 °C. e From Ref. [89]. f From Ref. [90]. g From Ref. [91]. h From Ref. [92]. i From Ref. [93]. j Temperature for transition from lamellar to hexagonal phase, from Avanti Polar Lipids web site, https:// avantilipids.com/tech-support/physical-properties/phase-transition-temps/. k From Ref. [94].

ARTICLE IN PRESS 34

Erik Strandberg and Anne S. Ulrich

4. Concluding remarks Solid-state NMR (SSNMR) is a very useful method to investigate the conformation, alignment, dynamics, oligomerization, and aggregation behaviour of membrane-bound peptides and proteins. We hope that the flow charts presented here, illustrating a systematic description of the different steps and the workflow in SSNMR studies of peptides in oriented membranes, can be useful for other groups who are interested in the method.

Acknowledgements We thank all present and former members of the group who have been involved in the solid state-NMR development over many years. We acknowledge financial support from the Helmholtz Association program BioInterfaces in Technology and Medicine (BIFTM) and thank the DFG-Center for Functional Nanostructures (TP E1.2) and the Deutsche Forschungsgemeinschaft (DFG) (INST 121384/58-1 FUGG) for supporting the NMR facility.

References [1] D. Maisch, P. Wadhwani, S. Afonin, C. B€ ottcher, B. Koksch, A.S. Ulrich, Chemical labeling strategy with (R)- and (S)-trifluoromethylalanine for solid state 19F NMR analysis of peptaibols in membranes, J. Am. Chem. Soc. 131 (2009) 15596–15597. [2] E. Strandberg, D. Tiltak, S. Ehni, P. Wadhwani, A.S. Ulrich, Lipid shape is a key factor for membrane interactions of amphipathic helical peptides, Biochim. Biophys. Acta 1818 (2012) 1764–1776. [3] E. Strandberg, N. Kanithasen, J. B€ urck, P. Wadhwani, D. Tiltak, O. Zwernemann, A.S. Ulrich, Solid state NMR analysis comparing the designer-made antibiotic MSI103 with its parent peptide PGLa in lipid bilayers, Biochemistry 47 (2008) 2601–2616. [4] P. Tremouilhac, E. Strandberg, P. Wadhwani, A.S. Ulrich, Conditions affecting the re-alignment of the antimicrobial peptide PGLa in membranes as monitored by solid state 2H-NMR, Biochim. Biophys. Acta 1758 (2006) 1330–1342. [5] P. Tremouilhac, E. Strandberg, P. Wadhwani, A.S. Ulrich, Synergistic transmembrane alignment of the antimicrobial heterodimer PGLa/magainin, J. Biol. Chem. 281 (2006) 32089–32094. [6] E. Strandberg, P. Wadhwani, P. Tremouilhac, U.H.N. D€ urr, A.S. Ulrich, Solid-state NMR analysis of the PGLa peptide orientation in DMPC bilayers: structural fidelity of 2 H-labels versus high sensitivity of 19F-NMR, Biophys. J. 90 (2006) 1676–1686. [7] R.W. Glaser, C. Sachse, U.H.N. D€ urr, S. Afonin, P. Wadhwani, E. Strandberg, A.S. Ulrich, Concentration-dependent realignment of the antimicrobial peptide PGLa in lipid membranes observed by solid-state 19F-NMR, Biophys. J. 88 (2005) 3392–3397. [8] R.W. Glaser, C. Sachse, U.H.N. D€ urr, P. Wadhwani, A.S. Ulrich, Orientation of the antimicrobial peptide PGLa in lipid membranes determined from 19F-NMR dipolar couplings of 4-CF3-phenylglycine labels, J. Magn. Reson. 168 (2004) 153–163. [9] S. Afonin, U.H.N. D€ urr, P. Wadhwani, J.B. Salgado, A.S. Ulrich, Solid state NMR structure analysis of the antimicrobial peptide gramicidin S in lipid membranes: concentration-dependent re-alignment and self-assembly as a β-barrel, Top. Curr. Chem. 273 (2008) 139–154.

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

35

[10] M. Ieronimo, S. Afonin, K. Koch, M. Berditsch, P. Wadhwani, A.S. Ulrich, 19F NMR analysis of the antimicrobial peptide PGLa bound to native cell membranes from bacterial protoplasts and human erythrocytes, J. Am. Chem. Soc. 132 (2010) 8822–8824. [11] K. Koch, S. Afonin, M. Ieronimo, M. Berditsch, A.S. Ulrich, Solid-state 19F-NMR of peptides in native membranes, Top. Curr. Chem. 306 (2012) 89–118. [12] S.L. Grage, S. Kara, A. Bordessa, V. Doan, F. Rizzolo, M. Putzu, T. Kubar, A.M. Papini, G. Chaume, T. Brigaud, S. Afonin, A.S. Ulrich, Orthogonal 19F-labeling for solid-state NMR reveals the conformation and orientation of short peptaibols in membranes, Chem. A Eur. J. 24 (2018) 4328–4335. [13] P. Wadhwani, E. Strandberg, J. van den Berg, C. Mink, J. B€ urck, R. Ciriello, A.S. Ulrich, Dynamical structure of the short multifunctional peptide BP100 in membranes, Biochim. Biophys. Acta 1838 (2014) 940–949. [14] P. Wadhwani, J. B€ urck, E. Strandberg, C. Mink, S. Afonin, A.S. Ulrich, Using a sterically restrictive amino acid as a 19F-NMR label to monitor and control peptide aggregation in membranes, J. Am. Chem. Soc. 130 (2008) 16515–16517. [15] S. Fangh€anel, P. Wadhwani, E. Strandberg, W.P.R. Verdurmen, J. B€ urck, S. Ehni, P.K. Mykhailiuk, S. Afonin, D. Gerthsen, I.V. Komarov, R. Brock, A.S. Ulrich, Structure analysis and conformational transitions of the cell penetrating peptide transportan 10 in the membrane-bound state, PLoS One 9 (2014) e99653. [16] S. Afonin, V. Kubyslikin, P.K. Mykhailiuk, I.V. Komarov, A.S. Ulrich, Conformational plasticity of the cell-penetrating peptide SAP as revealed by solid-state 19 F-NMR and circular dichroism spectroscopies, J. Phys. Chem. B 121 (2017) 6479–6491. [17] S. Afonin, U.H.N. D€ urr, R.W. Glaser, A.S. Ulrich, ’Boomerang’-like insertion of a fusogenic peptide in a lipid membrane revealed by solid-state 19F NMR, Magn. Reson. Chem. 42 (2004) 195–203. [18] D. Grasnick, U. Sternberg, E. Strandberg, P. Wadhwani, A.S. Ulrich, Irregular structure of the HIV fusion peptide in membranes demonstrated by solid-state NMR and MD simulations, Eur. Biophys. J. 40 (2011) 529–543. [19] S. Afonin, R.W. Glaser, M. Berditchevskaia, P. Wadhwani, K.H. Guhrs, U. Mollmann, A. Perner, A.S. Ulrich, 4-Fluorophenylglycine as a label for 19F-NMR structure analysis of membrane-associated peptides, ChemBioChem 4 (2003) 1151–1163. [20] S.L. Grage, J.B. Salgado, U.H.N. D€ urr, S. Afonin, R.W. Glaser, A.S. Ulrich, Solid state 19 F-NMR of biomembranes, in: S.R. Kiihne, H.J.M. deGroot (Eds.), Perspectives on Solid State NMR in Biology, vol. 1, Kluwer Academic Publishers, Dordrecht/Boston/ London, 2001, pp. 83–91. [21] A.S. Ulrich, P. Wadhwani, U.H.N. D€ urr, S. Afonin, R.W. Glaser, E. Strandberg, P. Tremouilhac, C. Sachse, M. Berditchevskaia, S.L. Grage, Solid-state 19F-nuclear magnetic resonance analysis of membrane-active peptides, in: A. Ramamoorthy (Ed.), NMR Spectroscopy of Biological Solids, CRC Press, Boca Raton, FL, 2006, pp. 215–236. [22] P. Wadhwani, E. Strandberg, Structure analysis of membrane-active peptides using 19 F-labeled amino acids and solid-state NMR, in: I. Ojima (Ed.), Fluorine in Medicinal Chemistry and Chemical Biology, Blackwell Publishing, London, 2009, pp. 463–493. [23] S. Afonin, P.K. Mikhailiuk, I.V. Komarov, A.S. Ulrich, Evaluating the amino acid CF3-bicyclopentylglycine as a new label for solid-state 19F-NMR structure analysis of membrane-bound peptides, J. Pept. Sci. 13 (2007) 614–623. [24] P.K. Mikhailiuk, S. Afonin, A.N. Chernega, E.B. Rusanov, M.O. Platonov, G.G. Dubinina, M. Berditsch, A.S. UIrich, I.V. Komarov, Conformationally rigid trifluoromethyl-substituted α-amino acid designed for peptide structure analysis by solid-state 19F NMR spectroscopy, Angew. Chem. Int. Ed. 45 (2006) 5659–5661.

ARTICLE IN PRESS 36

Erik Strandberg and Anne S. Ulrich

[25] E. Strandberg, S. Esteban-Martı´n, A.S. Ulrich, J. Salgado, Hydrophobic mismatch of mobile transmembrane helices: merging theory and experiments, Biochim. Biophys. Acta 1818 (2012) 1242–1249. [26] S.L. Grage, E. Strandberg, P. Wadhwani, S. Esteban-Martin, J. Salgado, A.S. Ulrich, Comparative analysis of the orientation of transmembrane peptides using solid-state 2 H- and 15N-NMR: mobility matters, Eur. Biophys. J. 41 (2012) 475–482. [27] S. Esteban-Martı´n, E. Strandberg, J. Salgado, A.S. Ulrich, Solid state NMR analysis of peptides in membranes: influence of dynamics and labeling scheme, Biochim. Biophys. Acta 1798 (2010) 252–257. [28] E. Strandberg, S. Esteban-Martı´n, J. Salgado, A.S. Ulrich, Orientation and dynamics of peptides in membranes calculated from 2H-NMR data, Biophys. J. 96 (2009) 3223–3232. [29] S. Esteban-Martı´n, E. Strandberg, G. Fuertes, A.S. Ulrich, J. Salgado, Influence of whole-body dynamics on 15N PISEMA NMR spectra of membrane peptides: a theoretical analysis, Biophys. J. 96 (2009) 3233–3241. [30] E. Strandberg, A.S. Ulrich, Dynamic structure analysis of peptides in membranes by solid-state NMR, in: F. Separovic, A. Naito (Eds.), Advances in Biological Solid-State NMR: Proteins and Membrane-Active Peptides, Royal Society of Chemistry, London, 2014, pp. 304–319. [31] E. Strandberg, J. Zerweck, P. Wadhwani, A.S. Ulrich, Synergistic insertion of antimicrobial magainin-family peptides in membranes depends on the lipid spontaneous curvature, Biophys. J. 104 (2013) L9–11. [32] S. Afonin, S.L. Grage, M. Ieronimo, P. Wadhwani, A.S. Ulrich, Temperaturedependent transmembrane insertion of the amphiphilic peptide PGLa in lipid bilayers observed by solid state 19F-NMR spectroscopy, J. Am. Chem. Soc. 130 (2008) 16512–16514. [33] P.C.A. Van der Wel, E. Strandberg, J.A. Killian, R.E. Koeppe II, Geometry and intrinsic tilt of a tryptophan-anchored transmembrane α-helix determined by 2H NMR, Biophys. J. 83 (2002) 1479–1488. € [34] E. Strandberg, S. Ozdirekcan, D.T.S. Rijkers, P.C.A. Van der Wel, R.E. Koeppe II, R.M.J. Liskamp, J.A. Killian, Tilt angles of transmembrane model peptides in oriented and non-oriented lipid bilayers as determined by 2H solid state NMR, Biophys. J. 86 (2004) 3709–3721. € [35] S. Ozdirekcan, D.T.S. Rijkers, R.M.J. Liskamp, J.A. Killian, Influence of flanking residues on tilt and rotation angles of transmembrane peptides in lipid bilayers. A solid-state 2 H NMR study, Biochemistry 44 (2005) 1004–1012. [36] V.V. Vostrikov, A.E. Daily, D.V. Greathouse, R.E. Koeppe II, Charged or aromatic anchor residue dependence of transmembrane peptide tilt, J. Biol. Chem. 285 (2010) 31723–31730. [37] A.E. Daily, D.V. Greathouse, P.C.A. van der Wel, R.E. Koeppe II, Helical distortion in tryptophan- and lysine-anchored membrane-spanning α-helices as a function of hydrophobic mismatch: a solid-state deuterium NMR investigation using the geometric analysis of labeled alanines method, Biophys. J. 94 (2008) 480–491. [38] N.J. Gleason, V.V. Vostrikov, D.V. Greathouse, C.V. Grant, S.J. Opella, R.E. Koeppe II, Tyrosine replacing tryptophan as an anchor in GWALP peptides, Biochemistry 51 (2012) 2044–2053. [39] P. Wadhwani, E. Strandberg, N. Heidenreich, J. B€ urck, S. Fangh€anel, A.S. Ulrich, Self-assembly of flexible β-strands into immobile amyloid-like β-sheets in membranes as revealed by solid-state 19F NMR, J. Am. Chem. Soc. 134 (2012) 6512–6515. [40] P. M€ uhlh€auser, P. Wadhwani, E. Strandberg, J. B€ urck, A.S. Ulrich, Structure analysis of the membrane-bound dermcidin-derived peptide SSL-25 from human sweat, Biochim. Biophys. Acta 1859 (2017) 2308–2318.

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

37

[41] V.S. Kubyshkin, P.K. Mykhailiuk, S. Afonin, A.S. Ulrich, I.V. Komarov, Incorporation of cis- and trans-4,5-difluoromethanoprolines into polypeptides, Org. Lett. 14 (2012) 5254–5257. [42] V. Kubyshkin, S. Afonin, S. Kara, N. Budisa, P.K. Mykhailiuk, A.S. Ulrich, γ-(S)Trifluoromethyl proline: evaluation as a structural substitute of proline for solid state 19 F-NMR peptide studies, Org. Biomol. Chem. 13 (2015) 3171–3181. [43] P.K. Mykhailiuk, S. Afonin, G.V. Palamarchuk, O.V. Shishkin, A.S. Ulrich, I.V. Komarov, Synthesis of trifluoromethyl-substituted proline analogues as 19F NMR labels for peptides in the polyproline II conformation, Angew. Chem. Int. Ed. 47 (2008) 5765–5767. [44] A.N. Tkachenko, D.S. Radchenko, P.K. Mykhailiuk, S. Afonin, A.S. Ulrich, I.V. Komarov, Design, synthesis, and application of a trifluoromethylated phenylalanine analogue as a label to study peptides by solid-state 19F NMR spectroscopy, Angew. Chem. Int. Ed. 52 (2013) 6504–6507. [45] R. Heinzmann, S.L. Grage, C. Schalck, J. B€ urck, Z. Banoczi, O. Toke, A.S. Ulrich, A kinked antimicrobial peptide from Bombina maxima. II. Behavior in phospholipid bilayers, Eur. Biophys. J. 40 (2011) 463–470. [46] A.N. Tkachenko, P.K. Mykhailiuk, S. Afonin, D.S. Radchenko, V.S. Kubyshkin, A.S. Ulrich, I.V. Komarov, A 19F NMR label to substitute polar amino acids in peptides: a CF3-substituted analogue of serine and threonine, Angew. Chem. 125 (2013) 1526–1529. [47] O.M. Michurin, S. Afonin, M. Berditsch, C.G. Daniliuc, A.S. Ulrich, I.V. Komarov, D.S. Radchenko, Delivering structural information on the polar face of membraneactive peptides: 19F-NMR labels with a cationic side chain, Angew. Chem. Int. Ed. 55 (2016) 14595–14599. [48] O.M. Michurin, K. Tolmachova, S. Afonin, O. Babii, S.L. Grage, A.S. Ulrich, I.V. Komarov, D.S. Radchenko, Conformationally constrained mono-fluorinated arginine as a cationic label for solid-state 19F NMR analysis of membrane-bound peptides, Eur. J. Org. Chem. (2018) 3826–3833. [49] E. Strandberg, P. Tremouilhac, P. Wadhwani, A.S. Ulrich, Synergistic transmembrane insertion of the heterodimeric PGLa/magainin 2 complex studied by solid-state NMR, Biochim. Biophys. Acta 1788 (2009) 1667–1679. [50] G.B. Fields, R.L. Noble, Solid-phase peptide synthesis utilizing 9-fluorenylmethoxycarbonyl amino acids, Int. J. Pept. Protein Res. 35 (1990) 161–214. [51] J. Salgado, S.L. Grage, L.H. Kondejewski, R.S. Hodges, R.N. McElhaney, A.S. Ulrich, Membrane-bound structure and alignment of the antimicrobial β-sheet peptide gramicidin S derived from angular and distance constraints by solid state 19F-NMR, J. Biomol. NMR 21 (2001) 191–208. [52] A.S. Ulrich, Solid state 19F-NMR methods for studying biomembranes, Prog. Nucl. Magn. Reson. Spectrosc. 46 (2005) 1–21. [53] C.R. Sanders, R.S. Prosser, Bicelles: a model membrane system for all seasons? Structure 6 (1998) 1227–1234. [54] A.A. De Angelis, S.J. Opella, Bicelle samples for solid-state NMR of membrane proteins, Nat. Protoc. 2 (2007) 2332–2338. [55] T.H. Walther, S.L. Grage, N. Roth, A.S. Ulrich, Membrane alignment of the poreforming component TatAd of the twin-arginine translocase from Bacillus subtilis resolved by solid-state NMR spectroscopy, J. Am. Chem. Soc. 132 (2010) 15945–15956. [56] O.V. Nolandt, T.H. Walther, S.L. Grage, A.S. Ulrich, Magnetically oriented dodecylphosphocholine bicelles for solid-state NMR structure analysis, Biochim. Biophys. Acta 1818 (2012) 1142–1147. [57] J. Seelig, 31P nuclear magnetic resonance and the head group structure of phospholipids in membranes, Biochim. Biophys. Acta 515 (1978) 105–140.

ARTICLE IN PRESS 38

Erik Strandberg and Anne S. Ulrich

[58] P.R. Cullis, B. de Kruijff, Lipid polymorphism and the functional roles of lipids in biological membranes, Biochim. Biophys. Acta 559 (1979) 399–420. [59] R.W. Glaser, A.S. Ulrich, Susceptibility corrections in solid-state NMR experiments with oriented membrane samples. Part I: applications, J. Magn. Reson. 164 (2003) 104–114. [60] S. Zhang, X.L. Wu, M. Mehring, Elimination of ringing effects in multiple-pulse sequences, Chem. Phys. Lett. 173 (1990) 481–484. [61] J.H. Davis, K.R. Jeffrey, M. Bloom, M.I. Valic, T.P. Higgs, Quadrupolar echo deuteron magnetic resonance spectroscopy in ordered hydrocarbon chains, Chem. Phys. Lett. 42 (1976) 390–394. [62] S.L. Grage, M.-A. Sani, O. Cheneval, S.T. Henriques, C. Schalck, R. Heinzmann, J.S. Mylne, P.K. Mykhailiuk, S. Afonin, I.V. Komarov, F. Separovic, D.J. Craik, A.S. Ulrich, Orientation and location of the cyclotide kalata B1 in phospholipid bilayers revealed by solid-state NMR, Biophys. J. 112 (2017) 630–642. [63] S. Esteban-Martı´n, J. Salgado, The dynamic orientation of membrane-bound peptides: bridging simulations and experiments, Biophys. J. 93 (2007) 4278–4288. [64] P. Wadhwani, J. Reichert, E. Strandberg, J. B€ urck, J. Misiewicz, S. Afonin, N. Heidenreich, S. Fangh€anel, P.K. Mykhailiuk, I.V. Komarov, A.S. Ulrich, Stereochemical effects on the aggregation and biological properties of the fibril-forming peptide [KIGAKI]3 in membranes, Phys. Chem. Chem. Phys. 15 (2013) 8962–8971. [65] E. Strandberg, D. Horn, S. Reißer, J. Zerweck, P. Wadhwani, A.S. Ulrich, 2H-NMR and MD simulations reveal membrane-bound conformation of magainin 2 and its synergy with PGLa, Biophys. J. 111 (2016) 2149–2161. [66] E. Strandberg, A. Grau-Campistany, P. Wadhwani, J. B€ urck, F. Rabanal, A.S. Ulrich, Helix fraying and lipid-dependent structure of a short amphipathic membrane-bound peptide revealed by solid-state NMR, J. Phys. Chem. B 122 (2018) 6236–6250. [67] H. Zamora-Carreras, E. Strandberg, P. M€ uhlh€auser, J. B€ urck, P. Wadhwani, M.A´. Jimenez, M. Bruix, A.S. Ulrich, Alanine scan and 2H NMR analysis of the membrane-active peptide BP100 point to a distinct carpet mechanism of action, Biochim. Biophys. Acta 1858 (2016) 1328–1338. [68] J.A. Killian, I. Salemink, M.R. de Planque, G. Lindblom, R.E. Koeppe II, D.V. Greathouse, Induction of nonbilayer structures in diacylphosphatidylcholine model membranes by transmembrane α-helical peptides: importance of hydrophobic mismatch and proposed role of tryptophans, Biochemistry 35 (1996) 1037–1045. [69] K.A. Sparks, N.J. Gleason, R. Gist, R. Langston, D.V. Greathouse, R.E. Koeppe II, Comparisons of interfacial Phe, Tyr, and Trp residues as determinants of orientation and dynamics for GWALP transmembrane peptides, Biochemistry 53 (2014) 3637–3645. [70] V.V. Vostrikov, R.E. Koeppe II, Response of GWALP transmembrane peptides to changes in the tryptophan anchor positions, Biochemistry 50 (2011) 7522–7535. [71] A.N. Martfeld, V. Rajagopalan, D.V. Greathouse, R.E. Koeppe II, Dynamic regulation of lipid-protein interactions, Biochim. Biophys. Acta 1848 (2015) 1849–1859. [72] V. Kubyshkin, S.L. Grage, J. B€ urck, A.S. Ulrich, N. Budisa, Transmembrane polyproline helix, J. Phys. Chem. Lett. 9 (2018) 2170–2174. [73] Deleted in review [74] A. Grau-Campistany, E. Strandberg, P. Wadhwani, F. Rabanal, A.S. Ulrich, Extending the hydrophobic mismatch concept to amphiphilic membranolytic peptides, J. Phys. Chem. Lett. 7 (2016) 1116–1120. [75] E. Strandberg, D. Tiltak, M. Ieronimo, N. Kanithasen, P. Wadhwani, A.S. Ulrich, Influence of C-terminal amidation on the antimicrobial and hemolytic activities of cationic α-helical peptides, Pure Appl. Chem. 79 (2007) 717–728.

ARTICLE IN PRESS Flow charts for 19F/2H-NMR structure analysis

39

[76] J. Zerweck, E. Strandberg, O. Kukharenko, J. Reichert, J. B€ urck, P. Wadhwani, A.S. Ulrich, Molecular mechanism of synergy between the antimicrobial peptides PGLa and magainin 2, Sci. Rep. 7 (2017) 13153. [77] A. Mortazavi, V. Rajagopalan, K.A. Sparks, D.V. Greathouse, R.E. Koeppe II, Juxta-terminal helix unwinding as a stabilizing factor to modulate the dynamics of transmembrane helices, ChemBioChem 17 (2016) 462–465. [78] J.M. Rankenberg, V.V. Vostrikov, C.D. DuVall, D.V. Greathouse, R.E. Koeppe II, C.V. Grant, S.J. Opella, Proline kink angle distributions for GWALP23 in lipid bilayers of different thicknesses, Biochemistry 51 (2012) 3554–3564. [79] U.H.N. D€ urr, S.L. Grage, R. Witter, A.S. Ulrich, Solid state 19F NMR parameters of fluorine-labeled amino acids. Part I: aromatic substituents, J. Magn. Reson. 191 (2008) 7–15. [80] S.L. Grage, U.H.N. D€ urr, S. Afonin, P.K. Mikhailiuk, I.V. Komarov, A.S. Ulrich, Solid state 19F NMR parameters of fluorine-labeled amino acids. Part II: aliphatic substituents, J. Magn. Reson. 191 (2008) 16–23. [81] J.H. Davis, The description of membrane lipid conformation, order and dynamics by 2 H-NMR, Biochim. Biophys. Acta 737 (1983) 117–171. [82] V.V. Vostrikov, C.V. Grant, S.J. Opella, R.E. Koeppe II, On the combined analysis of 2 H and 15N/1H solid-state NMR data for determination of transmembrane peptide orientation and dynamics, Biophys. J. 101 (2011) 2939–2947. [83] A. Grau-Campistany, E. Strandberg, P. Wadhwani, J. Reichert, J. B€ urck, F. Rabanal, A.S. Ulrich, Hydrophobic mismatch demonstrated for membranolytic peptides, and their use as molecular rulers to measure bilayer thickness in native cells, Sci. Rep. 5 (2015) 9388. [84] E. Strandberg, A.S. Ulrich, AMPs and OMPs: is the folding and bilayer insertion of β-stranded outer membrane proteins governed by the same biophysical principles as for α-helical antimicrobial peptides? Biochim. Biophys. Acta 1848 (2015) 1944–1954. [85] M.C. Gagnon, E. Strandberg, A. Grau-Campistany, P. Wadhwani, J. Reichert, J. B€ urck, F. Rabanal, M. Auger, J.F. Paquin, A.S. Ulrich, Influence of the length and charge on the activity of α-helical amphipathic antimicrobial peptides, Biochemistry 56 (2017) 1680–1695. [86] R. Koynova, M. Caffrey, Phases and phase transitions of the phosphatidylcholines, Biochim. Biophys. Acta 1376 (1998) 91–145. [87] P.W. Van Dijck, B. De Kruijff, L.L. Van Deenen, J. De Gier, R.A. Demel, The preference of cholesterol for phosphatidylcholine in mixed phosphatidylcholinephosphatidylethanolamine bilayers, Biochim. Biophys. Acta 455 (1976) 576–587. [88] H. Lindsey, N.O. Petersen, S.I. Chan, Physicochemical characterization of 1,2diphytanoyl-sn-glycero-3-phosphocholine in model membrane systems, Biochim. Biophys. Acta 555 (1979) 147–167. [89] A. Blume, H. Eibl, The influence of charge on bilayer membranes. Calorimetric investigations of phosphatidic acid bilayers, Biochim. Biophys. Acta 558 (1979) 13–21. [90] R. Koynova, M. Caffrey, Phases and phase transitions of the hydrated phosphatidylethanolamines, Chem. Phys. Lipids 69 (1994) 1–34. [91] T.P. McMullen, R.N. McElhaney, Differential scanning calorimetric studies of the interaction of cholesterol with distearoyl and dielaidoyl molecular species of phosphatidylcholine, phosphatidylethanolamine, and phosphatidylserine, Biochemistry 36 (1997) 4979–4986. [92] R. Koynova, Liquid crystalline phase metastability of phosphatidylglycerols, Chem. Phys. Lipids 89 (1997) 67–73. [93] J.L. Browning, J. Seelig, Bilayers of phosphatidylserine: a deuterium and phosphorus nuclear magnetic resonance study, Biochemistry 19 (1980) 1262–1270.

ARTICLE IN PRESS 40

Erik Strandberg and Anne S. Ulrich

[94] E.M. Tytler, J.P. Segrest, R.M. Epand, S.Q. Nie, R.F. Epand, V.K. Mishra, Y.V. Venkatachalapathi, G.M. Anantharamaiah, Reciprocal effects of apolipoprotein and lytic peptide analogs on membranes - cross-sectional molecular shapes of amphipathic α helixes control membrane stability, J. Biol. Chem. 268 (1993) 22112–22118. [95] D. Windisch, C. Ziegler, S.L. Grage, J. B€ urck, M. Zeitler, P.L. Gor’kov, A.S. Ulrich, Hydrophobic mismatch drives the interaction of E5 with the transmembrane segment of PDGF receptor, Biophys. J. 109 (2015) 737–749. [96] S. Kara, S. Afonin, O. Babii, A.N. Tkachenko, I.V. Komarov, A.S. Ulrich, Diphytanoyl lipids as model systems for studying membrane-active peptides, Biochim. Biophys. Acta 1859 (2017) 1828–1837.