Intnoduction

The interaction between small molecules and membrane proteins is a significant topic from the biological and pharmaceutical point of view. Detailed information on interaction mechanisms can be a guide for clarifying functions and structural features of membrane proteins, modifying small molecules as medicinal drugs, and searching for new drug targets on the protein. However, few successful interaction analyses at the level of molecular structure using full-length membrane proteins have been reported to date because there are still crucial problems. Namely, membrane proteins generally require individually optimized treatment, it is difficult to crystallize membrane proteins with phospholipid, and NMR spectra of membrane proteins usually have low resolution due to low mobility in lipid bilayer. To obtain information on interaction mechanisms while avoiding these problems, I tried interaction analysis using the molecular complex of a transmembrane fragment of a membrane protein and a natural product with NMR techniques.

In this research, domain IV segment 6 (DIVS6) of voltage-gated sodium channel (VGSC) and a neurotoxin veratridine (Figure 1) were chosen for the model system. Voltage-gated sodium channel is responsible for generating action potential in the nervous system. The primary structure of VGSC which consists of a (260 kDa), [31 (36 kDa), and [32 (33 kDa) subunits indicates four homologous domains of six a-helical segments each in a subunit. Veratridine, a lipid soluble alkaloid extracted from lily sabadilla seed, is considered to bind segment 6 of each domain and excites VGSC as a partial agonist. DIVS6 has the largest number of interactive residues among four segment 6 as suggested by point mutation experiments of VGSC.

Transmembrane model peptides (TMPs) corresponding to the sequence of DIVS6 (Table 1) were synthesized on solid phase. TMP1 and 2 dissolved in aqueous 2,2,2-trifluoroethanol (TFE) known as a helix-promoting solvent were analyzed firstly by solution NMR to confirm the interaction within the model system. I started my Ph.D. course study from NMR structure analysis of TMP1 and 2 in the solvent. Secondly, TMP3 reconstructed in dimyristoylphosphatidylcholine (DMPC) liposomes was analyzed by solid state NMR.

Analysis in solution

Peptide synthesis and NMR titration experiments with neurotoxic veratridine (VTD) and its non-toxic analog veracevine (VC) were performed in my master course study. Especially in the titration experiment on TMP 1, the observed chemical shift perturbation pattern was same between VTD and VC. VTD and VC share the common steroidal backbone structure but VTD has 3-acyl group which plays an important role for the neurotoxicity. The results suggest that the steroidal backbone of VTD interacts to DIVS6, and neurotoxicity of VTD is caused by interaction to another segment located near DIVS6 such as DIVS4 or DIS6 in the neuronal membrane.

In the next step, structures of TMP1 and 2 were calculated using NMR data and the simulated annealing software package XPLOR-NIH (version 2.21). Distance restraints derived from NOE peak intensities and dihedral-angle restraints derived from coupling constants were used to refine the structure. Refined and accepted structures of TMP 1 and 2 were given by calculations repeated 20 times and superimposed for each (Figure 2). Superimposed structures of TMP1 (Figure 2a) show that the center part of the peptide forms a-helix while N-terminus does not converge to a single conformation. Thus N-terminus of TMP1 does not take rigid conformation in the mixture of TFE/H20 = 1/1. In the titration experiment on TMP1, the amide proton signal of Lys8 disappeared. This disappearance indicates that structure of N-terminus was changed by interaction with VTD. The result from the structure calculation where N-terminus of TMP 1 fluctuates supports this consideration. Thus Lys8 and surrounding residues were eliminated from interaction site candidates. The structures of TMP2 (Figure 2b) shows that N-terminal three lysines do not form a-helix and slightly fluctuate as TMP 1.

Interacting residues suggested by titration experiments with VTD are shown on calculated structures in Figure 3. In TMP1, chemical shift perturbation was observed at the hydrophobic side of the helix, which is consistent with point mutation experiments. Therefore VTD is expected to interact with TMP1 in a similar manner to interaction with VGSC. Combined with the results in titration of VC, it is possible that this hydrophobic side faces other segment which is related to activation by VTD. Further interaction analysis on other segments and VTD is expected to give information on tertiary structure of VGSCs (Figure 4).

As for TMP2, VTD could interact to N-terminus region of the peptide. A noteworthy point is the chemical shift perturbation at the amide proton of Leul4 of both TMP1 and 2. Leul4 is thus suggested to be the key residue for interaction between VTD and TMPs. Leul4 of TMPs is L1580 in the sequence of a rat skeletal muscle VGSC, Nav l .4 Gil). In point mutation experiments, μ1.-L1580 was not included in required residues for binding. NMR experiments in this study indicated μ1-L1580 as a new possible VTD binding site.

Analysis in lipid bilayer

Peptide synthesis

Since native sequence of DIVS6 is too hydrophobic for normal conditions of peptide synthesis and purification, three lysines as hydrophilic amino acids are added on each terminus of the DIVS6 sequence at the design of TMP3 also to direct the helix perpendicular to lipid bilayer. While unlabeled TMP3 was synthesized on solid phase peptide synthesizer for CD measurement, (13)C, (15)N-labeled TMP3 for solid state NMR measurements was synthesized manually on solid phase to achieve high yield (approximately 10%, total 34 coupling cycles including a few repetitions).

Optimization of reconstitution conditions

The amino acid sequence of DIVS6 predicts that DIVS6 forms a-helix in lipid bilayer. To reconstitute TMP3 in lipid bilayer with a-helical conformation, conditions frequently used to reconstitute membrane associated peptides and membrane proteins were screened and evaluated with CD and solid state NMR analysis. The conditions can be separated into two methods, namely f dialysis and thin-film hydration. Conditions of dialysis resulted in the mixture of α-helix and β-strand in one sample. The sample where TMP3 forms a-helix uniformly in DMPC liposomes was obtained with one of the conditions of thin-film hydration as follows: TMP3 dissolved in hexafluoroisopropanol and DMPC dissolved in chloroform (molar ratio of peptide : lipid = 1 : 25) were mixed at the same volume, dried on vacuum line after solvent evaporation, and then hydrated with 5 mM phosphate buffer (pH 7.4) by vortex mixing.

Solid-state NMR measurement

First, cross polarization (CP) experiments at various temperatures were performed to confirm that TMP3 is reconstituted properly in lipid bilayer. CP which is one of the important techniques in solid state NMR enhances signals of dilute spins such as (13)C or (15)N by transferring polarization from abundant spins such as (1)H. This cross polarization produces the effect only to molecules with low mobility. So CP spectra can indicate the mobility of peptides in lipid bilayer.

1D CP spectra at three temperatures are shown in Figure 5. In the CP spectra, signals of TMP3 were observed at -60°C where sample is completely frozen, and at 10°C where DMPC is in gel phase and only molecules outside of lipid bilayer have high mobility. On the other hand, signals disappeared at 25°C where DMPC is liquid crystal phase. These CP experiments showed that mobility of TMP3 in the sample correlated with phase transition of lipid. To obtain more detailed information of TMP3 with lipid bilayer, (13)C (peptide)-observed (2)H (lipid)-selective (1)H (peptide)-demagnetization (CODSHD) experiment and water-edited CP experiment were performed. Signal intensities of a CODSHD spectrum reflects the distance between deuteriums of DMPC and protons of TMP3. The spectrum showed that most of TMP3 is located within lipid bilayer. A water-edited CP spectrum shows the distance between water and protons of TMP3. This experiment indicated that center part of TMP3 is located at the center of lipid bilayers. Thus it is confirmed that TMP 3 was reconstituted in lipid bilayer. Subsequently, 2D (13)C-(13)C (DARR) experiments were performed to assign carbon signals and to obtain constraint data for structure calculation. The 2D DARR spectrum with reasonable resolution for signal assignment were obtained at -60 °C with magic angle spinning (MAS) at 12.5 kHz and resonance at 175 MHz (Figure 6).

Structure calculation

Solid state NMR spectra of uniformly isotope labeled peptides or proteins have usually too low resolution to obtain structural constraints such as distances and dihedral angles. While site specific isotope labeling can simplify the spectra, a small number of structure constraints limits the calculative determination of whole molecular structure. Thus, combination of NMR experiments and computer simulation is needed. In this research, combination of NMR and replica exchange molecular dynamics (REMD) simulation with an implicit membrane model was applied to structure determination of TMP3 in DMPC liposomes.

REMD simulation generated transmembrane α-helical structure as major conformational state at 300 K. The generated structures were subjected to (13)C chemical shift prediction by SHIFTX. Backbone of NMR structures of TMP3 was selected from major conformational state by comparison of predicted and observed chemical shifts. Subsequently, orientation of the selected backbone structures with lipid bilayer was evaluated with spectral calculation of CODSHD and water-edited CP. The structure whose 1D simulated spectra was consistent with observed spectra was selected as the NMR structure. Finally structure of TMP3 in lipid bilayer was obtained from REMD simulation (Figure 7). The obtained structure and the crystal structure of pore segments of ion channels showed some similarities.

Conclusion

Analysis in solution

NMR titration experiments using TMP1 with VTD and VC indicated that the steroidal backbone of VTD can interact to DIVS6 and 3-acyl group of VTD can cause neurotoxicity by probably interaction to another segment. Structures of TMP1 and 2 in a helix-promoting solvent were obtained by simulated annealing method. Combination of the calculated structure and results of titration experiments indicated the perspective of interaction: a slight structure change can occur around Lys8 by interaction with VTD due to fluctuation at N-terminus of TMP 1; the possible interaction site of TMP1 shown on the calculated structure is located at hydrophobic side of α-helix; Leu14(μ1-L1580) in VGSC was newly suggested as VTD binding site. The interaction manner of VTD suggested by titration experiments agreed with the expected manner by competitive binding study and point mutation experiments reported before. Obtaining the interaction information by NMR analysis in solution phase using model system was achieved.

Analysis in lipid bilayer

Reconstitution of TMP3 to lipid bilayer was achieved by screening a broad array of conditions. CP, CODSHD and water-edited CP experiments on solid state NMR indicated that TMP3 was reconstituted within lipid bilayer. The reconstituted TMP3 gave 2D DARR spectra with assignable resolution. The NMR structure of TMP3 was obtained by using combination of NMR experiments and REMD simulation with implicit membrane model. Reconstituted TMP3 gave the structure information necessary for interaction analysis using model system.

Figure 1. Structures of veratridine and veracevine

Table 1. Amino acid sequences of DIVS6 and TMPs

underline : interaction site suggested by point mutation study

Figure 2. Superimposed structures of TMP 1 (a) and TMP2 (b)

Figure 3. Suggested interaction site (colored) on the calculated structure of TMPI

Figure 4. Suggested interaction model

Figure 5. 1D CP spectra of TMP3

Figure 6. 2D DARR spectrum of TMP3 with signal assignments

Figure 7. An NMR structure of TMP3 calculated by REMD

本論文は全5章からなり、第1章の序論では背景として本研究の主題となる電位作動性ナトリウムチャネルタンパク質とそこに結合し活性化または抑制することが知られる天然由来の低分子有機化合物が紹介されている。次にこれらの相互作用に関して脂質二重膜に結合したチャネルタンパク質と有機低分子との結合様式を直接観測することでの困難さと、この解決を目指した本研究での指針が記述され、本研究の位置付けと意義が明確になっている。本研究で用いたユリ科植物由来の神経毒であるベラトリジンの化学構造式を以下に示す。

第2章では上記ナトリウムチャネルとベラトリジンの相互作用に関してこれまでに報じられた前者の遺伝子操作(点変異)による実験結果に基づき24回膜貫通構造のうち一つの膜貫通ヘリックス構造(D4S6)に着目し、このモデルペプチド(20~26アミノ酸残基)を化学合成し溶液中でのベラトリジンとの相互作用を1H NMR(水素核核磁気共鳴)により調べた経緯が述べられている。溶媒にはペプチドのヘリックス構造を促進する50%トリフルオロエタノール水溶液を用いてアミド結合に関与する窒素に結合した水素核をすべて帰属しヘリックス構造を確認した上で、ベラトリジンとの相互認識を化学シフト摂動により観測した結果、双方の疎水性構造部位で結合していることを示している。この結果はすでに点変位実験により報告されている結果と概ね合致するが、ベラトリジンのアシル基(上記構造式の左部分)を書くことで毒性を示さない天然物であるベラセビンとも同様の結合が観測され、毒性の発現にはこのアシル基とD4S6近傍のヘリックス構造との相互作用が関与している可能性を提唱している。これはこれまでに報告のない新しい知見である。

第3章では前章で用いたモデルペプチドを脂質二重膜に再構成する際の条件検討が詳細に述べられており、最適化した再構成条件により得られたものに関して、重水素化したリン脂質とペプチドとの固体NMRによる磁化移動の実験データと計算化学的シミュレーションによりペプチドが脂質二重膜を一回貫通している構造を取っていることを示している。次いで前章と同様にベラトリジンと再構成ペプチドとの固体NMRによる化学シフト摂動による二重膜中での相互認識の観測を行ったところ、溶液中と同様の結果を得たことが述べられている。以上第2章、3章の実験内容と結果に関しては各章に詳細に記述されており、追試可能となっている。

第4章では以上の研究結果を踏まえた結論と、本研究手法により明らかにされるべき将来展望が述べられており、第5章は本論文での引用文献がリストアップされている。

以上、本論文の研究内容は従来困難であった脂質二重膜中でのタンパク質と低分子化合物との複合体に関するNMRによる化学構造解析に関して、ここでの問題を克服するための手段を示唆するとともに、今後の本分野での問題解決における指針を示しているとの判断が審査委員全員の賛同により認められた。

なお、本研究のうち第2章に関しては当初原田雅典、山垣亮との共同研究により遂行され、第3章のうち固体NMR実験の部分に関しては藤原敏道、池田啓介、および江川文子(以上大阪大学蛋白質研究所)との共同研究であるが、本論文に記された実験の立案と実施、およびこれらの結果の解析と考察は論文提出者のものであり、その寄与は十分であると判断できる。

従って、本論文提出者である吉中 藍は、博士(理学)の学位を授与できるものと認める。