The main focus of our BioNMR group are peptides and proteins interacting with biological membranes. We have selected a number of different systems that are either genetically or structurally or functionally related. Our aim is to compare the structures of these peptides and proteins on the primary (sequence), secondary (conformation), tertiary (folding), and quarternary (self-assembly) level, in order to correlate these properties with the respective functional mechanisms. The following list summarizes the different classes of molecules that are currently being studied and compared with one another:
· Fusogenic peptides: B18 (involved in sperm-egg fusion)
FP23 (from HIV-1 gp41 involved in virus-cell fusion)
· Antimicrobial peptides: Gramicidin S (from B. brevis)
PGLa (from X. laevis)
K3 (designer-made antibiotic)
MAP (model amphiphilic peptide)
Magainin-2 (from X. laevis)
· Cell penetrating peptides: HIV-TAT (sequence 47-60)
Oligo-arginine (R8), Oligo-lysine (K8)
MAP (model amphiphilic peptide)
Proline-rich peptides (from g-zein)
· Integral membrane proteins: Tat-A (bacterial protein export machinery)
Mscl (mechanosensitive channel protein)
E5-protein and PDGF-receptor (signal transduction)
· Biological fibres: Silk fibroin (from B. mori, S.c. ricini)
Cellulose (from Acetobacter Xylinum)
All of these membrane-active (or fibrous) molecules present a challenge to conventional structure analysis by X-ray diffraction of crystals or NMR in solution, hence we have developed and applied novel solid state NMR techniques to reveal their structures in a biologically relevant context, i.e. in the membrane-bound state under quasi-native conditions.
Starting with the genetic sequence and aiming at a functional interpretation, our group has to cover all steps from sample preparation with suitable isotope labels, to the development and application of NMR experiments, computer-aided data analysis, and biological activity tests. Over the last few years we have set up the necessary facilities and acquired the know-how to carry out the following procedures:
Sample preparation:
· Peptide synthesis with specific NMR isotope labels (including difficult 19F-amino acids)
· Cloning and overexpression of uniformly labelled peptides and membrane proteins
· Optimizing the preparation of uniaxially aligned membrane samples for NMR analysis
Novel methods to analyze biomolecular structure and dynamics:
· Highly sensitive 19F-NMR strategy to determine peptide structure from specific labels
· Use of non-perturbing 2H- and 15N-labels to supplement and verify the 19F-NMR method
· 2D MAS-technique RAI for chemical shift analysis of uniformly labelled biomolecules
· COSMOS-NMR molecular dynamics simulations with experimental NMR-constraints
Biological activity tests and supporting methods
· Membrane fusion assays by fluorescence spectroscopy, light scattering and DSC
· Antimicrobial activity (MIC) and hemolysis assays
· Oriented circular dichroism spectroscopy for peptide secondary structure and alignment
· Design and assembly of solid state NMR probes, filters, etc.
Biological results
The primary information we have been able to extract from the various membrane-active peptides concerns their alignment with respect to the membrane normal. That is, the orientation of a rigid α-helix or β-strand can be reconstructed from a set of individual orientational constraints determined by solid state NMR from the selectively isotope-labelled peptides (with 19F, 2H, 15N). Given a sufficient number of such NMR-constraints, the molecular conformation can be verified and insight is gained into the molecular mobility. For example, the dynamic averaging of a peptide and its ability to rotate freely about the membrane normal helps to discriminate monomeric peptides from small oligomers and from large aggregates. As we have discovered in virtually every system investigated so far, these different oligomeric states seem to play a decisive role in the fusogenic, antimicrobial and cell-penetrating activities of the respective peptides. In fact, our group has been the first to demonstrate experimentally and with quasi-atomic resolution that numerous membrane-bound peptides re-align as a function of concentration and/or lipid composition. The detailed description of the various functional states requires a systematic variation of the experimental conditions, which was only possible due to the exquisitely high sensitivity of the 19F-NMR approach we have developed and applied here. The following outline summarizes the structural and functional results on the different peptide systems that have been brought to a conclusion so far:
The peptide B18 from the fertilization protein Bindin had been extensively characterized by our group. It is involved in sperm-egg fusion and able to trigger fusion of lipid vesicles in vitro. Using solid state 19F-NMR and other biophysical methods, we determined its structure as a helix-loop-helix when bound to membranes in a functionally relevant, presumably pre-fusion state. Interestingly, the N-terminal helix is obliquely immersed into the lipid bilayer, while the C-terminal amphiphilic α-helix lies flat on the surface. In contrast, at a high peptide-concentration that is presumably representative of the post-fusion state, B18 self-assembles into inactive amyloid-like fibrils with a β-strand conformation. To address its transient conformation during the moment of fusion we compared the fusion kinetics of sterically restricted peptide analogues by fluorescence assays. Since fusion was found to proceed even when certain folding patterns were prohibited, the unusual conformational plasticity observed above appears reasonable, which indeed seems to be characteristic of fusogenic peptides in general.
A similar picture is currently being refined by 19F- and 2H-NMR for the fusion sequence FP23 from the HIV-1 protein gp41, which is involved in virus-cell fusion. Our first NMR constraints show that the N-terminal part of the a-helical peptide is also obliquely immersed in the lipid membrane. Likewise, FP23 is known to self-assemble into β-sheet fibrils, and our fusion assays show that the kinetics do not differ systematically for sterically restricted peptides. Hence, we may offer a first structural explanation as to why the genetically unrelated peptides B18 and FP23 have similar fusiogenic functions and appear to operate via a comparable mechanism in terms of their conformations and interactions with the target membrane.
The antimicrobial peptide Gramicidin S (GS) from B. brevis had been studied by our group using solid state 19F-NMR. Its cyclic β-sheet structure possesses a pronounced amphiphilic character, explaining why the peptide binds avidly to lipid bilayers and thereby kills bacteria by destabilizing their cellular membranes. Our structure analysis suggests that GS is able to re-align in membranes from a flat surface-bound S-state to an upright membrane-inserted I-state as a function of peptide concentration, lipid composition and temperature. This structural transition is accompanied by self-assembly of GS from a monomeric state to an oligomeric putative b-barrel pore. Biological activity tests of sterically constrained GS analogues showed that the desired antimicrobial activity of this potential drug molecule does not require self-assembly, but that the oligomeric structure instead appears to correlate with the hemolytic side effects.
The results on the β-sheet peptide GS call for a comparison with a representative α-helical antimicrobial peptide, such as PGLa from the Magainin family. Given their characteristic amphiphilic nature they are expected to operate by a similar mechanism, although the so-called “toriodal wormhole” model is preferred in the literature over the “barrel-stave” structure we found for GS. A thorough analysis of PGLa was thus carried out by 19F-NMR, and the resulting structures were verified by non-perturbing 2H- and 15N-labels (which also showed that conservatively introduced 19F-labelled amino acids do not lead to significant structural perturbations). The most intriguing result here was the observation that PGLa re-aligns with increasing peptide concentration from a flat surface-bound S-state to a tilted T-state, in which the α-helix is obliquely immersed in the lipid bilayer rather than assuming the expected transmembrane alignment. We explain this previously undescribed T-state by dimerization, as this class of peptides is known to form both homo- as well as hetero-dimers under certain conditions.
The designer-made peptides K3 and MAP, which are homologous to PGLa, are currently being characterized. They are found to undergo virtually the same kind of structural change in the membrane, albeit with different threshold concentrations. Hence, the slight differences in peptide sequence lead to significantly different equilibria in terms of (i) membrane-binding, (ii) peptide-dimerization, and (iii) assembly into extensive aggregates, all of which can be addressed by solid state NMR. Remarkably, our first comparative data suggest that antimicrobial activity (but not hemolysis) seems to correlate directly with the dimerization threshold, which may thus constitute the functionally relevant step in the mechanism of action. Finally, we studied the interaction between PGLa and Magainin-2, which have rather different sequences despite their related genomic origin (from the skin of X. laevis). They are known to act synergistically with enhanced efficiency at a 1:1 ratio, and our NMR data support the concept that a hetero-dimer is formed at a much lower threshold than for either of the pure peptides. Interestingly, this hetero-dimer seems to assume a transmembrane alignment as opposed to the tilted T-state observed so far for the homo-dimers. If this structure can be confirmed, it would for the first time provide experimental evidence for a stable toroidal wormhole structure in a lipid bilayer with quasi-atomic resolution.
In contrast to antimicrobial peptides that are meant to destroy membranes, cell penetrating sequences are supposedly able to traverse the lipid membrane without causing any damage. As they can transport large cargo molecules along with them, they constitute attractive targets in pharmacology for delivering drugs and other particles into the cytoplasm. Two different classes of cell-penetrating peptides are known, as represented by MAP (“model amphiphilic peptide”) and HIV-TAT (sequence 47-60) which are being studied here. MAP is very similar to the α-helical antimicrobial peptides discussed above, whereas the HIV-TAT peptide is rich in arginines and lysines (and without amphiphilic secondary structure). As expected, our NMR analysis of MAP revealed a structural transition from an S-state to a T-state just like PGLa and K3, even though its membrane-perturbing potential is supposed to be much less destructive. However, our hemolytic and antimicrobial assays on the same batches of erythrocytes and using the same bacterial strains showed that MAP is in fact highly lytic towards eukaryotic cells and weakly antibiotic against prokaryotes. These unfavourable activities render its suitability and previous applications as a cell penetrating peptide questionable.
Unlike MAP, the HIV-TAT peptide does not assume a well-defined secondary structure in lipid membranes, hence instead of characterizing the peptide structure we studied its effect on membranes by solid state 31P-NMR of the phospholipids. In the uncharged lipid DMPC it was observed that HIV-TAT induces an isotropic signal, which does not occur in negatively charged DMPG. We attribute this previously undescribed morphology to inverted rod-like micelles (as supported by electron microscopy), which would offer an explanation of the non-destructive cell penetrating mechanism across the lipid bilayer. The same but reduced effect was also observed for the less potent analogues oligo-arginine (R8) and oligo-lysine (K8). Finally, by CD spectroscopy we found the HIV-TAT peptide to assume a poly-proline (type II) conformation under certain conditions, hence we are now also including novel amphiphilic proline-rich peptides with cell penetrating properties in our comparative NMR analysis.
In addition to the biologically relevant results outline above, we have achieved significant progress in terms of developing several different spectroscopic methods. They are generally applicable to biomolecular structure analysis in the solid state, but with a specific focus on membranes and fibres. These techniques shall only be briefly summarized, even though they constitute our main set of tools to carry out all structural comparison and suggest functional models in the context of the “Comparative Genomics” programme.
· Synthesis of 19F-labelled peptides
In order to increase the intrinsically low sensitivity of conventional solid state NMR experiments based on selective 2H- and 15N-labels, and to avoid the difficulties associated with selective 13C-labelling, we have introduced a strategy based on selective 19F-labels. Several different non-natural amino acids have been used and assessed in terms of their ease of chemical synthesis and their suitability with regard to structural perturbations in the peptide: 4F-phenylglycine (F-Phg), 4-CF3-phenylglycine (CF3-Phg), 3F-alanine (F-Ala), 3,3,3-F3-alanine (F3-Ala), 3,3,3-F3-aminoisobutyric acid (F3-Aib), and CF3-bicyclopentaneglycine (CF3-Bcp). For most of these we have worked out protocols for their incorporation into synthetic peptides, by overcoming specific problems of racemization (F-Phg and CF3-Phg), HF-elimination (F-Ala and F3-Ala), as well as unresponsive coupling (F3-Aib).
The subsequent 19F-NMR structure analysis in macroscopically oriented samples is based on measuring either the anisotropic chemical shift of a single 19F-label or the homonuclear dipolar coupling of a CF3-group. By combining several such parameters from a series of labelled peptide analogues, the secondary structure this peptide can be confirmed and its alignment determined with the aid of self-written software. In practice, using 19F-NMR the effective improvement in sensitivity was found to be a factor of 10 and 100 respectively, compared to 2H- and 15N-NMR. This gain is very significant in terms of NMR time and the amount of material required, especially when very low peptide concentrations are addressed or when a series of different samples and experimental conditions are systematically screened (see examples above). The main challenge that has so far restricted a routine use of 19F-labels is the need for specialized 19F-NMR hardware such as an additional high-band channel and double/triple-tunable Teflon-free NMR probes. We have set up such commercial equipment in Karlsruhe and are currently building further specialized probes on our own (see below).
In a multiply or uniformly isotope-labelled sample it is generally impossible to resolve the overlapping chemical shift anisotropy (CSA) patterns of the individual sites. Nevertheless, it would be extremely useful to access these anisotropy parameters of especially 13C and 15N in biomolecules. The CSA values constitute valuable structural data that can be utilized by the COSMOS-NMR force-field to calculate and refine three-dimensional molecular structures from experimental NMR data (see below). We have thus applied a novel 2D solid state NMR experiment called RAI (Recoupling of Anisotropy Information) that is able to correlate the isotropic chemical shifts with the respective CSA powder pattern. A recent improvement of this magic angle spinning (MAS) technique has taken the rotation frequencies up to 35 kHz, which allows significantly better resolution besides a flexible scaling behaviour that enables even very broad CSA pattern to be addressed. For example, at very high spinning speed the 19F-NMR parameters were obtained for 19F-labelled amino acids and other biomolecules. The 2D iso-aniso experiment was successfully applied to (i) native silk fibroin from B. mori, in which all natural-abundance 13C resonances and CSA parameters of the hexarepeat sequence (GAGASA)n could be resolved and assigned, and (ii) to uniformly 13C-labelled cellulose from a bacterial source, in which all carbon atoms on both glucose rings in each unit cell of the crystal lattice could be analyzed.
· Molecular dynamics simulation using the COSMOS-NMR force-field
Using the isotropic and anisotropic 13C chemical shifts determined by the RAI experiment above, we have refined the low-resolution crystal structures of silk fibroin and cellulose that were available from fiber-diffraction analysis. To this aim, molecular dynamics simulations were performed using the group’s designated force-field COSMOS-NMR, which utilizes experimental NMR parameters as pseudo-forces. In the case of silk fibroin and cellulose, 13C chemical shifts were used as input and compared with predicted values that are calculated on-the-fly from a set of previously obtained 13C parameters. For further applications such sets of parameters are also being evaluated semi-empirically for 19F and 15N, using both crystal structure data and ab initio calculations. Besides obtaining the refined three-dimensional molecular structures, another major result from such an MD-supported biomolecular structure refinement is the ability to calculate the orientations of the principal axes for each individual chemical shift tensor within the molecular frame with high accuracy, which is a fundamental requirement for any further orientational analysis or correlation experiments.
We have recently extended the COSMOS-NMR force-field to utilize not only internal constraints (such as chemical shifts and distances) but also external constraints such as tensor components that are determined in oriented 19F-, 2H- and 15N-NMR experiments on the peptide systems described above. The first successful all-atom MD simulation using anisotropic NMR parameters was demonstrated on deuterated pyrene embedded in a lipid bilayer. The experimental 2H-NMR quadrupole splittings could be reproduced in a conventional MD simulation only by including the entire lipid bilayer and H2O after 20 ns and ensemble averaging. In contrast, COSMOS-NMR was able to align the molecule dynamically and reproduce the splittings accurately within a 1 ns MD, regarding only the few atoms of the pyrene. An analogous simulation worked out well for a membrane-embedded sterol, and is currently being applied to the PGLa peptide (see above).
· Design and assembly of solid state NMR probes
In order to carry
out NMR experiments which require unusual or more specialized hardware, such as
19F
Our current NMR approach using single labels on synthetic peptides in oriented samples still suffers from the work- and cost-intensive steps necessary to prepare and analyze a series of multiple related peptides analogues. Therefore, we have implemented oriented circular dichroism (OCD) as a complementary spectroscopic technique which can reveal the qualitative alignment of a peptide with high sensitivity and without the need for isotope labelling. Having built a rotatable sample holder for oriented membranes at controlled humidity, our first spectra on the peptides described above (GS, PGLa, K3, MAP, etc.) are promising and clearly reveal the same re-alignment that has been observed by solid state NMR under comparable conditions. With this new OCD method at hand we are thus in a position to speed up the process of screening the experimental conditions under which interesting and functionally relevant structural changes occur in our peptides. After identifying the key conditions by OCD, such peptide-lipid systems can then be specifically studied by solid state NMR with quasi-atomic resolution.
Future research plans
We intend to proceed with all successful projects in the same manner as outlined above using solid state NMR, to reach full structural and functional conclusions for each membraneous or fibrous system. The fusogenic and antimicrobial peptides will most likely be exhaustively described once the respective binding and oligomerization parameters have been quantitatively determined and correlated with genetic sequence and 3D structure. The cell-penetrating peptides and transport proteins will still require several more years of work, before we will have reached a similar level of understanding as for the aforementioned peptides.
It is also obvious that over the next few years our emphasis will shift more and more from synthetic peptides with selective labels to uniformly labelled recombinant membrane proteins, which will be analyzed with the aid of RAI and PISEMA NMR-experiments. The PISEMA technique developed by Opella and Cross makes it possible to determine the alignment of helical peptide segments from a single 2D spectrum. We have acquired such PISEMA spectra of our first 15N-labelled peptides in oriented samples. To find also the azimuthal rotation angle of a helix it is further necessary to assign some of the resonances in the PISEMA spectrum, for which an amino acid specific labelling strategy has been presented by Marassi. Such 15N-labelling scheme may also be necessary to assign the RAI spectra of the more complex membrane proteins that are currently under investigation in our group. Additionally, 15N-15N correlation spectra and PDSD experiments will also be useful for assignment and structure analysis. Once the chemical shift anisotropy parameters have been resolved in RAI experiments, these data are expected to provide a new route towards NMR-structure analysis of membrane proteins. Combined with the possibility to calculate and refine the structures of entire proteins by all-atom MD simulations from chemical shifts and orientational constraints, this should significantly extend the current capabilities of solid state NMR. We hope to be able soon to address more complex membrane proteins and fibers with multiple or uniform labels and complicated three-dimensional structures. Our group has spent some considerable effort on producing uniformly labelled samples either by fermentation, or by cloning and overexpression in E.coli. So far we have succeeded to prepare the following peptides and proteins recombinantly with 13C- and/or 15N-labels for solid state NMR investigation.
· Gramicidin S is produced with high yield in B. brevis (up to 800 mg per litre), having carefully optimized the growth conditions on specially labelled medium. First RAI experiments will be performed on these samples to carry out an assignment of the peptide in the solid state, and PISEMA will be conducted on the membrane-bound peptide to verify our previous 19F-NMR analysis by 15N-NMR.
· The HIV-TAT cell penetrating peptide shall be described in terms of its structure in the membrane-bound state by 13C- and 15N-NMR (as opposed to its effect on the lipid bilayer investigated so far). Specifically the possibility of bidentate complex formation between the arginine side chains and the lipid phosphate groups in DMPC and DMPG will be explored using 15N-31P REDOR distance measurements.
· The Tat-A protein is the major constituent of the “twin arginine translocation” protein export machinery, which secretes fully folded proteins (including co-factors) across the cellular membrane. The Tat-system from B. subtilis had been functionally reconstituted and overexpressed in E. coli, and we have now obtained the first uniformly 15N-labelled samples in mg amounts, for both the entire polypeptide sequence (70 amino acids) as well as of the extramembraneous portion alone. The first 15N-NOESY-HSQC spectra of this Tat-A segment in TFE look promising for a structure analysis in solution. This model structure shall then be docked onto the membrane to aid the analysis of a subsequent PISEMA experiment of the entire protein, whose spectrum may need to be simplified by considering the transmembrane segment and extramembraneous portion separately.
· The bacterial mechanosensitive channel protein MscL (136 amino acids) has been crystallized in the closed state, but its structure is yet unknown in the open state. A simple PISEMA experiment may reveal the changes in helix alignment upon channel opening (which can be triggered by the addition of lyso-lipids to the membrane-reconstituted protein). To this aim we have overexpressed MscL in mg amounts with uniform 15N-labels, and the first TROSY spectra in detergent micelles suggest an intact structure. Additionally, fully deuterated protein has been prepared for neutron scattering, which revealed a significant change in the overall shape of the protein upon the addition of lyso-lipids. The reconstitution of MscL into macroscopically oriented membranes for PISEMA is currently being optimized.
· The E5-protein from Papilloma virus is a small transmembrane protein (44 amino acids) that associates with the PDGF-receptor and causes its constitutive activation, which can lead to permanent signal transduction and thus oncogenesis. E5 is a dimeric protein that is linked by two disulfide bridges, and it is postulated to assemble via specific helix-helix interactions with the transmembrane segments of the PDGF-R. Having prepared synthetic E5 peptide rather inefficiently up to 2003, we recently succeeded to express the recombinant protein in E. coli with better yield. We intend to carry out a PISEMA analysis to examine the helix-helix interactions within E5 and between E5 and the PDGF-R segments, since very little is known about such molecular interfaces occurring within a hydrophobic lipid bilayer. Hence, monomeric E5 without disulfide bridges will be compared with dimeric E5, and finally with the ternary PDGF-R complex.
· Having completed the 3D structure refinement of silk fibroin from B.mori in the silk II conformation of the drawn fiber, we will next address the same type of protein from S.c. ricini, which has a similar overall amino acid composition but a different genetic sequence that resembles spider silk more than any other silk worm. Additionally, it will be attempted to refine the structure of silk fibroin in the silk I conformation, which corresponds to the state before spinning the fiber, a process in which intramolecular hydrogen-bonds between β-turns are extensively re-shuffeld into a pattern of intermolecular bridges connecting extended β-strands. Biological fibers will continue to be of interest, because natural fibers and other synthetic polymers are utilized as biocompatible scaffolds for cell cultures in tissue engineering. The chemical, conformational and mechanical properties of these materials determine the cell-adhesion properties and biodegradation. Hence, we will continue with the available biopolymers and include further extracellular matrix molecules and synthetic shape-memory polymers in our solid state NMR analysis.