Structural studies of membrane proteins using transmission electron microscopy

From The Royal Institute of Technology, School of Technology and Health Structural studies of membrane proteins using transmission electron microscop...
Author: Cory Freeman
0 downloads 0 Views 5MB Size
From The Royal Institute of Technology, School of Technology and Health

Structural studies of membrane proteins using transmission electron microscopy Qie Kuang

Stockholm 2015

I

All previously published papers were reproduced with permission from the publisher. Published by The Royal Institute of Technology. Printed by Universitetsservice US-AB.

© Qie Kuang, 2015 ISBN 978-91-7595-468-4

II

ABSTRACT Membrane proteins play important roles for living cells. They control transportation of ions, solutes, and nutrients across the membrane and catalyze metabolic reactions. Transmission electron microscopy has its advantages in convenient sample preparation, straightforward structural determination, and wide applications for diverse specimens. In this thesis, the structure of three membrane proteins are studied by this method. Kch, a potassium channel in Escherichia coli, has a transmembrane part and a cytosolic domain. Large and well-ordered two dimensional crystals were obtained from both a functional mutant (KchM240L) and a modified protein possessing only the transmembrane part (KchTM). Both samples crystallize as two symmetry-related overlapping layers. Furthermore, the KchTM structure was reconstructed which showed that the transmembrane part of the two adjacent proteins are involved in forming the crystal contacts. Thus, the cytosolic domains of Kch in crystals are deduced to expose to the solvent and do not interact with each other. MGST1 (microsomal glutathione transferase 1) is a detoxification enzyme. It was recombinantly over-expressed in the current study, instead of purified from rat liver as before. The crystallization condition was adjusted and isomorphic crystals were obtained. The refined model was built from a combined data set consisting of previous and new diffraction patterns. More residues at the C-terminus of the transmembrane helix 1 were assigned and the residues in the transmembrane helices 3 and 4 were remodeled. Several phospholipid molecules were observed and the ligand glutathione adopts an extended conformation in the refined model. The structure of MelB (a sugar/sodium symporter in Escherichia coli) was determined using a refined single particle reconstruction method. This novel method is aimed for processing small or locally distorted crystals. In comparison with the previously published single particle reconstruction protocol, the current method is improved in several aspects. A more reliable reconstruction of MelB was obtained and the resolution was increased. The docking experiment indicates that MelB adopts an open conformation under the present two dimensional crystallization condition. Electron microscopy has developed quickly recently with the help of modern instruments, techniques, and software. This method will without doubt play a more critical role in future structural biology.

III

ABSTRAKT Membranproteiner har viktiga funktioner i levande celler. De kontrollerar transport av joner, näringsämnen och andra lösta ämnen över membranet och katalyserar metaboliska reaktioner. Transmissionselektronmikroskopi är en analysmetod som kan användas för ett brett spektrum av tillämpningar för olika prover. Dess fördelar är bl.a. enkel provpreparering och möjligheten att bestämma fasinformationen från bilderna. I denna avhandling studeras tre membranproteinstrukturer med hjälp av elektronmikroskopi. Kch, en förmodad kaliumkanal i Escherichia coli, har en transmembrandel och en intracellulär domän. Stora och välordnade tvådimensionella kristaller kunde fås fram från både en funktionell mutant (KchM240L) och ett modifierat protein som utgörs endast av den transmembrana delen av proteinet (KchTM). Båda proverna kristalliserade som tvåsymmetrirelaterade överlappande lager. En tredimensionell karta från KchTM rekonstruerades, vilket visade att transmembrana delar av två intilliggande proteiner är involverade vid bildning av kristallkontakterna. Därför antas proteinets cytosoliska domäner vara exponerade till lösningsmedel och inte interagera med varandra. MGST1 (mikrosomalt glutationtransferas 1) är ett avgiftningsenzym. I den aktuella studien överuttrycktes MGST1 som ett rekombinant protein, i stället för att renas från råttlever som tidigare. Förhållanden för kristallisering justerades och de producerade kristallerna var isomorfa. En förfinad 3-D atommodell byggdes från ett kombinerat dataset bestående av tidigare och nya diffraktionsmönster. Fler aminosyror vid C-terminalen av transmembran alfahelix 1 kunde bestämmas och aminosyror i transmembran alfahelixar 3 och 4 modellerades om. Flera fosfolipidmolekyler observerades och liganden glutation antar en förlängd konformation i den förfinade modellen. Den tredimensionella kartan av MelB (ett transportprotein som är ansvarigt för i socker/natrium symport i Escherichia coli) bestämdes med en förbättrad metod för singlepartikelbearbetning av tvådimensionella kristalldata. Denna metod är idealisk för processning av små eller lokalt förvrängda kristaller. I jämförelse med den tidigare publicerade metoden, är den nuvarande förbättrad i flera avseenden. En mer pålitlig rekonstruktion av MelB skapades och upplösningen ökades. Dockning med en atomstruktur av ett besläktat protein indikerar att MelB antar en öppen konformation under de rådande villkoren för tvådimensionell kristallisering. Elektronmikroskopi har utvecklats snabbt påsenare tid med hjälp av moderna instrument, tekniker och programvara. Denna metod kommer utan tvekan ha en mer avgörande roll inom framtida strukturbiolog.

IV

LIST OF PUBLICATIONS This thesis includes the following articles, which are referred to by their Roman numerals I-IV. Published papers are permitted to be reprinted from the copyright owners. I Kuang, Q., Purhonen, P., Jegerschold, C., and Hebert, H. (2014). The projection structure of Kch, a putative potassium channel in Escherichia coli, by electron crystallography. Biochim. Biophys. Acta 1838, 237-243. II Kuang, Q., Purhonen, P., Jegerschold, C., Koeck, P.J., and Hebert, H. (2015). Free RCK Arrangement in Kch, a Putative Escherichia coli Potassium Channel, as Suggested by Electron Crystallography. Structure 23, 199-205. III Kuang, Q., Purhonen, P., Alander, J., Svensson, R., Hoogland, V., Winerdal, J., Spahiu, L., Wadlund, O.A., Armstrong, R., Jegerschold, C., Morgenstern, R., and Hebert, H. A refined atomic model for microsomal glutathione transferase 1 from electron crystallography. Manuscript. IV Kuang, Q., Purhonen, P., Pattipaka, T., Ayele, Y.H., Hebert, H., and Koeck, P.J. (2014). A refined single particle reconstruction procedure to process two dimensional crystal images from transmission electron microscopy. Microsc. Microanal., In revision.

V

CONTENT Part I Introduction 1 Membrane proteins ................................................................................................................... 1 1.1 Properties of membrane proteins ................................................................................... 1 1.2 Detergent and lipid ......................................................................................................... 1 1.3 Mutually adapted lipid and protein ................................................................................ 1 2 Structural studies ....................................................................................................................... 2 2.1 X-ray crystallography .................................................................................................... 2 2.2 Nuclear magnetic resonance spectroscopy ..................................................................... 2 2.3 Electron microscopy ...................................................................................................... 2 2.4 Computational modeling ................................................................................................ 3 2.5 Other methods ................................................................................................................ 3 2.6 Membrane protein structures.......................................................................................... 4 3 Electron microscopy ................................................................................................................. 4 3.1 The electron microscope ................................................................................................ 4 3.1.1 Source.................................................................................................................. 4 3.1.2 Lens and aperture ................................................................................................ 5 3.1.3 Stage .................................................................................................................... 5 3.1.4 Recording system ................................................................................................ 6 3.1.5 High voltage and high vacuum............................................................................ 6 3.2 Methods in electron microscopy .................................................................................... 6 3.2.1 Electron crystallography ..................................................................................... 7 3.2.2 Single particle reconstruction .............................................................................. 7 3.2.3 Electron tomography ........................................................................................... 8 3.2.4 Helical reconstruction ......................................................................................... 8 3.3 Special issues in electron microscopy ............................................................................ 8 3.3.1 Radiation damage ................................................................................................ 8 3.3.2 Image and diffraction .......................................................................................... 8 3.3.3 CTF effect ........................................................................................................... 9 4 Individual projects..................................................................................................................... 9 4.1 Potassium channel .......................................................................................................... 9 4.1.1 Pore-forming domain ........................................................................................ 10 4.1.1.1 Selectivity filter ...................................................................................... 10 4.1.1.2 Two gates................................................................................................ 11 4.1.2 Regulatory domains .......................................................................................... 12 4.1.2.1 Voltage gated sensor domain .................................................................. 13 4.1.3 Kch .................................................................................................................... 13 4.2 MAPEG........................................................................................................................ 15 4.2.1 Two main pathways ........................................................................................... 15 4.2.2 Structures of FLAP, LTC4S, and MPGES1 ....................................................... 16 4.2.3 MGST1, MGST2, and MGST3 ......................................................................... 17 4.2.4 Activity of MAPEG members ........................................................................... 18 4.3 Transporters.................................................................................................................. 19 4.3.1 Differences between channels and transporters ................................................ 19

VI

4.3.2 Alternating access model................................................................................... 20 4.3.3 Classification of transporters............................................................................. 21 4.3.4 The major facilitator superfamily fold .............................................................. 21 4.3.5 MelB.................................................................................................................. 22 Part II Methods 5 Techniques used in this thesis ................................................................................................. 23 5.1 Protein expression ........................................................................................................ 23 5.2 Purification ................................................................................................................... 24 5.3 Two dimensional crystallization................................................................................... 24 5.4 Sample preparation ...................................................................................................... 25 5.5 Electron crystallography image processing.................................................................. 27 5.5.1 Processing each individual image ..................................................................... 27 5.5.2 Map generation.................................................................................................. 28 5.6 Electron diffraction processing .................................................................................... 28 5.6.1 Steps in electron diffraction processing ............................................................ 28 5.6.2 Common crystallographic terminology ............................................................. 29 5.7 Single particle reconstruction processing..................................................................... 29 5.7.1 Steps in single particle reconstruction ............................................................... 29 5.7.2 Single particle reconstruction to process two dimensional crystal images ....... 30 Part III Results 6 Aim of investigation ................................................................................................................ 31 7 Summary of individual papers ................................................................................................ 31 7.1 Papers I and II: Projection structures of KchM240L and KchTM ............................... 32 7.2 Paper II: 3D structure of KchTM ................................................................................. 33 7.3 Paper III: A refined model of rMGST1 ........................................................................ 35 7.4 Paper IV: Single particle reconstruction processing of MelBec images........................ 36 Part IV Discussion and Conclusion 8 Future perspectives ................................................................................................................. 38 8.1 Kch ............................................................................................................................... 38 8.2 rMGST1 ....................................................................................................................... 38 8.3 MelBec .......................................................................................................................... 39 8.4 Single particle reconstruction processing of two dimensional crystal images ............. 39 8.5 Future of electron microscopy ..................................................................................... 40 9 Conclusion .............................................................................................................................. 40 Part V Epilogue 10 Acknowledgements ............................................................................................................... 41 11 References ............................................................................................................................. 44 Papers I-IV ................................................................................................................................. 56

VII

LIST OF ABBREVIATIONS 2D Two dimensional AA Arachidonic acid CDNB 2,4-Dinitrochlorobenzene CNBD Cyclic nucleotide-binding domain CTF Contrast transfer function EC Electron crystallography EM Electron microscopy FLAP 5-Lipoxygenase activating protein GSH Glutathione K2P Tandem pore domain potassium Kch Potassium channel in Escherichia coli Kligand Ligand gated potassium Kir Inwardly rectifying potassium Kv Voltage gated potassium LTC4S leukotriene C4 synthase MAPEG Membrane Associated Proteins in Eicosanoid and Glutathione metabolism MelB Melibiose permease MFS Major facilitator superfamily MGST Microsomal glutathione S-transferase MP Membrane protein MPGES Microsomal prostaglandin E synthase PDB Protein database PGE2 Prostaglandin E2 PSF Point spread function RCK The regulator of the conductance of potassium ion SF Selectivity filter SPR Single particle reconstruction TM Transmembrane α-helix TNB Trinitrobenzene VSD Voltage gated sensor domain

VIII

Part I Introduction 1 Membrane proteins Proteins perform all kinds of processes for organisms, e.g., maturation and senescence. There are two kinds of protein, depending on whether they are embedded in solution or lipid, soluble and membrane proteins, respectively. 1.1 Properties of membrane proteins About 30% genes in the human genome encode membrane proteins (MPs) (Wallin and von Heijne, 1998). MPs play vital roles in all biological processes, including transport of solutes and macromolecules across the membrane (channels and transporters), signaling in the cell (receptors), and carrying out enzymatic reactions (metabolic enzymes). In general, two forms of arrangement in the transmembrane part of MPs are observed: α-helical and β-barrels. The majority of transmembrane parts are α-helical, while β-barrels have been found only in bacterial outer membrane. In my thesis, I will present work on a channel (papers I and II), a metabolic enzyme (paper III), and a transporter (paper IV). 1.2 Detergent and lipid Since MPs prefer lipid environments, they tend to aggregate in solution and lose their functions. Therefore, detergents are used to shield the hydrophobic surface of MPs to mimic the lipids in the experiments. Detergent monomers are assembled to micelles above the critical micelle concentration, whereas below this threshold, the monomeric form of detergent remains. Detergents fall into one of the three categories: ionic, nonionic, and zwitterionic, depending on the properties of their headgroups (Garavito and Ferguson-Miller, 2001). The strong ionic ones may denature MPs (lose their native conformations). The micelles are fluid and exchange with the monomers in solvent rapidly (Garavito and Ferguson-Miller, 2001). Detergent micelles form sphere-shaped balls and phospholipids are constituted into planar layers in the cell membrane. These bilayer cell membranes are mosaic and fluidic (Engelman, 2005). Three types of lipids are defined, depending on the interactions with a MP: 1) annular lipids which are similar to the ones in the bilayer; 2) non-annular lipids which are frequently located in the cavities and clefts of the MP; and 3) integral lipids which specifically bind to the MP (Hunte, 2005). Phospholipids usually have two fatty acyl chains, whereas detergent has one hydrophobic tail. Cholesterol is enriched in lipid raft together with other phospholipids. It consists of four sterol conjugated rings and may mediate certain functions of the MP. 1.3 Mutually adapted lipid and protein It is expected that the length of each transmembrane α-helix (TM) matches the membrane thickness in order to avoid the exposure of its hydrophobic part to solution. However, proteins are dynamic and can adopt different conformations in their reaction cycles; the membranes are fluidic and may have different thicknesses at different locations and/or different stages. Many factors can affect the membrane thickness, including lipid composition, fatty acyl chain length, chain saturation, cholesterol content, and temperature. The lipid and protein are suggested to be mutually adapted with each other, where a MP can change the thickness of the surrounding lipids

1

and vice versa, a MP can modify itself to accommodate into the membrane (Killian, 1998). Several possible adaptations to mismatch are proposed (Killian, 1998), such as protein aggregation, acyl chain disordering, surface orientation of TM, helix tilt, and protein backbone conformational change. 2 Structural studies The natural proteins are made of 20 essential amino acids. These amino acids are arranged in a specific way to fold into a protein. Although the structure of a protein is dictated by its amino acid sequence, it is actually the folding of the amino acids which results in a functional protein (Anfinsen, 1973). The aim of the structural studies is to determine the conformation of the proteins. In other words, the structural studies provide direct views of the proteins and other macromolecules. With this information, the mechanism of reactions can be explained, which helps for designing drugs. Four main methods are applied to study a protein structure: x-ray crystallography, nuclear magnetic resonance spectroscopy (NMR), electron microscopy (EM, EM stands for transmission electron microscopy which is the sole tool used to study the protein structures in this thesis), and computational modeling. 2.1 X-ray crystallography X-ray crystallography determines a protein structure from a three dimensional (3D) crystal, where the protein molecules are ordered in 3D. The basic knowledge of crystallography is described in '5.6.2 Common crystallographic terminology'. The amplitudes can be measured directly from the diffraction patterns, whereas the phases are lost and can be determined by various methods including multiple isomorphous replacement, molecular replacement, and multiple-wavelength anomalous dispersion (Chiu, 1993). When both amplitude and phase information is known, the electron density map can be calculated followed by building the model of the protein structure. 2.2 Nuclear magnetic resonance spectroscopy In contrast to x-ray crystallography, this technique can study a low molecular weight (< 40 kiloDalton (kDa)) protein in solution (Poget and Girvin, 2007). The solid-state NMR method could be applied to the MP in an aligned lipid bilayer. NMR determines the physical and chemical properties of atoms. The cross-peaks in 2D spectra are interpreted as geometrical constraints (Chiu, 1993), thus a family of structures (ensemble) is generated. This technique is widely used in studying dynamics, kinetics, and chemical environment of the molecules. 2.3 Electron microscopy An electron microscope resembles a light microscope, but it can visualize the object in much more detail (at higher resolutions). In practice, resolution is defined as the smallest distance between two point-like objects at which they can still be distinguished as individual entities. According to the Rayleigh criterion, the resolution in the x-y plane (perpendicular to the optical axis) is dx,y = 0.61λ/NA, where λ is the wavelength of the beam and NA is the numerical aperture of the objective lens (Huang et al., 2009). In conventional light microscopy, the resolution limit in the

2

z-direction (along the optical axis) is 2-3 times (500 nm) as large as the one in the x-y plane (200 nm) due to the point spread function (PSF) (Huang et al., 2009). Since the wavelength is much shorter in electron microscope (λ is about 0.035 Å for 120 keV electrons; the higher the accelerating voltage of electrons, the shorter the wavelength), the obtainable resolution in EM is sufficient to distinguish atomic details. Because electrons have a very limited penetrating power, only thin objects (< 300 nm) can be analyzed in EM (Hohmann-Marriott et al., 2009). Thick samples, such as tissue, first need to be sectioned. One unique feature of EM in structural studies is its broad range of specimens, from atoms to cells. Different methods of EM can be applied for different objects including crystalline, helical, and single particle samples (see also '3 Electron microscopy'). Compared to x-ray crystallography, the convenience of EM is that the phases can be recorded from the collected images since the scattered electrons can be focused by the lenses in electron microscope. 2.4 Computational modeling The introduction of this section is adapted from a recent review (Werner et al., 2012). Protein structure prediction methods can be grouped into two classes: template-based and template-free. The template-based methods can be further divided into homology modeling and threading, whereas the template-free ones include the de novo and ab initio modeling. Homology modeling is based on the assumption that the protein in question (target) adopts a similar structure to its evolutionarily related homolog (template). Therefore, the target can be predicted if the template structure is experimentally determined. On the contrary, the threading modeling designs the target from the recognized folds (fold: a structural motif which includes a combination of several structural units) instead of from its homology, since the same fold is shared between different proteins. De novo modeling combines the principle of folding with the knowledge from references (previously solved structures). The structure of the target is chosen as the one having the lowest potential energy in the force field when the target structure changes. Ab initio modeling calculates the energy functions based on first principles of energy and atomic motion to predict the target structure without prior information from references. Since the all-atom models are often used, this method is limited to small molecules (< 10 kDa). 2.5 Other methods Various biochemical and biophysical methods complement the ones described above. They are also important tools and provide information on the proteins of interest in different aspects. Several of them related to the papers in this thesis are mentioned below. A). Electrophoretic gel-based separation methods Proteins can be separated based on their properties, such as their molecular weights and isoelectric points. The SDS-PAGE (sodium dodecyl sulfate-polyacrylamide gel electrophoresis) is often used to analyze the quality of the sample. In contrast, blue native-PAGE can analyze the sample in its native state since only mild detergents are used to solubilize the MP. Furthermore, 2D gel electrophoresis provides a better separation than 1D gel since the samples have an additional separation depending on their isoelectric points. B). Spectroscopy Plenty of spectrometry methods can detect the conformational changes of a MP, e.g., using circular dichroism, ultraviolet-visible spectrophotometry, and small-angle x-ray scattering. One of

3

them, mass spectrometry was used in paper I to validate the protein sample. Mass spectrometry identifies samples by their mass to charge ratios. Furthermore, the conformational alteration of a MP may lead to a change in its solvent accessibility, which can be recorded with the help of the mass exchange between the hydrogen and deuterium atoms. C). Fluorescence microscopy This technique can record the signal from the proteins labeled with fluorescent molecules. Its application includes protein-protein interactions, protein folding, and protein location. 2.6 Membrane protein structures More than 50% of the current drugs are targeted to MPs (Drews, 2000). However, MP structures occupy only approximately 2% among the ones deposited in the protein database (PDB), much less than the soluble proteins (98%). The majority of solved MP structures are determined by x-ray crystallography, with approximately 2% by EM. The first structure solved by x-ray crystallography was a soluble protein (Kendrew et al., 1958). Nearly 20 years later, the first MP structure was published to an intermediate resolution (Henderson and Unwin, 1975). After another 10 years, the first α-helical MP atomic structure was available (Deisenhofer et al., 1985) together with the β-barrels MP structure (Weiss et al., 1990). With the advent of cryo (data is collected from the specimens at low temperatures)-EM and modern microscopes, Henderson and his colleagues established the method of electron crystallography (EC) (Henderson et al., 1990; Henderson et al., 1986) and determined the first MP structure at atomic resolution using this method (Henderson et al., 1990). The highest resolution (1.9 Å) is obtained by EC (AQP0, PDB: 2B6O, (Gonen et al., 2005)) among all structures determined in EM so far. A MP (TRPV1, PDB: 3J5P and EMD-5778, (Liao et al., 2013)) was determined at 3.3 Å by single particle reconstruction (SPR) using a direct detector. 3 Electron microscopy Since the topic on this thesis is to use EM to analyze the MP structure, EM is discussed in the following text. 3.1 The electron microscope The electron microscope can provide high resolution information but it is costly. Fig. 1 shows the Jeol 3000SFF electron microscope. The main components are: electron source, lens system, specimen stage, and recording system. 3.1.1 Source Electrons can be generated by different kinds of source. A tungsten or lanthanum hexaboride filament is used in the thermal emission gun. The field emission gun can work at room temperature, where electrons are emitted by application of a strong electric potential to the tip. Since the field emission gun has advantages in providing a more coherent and brighter illumination which are beneficial for determination of the structure to a higher resolution, this kind of source is most frequently used in data collection. Thermal emission guns can work efficiently in screening negatively stained samples (negative staining technique is introduced in '5.4 Sample preparation').

4

3.1.2 Lens and aperture The electromagnetic lenses in electron microscope perform a similar function as the ones in light microscope, to focus the beam. In contrast, they are made of wire coils, instead of the glass in light microscope. The current flowing through the coils generates a magnetic field, which rotates the beam. The lenses can be divided into three categories: condensor, objective, and projector (including the intermediate lenses) lenses. 1) The condensor lenses determine the illuminated area; 2) The objective lens acts to primarily focus the beam and it initially magnifies the image as well; 3) The normal function of the projector and intermediate lenses is to magnify the image coming from the objective lens and to project the magnified image onto a recording system. Either a magnified image or a magnified electron diffraction pattern can be recorded by properly positioning the projector lens system. An aperture, a metal disk with a hole in the center, is always inserted into the lenses to limit the illuminated angles. No lens is perfect and the main aberrations are spherical and chromatic ones, which reduce the image resolution. On the other hand, the spherical aberration of the objective lens together with defocusing gives contrast to the samples which makes them visible in the image.

Fig. 1. The Jeol 3000SFF electron microscope. The schematic drawing (A) and the corresponding components of the microscope (B) are linked by arrows. For clarity, the intermediate and projector lenses are not shown. These lenses are for magnification purposes. The image plane is represented by the phosphorous screen, one of the recording systems. This figure is adapted from Martin Lindahl's course slides (HL2026), 2010. 3.1.3 Stage The specimen holder carrying the samples can be inserted into the stage, by either top- or side-entering. The axial symmetry design of the top-entry holder minimizes vibration and specimen drift induced by thermal expansion of the holder (Fujiyoshi, 1998), whereas the advantage of the side-entry holder is its convenience to tilt the specimen in data collection. The modern cryo-transfer device allows the exchange of specimen in less than 10 minutes (Fujiyoshi, 1998). 5

3.1.4 Recording system The exit electrons carrying sample information are visualized on the viewing screen or recorded on a detector such as photographic film or camera. Photographic film emulsions contain silver halide grains. When electrons hit the emulsion, silver halide is converted to metallic silver. After developing and fixing, the exposed regions appear as dark areas on the supporting film (Kitts, 1996). Charge coupled devices (CCDs) collect the data via converting the incident electrons to visible light followed by formation of an electronic read-out. Direct detectors avoid the intermediate light conversion step and record the incident electrons directly (Faruqi and McMullan, 2011). Detector performance can be evaluated quantitatively using the detective quantum efficiency (DQE), which describes how effectively the camera can produce an image. Compared to CCD, film has a higher DQE (meaning having a superior contrast performance), detects a larger number of available pixels, and works better for information at higher resolutions (Faruqi and McMullan, 2011). On the other hand, CCD offers much greater linearity in a dynamic range (Wang and Downing, 2011) and works better than film for information at low resolutions (Faruqi and Henderson, 2007). Besides that, CCD provides on-line results and avoids the tediously digitizing process of films. The K2 Summit in super-resolution mode (the new generation of a direct detector from Gatan) has the highest DQE among the three detectors mentioned above (Ruskin et al., 2013). The properties of having very low noise and a high DQE in the direct detector make it possible to detect small and low-contrast samples (Li et al., 2013). Furthermore, since a direct detector records the data at high speeds, the beam-induced motion of the particles in data collection can be corrected (Li et al., 2013). This rapidly developing technology has already been applied to determine the structure of a number of proteins (Allegretti et al., 2014; Amunts et al., 2014; Cao et al., 2013a; Li et al., 2013; Liao et al., 2013; Lu et al., 2014). 3.1.5 High voltage and high vacuum Since the wavelength is inversely proportional to the square root of the accelerating voltage, a higher voltage results in a shorter wavelength, which can theoretically enhance detection of the specimen at a higher resolution. Choosing a higher voltage also improves the data collection in several aspects: a flatter Ewald sphere, decreased events of double-scattering, and minimization of the influence of specimen charging. Thus it makes data collection possible from thicker specimens (Henderson, 1995; Massover, 2007). When electrons travel through the microscope, they are easily scattered. In order to obtain a coherent electron beam, it is necessary to maintain a high vacuum in microscope. Several pumps work together, each with its own efficiency. The high vacuum in microscope makes the observation of live samples impossible. Therefore, the samples are either directly frozen in vitrified ice or embedded in other media which are discussed in ' 5.4 Sample preparation'. 3.2 Methods in electron microscopy Only thin samples can be analyzed in EM and these samples usually can only sustain one exposure due to the high radiation damage. Therefore, only one image, a projection of a 3D object along the beam axis is collected each time. The 'Central Section Theorem' (De Rosier and Klug, 1968) states that the Fourier transform of a projection image forms a central section in the 3D Fourier transform of the object. Therefore, by combination of images collected at different directions to

6

fill Fourier space, the object can be reconstructed (Fig. 2). EC, SPR, and electron tomography (ET) are three major methods in EM. Although these methods are suitable for different kinds of sample and differ in detail, the principles (the Central Section Theorem and reconstruction process) are similar.

Fig. 2. 3D reconstruction in EM. The images of an object (represented by a duck) are collected at different directions. These images are Fourier transformed and combined together to reconstruct back to the duck. The reconstruction process can be performed either in reciprocal space or in real space. The black arrows (indicating the beam directions in the left panel), thick lines (representing recorded images in the left panel), and thin lines (corresponding to central sections in the middle panel) are depicted.

3.2.1 Electron crystallography 2D crystals, where the protein molecules are ordered in 2D, are analyzed in EC. This method is particularly suitable for studying of MPs, which are reconstituted into lipid membranes (see also '5.3 Two dimensional crystallization' and '5.5 Electron crystallography image processing'). Since the tilt angles accessible in microscope are limited to around ± 70°, a certain part of data is missing. This so called missing cone problem leads to elongation of the structure along the beam direction (z-axis, perpendicular to the x-y plane) and the resolution along the z-axis is worse, as compared to the one in the x-y plane (anisotropic resolution). However, missing cone may not be a serious problem when the images tilted to high angles are included (Chiu, 1993; Ford and Holzenburg, 2008). EC is used to study MP structures in papers I-III and 2D crystals from different MPs are analyzed in all papers included in this thesis. 3.2.2 Single particle reconstruction SPR is a method to reconstruct the object from images of many identical protein molecules. These protein molecules are separated (thus called single particles), instead of forming 2D crystals as in EC. Ideally, the particles analyzed in SPR are oriented randomly on the grid, although in some cases, they have preferred orientations (Special methods are applied to solve the structure of the objects in these cases, such as the random conical tilt reconstruction technique (Radermacher et al., 1987)). The orientation of each particle is assigned by either the common line method if there is no reference or projection matching with a reference. Since the particles are randomly oriented in the majority of cases, there is no missing cone problem in SPR. The advantages of SPR are: 1) crystal formation is not needed which is still the bottleneck in EC; 2) the data collection can be automated; and 3) sample preparation and data collection are easier as compared to EC. The main

7

limitation for this method is that only proteins with large molecular weights (the cutoff is 250 kDa, (Unger, 2001)) are suitable, since these particles are large enough to be precisely aligned to each other to increase the signal-to-noise ratio. I applied a SPR procedure to process the 2D crystal images in paper IV. 3.2.3 Electron tomography ET allows determining the structure of individual cell organelles and bacterial cells to nanometer resolutions (Unger, 2001). One unique feature of ET is that the whole data set is built from a tilt series of images taken of a single copy of the object. Therefore, heterogeneous samples can be studied using this method (Zhang and Ren, 2012). Since the same object is exposed several times in a tilt series, the sample would be damaged if the same electron dose was used as in EC or SPR. Therefore, each image is collected at a lower dose, making the total dose over the entire tilt series comparable to the one used in low dose imaging (Subramaniam and Milne, 2004; Unger, 2001). When the sample is collected from a single-axis tilt, the data has a missing wedge. The improvement using double-tilt can alleviate this problem and gives a missing pyramid in the data. Fiducial markers are used for the alignment purpose due to the low contrast and possible sample movement induced by the beam (Subramaniam and Milne, 2004). 3.2.4 Helical reconstruction Some proteins are arranged to have a helical symmetry, in which the protein molecules can be superimposed to each other by rotational and translational movements along the screw axis. Therefore in principle, one image contains all views of the protein molecule which are sufficient for reconstructing the object (Unger, 2001). Furthermore, there is no missing part in the data. The EC and SPR methods can be combined together to solve their structures (Unger, 2001). Several MP structures solved by this method have reached near atomic resolutions, such as the acetylcholine receptor (Miyazawa et al., 2003). 3.3 Special issues in EM Several issues specific to EM are discussed in this section. 3.3.1 Radiation damage During data collection, the electrons interact with the sample and inevitably damage it. The damage can be caused by heating, mass loss, breakage of covalent bonds, charge alteration and ionization, and generation of free-radicals (Massover, 2007). Several approaches have been developed to reduce the electron beam damage. Low-dose exposures ( ~15 e−/Å2) and 'minimum dose systems' (three modes of operation: search, focus, and photo modes) are used for data collection (Fujiyoshi, 1998). The samples are usually protected at low temperatures during the entire procedure. The lower temperature of liquid helium (4K) provides at least two times more protection than using liquid nitrogen (77K) (Fujiyoshi, 1998). The higher voltage reduces the radiation damage as well (Henderson, 1995; Massover, 2007). 3.3.2 Image and diffraction When the electron beam passes through the sample in microscope, it interacts with the sample and the interactions can be divided into two classes: elastic and inelastic scattering. In the former, the

8

energy (wavelength) of the electrons does not change but the direction of them changes; in the latter, the energy of the electrons changes. The transferred energy in inelastic scattering changes the state of the sample and then damages it. Only elastic scattering waves contribute to the sample signal in an image and a diffraction pattern, whereas inelastic scattering gives rise to background noise (Amos et al., 1982). One great advantage of EM is that both images and diffraction patterns can be collected, therefore, the amplitudes extracted from the diffraction pattern and phases obtained from the image can be combined together to solve the structure. The image formation can be regarded as the interference of the elastically scattered waves and the unscattered waves (Amos et al., 1982). On the other hand, the diffraction formation does not rely on interference, and only the reflections (spots in the diffraction pattern) are recorded. The higher frequency waves give the diffraction spots further away from the origin (thus the diffraction extends to a higher resolution). Since only intensity can be recorded in the recording system, the phase information in diffraction is lost. 3.3.3 CTF effect In reality, image formation is influenced by the phase contrast transfer function (CTF). The final image obtained in microscope can be described as the projection of the object convoluted with the PSF. Regarding that convolution in real space corresponds to multiplication in Fourier space (reciprocal space) mathematically, the previous sentence can be described as: Fourier transform of the final image equals Fourier transform of the projection multiplied by the CTF (Wade, 1992). The CTF effect is generated by defocusing and the spherical aberration of the objective lens, and can be calculated according to the formula: CTF (θ) = -2 sin [(2π/λ)(-zθ2/2 + Csθ4/4)], where θ is the scattering angle; λ is the wavelength; Cs is the spherical aberration; and z is the defocus value, positive if the image is taken at underfocus (Wade, 1992). In practice, the CTF effect is modified by several other optical factors, including the chromatic aberration of the objective lens, partial coherence of the illuminating beam, astigmatism, and amplitude contrast (Amos et al., 1982). Due to the sine expression of the CTF effect, the phases are shifted by 180°for the odd-numbered Thon ring regions. The CTF effect should be corrected for all images recorded in microscope. 4 Individual projects In this thesis, three MP structures are analyzed using EM. They are a putative potassium channel (Kch), a metabolic enzyme (MGST1), and a secondary transporter (MelBec). Several important issues are picked up, since all these proteins are actively studied. 4.1 Potassium channel Potassium (K+) channels ubiquitously exist in all kingdoms of life, except for some parasites and inner organelles (Kuo et al., 2005). The main function of K+ channels is to transport K+ ions between the extracellular environment and the cell, since the cell membrane is almost impermeable to K+ ions. Depending on the number of TMs in the protein and different properties of the protein, K+ channels can be divided into three categories: voltage gated potassium (Kv), inwardly rectifying potassium (Kir), and tandem pore domain potassium (K2P) channels (Buckingham et al., 2005). Kv channels (6 TMs) sense the membrane potential change (Long et al., 2005a, b; Long et al., 2007); Kir channels (2 TMs) transport K+ ions from the extracellular environment to the cell (Kuo et al, 2003); and K2P channels (4 TMs) are usually constitutively

9

open and set the membrane potential (Butterwick and MacKinnon, 2010; Miller and Long, 2012). Furthermore, K+ channels can be stimulated by various kinds of ligand. This group of channel is called Kligand (ligand gated potassium) channel, which contains either 2 or 6 TMs (Kuo et al., 2003a). Structurally, a K+ channel can be divided into two parts: the pore-forming domain and the regulatory domain. The pore-forming domain contains 2 TMs in each monomer and the regulatory domain is different in each class of K+ channels. The pore-forming domains are in charge of conduction of K+ ions and are embedded in lipid membranes. On the other hand, the regulatory domains sense various kinds of stimulus, which control the closing and opening of the pore-forming domains (called gating (Jiang et al., 2003)). Four K+ channels assemble together to form a tetramer and the tetramerization is required for channel functioning (Doyle et al., 1998). 4.1.1 Pore-forming domain Pore-forming domain is the smallest functional unit in the channel that can conduct K+ ions efficiently (107 ions channel-1s-1) (Sansom et al., 2002). K+ ions are selected among other ions when transported through the K+ channel (the selectivity ratio of K+ ions to sodium (Na+) ions is more than 10000, (Doyle et al., 1998)). The pore-forming domain contains an outer helix, a pore helix, the selectivity filter (SF), and an inner helix from the N- to C-terminus (Fig. 3). K+ ions move down the electrochemical gradient from the helical bundle in the intracellular side, to the central water-filled cavity, next through the SF, and eventually to the extracellular side. K+ ions occupy several positions in the SF. 4.1.1.1 Selectivity filter The conventional K+ channel has a conserved sequence TVGYG75-79 (the sequence is based on KcsA, a bacterial K+ channel from Streptomyces lividans), called SF, where K+ ions are conducted efficiently and selectively (Doyle et al., 1998). In the ion conduction pathway shown in Fig. 3, K+ ions are only dehydrated in the SF (S1-S4) and kept hydrated in other regions (Sc at the intracellular side and S0 and Sext at the extracellular side). Four oxygens in the main chain of TVGY75-78 together with the one in the side chain of T75 point to the ion conduction pathway and surround the dehydrated K+ ions. Each K+ ion sits in the middle of two oxygen layers and four K+ ions can bind to four such evenly spaced K+ binding sites (S1-S4) in the SF, with one K+ ion in each binding site. Since the free energy difference between the K+ ions surrounded by waters and the ones surrounded by oxygens in the SF is small, K+ ions can be transported across the membrane efficiently. On the other hand, binding to the SF for Na+ ions is energetically unfavorable, which partially explains the selectivity of K+ ions (Doyle et al., 1998; Lockless et al., 2007). Four K+ ions adopt two configurations in the SF, either 1,3 or 2,4 configuration. In the 1,3 configuration, two K+ ions occupy the S1 and S3 sites with two waters in the S2 and S4 sites. In the 2,4 configuration, two K+ ions occupy the S2 and S4 sites with two waters in the S1 and S3 sites (Morais-Cabral et al., 2001; Zhou and MacKinnon, 2004; Zhou and MacKinnon, 2003; Zhou et al., 2001). The shift between two configurations can be achieved when a third ion enters from one side and pushes one ion away from another side or by movement of the ion-water queue (MacKinnon, 2003; Morais-Cabral et al., 2001; Zhou and MacKinnon, 2003; Zhou et al., 2001). The energy transfer cost between these two configurations is low (Morais-Cabral et al., 2001). The

10

repulsion of two occupied K+ ions in each configuration further facilitates the conduction of ions (MacKinnon, 2003; Zhou and MacKinnon, 2003). Other factors maintaining the integrity of the SF (such as the proteinous environment near the SF and the bound ions) are also important to keep conduction of K+ ions efficient and selective (Alam and Jiang, 2011; MacKinnon, 2003; Nimigean and Allen, 2011).

Fig. 3 The structure of KcsA, a prototype of K+ channel. The transmembrane part of KcsA is homologous to the pore-forming domain of other K+ channels. The conductive state of KcsA (PDB: 1K4C) is shown here. K+ ions (purple balls) and water (red balls) are depicted. The SF together with observed K+ ion binding sites (Sext, S0, S1-S4, and Sc) in the ion conduction pathway and the glycine hinge (Gly99) are labeled. EC: extracellular side; IC: intracellular side; OH: outer helix; IH: inner helix; PH: pore helix; HB: helical bundle. Only two diagonal monomers in the tetramer are shown.

4.1.1.2 Two gates The majority of K+ channels are closed in the resting state, opened with certain stimuli, and then enter into a nonconductive state (Norton and Gulbis, 2010). Except for some K2P members that are constitutively open, gating is tightly regulated in other K+ channels (Brohawn et al., 2012; Miller and Long, 2012). Two kinds of gate exist, where the extracellular one includes the SF and the intracellular one is at the position where the inner helices bend (Imai et al., 2010). The intracellular one was proposed originally by a structural comparison of KcsA (closed state) and MthK (a calcium induced K+ channel from Methanobacterium thermoautotrophicum, in an

11

open state). The inner helices cross to form a helical bundle (Fig. 3) to block the conduction of K+ ions in KcsA; on the other hand, the inner helices are bent and splayed open after a glycine in MthK (Jiang et al., 2002b) (Fig. 4A). These glycine hinges are conserved and located in a similar position in several bacterial K+ channels (such as G99 in KcsA and G83 in MthK shown in Fig. 4A). In eukaryotic Kv channels, PXP (P is proline and X is any residue) replaces the glycine hinge to bend the sixth helices (corresponding to the inner helices in the 2 TMs channel) for conduction of K+ ions (Long et al., 2005b).

Fig. 4 Two gates in K+ channels controlling conduction of K+ ions. (A) The intracellular gate where the inner helices bend. KcsA in a closed state (PDB: 1K4C, blue), KcsA in an open state (PDB: 3F5W, red), and MthK in an open state (PDB: 3LDC, orange) are viewed from the intracellular side. The glycine hinges are located in a similar position in these channels. IH: inner helix; OH: outer helix. (B) The extracellular gate in the SF. The conduction (PDB: 1K4C, blue), C-type inactivation (PDB: 3F5W, magenta), and flipped (PDB: 2ATK, gray) states of KcsA are depicted. Only two diagonal monomers are displayed in (B). The extracellular gate is located in the SF and a subtle conformational change in this gate can result in nonconduction. Several distorted SF structures have been observed in KcsA (Fig. 4B), including high Na+/low K+ (PDB: 1K4D, (Zhou et al., 2001)), C-type inactivation (PDB: 3F5W and 3F7V, (Cuello et al., 2010)), and flipped (PDB: 2ATK, (Cordero-Morales et al., 2006) and PDB: 3OGC (Cheng et al., 2011)) ones. When the SF adopts a structure that deviates from the conductive state (PDB: 1K4C, Figs. 3 and 4B), the channel may not be able to conduct K+ ions, regardless whether the intracellular gate is open or not (Cuello et al., 2010). These two gates are regulated by specific regulatory domains located N- and/or C- terminally to the pore-forming domain. 4.1.2 Regulatory domains Besides the pore-forming domains, Kv, Kir, and Kligand channels possess their own regulatory domains which regulate the gating of the channels. All regulatory domains have some freedom relative to the pore-forming domains and a channel may have more than one regulatory domains. The voltage gated sensor domains (VSDs) are located at the periphery of the Kv channel (Chen et

12

al., 2010; Clayton et al., 2008; Long et al., 2005a; Long et al., 2007) and sense the electron change in the membrane (Aggarwal and MacKinnon, 1996; Tombola et al., 2006). The cytosolic domains in Kir channels are located below their pore-forming domains and provide the binding sites for diverse intracellular regulatory mediators to interact (Hibino et al., 2010). Kligand channels have diverse kinds of cytosolic domain for their respective ligands, e.g., CNBD (cyclic nucleotide-binding domain, (Clayton et al., 2004)) and RCK (the regulator of the conductance of K+ ion, (Jiang et al., 2001)) domains. 4.1.2.1 Voltage gated sensor domain Since Kch studied in this thesis has a domain that corresponds to the VSD in a Kv channel, VSD is introduced in this section. The VSDs have been found in other voltage gated channels and enzymes as well (Catterall, 2010b; Murata et al., 2005; Ramsey et al., 2006; Sasaki et al., 2006). The VSD separates from its pore-forming domain and weakly attaches to the pore-forming domain from the adjacent subunit (Fig. 5), which indicates that the VSD has mobility relative to the pore-forming domain. Fig. 5 Extracellular view of MlotiK1 (a non-voltage gated K+ channel from Mesorhizobium loti, PDB: 3BEH). Four monomers (in different colors) form a biological unit. The VSD (in an ellipse) is composed of TMs 1 to 4 (S1-S4) and the pore-forming domain (in a rectangle) consists of S5 (corresponds to the outer helix in KcsA shown in Fig. 3) and S6 (corresponds to the inner helix in KcsA shown in Fig. 3). The pore helix is labeled as PH.

Up to eight positively charged residues have been found in the fourth helix (S4) in the VSD (Kuo et al., 2005), and among them, four residues (R1-R4) at the N-terminus of S4 are the most crucial ones responsible for charge movement during activation (Aggarwal and MacKinnon, 1996; Swartz, 2008). These positive residues are intercalated by hydrophobic residues (Kuo et al., 2005). In fact, this pattern rather than the positive residues determines electron charge translocation (Xu et al., 2010). The positive charges are counter-balanced by several negatively charged residues (E183, E226, E154, E236, and D259, based on PDB: 2R9R) located in other helices in the VSD (Chen et al., 2010; Long et al., 2005b; Long et al., 2007). These interactions are believed to assist the movement of S4 in the focused electric field across the membrane (a hydrophobic region of approximately 10 Å thickness, separated by the water assessable crevices at both ends) during gating (Catterall, 2010a; Chen et al., 2010; Long et al., 2007; Tombola et al., 2006). 4.1.3 Kch Kch, a putative ligand K+ channel in Escherichia coli (E. coli), was first reported by Milkman

13

(Milkman, 1994). It has a conserved signature sequence-TVGYG as in other K+ channels, but it doesn’t have any positively charged residue in S4 (Milkman, 1994). Besides its non-voltage gated sensor domain and pore-forming domain, it has a cytosolic RCK domain (Jiang et al., 2001). As in MthK, the kch gene encodes both the full-length channel and a soluble RCK domain protein (Jiang et al., 2002a; Jiang et al., 2001; Lundback et al., 2009). Free RCK monomers form a RCK dimer due to the strong interactions in the dimer interface (Jiang et al., 2001). The RCK dimer structure of Kch (PDB: 1ID1) in solution shares a similar fold as the dimer structure of MthK in solution and the dimer structures of MthK, BKca (a large conductance K+ channel, both voltage and calcium gated, from Homo sapiens), KtrA (a component of a prokaryotic K+ uptake system), and TrkA (a component of a major K+ uptake system in prokaryotes) in the octameric gating rings, although the RCK domain of Kch does not have a C-terminal subdomain as the one in MthK, BKca, or TrkA. Four soluble RCK proteins and four RCK domains in the membrane-bound form were proposed to assemble to form an octameric gating ring structure (Fig. 6) (Jiang et al., 2002a).

Fig. 6 Demonstration of the octameric gating ring arrangement. (A) Top view. The RCK monomer in one layer (blue box) interact with the adjacent RCK monomer in another layer (red box) to form a dimer through the flexible interface. The dimer interaction in the gating ring corresponds to the one in the isolated dimer in solution. Four such dimers are arranged to form an octameric gating ring through the assembly interfaces. (B) Side view, rotated 90°from (A). The function and regulation of Kch in E. coli are not yet completely understood, neither has the ligand for the RCK domain been identified yet (Jiang et al., 2001). Until now, no direct evidence shows that Kch forms a functional channel in vivo or in vitro: no discernible phenotype has been reported from the mutants, including deletion of the kch gene (Epstein, 2003; Kuo et al., 2003b); no report of electric conductance of Kch has been published by reconstitution or by heterologous expression. It has been suggested that Kch opens on certain environmental or metabolic conditions to maintain the cellular proton (H+) motive force constant from a gain of function experiment (Kuo et al., 2005; Kuo et al., 2003b). Kch may function to adjust the membrane potential, instead of taking up K+ ions (Kuo et al., 2005; Kuo et al., 2003b).

14

4.2 MAPEG Membrane Associated Proteins in Eicosanoid and Glutathione metabolism (MAPEG) superfamily is a group of enzymes embedded in membranes. MAPEG consists of six members in human: microsomal glutathione S-transferase 1 (MGST1), MGST2, MGST3, microsomal prostaglandin E synthase 1 (MPGES1), leukotriene C4 synthase (LTC4S), and 5-Lipoxygenase activating protein (FLAP) (Jakobsson et al., 1999). These enzymes are associated with production of various 20-carbon lipid (eicosanoid) mediators which are involved in a diversity of human physiological activities. MAPEG members ubiquitously exist in all kingdoms of life, except in archaea (Bresell et al., 2005). Glutathione (GSH) transferase activity or glutathione-dependent peroxidase activity has been demonstrated in most of these members, except for FLAP which has no enzymatic activity (Hebert and Jegerschold, 2007). Although the enzymes have different functions, they share a common fold where three subunits are arranged into a trimer. 4.2.1 Two main pathways The MAPEG members are involved in generation of two groups of eicosanoid mediator, prostaglandins and leukotrienes (Fig. 7). With certain stimuli, the phospholipases move to the nuclear envelope, endoplasmic reticulum, and Golgi apparatus (Funk, 2001). Several kinds of phospholipase can release arachidonic acid (AA), however, the Ca2+ activated type IV cytosolic phospholipase A2 (cPLA2) is particularly important for production of eicosanoid mediators (Uozumi et al., 1997). AA can be converted to a number of lipid mediators in different pathways. Fig. 7 Two pathways of generating prostaglandins and leukotrienes. AA is released from the membrane mainly by cPLA2, and then metabolized to prostaglandins (left) and leukotrienes (right). Different prostanoids can be generated from PGH2 by different terminal synthases in the prostaglandin pathway. For clarity, only the MAPEG members (FLAP, MPGES1, and LTC4S are in purple) and their respective lipid mediators are depicted. MGST1, MGST2, and MGST3 are introduced in '4.2.3 MGST1, MGST2, and MGST3'.

In the cyclooxygenase pathway, prostaglandin H2 (PGH2) is generated from AA through cyclooxygenase (COX1 and COX2). PGH2 serves as an intermediate product and is further converted to different prostanoids, including prostaglandins, prostacyclin, and thromboxane by their respective terminal synthases. One of the prostaglandins, PGE2, is catalyzed by MPGES1, a member of MAPEG. PGE2 has diverse functions, including mediating inflammation, pain, and fever (Samuelsson et al., 2007). In the lipoxygenase pathway, AA is first converted to leukotriene A4 (LTA4) by 5-lipoxygenase (5-LO) with the help of FLAP. The generated LTA4 can be either

15

hydrolyzed to produce leukotriene B4 (LTB4) by LTA4 hydrolase or conjugated with GSH to produce leukotriene C4 (LTC4) by LTC4S. LTC4 and its further metabolized products, leukotriene D4 (LTD4) and leukotriene E4 (LTE4), comprise the cysteinyl leukotrienes, which are involved in the inflammation processes, such as asthma (Evans et al., 2008; Martinez Molina et al., 2007). Besides these two well-studied pathways, AA can be metabolized through a third cytochrome P450-dependent pathway, where epoxyeicosatrienoic acids are produced (Evans et al., 2008; Zeldin, 2001). Since no MAPEG member exits in the third pathway, the enzymes in the first two pathways are discussed. 4.2.2 Structures of FLAP, LTC4S, and MPGES1 The MAPEG members share a common fold. The structure of FLAP (PDB: 2Q7M, (Ferguson et al., 2007)) aligns well with the structure of LTC4S (PDB: 2UUH, (Martinez Molina et al., 2007) and PDB: 2PNO, (Ago et al., 2007)) as shown in Fig. 8A. The GSH binding site was observed and the secondary substrate (LTA4) binding site was proposed from a structurally similar detergent molecule (n-dodecyl β-maltoside (DDM)) bound in the LTC4S structure (Martinez Molina et al., 2007). On the other hand, the FLAP structure (PDB: 2Q7M) provided the first view of a MAPEG member with its inhibitor (MK591, depicted in red in Figs. 8A and B) which competes with AA (Mancini et al., 1993). The GSH binding site (Fig. 8C) and the inhibitor binding site (Fig. 8B) do not overlap. However, the long acyl chain in the secondary substrate may insert into the inhibitor binding site (Fig. 8C). Another special feature in FLAP is its long C2 loop between the helices 3 and 4. Considering its closeness to the proposed secondary substrate binding site and the inhibitor binding site, as well as the results from the mutagenesis data (Ferguson et al., 2007), it is possible that this loop plays a role in AA transport. MPGES1 carrying out the isomerization reaction of PGH2 to prostaglandin E2 (PGE2), is a member of MAPEG superfamily. The overall MPGES1 structures in an open conformation (PDB: 4AL0 and 4AL1, (Sjogren et al., 2013) and PDB: 4BPM, (Li et al., 2014)) are similar to the LTC4S structure (PDB: 2PNO and 2UUH). However, the cytosolic half of TM 1 and TM 2 is displaced in the closed MPGES1 conformation (PDB: 3DWW, (Jegerschold et al., 2008)) as shown in Fig. 8D. Although the GSH binding sites of these two MAPEG members in open conformations are conserved (the GSH binding site of LTC4S is shown in Fig. 8C), different residues, S127 in MPGES1 and R104 in LTC4S (corresponding to R126 in MPGES1) are proposed to stabilize the thiol group of GSH (Hammarberg et al., 2009; Jegerschold et al., 2008; Sjogren et al., 2013).

16

Fig. 8 Structural comparison of FLAP, LTC4S, and MPGES1. (A) Alignment of FLAP (PDB: 2Q7M, blue) and LTC4S (PDB: 2UUH, brown). (B) Inhibitor binding site in FLAP. (C) GSH and DDM binding sites in LTC4S. DDM is proposed to occupy the secondary substrate binding site. Adjacent subunits in the trimer are depicted in different colors in (B and C). (D) MPGES1 (in a closed state, PDB: 3DWW, purple) is compared with LTC4S. The displacement of the cytosolic half of TM 1 and TM 2 is via the K26-D75 salt bridge. 4.2.3 MGST1, MGST2, and MGST3 Glutathione transferases (GSTs) are a group of phase II detoxification enzymes, which conjugate the electrophilic substrates with GSH (Fig. 9A) and make these substrates more water soluble to be excreted easily (Hayes et al., 2005). They ubiquitously exist in most life forms and can be divided into soluble and membrane-bound families. Although the members in these two classes

17

perform similar reactions, the soluble ones are dimers and the membrane-bound ones are trimers (Hayes et al., 2005).

Fig. 9 GST activities and the rat MGST1 (rMGST1) structure. (A) GSH transferase reaction, CDNB (2,4-Dinitrochlorobenzene) conjugates with GSH. (B) The 'dead-end' Meisenheimer complex formation with TNB (trinitrobenzene). The crossover arrow indicates that the product does not form. (C) Lipid peroxidase reaction. (D) Different GSH positions in MGST1 (newly refined model in paper III, red) and MPGES1 (PDB: 4AL0, cyan). The GSH positions in MPGES1 and LTC4S are similar and shown in Fig. 8C. The entrance of the secondary substrate in rMGST1 is suggested to be located in the interface between two subunits (Holm et al., 2006). MGST1, MGST2, and MGST3 are membrane-bound GSTs and all of them belong to the MAPEG superfamily. MGST1 is abundant in human liver as well as in rat (McLellan et al., 1989; Morgenstern et al., 1984). Besides the GSH transferase activity, it can reduce the oxidized lipids by its peroxidase activity as well (Fig. 9C) (Mosialou et al., 1993). MGST1 has a special feature that it can be activated to a large degree by various treatments (Morgenstern, 2005). One treatment, alkylation of C49 with sulfhydryl reagents, e.g., N-ethylmaleimide can activate MGST1 up to 30-fold (Svensson et al., 2004). Although MGST1 has a high sequence identity to MPGES1 (38%), it does not catalyze the isomerization of PGH2 as in MPGES1. GSH in MGST1 adopts an extended conformation (red GSH in Fig. 9D), rather than the horseshoe-shaped conformations in MPGES1 (PDB: 3DWW, 4AL0, and 4BPM, cyan in Fig. 9D) and LTC4S (PDB: 2PNO and 2UUH). The functions of MGST2 and MGST3 are less understood in vivo. Both proteins may be involved in the lipoxygenase pathway and can produce LTC4 from LTA4 in vitro (Jakobsson et al., 1996; Jakobsson et al., 1997). It is suggested that MGST2 and MGST3 are structurally more close to LTC4S than to MGST1, although their structures remain to be determined (Martinez Molina et al., 2008). 4.2.4 Activity of MAPEG members A). FLAP does not bind to GSH, nor does it have any enzymatic activity (Ferguson et al., 2007). B). LTC4S does not react with CDNB, a typical substrate for measuring the GSH transferase activity, nor does it have a lipid peroxidase activity (Ahmad et al., 2013). It catalyzes LTC4 18

production from LTA4. C). MPGES1 has both GSH transferase and peroxidase activities (Thoren et al., 2003). Its main function is to convert PGH2 to PGE2. D). MGST1 has both GSH transferase and peroxidase activities (Morgenstern, 2005). It has broad substrate specificities and mainly acts as a detoxification enzyme. It may protect the membrane from oxidative stress in cell as well. In the GSH transferase reaction, the Meisenheimer complex is formed as an intermediate step followed by leaving of the halogen group (Fig. 9A). The Meisenheimer complex is accumulated due to the absence of a leaving group when GSH reacts with TNB (trinitrobenzene) (Fig. 9B). E). MGST2 reacts with CDNB and it also catalyzes a lipid peroxidase reaction (Ahmad et al., 2013). It displays LTC4 production from LTA4 as well (Ahmad et al., 2013; Jakobsson et al., 1996). F). MGST3 shows a peroxidase activity, however, it does not react with CDNB (Jakobsson et al., 1997). 4.3 Transporters Cells have their plasma membranes to separate the inner parts of them from the extracellular environments. Membranes are impermeable to most ions and substances, which require certain proteins for their transportation. These proteins can be classified as channels (see also '4.1 Potassium channel') and transporters (Gadsby, 2009). 4.3.1 Differences between channels and transporters Channels and transporters carry out a similar task for living cells, transporting substrates across the membrane. The structure of channels may or may not resemble the structure of transporters. The fundamental differences distinguishing them (Gadsby, 2009) are summarized as follows: A). The direction of transportation Substrates are diffused through channels from the high electrochemical gradient side to the low gradient side (downhill), whereas substrates are moved by transporters in an uphill direction. Although uniporters transport the substrates down the solution gradient as in channels, the substrates bind to the protein on one side and are translocated to the other side. This property makes the structures of uniporters and transporters resemble each other. B). Energy consumption Downhill movement does not require energy, whereas uphill movement consumes energy. Thus, either the primary energy (adenosine triphosphate, ATP) or the free energy stored in the counter-transported substrate gradient could be utilized for transporters. C). Opening in the conduction pathway The essential difference between channels and transporters is whether the substrate conduction pathway can be opened at both sides or allow opening at one side at a time. In a channel, the conduction pathway is permeable at both ends when the gate is open. However, the conduction pathway is never opened at both sides in a transporter. Instead, it is always closed at least at one side. The transport usually develops occluded states, where the substrates are not accessible from either end. D). Translocation speed The translocation speed for channels is fast, e.g., the K+ ions are conducted very efficiently, at near

19

diffusion-limited rates (107 ions channel-1s-1) (Sansom et al., 2002), whereas the speed for transporters is several hundred ions per channel per second. The main reasons for a slower conduction of the substrates can be due to large conformational changes, substrates binding, and closure of substrate conduction pathway in a transport cycle for transporters. E). Current recording For ion passengers, the current can be recorded from a single channel. However, the current is far too small to be detected from a single transporter, since the translocation speed is several orders of magnitude slower in transporters as compared to channels. 4.3.2 Alternating access model The mechanism of transportation can be well explained by the alternating access model, where the essence is that the substrate binding sites are accessible to one side of the membrane at a time in one transport cycle (Jardetzky, 1966). The cartoon in Fig. 10 demonstrates a simplified transport cycle to explain how the model works. Fig. 10 Alternating access model in a transport cycle based on Jardetzky (1966) and Gadsby (2009). Three main states are identified as inward-facing (left), occluded (middle), and outward-facing (right). The substrate transported from the intracellular side of the cell to the extracellular side is depicted as a yellow ball and the counter-transported substrate in a reversed direction is shown as a red ball. The curved arrows indicate that both kinds of substrates are accessible when the intracellular gate is open (left) or the extracellular gate is open (right). The N- and C-terminal domains in a transporter are depicted as two ellipses in different colors. Notice that a transporter is never accessible to both sides and all steps are reversible in the cycle. A transporter has three main states: inward-facing, occluded, and outward-facing. The substrate conduction pathway is opened at the intracellular side in the inward-facing conformation (left icon) and is opened at the extracellular side in the outward-facing conformation (right icon), with both sides closed in the occluded conformation (middle icons). In the inward-facing conformation, the transporter has a high affinity to the transported substrates, but a low affinity to the counter-transported substrates. Then, the transported substrates can bind to the protein. The binding triggers the structural rearrangements from the inward-facing to the occluded conformation (middle up icon), which is followed by entering into the outward-facing conformation. In the outward-facing conformation, the transported substrates are released into the extracellular side due to their low affinities to the protein and a downhill gradient. Meanwhile, the protein has a high affinity to the counter-transported substrates. After binding of the

20

counter-transported substrates, the protein turns to another occluded state (middle down icon), followed by turning to the initial inward-facing state. This cycle can be reversed and the different protein conformations and the corresponding affinities to different substrates determine the direction of transportation. The alternating access model was proposed to explain the P-type ATPases (primary transporters) initially and phosphorylation/dephosphorylation by ATP induces the structural rearrangements of it. The original model with small adjustments can be applied to other transporters as well. 4.3.3 Classification of transporters The bacterial transporters can be classified into three main categories: primary transporters, secondary transporters, and group translocation systems (Law et al., 2008). 1) The primary transporters, such as ATPases and ATP-binding cassettes use the energy of hydrolysis of ATP for pumping the substrates across the membrane; 2) The secondary transporters employ two kinds of substrates, in which one kind of substrates travel downhill the gradient and another transport uphill. ATP hydrolysis is not involved in this class, instead, secondary transporters use the free energy from the movement of the substrates downhill to pump the other substrates uphill; 3) The group translocation systems first allow the substrates to bind to the protein at one side of the membrane, and then chemically modify them, followed by releasing them to the other side. Secondary transporters pump a wide range of substrates and can be divided into symporter and antiporter classes, depending on whether the substrates and counter-transported substrates move in the same (symporter) or opposite (antiporter) directions across the membrane. Furthermore, secondary transporters can be divided into over 100 families and superfamilies (Tsai and Ziegler, 2005). Since the structural studies of the transporter included in this thesis is from a secondary transporter having a major facilitator superfamily (MFS) fold, this fold is discussed in the following section. The MFS presents the largest superfamily of the secondary transporters (http://www.tcdb.org/search/result.php?tc=2.A.1, (Saier, 2000)) and accounts for about 25% of all transporter proteins (Saier et al., 1999). The MFS members are ubiquitous in all kingdoms of life and play vital roles in numerous physiological processes, e.g., glucose transporters in sugar transportation (Shi, 2013; Yan, 2013a). 4.3.4 The major facilitator superfamily fold Currently, a number of 3D structures of MFS members have been determined to atomic resolutions, including uniporters, symporters, and antiporters (The information can be found at http://blanco.biomol.uci.edu/mpstruc/. Several papers (e.g., (Yan, 2013a, b)) have listed some of them, in addition to some intermediate resolution structures from EM (Tsai and Ziegler, 2010)). Despite low sequence similarities and different kinetic mechanisms, the MFS members share a common fold. The following description of the MFS fold is adapted from Shi (2013) and Yan (2013a). Fig. 11A illustrates the general feature of the MFS fold. All MFS members appear to function as monomers and the core of the MFS fold is composed of 12 TMs. The fold can be separated into two parts, the N- and C-terminal domains, with 6 consecutive TMs in each domain. These two domains, despite in a low sequence similarity, are related to each other by a pseudo-twofold symmetry axis that is perpendicular to the membrane. Furthermore, each domain is evolved from a pair of inverted 3 TMs repeats, which can be

21

matched to each other by an approximate 180°rotation around an axis parallel to the membrane (Madej et al., 2013). These two repeats are intertwined with one helix from one pair inserted into the center of another pair. Due to the pseudo-twofold symmetry and the 180°rotation, TMs 1, 4, 7, and 10 locate in a similar position (in the center); TMs 2, 5, 8, and 11 are also observed in a similar position (on the minor axis side); TMs 3, 6, 9, and 12 occupy in a similar position (on the major axis side). A long linker between helices 6 and 7 usually exists to connect the N- and C-terminal domains and this linker is expected to help the large degree movement between two domains in an alternating access cycle.

Fig. 11 The MelBst structure (PDB: 4M64, Mol-A). (A) Extracellular side view shows a typical MFS fold. The TM 1-3 and TM 7-9 are in yellow; the TM 4-6 and TM 10-12 are in red; and the loops are in blue. TM 1-6 and TM 7-12 compose the N- and C-terminal domains, respectively. (B) The proposed cation binding (left to the K377 and W128 part) and sugar binding (right to K377 and W128) sites.

4.3.5 MelB Melibiose permease (MelB) is a secondary transporter, belonging to a glycoside-pentoside-hexuronide cation symporter family (TC 2.A.2). It utilizes the free energy released from the downhill movement of Na+, lithium (Li+) or H+ to uptake galactopyranosides. Although this process can be reversed, in which the protein can take up the monovalent cations when moving the sugar downhill (Guan et al., 2011), the cell usually uses this protein to accumulate the sugars. The configuration of the sugar affects the coupling of counter-transported cations. For instance, α-galactosides (e.g., melibiose and raffinose) can be counter-transported with all of the three kinds of cation, whereas β-galactosides (e.g., lactose) are counter-transported with Na+ and Li+ but not H+ (Wilson and Wilson, 1987). Recently, the 3D structure of MelB of Salmonella typhimurium (MelBst) was determined (PDB: 4M64, Mol-A and B) (Ethayathulla et al., 2014). A pocket formed by D55, D59, and D124 is suggested as the binding site for all of the three kinds of cation, and D59 is essential for H+ binding (Fig. 11B). The sugar binding site is close to the cation binding site. The closeness facilitates the cooperative binding and transportation. The water cavity formed by D19, R149, Y120, D124, W128, and K377 is suggested to be the sugar binding site (Fig. 11B). Furthermore, the authors determined an inactive structure of MelBst (PDB: 4M64, Mol-B) as well, in which both the sugar binding and the cation binding sites are collapsed. Both structures represent the 22

outward-facing occluded states, and the inactivate structure opens larger on the extracellular side. Considering its high primary sequence identity to MelBec, MelB of E. coli, (85% identity, (Mizushima et al., 1992)) the MelBst structure was used to dock into the MelBec map obtained in paper IV. Part II Method 5 Techniques used in this thesis This section covers the main techniques used in this thesis. They are: protein expression, purification, 2D crystallization, sample preparation, and image processing. The experimental procedure from sections 5.1 to 5.4 is depicted in Fig. 12. The image processing procedure is discussed in sections 5.5 to 5.7.

Fig. 12 Flowchart of the experimental steps from sections 5.1 to 5.4. After reconstruction of the plasmid harboring the target gene, the target protein can be over-expressed in various kinds of system. Purification is normally performed on columns to enrich the target protein. Next, a suitable crystallization condition is searched for the target protein to form an ordered 2D array in the presence of lipid molecules. The specimen (several microliters) is loaded on the grid followed by insertion into microscope. The target gene/proteins (red), impurities (brown), purification column (blue), lipids (one head group and two tails), pipette (triangle), grids (yellow), and holder (T shape) are depicted. All of the cartoon components are not drawn in their real sizes. 5.1 Protein expression Except for some rare cases where membrane proteins are abundantly rich in their native membrane environments (such as bacteriorhodopsin of Halicobacterium salinarium (Henderson and Unwin, 1975), water channel 0 (Gonen et al., 2004), and Na+,K+-ATPase (Jorgensen, 1988)), the majority of proteins are usually over-expressed to amplify their amounts. Different expression 23

systems include: bacterium, yeast, insect, mammalian cells, and cell-free expression (Derewenda, 2004). Each system has its advantages and disadvantages. Among them, the bacterial system is easiest to work with and all proteins included in this thesis are over-expressed in E. coli, followed by structural studies. The gene of the target protein is amplified by polymerase chain reaction and then the resulting plasmids are transformed into the competent E. coli cells. Under proper conditions, the target protein is stimulated to be produced. However, the over-expression efficiency varies in different conditions for each protein. The generally tunable parameters include host strain, plasmid, culture medium, inducer concentration, duration of induction, culture temperature, and additive (Papaneophytou and Kontopidis, 2014). Kch (papers I and II) and rMGST1 (paper III) were over-expressed in C43 and BL21DE3 (the original strain of C43) strains, respectively. The same kind of plasmid (pSP19T7LT) and medium (terrific broth) were used for both proteins, but the expression conditions were different. 5.2 Purification After over-expression of the target protein in the host (including all five kinds of the expression systems mentioned above), a purification step is performed to enrich the target protein and to get rid of impurities. The separation is based on the properties of the target protein, e.g., affinity interaction (affinity purification), molecular weight (size exclusion purification), surface charge (ion exchange purification), or hydrophobic interaction (hydrophobic purification) (Saraswat et al., 2013). An extra tag is often exploited to facilitate the recombinant protein expression and/or purification (Young et al., 2012). These tags include large (several dozens of kDa, such as Mistic (Membrane-Integrating Sequence for Translation of Integral membrane protein Constructs), (Kefala et al., 2007; Roosild et al., 2005)) and small (a few amino acids, such as polyhistidine tags) ones. Since the large tags may interfere with the oligomeric state, activity, and structure of the target protein, they are most frequently cleaved off from the tagged proteins. Different kinds of tag and their positions (at the N- or C-terminus) give different expression and purification efficiency. All proteins studied in this thesis were over-expressed with additional histidine tags. Size exclusion purification was performed to further improve the purity of the Kch proteins (papers I and II) and ion exchange purification was applied for rMGST1 (paper III). 5.3 Two dimensional crystallization Once a considerable amount of pure sample is obtained after purification, 2D crystallization experiments can be performed. In 2D crystals, the protein molecules are ordered in register in two dimensions (x-y plane/membrane plane). Since one crystal has one or up to several layers (but should be thin enough for electrons to penetrate) in the third dimension, it is suitable for electron crystallographic study. 2D crystal is formed since a net entropy is gained when a detergent solubilized sample is embedded in a lipid environment (Kuhlbrandt, 1992). The crystallization experiment is carried out by mixing the target protein surrounded by detergent molecules with the lipid-detergent micelles and thereafter removing the detergent. Dialysis, dilution, hydrophobic adsorption, and lipid monolayer are four common methods to clear away the detergent (Mosser, 2001). Dialysis was performed to reconstitute 2D crystals for Kch (papers I and II) and rMGST1 (paper III).

24

Successful crystallization experiment can result in different kinds of crystal, e.g., single layer, stacked sheets, tubular, and vesicular types (Mosser, 2001). The majority of Kch crystals were stacked sheets (papers I and II) and rMGST1 protein formed stacked sheets and separated single layer sheets (paper III). The previously obtained MelBec crystals were of the long tubular type (Hacksell et al., 2002). Crystal quality varies under different conditions (De Zorzi et al., 2013; Schmidt-Krey et al., 1998). In order to obtain large and well-ordered crystals, different parameters are screened. Protein, detergent, lipid, buffer, additive, crystal growth temperature, and crystallization procedure are general parameters to be considered (Mosser, 2001). However, all of these factors work together. For instance, the crystallization condition of rMGST1 from the sample over-expressed in E. coli was adjusted. Only small crystals were obtained using the previously published protocol (Schmidt-Krey et al., 1998) (Fig. 13A). Additional CaCl2 (Fig. 13B) or crystal growth at 30ᵒC (Fig. 13C) increases the crystal size, but the obtained crystals were heavily stacked and not sufficiently good for data collection. By combining these two factors, much larger crystals were obtained and the stacking problem was less severe (Fig. 13D). The electron diffraction data set was collected from rMGST1 grown in the adjusted condition (paper III). The crystallization conditions were screened manually in the current study. However, automatic screening by robots can facilitate the speed and success rate in each project (Hu et al., 2010; Iacovache et al., 2010; Wisedchaisri et al., 2011). 5.4 Sample preparation As mentioned in '3.1.5 High voltage and high vacuum', biological samples are either dehydrated or frozen to avoid water evaporation when they are analyzed in electron microscope. Accordingly, they can be prepared by either negative staining or cryo, respectively. In negative staining, the sample is embedded in heavy metals, in which the background is darker than the specimen; whereas the sample is embedded in either vitrified ice or sugar in the cryo method, where the contrast is reversed as compared to the negative staining method (Schmidt-Krey and Rubinstein, 2011). The negative staining method is usually used for analysis of the quality of the protein and crystals under different conditions. Since it gives a much higher contrast than the cryo method, the samples can be easily identified in the negative staining method. On the other hand, the stain may distort the protein molecule (De Carlo and Harris, 2011; Stoylova et al., 1998) and only the contour of the protein molecule is displayed. The heavy metal grain size limits the final resolution (Unger, 2000). The rMGST1 crystals after negative staining by uranyl acetate are shown in Fig. 13. Data is usually collected from cryo samples, where the proteins are preserved in their native states. However, this method is much more difficult to perform than the negative staining method, since: 1) the contrast formed between the specimen (protein or crystal) and the background (ice or carbon film) is low; 2) native samples do not tolerate the same amount of radiation damage; 3) specimens are easily contaminated by ice crystals; and 4) ice thickness should be proper. The protein molecules are often directly plunge-frozen in liquid ethane (Dubochet et al., 1988; Knapek and Dubochet, 1980) and the 2D crystals are commonly embedded in sugar medium, which mimics the effect of water by hydrogen bonding to the crystals (Wang and Downing, 2011). Successful preserving media include glucose (Henderson et al., 1990), tannin (Kuhlbrandt et al.,

25

1994), trehalose (Hirai et al., 1999), and mixture of these (Nogales et al., 1998). The 2D crystals of all projects in this thesis were preserved in trehalose. Some MelBec crystals were directly frozen in liquid ethane as well.

Fig. 13. rMGST1 crystals in different conditions. (A) initial condition (Schmidt-Krey, 1998). (B) same condition as in (A) with additional 50 mM CaCl2. (C) same condition as in (A), but crystallization was performed at 30ᵒC. (D) adjusted condition with 50 mM CaCl2 and crystals were grown at 30ᵒC. The crystal size increased significantly (some reach 5 μm in (D)) as compared to the initial condition (A). Although stacked crystals were grown on top of each other, one layer crystalline areas were found as indicated by the arrow in (D). The crystals in these four conditions were embedded in 1% uranyl acetate and all images were taken at the same magnification. The scale bar is 2 μm. Cryo-negative staining method was proposed to take advantages of both having a high contrast (from negative staining) and maintaining the native state of the specimen (from cryo) (Adrian et al., 1998; De Carlo and Harris, 2011). Diverse stains can lead to significantly different results for some proteins (such as the apolipoprotein E (Zhang et al., 2010)). Besides the EC study, the SPR method was applied to the Kch sample as well. The single Kch molecules were plunge-frozen, but 26

they were not discernible in the image. Cryo-negative staining of Kch sample did not work either. Different kinds of grid and stain may be tested in future (De Carlo and Harris, 2011; Quispe et al., 2007). The specimen is loaded on the grid followed by insertion into microscope. Glow discharge procedure is a common step to make the grid more hydrophilic for a better specimen adherence (Wang and Downing, 2011). However, the efficiency of glow discharge largely depends on how it is performed and different proteins may need different discharging treatments (Grassucci et al., 2007). For Kch (papers I and II) and rMGST1 (paper III), the glow discharge procedure was not used since it did not help crystals adsorb to the copper grid. The grids were prepared by the back injection method (Wang and Kuhlbrandt, 1991), which helps to preserve the high resolution data (Schmidt-Krey and Rubinstein, 2011), and then transferred into microscope for cryo data collection. 5.5 Electron crystallography image processing The experimental procedure of processing 2D crystalline images is introduced in this section. MRC (Crowther et al., 1996) and 2dx (Gipson et al., 2007) program suites were used for processing Kch 2D crystal images in papers I and II. 5.5.1 Processing each individual image The steps of processing 2D crystal images are summarized in the diagram in Fig. 14. The image of 2D crystals is collected on photographic film or by other detectors. In the first step, the crystalline areas on film are selected by optical diffraction and these areas are digitized on scanner at a step size, which should be appropriate to the resolution of the image. On the other hand, the digitization is done when the image is recorded in CCD. Next, defocus and tilt geometry values are determined. Several programs can perform this task, e.g., CTFFIND3 (Mindell and Grigorieff, 2003). The defocus values determined by the program are recommended to be evaluated by comparing the simulated Thon rings with the experimentally obtained Thon rings. The Fourier transform of the image shows the computed diffraction pattern, in which the lattice should be correctly indexed. Then, the translational distortion of the crystal can be corrected (called unbending, (Henderson et al., 1986)). CTF correction (see also '3.3.3 CTF effect') can be done by phase flipping. For high tilted images, the TTF (tilted transfer function) effect, instead of the CTF effect is corrected (Henderson et al., 1990). The space group can be determined by the ALLSPACE program (Valpuesta et al., 1994) from the listed amplitude and phase values after CTF or TTF correction.

Fig. 14 Flowchart showing steps in processing 2D crystal images.

27

5.5.2 Map generation After processing all images individually, they can be merged (Figs. 14 and 15). These images are moved to the same origin and then aligned. The processing procedure can be iteratively performed to obtain a consistent data set. Usually, the projection structure from untilted images is generated first. The projection structure only shows the 2D information, where the information along the beam axis is added together. Furthermore, a 3D map can be reconstructed by inverse Fourier transform of the structure factors of the data set including tilted images. Since there is no periodicity in the z-direction (along the beam axis), 2D crystals show continuous lines (called lattice lines) along this dimension in reciprocal space. The points where a central section intersects the lattice lines are recorded (Amos et al., 1982). Once a sufficient amount of these points are collected, a smooth curve can be adapted to generate each lattice line (Agard, 1983). The discrete points at equidistant sampling (spaced at 1/thickness) in each generated lattice line are sinc functions interpolated from the recorded points nearby (Amos et al., 1982). Then, the structure factors can be extracted from these discrete points and used to calculate the 3D map. 5.6 Electron diffraction processing A structure can be reconstructed from the data consisting of pure images, or it can be obtained from combination of phase information from images and amplitude information from diffraction patterns (Fig. 15) (Fujiyoshi, 1998). The amplitude information calculated from diffraction patterns is more accurate than the one from images, since the diffraction does not suffer from the CTF effect (Amos et al., 1982).

Fig. 15 Map generation in EC. (Top) The map is calculated from images only. (Bottom) The phase information from images and the amplitude information from diffraction patterns is combined together to generate the map. Kch is used as an example for image processing and rMGST1 is illustrated for diffraction pattern processing.

5.6.1 Steps in electron diffraction processing The crystal needs to be larger for electron diffraction purpose as compared to the one for image processing (Unger, 2000). After indexing and tilt geometry determination, the intensities of the reflection are integrated. Then various corrections of amplitudes are performed, e.g., deletion of 28

extremely strong densities from 'blooming' and subtraction of the background (Brink and Wei Tam, 1996). The listed amplitudes can be merged together and refined by the MERGEDIFF program, where the initially determined parameters are tuned (Ceska and Henderson, 1990). The quality of the data set is measured by Rsym (R-factor between the Friedel pairs in each diffraction pattern) and Rmerge (R-factor between the newly added diffraction pattern and the reference) (Ceska and Henderson, 1990). Electron diffraction patterns were collected to solve the rMGST1 structure in paper III. 5.6.2 Common crystallographic terminology Single sheets of 2D crystals can stack together to form the type I 3D crystal (Michel, 1983). The basic theories shared by x-ray diffraction and electron diffraction are summarized as follows: A). Building block The unit cell is the smallest repeating unit to build the entire crystal by only translational movements. A unit cell can further be divided into several asymmetric units, which can build one unit cell by application of the space group symmetry operation, including both translational and rotational movements. B). Bragg diffraction When the incident beam hits the atoms in the crystal, it scatters in all directions. Only those fulfilling the Bragg's law (2dsinθ = nλ, where d is the distance between two crystal planes, θ is the angle of the incident beam, n is an integer, and λ is the wavelength of the incident beam) can constructively interfere and then a reflection is observed. Each reflection has contributions from all atoms in every unit cell. C). Structural determination Both amplitude and phase information is required to calculate the protein structure. Although the phase information can be extracted from images, the molecular replacement method used in x-ray crystallography can also be applied in EC. Since the images are distorted by the CTF effect (see also '3.3.3 CTF effect'), the phase information obtained from them may not be very accurate. Thus, molecular replacement or other phase extension methods may lead to better quality results (Wisedchaisri and Gonen, 2011). Besides that, when the structure determined in EC reaches atomic resolutions, the differences between electron and x-ray scattering factors are small, thus the programs for structural determination in x-ray crystallography can be used in a similar way to process the EC data (Gonen et al., 2004; Holm et al., 2006). 5.7 Single particle reconstruction processing SPR is most frequently used to determine the structure of complexes (particles with large molecular weight) and nowadays, using the advanced direct detector, the reconstructed structures can reach atomic resolutions (such as in TRPV1: EMD-5778) (see also '3.2.2 Single particle reconstruction'). 5.7.1 Steps in single particle reconstruction Fig. 16 summarizes the main steps in SPR using Kch as an example. First of all, images are taken from the negative staining or cryo samples. Then, the single particles in each image are picked and stacked together, followed by CTF correction. The defocus parameters are determined by Thon ring fitting. After combining all picked particles together, 2D classification is performed to group

29

them based on their individual orientations. 2D classification includes several steps in sequence: 1) the particles are aligned to each other and centered; 2) different algorithms of classification can be applied to the aligned particles, e.g., hierarchical ascendant or κ-means classification; and 3) the particles in each class are averaged to generate the class average. Next, the corresponding Euler angle is assigned to each class average by either common lines in the reference-free method or projection matching with a reference. Then, reconstruction is done by inverse Fourier transform in reciprocal space or back-projection in real space. The initial model can be iteratively refined by the reprojection-classification-reconstruction cycle. The cycle is repeated until the reconstructions no longer change and the variation of the Euler angles is below a threshold. Lastly, the external model such as the one determined by x-ray crystallography is docked into the entire or part of the Coulomb potential map, if the resolution of the map does not allow model building directly.

Fig. 16 Processing of separated single particles in conventional SPR. The functional mutant, KchM240L is used as an example to illustrate the procedure. The defocus values are determined by Thon ring fitting using the 2dx program (right power spectrum). The class averages are generated using EMAN2. The Kch 3D map (Lundback et al., 2009) is docked with the atomic structures of Kv1.2 (PDB: 2A79, model in blue) and KtrA gating ring (PDB: 2HMW, model in yellow). A broad spectrum of SPR programs are available at present, e.g., EMAN1 (Ludtke et al., 1999), EMAN2 (Tang et al., 2007), Spider (Frank et al., 1996), Imagic (van Heel et al., 1996), BSoft (Heymann, 2001), Xmipp (Sorzano et al., 2004), Sparx (Hohn et al., 2007), Frealign (Grigorieff, 2007), Relion (Scheres, 2012), and Simple (Elmlund and Elmlund, 2012). Each program has its own algorithm and may result in different reconstructions from the same data (Murray et al., 2013). Furthermore, different programs may have slightly modified procedures as described above. For example, the CTF effect is corrected on reconstructions in Spider but on picked particles in EMAN1 and EMAN2. 5.7.2 Single particle reconstruction to process two dimensional crystal images Although normally the separated particles are analyzed in SPR and periodic arrays of unit cells are studied in EC, the general principle of these two methods is similar. In addition, since SPR 30

determines the Euler angle for each particle, it can potentially correct for variations of tilt angle, rotations of individual unit cells and large translational errors which are not taken into account by EC. Therefore, SPR can be adapted to process 2D crystal images (Koeck et al., 2007; Scherer et al., 2014). Paper IV presented such a SPR procedure developed from the previous one (Koeck et al., 2007). EMAN programs were used for all steps, except for CTF determination (performed by CTFFIND3) and correction (performed by Spider). The resolution estimation is measured at the point where the Fourier shell correlation (FSC) is at 0.5 (FSC = 0.5 criterion) (Penczek, 2010). MelBec 2D crystal images were processed by the SPR procedure in paper IV. Part III Results 6 Aim of investigation Membrane proteins play important roles in all processes for living cells. Many of them have clinical and pharmaceutical interests. The aim of this investigation is to study structure of membrane proteins using electron microscopy in order to better understand their function. Three proteins are analyzed: Kch (a putative K+ channel in E. coli, papers I and II), rMGST1 (a detoxification enzyme from rat, paper III), and MelBec (a secondary transporter in E. coli, paper IV). Apart from structural determination, a SPR procedure is developed for analysis of small or locally distorted 2D crystals. Specific aims in each project 1). Kch project A). To study a K+ channel in a lipid environment B). To identify the position of the RCK domain C). To compare the structures obtained from EC and SPR 2). rMGST1 project A). To refine the previously experimental procedure B). To build a more accurate model C). To analyze the ligand conformation 3). MelBec project A). To accurately determine and correct the CTF effect from tilted images B). To refine the previous SPR procedure C). To use both EC and SPR methods to process MelBec images 7 Summary of individual papers The experimental steps and the corresponding results are summarized as follows: A). Protein expression, purification, and 2D crystallization The aim of these experimental steps is to prepare the samples for data collection. These steps are included in all projects in this thesis. B). Projection structure determination It is an intermediate step in image processing. The projection structure in paper I suggests a new

31

assembly of Kch. C). 3D map generation The 3D map is generated after calculation of the projection structure. The 3D map of the transmembrane part of Kch (KchTM, paper II) suggests that the RCK domain is exposed to the solvent. D). Electron diffraction process Electron diffraction patterns were collected from large rMGST1 2D crystals and this data was used to determine its atomic structure in paper III. E). Single particle reconstruction process Besides the standard EC method, a SPR method was used to process the 2D crystal images of MelBec in paper IV. F). Structural determination The Kch structure was obtained by processing 2D crystal images in paper II. The atomic structure of rMGST1 was presented in paper III. The MelBec structure was generated by SPR in paper IV. 7.1 Papers I and II: Projection structures of KchM240L and KchTM Kch contains 6 TMs and an RCK domain in the cytosol. The previously x-ray crystallographic studies showed that the RCK domain forms an octameric gating ring structure in MthK (Jiang et al., 2002a; Ye et al., 2006), BKca (Leonetti et al., 2012; Wu et al., 2010; Yuan et al., 2012; Yuan et al., 2010), KtrA (Albright et al., 2006), and TrkA (Cao et al., 2013b). The gating ring in Kch was observed in the previous SPR study as well (Lundback et al., 2009). However, the majority of studies concerned isolated gating ring domains without their transmembrane parts or analyzed the samples surrounded by detergent. Thus, it is interesting to study such a kind of channel in a lipid environment. Large and well-ordered 2D crystals of both KchM240L, the full-length protein (paper I), and KchTM, which lacks the RCK domain (paper II) were obtained. KchM240L contains one point mutation at position 240 (methionine is replaced with leucine) and is a functional mutant of Kch (Kuo et al., 2003b). The projection structure of KchM240L was merged from the twelve best images with Fourier components extending to 6 Å resolution (Fig. 17B) and the projection structure of KchTM was averaged from nine images with a resolution cutoff at 8 Å (Fig. 17A). Both crystals were assigned the c12 two-sided plane group (Supplementary Fig. S1 for KchM240L in paper I and Figure S1 for KchTM in paper II) and the unit cell dimensions were similar: a = 144 ±1.9 Å, b = 84 ±1.0 Å, γ = 90° for KchM240L (Table 1 in paper I) and a = 143 ± 0.9 Å, b = 82 ± 0.6 Å, γ = 90° for KchTM (Table 1 in paper II). As the two projection structures resembled each other, it indicates that a similar crystal packing exists in both samples and RCK is probably flexible thus not affecting the crystal packing. In both projection structures, only two kinds of density zone appear which are outlined in a square and two circles in Figs. 17A and B. The square region is where the two-fold rotation symmetry element is situated and the circle regions are where the two-fold screw symmetry element goes through. Based on that, we generated two different models (Supplementary Fig. S4A in paper I and Fig. 17C) using a structurally similar K+ channel (MlotiK1, PDB: 3BEH) to overlay either of these two distinct zones. As explained in paper I, only the model shown in Fig. 17C (corresponding to Fig. 6C in paper I) is possible. In this model, the pore-forming domain of MlotiK1 is centered on the circle regions and the sensor domain is on the square regions. Both

32

KchM240L and KchTM crystallize in shifted double layers, as indicated in Fig. 17C. The tetramers in blue, red, and magenta are in the top layer and those in yellow, cyan, and green are in the bottom layer. The arrangement of KchM240L under the present 2D crystallization condition is not compatible with formation of a dimer of molecules through the RCK gating ring. Fig. 17 The c12 symmetrized projection structures of KchTM (A) and KchM240L (B and C). One unit cell (in a rectangle), the two-fold rotation symmetry elements (full arrows), the two-fold screw symmetry elements (half arrows), a and b axes, and two density zones (in a square and two circles) are depicted in (A and B). In (C), Tetramers of MlotiK1 (PDB: 3BEH, each tetramer has one color) are superimposed to the KchM240L projection structure (B). A temperature factor of B = -500 Å2 was applied to boost the fall-off of the amplitudes at intermediate and high resolutions in both (B and C). The positive contours (continuous lines) indicate the peaks of density above the average and the negative contours (dashed lines) are below the average.

7.2 Paper II: 3D structure of KchTM Although the projection structure of KchM240L indicates that the RCK gating ring does not form under the present 2D crystallization condition, it is not possible to determine the orientation of the

33

channel molecules from untilted images. The proceeding study (paper II) was carried out to analyze the 3D structure of Kch. One computed diffraction pattern of an image of the Kch 2D crystals can be indexed in three ways. Among them, only one alternative has a correct indexing, where the square regions (Figs. 17A and B) in the newly added image match with the corresponding zones in the reference. The other two alternatives compare their regions to the different zones, where the circle regions (Figs. 17A and B) in the newly added image are aligned to the square regions in the reference, thus they should be excluded. However, due to the similarity, but not identity, between the circle and square regions in Fig. 17B, choosing the correct indexing in KchM240L is not obvious. This ambiguity becomes more severe in particular for processing of tilted images. The projection structures with different indexing in each individual image were compared and the ones in which the two-fold rotation symmetry elements are located in the square region were considered as having a correct indexing. Furthermore, the merging process was also judged by the phase residuals. On the other hand, the indexing and merging of the projection images of KchTM were easier than for KchM240L, since the two density zones were significantly different (Fig. 17A). However, the visibility of the c12 projection symmetry disappeared when tilted images were processed, making the choice of correct indexing less certain. Thus the strategy was purely based on choosing the alternative that gives a significantly lower phase residual than the other two possibilities after origin refinement. In contrast to KchM240L, a 3D map was reconstructed for KchTM. Based on the position of the pore-forming domain as shown in Fig. 18 and adjacent molecules imposed by the unit cell dimensions and the symmetry elements, the molecular packing in the crystal is generated. Although KchTM does not contain an RCK domain, the position of the RCK domain can be deduced from the KchTM crystals, since the same crystal packing exists in these two crystals. As shown in Fig. 19, the RCK domains are exposed to the solvent and the RCK gating ring does not form.

Fig. 18 The 3D map of KchTM with the MlotiK1 model (PDB: 3BEH). Both the complete map (A) and the slices ((B) is in a vertical position and (C) is in a horizontal position) are shown. (A) and (B) are visualized along the membrane plane with the intracellular side up and the extracellular side down. (C) is viewed from the intracellular side. Only the pore-forming domain of the model (magenta) and its corresponding map (cyan) are displayed, since there is no convincing density of the sensor domain in the map.

34

Fig. 19 Crystal packing in KchTM. The double layered packing is visualized along the membrane plane, the same view in Figs. 18A and B. As it is rotated 90° from the view in Fig. 17, the MlotiK1 tetramers in the top layer are in blue and red while the tetramer in the bottom layer is in yellow. The crystal contact interfaces are boxed in a rectangle, which includes the pore-forming domain, the sensor domain, and the loops between S5 and the pore helix (gray). The RCK domains are deduced to be located at both leaflets of the crystal.

7.3 Paper III: A refined model of rMGST1 MGST1 was the first MAPEG member for which an atomic model was determined. This structure was solved by EC (Holm et al., 2006). In the present work, the previous procedure was refined in several aspects. First of all, the enzyme was previously purified from rat liver and the present source for 2D crystallization was from the protein over-expressed in E. coli. In addition, the crystallization parameters were adjusted. The crystals grown at the present work were isomorphic with those obtained earlier (same unit cell parameters (a = b = 81.8 Å and γ = 120°) and the amplitudes showed low deviation over the entire resolution range when compared). Thus the newly collected electron diffraction data was combined to the old data. The total number of merged electron diffraction patterns extended from 100 to 225. In the final model, all residues were in the allowed region in the Ramachandran plot and the R factors of Rwork and Rfree were 23.6% and 22.4%, respectively. Although the resolution did not increase from the previous work, the crystallographic statistics was improved to allow us build a more reliable model. As compared to the previous rMGST1 model (PDB: 2H8A), there were frame shifts in both TM 3 and TM 4 in the present model. Furthermore, more residues at the C-terminal end of TM 1 towards the cytoplasmic side were modeled. The final model was built from N10 (the N-terminal residue) to L151 (the C-terminal residue), with the absence of the loop between TM 1 and TM 2 (Figs. 20A and B). Several phospholipid molecules could be modeled as well. Three GSH molecules were built in one trimer, where each one was situated in a similar position as in PDB: 2H8A. As shown in Figs. 20C and D, each GSH adopts an extended conformation and is located in a pocket comprised of R38, K42, D66, and R73 from one subunit and R74, N78, Y121, Q128, N130, and R131 from a neighboring subunit in the trimer. Arginine residues at positions 38, 74, and 131 are conserved and have been suggested to play key roles in catalyzing the conjugation of GSH to second substrates in other MAPEG members (Ago et al., 2007; Jegerschold et al., 2008; Martinez Molina et al., 2007; Sjogren et al., 2013). The extended conformation of GSH in rMGST1 is unique among the MAPEG members and resembles the one in soluble GSTs (Dirr et al., 1994).

35

Fig. 20 The refined model of rMGST1. (A) Top view from the cytoplasmic side. (B) is visualized along the membrane plane, rotated 90°from (A). (C) and (D) Enlarged view of the boxed area in (B). (D) is rotated 90°from (C). The residues in the GSH binding pocket (C and D) and the phospholipid molecules (A and B) are labeled. 7.4 Paper IV: Single particle reconstruction processing of MelBec images In order to study small or locally distorted crystals and take advantage of SPR, a processing procedure was developed in paper IV (see also '5.7.2 Single particle reconstruction to process two dimensional crystal images'). The images are processed in a conventional EC procedure described in '5.5 Electron crystallography image processing' to generate the 3D map. This reconstruction can be used as the initial model in SPR processing. Firstly, the crystalline area in each image is identified based on the cross-correlation profiles. Secondly, the defocus parameters are determined and the CTF effect is corrected for each image. Since different regions in the grid have different heights in microscope when the grid is tilted, the corresponding strips in the image are corrected individually. Thirdly, various projections are calculated from the initial model and then used to pick up the unit cells in the image. Fourthly, the Euler angles of these cells are assigned by comparison with the initial model. If the Euler angle distribution of the picked unit cells matches the tilt angle of the recorded image, these cells are stacked together for reconstruction. Otherwise, they are discarded. Fifthly, the reconstruction is built up from all picked unit cells passing the Euler angle distribution test. Sixthly, the cycle from steps 3-5 is repeated, which reduces the bias from the initial model. Lastly, various post-reconstruction processes are performed, such as resolution estimation and map sharpening. 36

As compared with the previously published procedure (Koeck et al., 2007), the current one was refined in several aspects. The improvements contained several new steps, a more accurate CTF determination and correction, an automatic unit cell picking, and adjustment of the refinement parameters. MelBec 2D crystal images were processed with the current SPR procedure. 67.8% of the picked unit cells have their assigned angles equal or below 45°(Fig. 21A); the resolution based on the 'gold standard' test (Grigorieff, 2000) is 13 Å (Fig. 21B), where the split of the entire data set into two halves is carried out before classification. Docking of the MelBst map (MelBst has a sequence identity of 85% to MelBec (Mizushima et al., 1992)) generated from the available atomic model (PDB: 4M64, Mol-B) suggests that MelBec in the 2D crystals adopts an open conformation (Figs. 21C and D).

Fig. 21 The assessments of reconstructed MelBec using the refined SPR procedure. (A) Euler angle distribution. The unit cells having similar Euler angles are grouped together and each class is represented by a white dot. The brighter the dot, the more unit cells in each class. The altitude covers 0–90° tilt angles measured from the top to the bottom. The longitude covers 0–180° tilt-axis angles measured from the left to the right. The horizontal line marks the nominally maximum tilt angle (45°) in data collection. (B) The resolution curve depicted in red. The resolution is chosen at the point where the FSC equals 0.5 (labeled at the cross between two blue lines). (C) and (D) Fitting of the MelBst map at a resolution cutoff of 13 Å (purple, surface view) into the map obtained by the current SPR procedure (green, mesh view). The threshold values are adjusted to keep a similar contouring level for both maps in (C and D). The views are perpendicular to the membrane plane (C) and along the membrane plane (D). 37

Part IV Discussion and Conclusion 8 Future perspectives Multiple types of protein are analyzed in this thesis using EM. The future work for each individual project is discussed. 8.1 Kch It would be interesting to determine the 3D structure of the full-length Kch, which possesses both the transmembrane part and the RCK domain. Indeed, a number of tilted images of KchM40L have been collected and processed. However, the data did not give a reasonable reconstruction. The failure may be due to an uncertainty in choosing correct indexing of these images. One possible solution is to collect more tilted images with a smaller tilt angle increment. When images collected in this way are compared to the reference, the phase residual may reflect the effect of indexing rather than from tilting. The low tilted images (10° for instance) can be merged proceeding from the untitled projection structure and the resulting data can be used as a reference to search for correct indexing of 20°tilted images. This procedure can be iterated to include higher tilted images. On the other hand, choosing correct indexing and merging KchTM images are much more certain, since the two density zones are significantly different (Fig. 17A). Thus, it is highly possible to reconstruct a more accurate 3D map of KchTM when more images are included. Different crystallization parameters, e.g., detergent, lipid, pH, salt, additive, temperature, and crystallization procedure were screened. The majority of obtained 2D crystals were of the stacked sheets type with a few separated single sheets. Other parameters can be adjusted; one of them, various kinds of lipid, is of a particular interest. Apart from keeping the integrity of the K+ channels (Jiang et al., 2003; Lee et al., 2005), some lipids are ligands to modulate the function of channels (Suh and Hille, 2008). Thus, it is worth trying a mixture of lipids in crystallization. With more separated single sheet type crystals, it is possible to analyze the specimen by shadowing or atomic force microscopy which can provide additional information of the structure of the RCK domain. The physiological function of Kch could be studied more closely in the E. coli strain (TK2420) which is defective in three major K+ ion uptake proteins: Trk/Kdp/Kup (Epstein et al., 1993). Considering the indiscernible phenotype of the mutant strain where the kch gene is knocked out (Epstein, 2003; Kuo et al., 2003b), harsh conditions (extreme pH and K+ concentrations in the culture media) may be investigated (Kuo et al., 2005; Kuo et al., 2003b). 8.2 rMGST1 Although each GSH molecule in rMGST1 adopts an extended conformation, the exact conformation, position, and orientation of GSH are not completely certain. The functional studies have shown that rMGST1 has one-third-of-the-sites-reactivity, in which three GSH molecules can bind to each protein trimer with their own binding affinities (Kd = 20 μM, 0.4-2.5 mM, and 2.5 mM, respectively), but only one GSH is active (Alander et al., 2009). Since the rMGST1 crystals were grown at 1 mM GSH, it is possible that only one or two GSH binding sites among the three sites are occupied. Thus, the GSH density would be averaged out and become weaker. One solution is to determine the rMGST1 structure from the crystals grown at a higher GSH concentration. The initial result showed that the isomorphic crystals could also be obtained at 5

38

mM concentration. In future, a whole data set of rMGST1 can be collected at this concentration. Furthermore, the structural and functional studies of rMGST1 can serve as a model for analysis of other MAPEG members. MGST2 and MPGES1 own the 'one-third-of-the-sites-reactivity' property as well (Ahmad et al., 2013; He et al., 2011), whereas LTC4S has one active GSH in each monomer (Ahmad et al., 2013). Another solution is to co-crystallize with an inhibitor and/or a substrate. The experiment of co-crystallization with TNB, which results in a dead-end product (Fig. 9B), can provide the information of an intermediate step in the reaction cycle as well as result in an easier assignment of the GSH density due to a bulk benzene ring in TNB. The initial experiments showed that soaking the 2D crystals into the TNB solution tended to make them fragile, although diffraction patterns have been collected from the intact ones. These TNB soaked crystals were isomorphic with the original ones, which were grown at 1 mM GSH. The in-plane resolution of the TNB data also extended to 3 Å. However, the data is awaiting further refinement and interpretation. 8.3 MelBec Docking of the available atomic structure of MelBst into the obtained map is presented in Figs. 21C and D. Considering the inner symmetry of the transporter (Fig. 11A) and many intermediate steps in the entire reaction cycle (Fig. 10), it is worth trying to test docking of the models in different conformations. A number of structures crystallized in various conformations (different states, with or without ligand) from related MFS members are available now. Thus, homology modeling using these structures as the templates (see also '2.4 Computational modeling') may improve the docking efficiency and give further information of transition between different states. It would also be interesting to grow 2D crystals without sugar. By combination of the apo data (without ligand) and the data with sugar (paper IV), it is possible to demonstrate the conformational change after sugar binding. The MelBec structures can be further compared with the LacY structures (lactose permease of E. coli, homologue to MelBec, (Yousef and Guan, 2009)) with (PDB: 4OAA, (Kumar et al., 2014)) and without (PDB: 2CFQ, (Mirza et al., 2006)) ligand. The developed SPR method can be applied to process the apo data. The reconstruction obtained from the sugar data in paper IV is possible to be used as the initial model for unit cell picking from the apo data. Low-pass filtering of the initial model may help to reduce the basis from the sugar data. 8.4 Single particle reconstruction processing of two dimensional crystal images The developed SPR procedure is semi-automated; commands are typed to run the scripts in each step. However, these commands are not based on the choices of users. Therefore, the SPR procedure can be improved to be fully automated in future. This SPR procedure is aimed for small or locally distorted crystals. Since testing each crystallization condition consumes approximately 40 μl concentrated samples in a conventional dialysis experiment in the beaker (microdialysis uses less materials, however the result may be less reproducible), the total number of screened conditions is normally not very large. Thus, many resulting crystals are small and the resolutions of them remain around 10 Å. The 2D crystal of Na+,K+-ATPase from pig kidney (a primary transporter moves K+ and Na+ ions in opposite directions across the plasma membrane) is such a case. The small crystals usually contain 300-500 unit cells (0.2 μm, (Hebert et al., 2001)), thus they are suitable to be evaluated by the SPR method.

39

A potential benefit of using the SPR procedure is to include unmerged images in EC. Therefore, application of this method to process the KchM240L and KchTM images can avoid the indexing problem and the resulting reconstruction is expected to be less affected by personal judgments. rMGST1 is also an interesting data to be tested. Previous studies showed that rMGST1 purified from rat liver could crystallize in two ways: p6 and p21212 two-sided plane group symmetries (Schmidt-Krey et al., 1998). The 3D maps reconstructed from both crystals reached 3.5 Å (Holm et al., 2006). However, due to a higher symmetry and completeness in the p6 type, the structure of rMGST1 was determined from the p6 type crystals whereas the p21212 data was used as an independent control (Holm et al., 2006). On the other hand, the number of p21212 type images was more than the p6 type ones (Holm et al., 2006). Thus, the developed SPR method could be applied to the p21212 type images to validate whether it works for a near atomic resolution data or not. Furthermore, it is possible to distinguish the bound GSH molecules in multiple conformations from both data sets using the SPR method. 8.5 Future of electron microscopy Until now, x-ray crystallography is the main method to determine a protein's structure. However, it still demands suitable well-ordered 3D crystals for data collection, which might be difficult. On the other hand, NMR has a limitation on the size of the analyzed sample. EM is a powerful tool suitable for a broad range of specimens which are generally easier to be prepared as compared to the specimens for x-ray crystallography. The great advantage of EM is that phase information can be extracted from an image, whereas additional work is required to get the phase information in x-ray crystallography. MPs are suitable for EC, where they are embedded in more natural lipid environments. Lipid environment plays critical roles at least in some proteins (Zhou and Cross, 2013). Although EC still requires repeating work and is a slow process as compared to the other methods, automation is expected to improve the speed and success of each project (Hu et al., 2010; Iacovache et al., 2010; Wisedchaisri et al., 2011). In cells, MPs are often assembled together to form protein complexes (Alberts, 1998). These super-complexes are of central importance for cellular function and they are ideal targets to be studied by SPR, e.g., COX2/MPGES1 (Yamagata et al., 2001) and BKca/voltage gated calcium channels (Berkefeld et al., 2006). On the other hand, SPR faces the problem of processing low signal-to-noise ratio images. However, with the development of modern electron microscopes, in particular the direct detectors, and processing software, biological samples can reach near atomic resolutions in EM now, which could be obtained only in x-ray crystallography in the past. Another advent is to combine fluorescence microscopy and EM together, which provides both the functional and structural views of the protein of interest in its environment (Agronskaia et al., 2008). 9 Conclusion The structure of three MPs: Kch, rMGST1, and MelBec are studied in this thesis. For this, EC is mainly used. In addition, a SPR procedure is developed to process 2D crystal images. For the Kch project, large and well-ordered 2D crystals were grown for both the full-length protein (KchM240L, a functional mutant of Kch) and the channel containing only the transmembrane part (KchTM, lacking the RCK domain at the C-terminus). The projection structure of KchM240L was calculated at 6 Å resolution and 8 Å for KchTM. These two

40

projection structures resemble each other, indicating a similar crystal packing in both cases. The superimposing of a structurally comparable atomic model of a K+ channel showed that Kch crystallizes as two symmetry-related overlapping layers. The determined 3D map of KchTM showed that the crystal contacts are from the extracellular surfaces of the protein. Thus, the RCK domain in the intracellular side of Kch is exposed to solvent. This conclusion stands in contrast to previous studies which predicted that the RCK domains interact to form an octameric gating ring structure. The discrepancy between the current electron crystallographic and previously structural studies, including the x-ray crystallographic and single particle reconstruction ones, probably reflects the important effect of lipid environment on MPs. rMGST1 was over-expressed in E. coli and the 2D crystallization condition was adjusted as compared to the previous protocol. In contrast to the previous result, only the p6 two-sided plane group symmetry crystals were detected. The refined structure was determined from 225 diffraction patterns and calculated at 3.5 Å. The new model contains 123 out of 155 amino acid residues, two structured phospholipid molecules, two hydrocarbon chains, and one GSH molecule. Compared with the previously determined model (PDB: 2H8A), the present one extends the region at the C-terminal end of TM 1 and corrects the frame shifts in both TM 3 and TM 4. Two phospholipid molecules glue two adjacent protein trimers together. In addition, one hydrocarbon chain is located at the crystallographic symmetry axis and the other is between TM 1 and TM 3 on the cytoplasmic side. In contrast to other MAPEG members, the GSH molecules bind in an extended conformation at the interfaces between two subunits of the trimer. Each GSH is located in a pocket comprised of R38, K42, D66, and R73 from one subunit and R74, N78, Y121, Q128, N130, and R131 from a neighboring subunit in the trimer. The location of GSH is supported by the mutagenesis data in vitro. A refined SPR procedure for processing of 2D crystal images was developed. This method combines the advantage of both SPR and EC. It is aimed at processing images of MPs that form small or locally distorted 2D crystals. The reconstruction obtained from EC is used as the initial model to pick up similar unit cells based on the cross-correlation profiles. The picked unit cells in each image are analyzed by their respective Euler angle distributions to sort out the ones which do not match their expected tilt angles. The reconstruction is calculated from the newly picked unit cells and the resulting map serves as an initial model to repeat the entire procedure in order to reduce the model bias in the first run. Other improvements include a strip based CTF determination and correction, adjustment of the refinement parameters, and several refined post-reconstruction processes. The tested data, MelBec showed that a reliable reconstruction was obtained at 13 Å resolution (based on the 'gold standard' test). The docking of the atomic structure of MelB from another species into the final map indicated that MelBec adopts an open conformation under the present 2D crystallization condition. Part V Epilogue 10 Acknowledgements The work was financed by the Swedish Research Council and the Karolinska Institutet Center for Innovative Medicine. Many thanks to the Department of Biosciences and Nutrition, Karolinska Institutet and School of Technology and Health, Royal Institute of Technology supporting me to carry out the investigation in my Ph.D. study. I would like to express to my gratitude to my

41

colleagues, family, and friends who have helped me complete the work. First of all, my supervisor Hans Hebert, for accepting me in your group, for giving me the opportunity to work with EM, and for many supports in all aspects. My supervisor Caroline Jegerschöld, for your endless enthusiasm in science and never giving up spirit. My supervisor Per-Johan Jakobsson, for assisting me at the beginning of my study and the guidance in the MAPEG project. Current members in Hans Hebert's group: Pasi Purhonen, for your help in data collection and translating Swedish letters for me. Philip Koeck, for explaining SPR knowledge. Martin Lindahl, for your teaching the EM course and discussing about the projects. Johan Härmark, for making the website for the group and introducing the Swedish society. Harriet Nilsson, for your encouragement all the time. Ramakrishnan Balakrishnan Kumar, for sharing your experimental technology and showing me how to work with yeast. Lin Zhu, for your concern about my life and discussion about the projects. Rampradeep Samiappan, for your good mood and inviting me to play cricket. Many thanks to the previous members: Kimberley Cheng, Anna-Karin Lundbäck, Hans Elmlund, Dominika Elmlund, and Peter Holm. Special thanks to Thirupathi Pattipaka, for writing the script to pick up the unit cells automatically and Yohannes Haileyesus Ayele, for guiding me in processing crystal images and discussing the problems you met. Colleagues in STH, KTH: Xiaogai Li, for inviting me to your apartment and cooking delicious food. Also thanks for the concern from you and your husband, Weidong Lian. Sicong Yu, for your help in my routine life and sharing working experiences with me. Eva-Rut Lindberg and Fredrik Häggström, for handling the academic documents. Current members in Department of Biosciences and Nutrition, KI: Sharif Hasni, for sharing your life experience in UK. Min Jia, Zhiqiang Huang, Ning Liang, Jian Zhu, Ting Zhuang, Miao Zhao, Xiaowei Gong, Zhilun Li, Mathilda Sjöberg, You Xu, Erik Lundgren, Anders Lindholm, Punit Prasad, and Kashyap Dave for a nice atmosphere. Previous members in Department of Biosciences and Nutrition, KI: Xiaohua Lou, for teaching me how to use the AKTA machine and the knowledge of x-ray crystallography. Wei Liu, for answering my questions in x-ray crystallography and providing technical assistance during my study. Also thanks to your wife, Jun Wang for her optimistic attitude. Shangrung Wu, for assisting me to use electron microscope and discussing SPR. Chiounan Shiue, for your introduction of the Swedish society and the situation in Taiwan. Colleagues in other departments, KI: Ralf Morgenstern, for your guidance, support, and effort in the rMGST1 project.

42

Johan Ålander, for showing me how to express and purify rMGST1 at the beginning of my study. Linda Spahiu, for organizing the MAPEG meeting in Djurönäset. Sven-Christian Pawelzik, for instructing me how to do science in an efficient and organized way. Serhiy Souchelnytskyi, for helping me to use mass spectrometry. Christina Hebert, for inviting me to your house and delicious Swedish salmon. Sylvie-Le Guyader, for maintainance of the fluorescence microscope and organizing the cookie club. Pablo-Hernandez Varas, for generously sharing me the protocol and materials for protein labeling. Anna Magnusson, for your guidance in the preliminary experiment of working with rat, for your kind invitation to your house, and for your positive attitude. Colleagues in other universities: Richard Armstrong, for your many times discussion about the rMGST1 structure and H/D exchange experiments. Lin Zhu, for your introduction of the situation in China and suggestions for my future. Henning Stahlberg, for organizing the 2dx workshop and your active discussion in the EM community. Sanjeewani Sooriyaarachchi, for being a good friend and always encouraging me. Wimal ubhayaseker, for helping me in the rMGST1 project and encouraging me to work with EM. Adrian Suarez Covarrubias, for telling me the job opportunities and assisting me in expression and purification when I started to perform experiments. Lars Liljas, for your concern all the time and introducing me to the structural study. My parents: Chengbu Kuang, for your care and love, always listening to me, and encouraging me to start my Ph.D. study in Sweden. Kang Li, for you concern and guidance, reminding me things which I forgot, and supporting me during my entire life. Also many thanks to my other relatives, friends, and colleagues, who are not mentioned due to the limited space in this acknowledgements.

43

11 References Adrian, M., Dubochet, J., Fuller, S.D., and Harris, J.R. (1998). Cryo-negative staining. Micron 29, 145-160. Agard, D.A. (1983). A least-squares method for determining structure factors in three-dimensional tilted-view reconstructions. J. Mol. Biol. 167, 849-852. Aggarwal, S.K., and MacKinnon, R. (1996). Contribution of the S4 segment to gating charge in the Shaker K+ channel. Neuron 16, 1169-1177. Ago, H., Kanaoka, Y., Irikura, D., Lam, B.K., Shimamura, T., Austen, K.F., and Miyano, M. (2007). Crystal structure of a human membrane protein involved in cysteinyl leukotriene biosynthesis. Nature 448, 609-612. Agronskaia, A.V., Valentijn, J.A., van Driel, L.F., Schneijdenberg, C.T., Humbel, B.M., van Bergen en Henegouwen, P.M., Verkleij, A.J., Koster, A.J., and Gerritsen, H.C. (2008). Integrated fluorescence and transmission electron microscopy. J. Struct. Biol. 164, 183-189. Ahmad, S., Niegowski, D., Wetterholm, A., Haeggstrom, J.Z., Morgenstern, R., and Rinaldo-Matthis, A. (2013). Catalytic characterization of human microsomal glutathione S-transferase 2: identification of rate-limiting steps. Biochemistry (Mosc). 52, 1755-1764. Alam, A., and Jiang, Y. (2011). Structural studies of ion selectivity in tetrameric cation channels. J. Gen. Physiol. 137, 397-403. Alander, J., Lengqvist, J., Holm, P.J., Svensson, R., Gerbaux, P., Heuvel, R.H., Hebert, H., Griffiths, W.J., Armstrong, R.N., and Morgenstern, R. (2009). Microsomal glutathione transferase 1 exhibits one-third-of-the-sites-reactivity towards glutathione. Arch. Biochem. Biophys. 487, 42-48. Alberts, B. (1998). The cell as a collection of protein machines: preparing the next generation of molecular biologists. Cell 92, 291-294. Albright, R.A., Ibar, J.L., Kim, C.U., Gruner, S.M., and Morais-Cabral, J.H. (2006). The RCK domain of the KtrAB K+ transporter: multiple conformations of an octameric ring. Cell 126, 1147-1159. Allegretti, M., Mills, D.J., McMullan, G., Kuhlbrandt, W., and Vonck, J. (2014). Atomic model of the F420-reducing [NiFe] hydrogenase by electron cryo-microscopy using a direct electron detector. eLife 3, e01963. Amos, L.A., Henderson, R., and Unwin, P.N. (1982). Three-dimensional structure determination by electron microscopy of two-dimensional crystals. Prog. Biophys. Mol. Biol. 39, 183-231. Amunts, A., Brown, A., Bai, X.C., Llacer, J.L., Hussain, T., Emsley, P., Long, F., Murshudov, G., Scheres, S.H., and Ramakrishnan, V. (2014). Structure of the yeast mitochondrial large ribosomal subunit. Science 343, 1485-1489. Anfinsen, C.B. (1973). Principles that govern the folding of protein chains. Science 181, 223-230. Berkefeld, H., Sailer, C.A., Bildl, W., Rohde, V., Thumfart, J.O., Eble, S., Klugbauer, N., Reisinger, E., Bischofberger, J., Oliver, D., et al. (2006). BKCa-Cav channel complexes mediate rapid and localized Ca2+-activated K+ signaling. Science 314, 615-620. Bresell, A., Weinander, R., Lundqvist, G., Raza, H., Shimoji, M., Sun, T.H., Balk, L., Wiklund, R., Eriksson, J., Jansson, C., et al. (2005). Bioinformatic and enzymatic characterization of the MAPEG superfamily. FEBS J. 272, 1688-1703. Brink, J., and Wei Tam, M. (1996). Processing of electron diffraction patterns acquired on a slow-scan CCD camera. J. Struct. Biol. 116, 144-149.

44

Brohawn, S.G., del Marmol, J., and MacKinnon, R. (2012). Crystal structure of the human K2P TRAAK, a lipid- and mechano-sensitive K+ ion channel. Science 335, 436-441. Buckingham, S.D., Kidd, J.F., Law, R.J., Franks, C.J., and Sattelle, D.B. (2005). Structure and function of two-pore-domain K+ channels: contributions from genetic model organisms. Trends Pharmacol. Sci. 26, 361-367. Butterwick, J.A., and MacKinnon, R. (2010). Solution structure and phospholipid interactions of the isolated voltage-sensor domain from KvAP. J. Mol. Biol. 403, 591-606. Cao, E., Liao, M., Cheng, Y., and Julius, D. (2013a). TRPV1 structures in distinct conformations reveal activation mechanisms. Nature 504, 113-118. Cao, Y., Pan, Y., Huang, H., Jin, X., Levin, E.J., Kloss, B., and Zhou, M. (2013b). Gating of the TrkH ion channel by its associated RCK protein TrkA. Nature 496, 317-322. Catterall, W.A. (2010a). Ion channel voltage sensors: structure, function, and pathophysiology. Neuron 67, 915-928. Catterall, W.A. (2010b). Signaling complexes of voltage-gated sodium and calcium channels. Neurosci. Lett. 486, 107-116. Ceska, T.A., and Henderson, R. (1990). Analysis of high-resolution electron diffraction patterns from purple membrane labelled with heavy-atoms. J. Mol. Biol. 213, 539-560. Chen, X., Wang, Q., Ni, F., and Ma, J. (2010). Structure of the full-length Shaker potassium channel Kv1.2 by normal-mode-based X-ray crystallographic refinement. Proc. Natl. Acad. Sci. U. S. A. 107, 11352-11357. Cheng, W.W., McCoy, J.G., Thompson, A.N., Nichols, C.G., and Nimigean, C.M. (2011). Mechanism for selectivity-inactivation coupling in KcsA potassium channels. Proc. Natl. Acad. Sci. U. S. A. 108, 5272-5277. Chiu, W. (1993). What does electron cryomicroscopy provide that X-ray crystallography and NMR spectroscopy cannot? Annu. Rev. Biophys. Biomol. Struct. 22, 233-255. Clayton, G.M., Altieri, S., Heginbotham, L., Unger, V.M., and Morais-Cabral, J.H. (2008). Structure of the transmembrane regions of a bacterial cyclic nucleotide-regulated channel. Proc. Natl. Acad. Sci. U. S. A. 105, 1511-1515. Clayton, G.M., Silverman, W.R., Heginbotham, L., and Morais-Cabral, J.H. (2004). Structural basis of ligand activation in a cyclic nucleotide regulated potassium channel. Cell 119, 615-627. Cordero-Morales, J.F., Cuello, L.G., Zhao, Y., Jogini, V., Cortes, D.M., Roux, B., and Perozo, E. (2006). Molecular determinants of gating at the potassium-channel selectivity filter. Nat. Struct. Mol. Biol. 13, 311-318. Crowther, R.A., Henderson, R., and Smith, J.M. (1996). MRC image processing programs. J. Struct. Biol. 116, 9-16. Cuello, L.G., Jogini, V., Cortes, D.M., and Perozo, E. (2010). Structural mechanism of C-type inactivation in K(+) channels. Nature 466, 203-208. De Carlo, S., and Harris, J.R. (2011). Negative staining and cryo-negative staining of macromolecules and viruses for TEM. Micron 42, 117-131. De Rosier, D.J., and Klug, A. (1968). Reconstruction of three dimensional structures from electron micrographs. Nature 217, 130-134. De Zorzi, R., Nicholson, W.V., Guigner, J.M., Erne-Brand, F., and Venien-Bryan, C. (2013). Growth of large and highly ordered 2D crystals of a K(+) channel, structural role of lipidic environment. Biophys. J. 105, 398-408.

45

Deisenhofer, J., Epp, O., Miki, K., Huber, R., and Michel, H. (1985). Structure of the protein subunits in the photosynthetic reaction centre of Rhodopseudomonas viridis at 3A resolution. Nature 318, 618-624. Derewenda, Z.S. (2004). The use of recombinant methods and molecular engineering in protein crystallization. Methods 34, 354-363. Dirr, H., Reinemer, P., and Huber, R. (1994). X-ray crystal structures of cytosolic glutathione S-transferases. Implications for protein architecture, substrate recognition and catalytic function. Eur. J. Biochem. 220, 645-661. Doyle, D.A., Morais Cabral, J., Pfuetzner, R.A., Kuo, A., Gulbis, J.M., Cohen, S.L., Chait, B.T., and MacKinnon, R. (1998). The structure of the potassium channel: molecular basis of K+ conduction and selectivity. Science 280, 69-77. Drews, J. (2000). Drug discovery: a historical perspective. Science 287, 1960-1964. Dubochet, J., Adrian, M., Chang, J.J., Homo, J.C., Lepault, J., McDowall, A.W., and Schultz, P. (1988). Cryo-electron microscopy of vitrified specimens. Q. Rev. Biophys. 21, 129-228. Elmlund, D., and Elmlund, H. (2012). SIMPLE: Software for ab initio reconstruction of heterogeneous single-particles. J. Struct. Biol. 180, 420-427. Engelman, D.M. (2005). Membranes are more mosaic than fluid. Nature 438, 578-580. Epstein, W. (2003). The roles and regulation of potassium in bacteria. Prog. Nucleic Acid Res. Mol. Biol. 75, 293-320. Epstein, W., Buurman, E., McLaggan, D., and Naprstek, J. (1993). Multiple mechanisms, roles and controls of K+ transport in Escherichia coli. Biochem. Soc. Trans. 21, 1006-1010. Ethayathulla, A.S., Yousef, M.S., Amin, A., Leblanc, G., Kaback, H.R., and Guan, L. (2014). Structure-based mechanism for Na(+)/melibiose symport by MelB. Nat. Commun. 5, 3009. Evans, J.F., Ferguson, A.D., Mosley, R.T., and Hutchinson, J.H. (2008). What's all the FLAP about?: 5-lipoxygenase-activating protein inhibitors for inflammatory diseases. Trends Pharmacol. Sci. 29, 72-78. Faruqi, A.R., and Henderson, R. (2007). Electronic detectors for electron microscopy. Curr. Opin. Struct. Biol. 17, 549-555. Faruqi, A.R., and McMullan, G. (2011). Electronic detectors for electron microscopy. Q. Rev. Biophys. 44, 357-390. Ferguson, A.D., McKeever, B.M., Xu, S., Wisniewski, D., Miller, D.K., Yamin, T.T., Spencer, R.H., Chu, L., Ujjainwalla, F., Cunningham, B.R., et al. (2007). Crystal structure of inhibitor-bound human 5-lipoxygenase-activating protein. Science 317, 510-512. Ford, R.C., and Holzenburg, A. (2008). Electron crystallography of biomolecules: mysterious membranes and missing cones. Trends Biochem. Sci. 33, 38-43. Frank, J., Radermacher, M., Penczek, P., Zhu, J., Li, Y., Ladjadj, M., and Leith, A. (1996). SPIDER and WEB: processing and visualization of images in 3D electron microscopy and related fields. J. Struct. Biol. 116, 190-199. Fujiyoshi, Y. (1998). The structural study of membrane proteins by electron crystallography. Adv. Biophys. 35, 25-80. Funk, C.D. (2001). Prostaglandins and leukotrienes: advances in eicosanoid biology. Science 294, 1871-1875. Gadsby, D.C. (2009). Ion channels versus ion pumps: the principal difference, in principle. Nat. Rev. Mol. Cell Biol. 10, 344-352.

46

Garavito, R.M., and Ferguson-Miller, S. (2001). Detergents as tools in membrane biochemistry. J. Biol. Chem. 276, 32403-32406. Gipson, B., Zeng, X., Zhang, Z.Y., and Stahlberg, H. (2007). 2dx--user-friendly image processing for 2D crystals. J. Struct. Biol. 157, 64-72. Gonen, T., Cheng, Y., Sliz, P., Hiroaki, Y., Fujiyoshi, Y., Harrison, S.C., and Walz, T. (2005). Lipid-protein interactions in double-layered two-dimensional AQP0 crystals. Nature 438, 633-638. Gonen, T., Sliz, P., Kistler, J., Cheng, Y., and Walz, T. (2004). Aquaporin-0 membrane junctions reveal the structure of a closed water pore. Nature 429, 193-197. Grassucci, R.A., Taylor, D.J., and Frank, J. (2007). Preparation of macromolecular complexes for cryo-electron microscopy. Nat. Protoc. 2, 3239-3246. Grigorieff, N. (2000). Resolution measurement in structures derived from single particles. Acta Crystallogr. D: Biol. Crystallogr. 56, 1270-1277. Grigorieff, N. (2007). FREALIGN: high-resolution refinement of single particle structures. J. Struct. Biol. 157, 117-125. Guan, L., Nurva, S., and Ankeshwarapu, S.P. (2011). Mechanism of melibiose/cation symport of the melibiose permease of Salmonella typhimurium. J. Biol. Chem. 286, 6367-6374. Hacksell, I., Rigaud, J.L., Purhonen, P., Pourcher, T., Hebert, H., and Leblanc, G. (2002). Projection structure at 8 A resolution of the melibiose permease, an Na-sugar co-transporter from Escherichia coli. EMBO J. 21, 3569-3574. Hammarberg, T., Hamberg, M., Wetterholm, A., Hansson, H., Samuelsson, B., and Haeggstrom, J.Z. (2009). Mutation of a critical arginine in microsomal prostaglandin E synthase-1 shifts the isomerase activity to a reductase activity that converts prostaglandin H2 into prostaglandin F2alpha. J. Biol. Chem. 284, 301-305. Hayes, J.D., Flanagan, J.U., and Jowsey, I.R. (2005). Glutathione transferases. Annu. Rev. Pharmacol. Toxicol. 45, 51-88. He, S., Wu, Y., Yu, D., and Lai, L. (2011). Microsomal prostaglandin E synthase-1 exhibits one-third-of-the-sites reactivity. Biochem. J. 440, 13-21. Hebert, H., and Jegerschold, C. (2007). The structure of membrane associated proteins in eicosanoid and glutathione metabolism as determined by electron crystallography. Curr. Opin. Struct. Biol. 17, 396-404. Hebert, H., Purhonen, P., Vorum, H., Thomsen, K., and Maunsbach, A.B. (2001). Three-dimensional structure of renal Na,K-ATPase from cryo-electron microscopy of two-dimensional crystals. J. Mol. Biol. 314, 479-494. Henderson, R. (1995). The potential and limitations of neutrons, electrons and X-rays for atomic resolution microscopy of unstained biological molecules. Q. Rev. Biophys. 28, 171-193. Henderson, R., Baldwin, J.M., Ceska, T.A., Zemlin, F., Beckmann, E., and Downing, K.H. (1990). Model for the structure of bacteriorhodopsin based on high-resolution electron cryo-microscopy. J. Mol. Biol. 213, 899-929. Henderson, R., Baldwin, J.M., Downing, K.H., Lepault, J., and Zemlin, F. (1986). Structure of Purple Membrane from Halobacterium-Halobium - Recording, Measurement and Evaluation of Electron-Micrographs at 3.5 a Resolution. Ultramicroscopy 19, 147-178. Henderson, R., and Unwin, P.N. (1975). Three-dimensional model of purple membrane obtained by electron microscopy. Nature 257, 28-32. Heymann, J.B. (2001). Bsoft: image and molecular processing in electron microscopy. J. Struct.

47

Biol. 133, 156-169. Hibino, H., Inanobe, A., Furutani, K., Murakami, S., Findlay, I., and Kurachi, Y. (2010). Inwardly rectifying potassium channels: their structure, function, and physiological roles. Physiol. Rev. 90, 291-366. Hirai, T., Murata, K., Mitsuoka, K., Kimura, Y., and Fujiyoshi, Y. (1999). Trehalose embedding technique for high-resolution electron crystallography: application to structural study on bacteriorhodopsin. J. Electron Microsc. (Tokyo). 48, 653-658. Hohmann-Marriott, M.F., Sousa, A.A., Azari, A.A., Glushakova, S., Zhang, G., Zimmerberg, J., and Leapman, R.D. (2009). Nanoscale 3D cellular imaging by axial scanning transmission electron tomography. Nat. Methods 6, 729-731. Hohn, M., Tang, G., Goodyear, G., Baldwin, P.R., Huang, Z., Penczek, P.A., Yang, C., Glaeser, R.M., Adams, P.D., and Ludtke, S.J. (2007). SPARX, a new environment for Cryo-EM image processing. J. Struct. Biol. 157, 47-55. Holm, P.J., Bhakat, P., Jegerschold, C., Gyobu, N., Mitsuoka, K., Fujiyoshi, Y., Morgenstern, R., and Hebert, H. (2006). Structural basis for detoxification and oxidative stress protection in membranes. J. Mol. Biol. 360, 934-945. Hu, M., Vink, M., Kim, C., Derr, K., Koss, J., D'Amico, K., Cheng, A., Pulokas, J., Ubarretxena-Belandia, I., and Stokes, D. (2010). Automated electron microscopy for evaluating two-dimensional crystallization of membrane proteins. J. Struct. Biol. 171, 102-110. Huang, B., Bates, M., and Zhuang, X. (2009). Super-resolution fluorescence microscopy. Annu. Rev. Biochem. 78, 993-1016. Hunte, C. (2005). Specific protein-lipid interactions in membrane proteins. Biochem. Soc. Trans. 33, 938-942. Iacovache, I., Biasini, M., Kowal, J., Kukulski, W., Chami, M., van der Goot, F.G., Engel, A., and Remigy, H.W. (2010). The 2DX robot: a membrane protein 2D crystallization Swiss Army knife. J. Struct. Biol. 169, 370-378. Imai, S., Osawa, M., Takeuchi, K., and Shimada, I. (2010). Structural basis underlying the dual gate properties of KcsA. Proc. Natl. Acad. Sci. U. S. A. 107, 6216-6221. Jakobsson, P.J., Mancini, J.A., and Ford-Hutchinson, A.W. (1996). Identification and characterization of a novel human microsomal glutathione S-transferase with leukotriene C4 synthase activity and significant sequence identity to 5-lipoxygenase-activating protein and leukotriene C4 synthase. J. Biol. Chem. 271, 22203-22210. Jakobsson, P.J., Mancini, J.A., Riendeau, D., and Ford-Hutchinson, A.W. (1997). Identification and characterization of a novel microsomal enzyme with glutathione-dependent transferase and peroxidase activities. J. Biol. Chem. 272, 22934-22939. Jakobsson, P.J., Morgenstern, R., Mancini, J., Ford-Hutchinson, A., and Persson, B. (1999). Common structural features of MAPEG -- a widespread superfamily of membrane associated proteins with highly divergent functions in eicosanoid and glutathione metabolism. Protein Sci. 8, 689-692. Jardetzky, O. (1966). Simple allosteric model for membrane pumps. Nature 211, 969-970. Jegerschold, C., Pawelzik, S.C., Purhonen, P., Bhakat, P., Gheorghe, K.R., Gyobu, N., Mitsuoka, K., Morgenstern, R., Jakobsson, P.J., and Hebert, H. (2008). Structural basis for induced formation of the inflammatory mediator prostaglandin E2. Proc. Natl. Acad. Sci. U. S. A. 105, 11110-11115. Jiang, Y., Lee, A., Chen, J., Cadene, M., Chait, B.T., and MacKinnon, R. (2002a). Crystal structure

48

and mechanism of a calcium-gated potassium channel. Nature 417, 515-522. Jiang, Y., Lee, A., Chen, J., Cadene, M., Chait, B.T., and MacKinnon, R. (2002b). The open pore conformation of potassium channels. Nature 417, 523-526. Jiang, Y., Lee, A., Chen, J., Ruta, V., Cadene, M., Chait, B.T., and MacKinnon, R. (2003). X-ray structure of a voltage-dependent K+ channel. Nature 423, 33-41. Jiang, Y., Pico, A., Cadene, M., Chait, B.T., and MacKinnon, R. (2001). Structure of the RCK domain from the E. coli K+ channel and demonstration of its presence in the human BK channel. Neuron 29, 593-601. Jorgensen, P.L. (1988). Purification of Na+,K+-ATPase: enzyme sources, preparative problems, and preparation from mammalian kidney. Methods Enzymol. 156, 29-43. Kefala, G., Kwiatkowski, W., Esquivies, L., Maslennikov, I., and Choe, S. (2007). Application of Mistic to improving the expression and membrane integration of histidine kinase receptors from Escherichia coli. J. Struct. Funct. Genomics 8, 167-172. Kendrew, J.C., Bodo, G., Dintzis, H.M., Parrish, R.G., Wyckoff, H., and Phillips, D.C. (1958). A three-dimensional model of the myoglobin molecule obtained by x-ray analysis. Nature 181, 662-666. Killian, J.A. (1998). Hydrophobic mismatch between proteins and lipids in membranes. Biochim. Biophys. Acta 1376, 401-415. Kitts, E.I., Jr. (1996). Physics and chemistry of film and processing. Radiographics 16, 1467-1479; quiz 1464-1465. Knapek, E., and Dubochet, J. (1980). Beam damage to organic material is considerably reduced in cryo-electron microscopy. J. Mol. Biol. 141, 147-161. Koeck, P.J.B., Purhonen, P., Alvang, R., Grundberg, B., and Hebert, H. (2007). Single particle refinement in electron crystallography: a pilot study. J. Struct. Biol. 160, 344-352. Kuhlbrandt, W. (1992). Two-dimensional crystallization of membrane proteins. Q. Rev. Biophys. 25, 1-49. Kuhlbrandt, W., Wang, D.N., and Fujiyoshi, Y. (1994). Atomic model of plant light-harvesting complex by electron crystallography. Nature 367, 614-621. Kumar, H., Kasho, V., Smirnova, I., Finer-Moore, J.S., Kaback, H.R., and Stroud, R.M. (2014). Structure of sugar-bound LacY. Proc. Natl. Acad. Sci. U. S. A. 111, 1784-1788. Kuo, A., Gulbis, J.M., Antcliff, J.F., Rahman, T., Lowe, E.D., Zimmer, J., Cuthbertson, J., Ashcroft, F.M., Ezaki, T., and Doyle, D.A. (2003a). Crystal structure of the potassium channel KirBac1.1 in the closed state. Science 300, 1922-1926. Kuo, M.M., Haynes, W.J., Loukin, S.H., Kung, C., and Saimi, Y. (2005). Prokaryotic K(+) channels: from crystal structures to diversity. FEMS Microbiol. Rev. 29, 961-985. Kuo, M.M., Saimi, Y., and Kung, C. (2003b). Gain-of-function mutations indicate that Escherichia coli Kch forms a functional K+ conduit in vivo. EMBO J. 22, 4049-4058. Law, C.J., Maloney, P.C., and Wang, D.N. (2008). Ins and outs of major facilitator superfamily antiporters. Annu. Rev. Microbiol. 62, 289-305. Lee, S.Y., Lee, A., Chen, J., and MacKinnon, R. (2005). Structure of the KvAP voltage-dependent K+ channel and its dependence on the lipid membrane. Proc. Natl. Acad. Sci. U. S. A. 102, 15441-15446. Leonetti, M.D., Yuan, P., Hsiung, Y., and Mackinnon, R. (2012). Functional and structural analysis of the human SLO3 pH- and voltage-gated K+ channel. Proc. Natl. Acad. Sci. U. S. A. 109,

49

19274-19279. Li, D., Howe, N., Dukkipati, A., Shah, S.T., Bax, B.D., Edge, C., Bridges, A., Hardwicke, P., Singh, O.M., Giblin, G., et al. (2014). Crystallizing Membrane Proteins in the Lipidic Mesophase. Experience with Human Prostaglandin E2 Synthase 1 and an Evolving Strategy. Cryst. Growth Des., 14, 2034-2047. Li, X., Mooney, P., Zheng, S., Booth, C.R., Braunfeld, M.B., Gubbens, S., Agard, D.A., and Cheng, Y. (2013). Electron counting and beam-induced motion correction enable near-atomic-resolution single-particle cryo-EM. Nat. Methods 10, 584-590. Liao, M., Cao, E., Julius, D., and Cheng, Y. (2013). Structure of the TRPV1 ion channel determined by electron cryo-microscopy. Nature 504, 107-112. Lockless, S.W., Zhou, M., and MacKinnon, R. (2007). Structural and thermodynamic properties of selective ion binding in a K+ channel. PLoS Biol. 5, e121. Long, S.B., Campbell, E.B., and Mackinnon, R. (2005a). Crystal structure of a mammalian voltage-dependent Shaker family K+ channel. Science 309, 897-903. Long, S.B., Campbell, E.B., and Mackinnon, R. (2005b). Voltage sensor of Kv1.2: structural basis of electromechanical coupling. Science 309, 903-908. Long, S.B., Tao, X., Campbell, E.B., and MacKinnon, R. (2007). Atomic structure of a voltage-dependent K+ channel in a lipid membrane-like environment. Nature 450, 376-382. Lu, P., Bai, X.C., Ma, D., Xie, T., Yan, C., Sun, L., Yang, G., Zhao, Y., Zhou, R., Scheres, S.H., et al. (2014). Three-dimensional structure of human gamma-secretase. Nature 512, 166-170. Ludtke, S.J., Baldwin, P.R., and Chiu, W. (1999). EMAN: semiautomated software for high-resolution single-particle reconstructions. J. Struct. Biol. 128, 82-97. Lundback, A.K., Muller, S.A., Engel, A., and Hebert, H. (2009). Assembly of Kch, a putative potassium channel from Escherichia coli. J. Struct. Biol. 168, 288-293. MacKinnon, R. (2003). Potassium channels. FEBS Lett. 555, 62-65. Madej, M.G., Dang, S., Yan, N., and Kaback, H.R. (2013). Evolutionary mix-and-match with MFS transporters. Proc. Natl. Acad. Sci. U. S. A. 110, 5870-5874. Mancini, J.A., Abramovitz, M., Cox, M.E., Wong, E., Charleson, S., Perrier, H., Wang, Z., Prasit, P., and Vickers, P.J. (1993). 5-lipoxygenase-activating protein is an arachidonate binding protein. FEBS Lett. 318, 277-281. Martinez Molina, D., Eshaghi, S., and Nordlund, P. (2008). Catalysis within the lipid bilayer-structure and mechanism of the MAPEG family of integral membrane proteins. Curr. Opin. Struct. Biol. 18, 442-449. Martinez Molina, D., Wetterholm, A., Kohl, A., McCarthy, A.A., Niegowski, D., Ohlson, E., Hammarberg, T., Eshaghi, S., Haeggstrom, J.Z., and Nordlund, P. (2007). Structural basis for synthesis of inflammatory mediators by human leukotriene C4 synthase. Nature 448, 613-616. Massover, W.H. (2007). Radiation damage to protein specimens from electron beam imaging and diffraction: a mini-review of anti-damage approaches, with special reference to synchrotron X-ray crystallography. J. Synchr. Rad. 14, 116-127. McLellan, L.I., Wolf, C.R., and Hayes, J.D. (1989). Human microsomal glutathione S-transferase. Its involvement in the conjugation of hexachlorobuta-1,3-diene with glutathione. Biochem. J. 258, 87-93. Michel, H. (1983). Crystallization of Membrane-Proteins. Trends Biochem. Sci. 8, 56-59. Milkman, R. (1994). An Escherichia coli homologue of eukaryotic potassium channel proteins.

50

Proc. Natl. Acad. Sci. U. S. A. 91, 3510-3514. Miller, A.N., and Long, S.B. (2012). Crystal structure of the human two-pore domain potassium channel K2P1. Science 335, 432-436. Mindell, J.A., and Grigorieff, N. (2003). Accurate determination of local defocus and specimen tilt in electron microscopy. J. Struct. Biol. 142, 334-347. Mirza, O., Guan, L., Verner, G., Iwata, S., and Kaback, H.R. (2006). Structural evidence for induced fit and a mechanism for sugar/H+ symport in LacY. EMBO J. 25, 1177-1183. Miyazawa, A., Fujiyoshi, Y., and Unwin, N. (2003). Structure and gating mechanism of the acetylcholine receptor pore. Nature 423, 949-955. Mizushima, K., Awakihara, S., Kuroda, M., Ishikawa, T., Tsuda, M., and Tsuchiya, T. (1992). Cloning and sequencing of the melB gene encoding the melibiose permease of Salmonella typhimurium LT2. Mol. Gen. Genet. 234, 74-80. Morais-Cabral, J.H., Zhou, Y., and MacKinnon, R. (2001). Energetic optimization of ion conduction rate by the K+ selectivity filter. Nature 414, 37-42. Morgenstern, R. (2005). Microsomal glutathione transferase 1. Methods Enzymol. 401, 136-146. Morgenstern, R., Lundqvist, G., Andersson, G., Balk, L., and DePierre, J.W. (1984). The distribution of microsomal glutathione transferase among different organelles, different organs, and different organisms. Biochem. Pharmacol. 33, 3609-3614. Mosialou, E., Ekstrom, G., Adang, A.E., and Morgenstern, R. (1993). Evidence that rat liver microsomal glutathione transferase is responsible for glutathione-dependent protection against lipid peroxidation. Biochem. Pharmacol. 45, 1645-1651. Mosser, G. (2001). Two-dimensional crystallogenesis of transmembrane proteins. Micron 32, 517-540. Murata, Y., Iwasaki, H., Sasaki, M., Inaba, K., and Okamura, Y. (2005). Phosphoinositide phosphatase activity coupled to an intrinsic voltage sensor. Nature 435, 1239-1243. Murray, S.C., Flanagan, J., Popova, O.B., Chiu, W., Ludtke, S.J., and Serysheva, II. (2013). Validation of cryo-EM structure of IP(3)R1 channel. Structure 21, 900-909. Nimigean, C.M., and Allen, T.W. (2011). Origins of ion selectivity in potassium channels from the perspective of channel block. J. Gen. Physiol. 137, 405-413. Nogales, E., Wolf, S.G., and Downing, K.H. (1998). Structure of the alpha beta tubulin dimer by electron crystallography. Nature 391, 199-203. Norton, R.S., and Gulbis, J.M. (2010). Potassium channel gating: not an open and shut case. Proc. Natl. Acad. Sci. U. S. A. 107, 7623-7624. Papaneophytou, C.P., and Kontopidis, G. (2014). Statistical approaches to maximize recombinant protein expression in Escherichia coli: a general review. Protein Expr. Purif. 94, 22-32. Penczek, P.A. (2010). Resolution measures in molecular electron microscopy. Methods Enzymol. 482, 73-100. Poget, S.F., and Girvin, M.E. (2007). Solution NMR of membrane proteins in bilayer mimics: small is beautiful, but sometimes bigger is better. Biochim. Biophys. Acta 1768, 3098-3106. Quispe, J., Damiano, J., Mick, S.E., Nackashi, D.P., Fellmann, D., Ajero, T.G., Carragher, B., and Potter, C.S. (2007). An improved holey carbon film for cryo-electron microscopy. Microsc. Microanal. 13, 365-371. Radermacher, M., Wagenknecht, T., Verschoor, A., and Frank, J. (1987). Three-dimensional reconstruction from a single-exposure, random conical tilt series applied to the 50S ribosomal

51

subunit of Escherichia coli. J. Microsc. 146, 113-136. Ramsey, I.S., Moran, M.M., Chong, J.A., and Clapham, D.E. (2006). A voltage-gated proton-selective channel lacking the pore domain. Nature 440, 1213-1216. Roosild, T.P., Greenwald, J., Vega, M., Castronovo, S., Riek, R., and Choe, S. (2005). NMR structure of Mistic, a membrane-integrating protein for membrane protein expression. Science 307, 1317-1321. Ruskin, R.S., Yu, Z., and Grigorieff, N. (2013). Quantitative characterization of electron detectors for transmission electron microscopy. J. Struct. Biol. 184, 385-393. Saier, M.H., Jr. (2000). A functional-phylogenetic classification system for transmembrane solute transporters. Microbiol. Mol. Biol. Rev. 64, 354-411. Saier, M.H., Jr., Beatty, J.T., Goffeau, A., Harley, K.T., Heijne, W.H., Huang, S.C., Jack, D.L., Jahn, P.S., Lew, K., Liu, J., et al. (1999). The major facilitator superfamily. J. Mol. Microbiol. Biotechnol. 1, 257-279. Samuelsson, B., Morgenstern, R., and Jakobsson, P.J. (2007). Membrane prostaglandin E synthase-1: a novel therapeutic target. Pharmacol. Rev. 59, 207-224. Sansom, M.S., Shrivastava, I.H., Bright, J.N., Tate, J., Capener, C.E., and Biggin, P.C. (2002). Potassium channels: structures, models, simulations. Biochim. Biophys. Acta 1565, 294-307. Saraswat, M., Musante, L., Ravida, A., Shortt, B., Byrne, B., and Holthofer, H. (2013). Preparative purification of recombinant proteins: current status and future trends. Biomed. Res. Int. 2013, 312709. Sasaki, M., Takagi, M., and Okamura, Y. (2006). A voltage sensor-domain protein is a voltage-gated proton channel. Science 312, 589-592. Scherer, S., Arheit, M., Kowal, J., Zeng, X., and Stahlberg, H. (2014). Single particle 3D reconstruction for 2D crystal images of membrane proteins. J. Struct. Biol. 185, 267-277. Scheres, S.H. (2012). RELION: implementation of a Bayesian approach to cryo-EM structure determination. J. Struct. Biol. 180, 519-530. Schmidt-Krey, I., Lundqvist, G., Morgenstern, R., and Hebert, H. (1998). Parameters for the two-dimensional crystallization of the membrane protein microsomal glutathione transferase. J. Struct. Biol. 123, 87-96. Schmidt-Krey, I., and Rubinstein, J.L. (2011). Electron cryomicroscopy of membrane proteins: specimen preparation for two-dimensional crystals and single particles. Micron 42, 107-116. Shi, Y. (2013). Common folds and transport mechanisms of secondary active transporters. Annu. Rev. Biophys. 42, 51-72. Sjogren, T., Nord, J., Ek, M., Johansson, P., Liu, G., and Geschwindner, S. (2013). Crystal structure of microsomal prostaglandin E2 synthase provides insight into diversity in the MAPEG superfamily. Proc. Natl. Acad. Sci. U. S. A. 110, 3806-3811. Sorzano, C.O., Marabini, R., Velazquez-Muriel, J., Bilbao-Castro, J.R., Scheres, S.H., Carazo, J.M., and Pascual-Montano, A. (2004). XMIPP: a new generation of an open-source image processing package for electron microscopy. J. Struct. Biol. 148, 194-204. Stoylova, S.S., Flint, T.D., Kitmitto, A., Ford, R.C., and Holzenburg, A. (1998). Comparison of photosystem II 3D structure as determined by electron crystallography of frozen-hydrated and negatively stained specimens. Micron 29, 341-348. Subramaniam, S., and Milne, J.L. (2004). Three-dimensional electron microscopy at molecular resolution. Annu. Rev. Biophys. Biomol. Struct. 33, 141-155.

52

Suh, B.C., and Hille, B. (2008). PIP2 is a necessary cofactor for ion channel function: how and why? Annu. Rev. Biophys 37, 175-195. Swartz, K.J. (2008). Sensing voltage across lipid membranes. Nature 456, 891-897. Svensson, R., Alander, J., Armstrong, R.N., and Morgenstern, R. (2004). Kinetic characterization of thiolate anion formation and chemical catalysis of activated microsomal glutathione transferase 1. Biochemistry (Mosc). 43, 8869-8877. Tang, G., Peng, L., Baldwin, P.R., Mann, D.S., Jiang, W., Rees, I., and Ludtke, S.J. (2007). EMAN2: an extensible image processing suite for electron microscopy. J. Struct. Biol. 157, 38-46. Thoren, S., Weinander, R., Saha, S., Jegerschold, C., Pettersson, P.L., Samuelsson, B., Hebert, H., Hamberg, M., Morgenstern, R., and Jakobsson, P.J. (2003). Human microsomal prostaglandin E synthase-1: purification, functional characterization, and projection structure determination. J. Biol. Chem. 278, 22199-22209. Tombola, F., Pathak, M.M., and Isacoff, E.Y. (2006). How does voltage open an ion channel? Annu. Rev. Cell Dev. Biol. 22, 23-52. Tsai, C.J., and Ziegler, C. (2005). Structure determination of secondary transport proteins by electron crystallography: two-dimensional crystallization of the betaine uptake system BetP. J. Mol. Microbiol. Biotechnol. 10, 197-207. Tsai, C.J., and Ziegler, C. (2010). Coupling electron cryomicroscopy and X-ray crystallography to understand secondary active transport. Curr. Opin. Struct. Biol. 20, 448-455. Unger, V.M. (2000). Assessment of electron crystallographic data obtained from two-dimensional crystals of biological specimens. Acta Crystallogr. D: Biol. Crystallogr. 56, 1259-1269. Unger, V.M. (2001). Electron cryomicroscopy methods. Curr. Opin. Struct. Biol. 11, 548-554. Uozumi, N., Kume, K., Nagase, T., Nakatani, N., Ishii, S., Tashiro, F., Komagata, Y., Maki, K., Ikuta, K., Ouchi, Y., et al. (1997). Role of cytosolic phospholipase A2 in allergic response and parturition. Nature 390, 618-622. Wade, R.H. (1992). A Brief Look at Imaging and Contrast Transfer. Ultramicroscopy 46, 145-156. Wallin, E., and von Heijne, G. (1998). Genome-wide analysis of integral membrane proteins from eubacterial, archaean, and eukaryotic organisms. Protein Sci. 7, 1029-1038. Valpuesta, J.M., Carrascosa, J.L., and Henderson, R. (1994). Analysis of electron microscope images and electron diffraction patterns of thin crystals of phi 29 connectors in ice. J. Mol. Biol. 240, 281-287. van Heel, M., Harauz, G., Orlova, E.V., Schmidt, R., and Schatz, M. (1996). A new generation of the IMAGIC image processing system. J. Struct. Biol. 116, 17-24. Wang, D.N., and Kuhlbrandt, W. (1991). High-resolution electron crystallography of light-harvesting chlorophyll a/b-protein complex in three different media. J. Mol. Biol. 217, 691-699. Wang, H., and Downing, K.H. (2011). Specimen preparation for electron diffraction of thin crystals. Micron 42, 132-140. Weiss, M.S., Wacker, T., Weckesser, J., Welte, W., and Schulz, G.E. (1990). The three-dimensional structure of porin from Rhodobacter capsulatus at 3 A resolution. FEBS Lett. 267, 268-272. Werner, T., Morris, M.B., Dastmalchi, S., and Church, W.B. (2012). Structural modelling and dynamics of proteins for insights into drug interactions. Adv. Drug Deliv. Rev. 64, 323-343. Wilson, D.M., and Wilson, T.H. (1987). Cation specificity for sugar substrates of the melibiose carrier in Escherichia coli. Biochim. Biophys. Acta 904, 191-200.

53

Wisedchaisri, G., and Gonen, T. (2011). Fragment-based phase extension for three-dimensional structure determination of membrane proteins by electron crystallography. Structure 19, 976-987. Wisedchaisri, G., Reichow, S.L., and Gonen, T. (2011). Advances in structural and functional analysis of membrane proteins by electron crystallography. Structure 19, 1381-1393. Wu, Y., Yang, Y., Ye, S., and Jiang, Y. (2010). Structure of the gating ring from the human large-conductance Ca(2+)-gated K(+) channel. Nature 466, 393-397. Xu, Y., Ramu, Y., and Lu, Z. (2010). A shaker K+ channel with a miniature engineered voltage sensor. Cell 142, 580-589. Yamagata, K., Matsumura, K., Inoue, W., Shiraki, T., Suzuki, K., Yasuda, S., Sugiura, H., Cao, C., Watanabe, Y., and Kobayashi, S. (2001). Coexpression of microsomal-type prostaglandin E synthase with cyclooxygenase-2 in brain endothelial cells of rats during endotoxin-induced fever. J. Neurosci. 21, 2669-2677. Yan, N. (2013a). Structural advances for the major facilitator superfamily (MFS) transporters. Trends Biochem. Sci. 38, 151-159. Yan, N. (2013b). Structural investigation of the proton-coupled secondary transporters. Curr. Opin. Struct. Biol. 23, 483-491. Ye, S., Li, Y., Chen, L., and Jiang, Y. (2006). Crystal structures of a ligand-free MthK gating ring: insights into the ligand gating mechanism of K+ channels. Cell 126, 1161-1173. Young, C.L., Britton, Z.T., and Robinson, A.S. (2012). Recombinant protein expression and purification: a comprehensive review of affinity tags and microbial applications. Biotechnol. J., 7, 620-634. Yousef, M.S., and Guan, L. (2009). A 3D structure model of the melibiose permease of Escherichia coli represents a distinctive fold for Na+ symporters. Proc. Natl. Acad. Sci. U. S. A. 106, 15291-15296. Yuan, P., Leonetti, M.D., Hsiung, Y., and MacKinnon, R. (2012). Open structure of the Ca2+ gating ring in the high-conductance Ca2+-activated K+ channel. Nature 481, 94-97. Yuan, P., Leonetti, M.D., Pico, A.R., Hsiung, Y., and MacKinnon, R. (2010). Structure of the human BK channel Ca2+-activation apparatus at 3.0 A resolution. Science 329, 182-186. Zeldin, D.C. (2001). Epoxygenase pathways of arachidonic acid metabolism. J. Biol. Chem. 276, 36059-36062. Zhang, L., and Ren, G. (2012). IPET and FETR: experimental approach for studying molecular structure dynamics by cryo-electron tomography of a single-molecule structure. PloS one 7, e30249. Zhang, L., Song, J., Newhouse, Y., Zhang, S., Weisgraber, K.H., and Ren, G. (2010). An optimized negative-staining protocol of electron microscopy for apoE4 POPC lipoprotein. J. Lipid Res. 51, 1228-1236. Zhou, H.X., and Cross, T.A. (2013). Influences of membrane mimetic environments on membrane protein structures. Annu. Rev. Biophys. 42, 361-392. Zhou, M., and MacKinnon, R. (2004). A mutant KcsA K(+) channel with altered conduction properties and selectivity filter ion distribution. J. Mol. Biol. 338, 839-846. Zhou, Y., and MacKinnon, R. (2003). The occupancy of ions in the K+ selectivity filter: charge balance and coupling of ion binding to a protein conformational change underlie high conduction rates. J. Mol. Biol. 333, 965-975. Zhou, Y., Morais-Cabral, J.H., Kaufman, A., and MacKinnon, R. (2001). Chemistry of ion

54

coordination and hydration revealed by a K+ channel-Fab complex at 2.0 A resolution. Nature 414, 43-48.

55

Suggest Documents