The advantages of being small Glycosyltransferases in many dimensions and glycolipid synthesis in

Mycoplasma pneumoniae

Maria Rosén Klement

Stockholm University

Doctoral thesis ©Maria Rosén Klement, Stockholm 2007 ISBN 978-91-7155-503-8 Pages 1-56 Printed in Sweden by US-AB, Stockholm 2007 Distributor: Stockholm University Libary

Till min familj, Göran, Emmy och Thyra

List of Publications

This thesis is based on the following publications: I. Rosén M.L., Edman M., Sjöström M., and Wieslander Å. Recognition

of fold and sugar linkage for glycosyltransferases by multivariate sequence analysis. J. Biol. Chem., 2004, 279(37):38683-92.

II. Lind J1., Rämö T1., Rosén Klement M.L., Bárány-Wallje E., Epand

R.M., Epand R.F., Mäler L., and Wieslander Å. High cationic charge and bilayer interface-binding helices in a regulatory lipid glycosyltransferase. Biochemistry, 2007, 46(19):5664-77

III. Rosén Klement M.L, Öjemyr L., Tagscherer K., Widmalm G., and

Wieslander Å. A processive lipid glycosyltransferase in the small human pathogen Mycoplasma pneumoniae: Involvement in host immune response. Mol. Microbiol., 2007, 65(6):1444-57

IV. Wikström M., Kelly A., Eriksson H., Georgiev, A., Rosén Klement

M.L., Dowhan W., and Wieslander Å. Curvature-engineered Escherichia coli bilayers reveal critical lipid head-group size for membrane protein function in vivo. Manuscript

1These authors have contributed equally Reprints were made with permission from the publisher

Additional publication Wieslander, Å. Rosén, M. (2001) Cell membranes and transport. In “Molecular Biology and Pathogenicity of Mycoplasmas” (S. Razin & R. Herrmann, eds), Kluwer Academic/Plenum Press

Rajalahti T, Huang F, Rosén Klement M.L., Pisareva T, Edman M, Sjöström M, Wieslander A, Norling B. Proteins in different Synechocystis compartments have distinguishing N-terminal features: A combined proteomics and multivariate sequence analysis. J. Proteome. Res. 2007 6(7):2420-34

Contents

Introduction .....................................................................................................1

The biological membrane................................................................................3 Lipid properties ................................................................................................................4

Lipid shape.................................................................................................................4 Glycolipids .......................................................................................................................7

Mollicutes and Mycoplasmas ..........................................................................9 Lipid content and synthesis ...........................................................................................10 Cholesterol ....................................................................................................................11 Immune response in a Mycoplasma infection...............................................................12 Autoimmunity.................................................................................................................13

Glycosyltransferases.....................................................................................14 Glycosyltransferase classification .................................................................................14 Glycosyltransferase fold types ......................................................................................16 GT-A ..............................................................................................................................17

Inverting reaction mechanism..................................................................................18 Retaining reaction mechanism.................................................................................20

GT-B ..............................................................................................................................21 Inverting mechanism................................................................................................22 Retaining mechanism ..............................................................................................24

Membrane binding ........................................................................................25

Sequence analysis and protein classification ...............................................27 Multivariate data analysis ..............................................................................................27

Previous studies.......................................................................................................28

Multivariate projection methods ....................................................................29 Principal component analysis ..................................................................................29

Partial least squares projection to latent structures ......................................................30

Summary of papers.......................................................................................32 Paper I ...........................................................................................................................32 Paper II ..........................................................................................................................34 Paper III .........................................................................................................................36 Paper IV.........................................................................................................................37

Concluding remarks ......................................................................................40

Acknowledgment...........................................................................................42

References....................................................................................................42

Abbreviations

Gal Galactose Glc Glucose GalGalDAG 1,2-diacyl-3-O-[α-D-galactopyranosyl-(1→6)-O-β-

D-galactopyranosyl]-sn-glycerol GlcGlcDAG 1,2-diacyl-3-O-[α-D-glucopyranosyl-(1→2)-O-α-

D-glucopyranosyl]-sn-glycerol GalDAG 1,2-diacyl-3-O-( β -D-galactopyranosyl)-sn-glycerol GlcDAG 1,2-diacyl-3-O-(α-D-glucopyranosyl)-sn-glycerol GT Glycosyltransferase LPS Lipopolysaccharide PA Phosphatidic acid PC Phosphatidylcholine PE Phosphatidylethanolamine PG Phosphatidylglycerol CL Cardiolipin GM1 monosialotetrahexosylganglioside GA-1 asialo GM1 TM Trans membrane APC Antigen presenting cell GBS Guillian Barré syndrome ACC Auto cross covariances PCA Principal component analysis PLS Partial least squares projections to latent structures PLS-DA PLS discriminant analysis

1

Introduction

Since the 1950´s the minimal genome concept has been a challenging idea for many researchers, that is the genome believed to describe life. What is life? As stated in many textbooks, living organisms should be able to grow by themselves, react to stimuli, and self replicate, but what genes and proteins are needed for this? Mollicutes are the smallest self replicating bacteria known today. After the publication of the genome sequence of Mycoplasma genitalium (Fraser et al., 1995) the scientific world has questioned if this bacterium contains the minimal self sufficient gene set or if even a smaller genome is possible? By comparing M. genitalium, with 468 ORFs with Haemophilus influenzae containing 1703 ORFs 240 genes were found to be homologous in the two species (Mushegian and Koonin, 1996). These genes were believed to be the minimal genome set, since the two bacterial species belong to two ancient bacterial lineages, Gram-positive and Gram-negative, respectively. However, when analysing the 240 orthologous genes it became evident that some essential pathways were missing (Mushegian and Koonin, 1996). This is explained by the presence of non-orthologous genes coding for enzymes performing the same reaction. Hence, estimating the minimal genome becomes difficult and the metabolome is also of importance and will vary in different growth conditions. When taking non-orthologous genes into account a minimal genome set was predicted to be 256 genes (Mushegian and Koonin, 1996). In 2006 Glass and colleagues randomly inactivated all genes in M. genitalium to investigate the minimal number of essential genes in this minimal bacterium. It was found that 382 out of the 482 were essential (Glass et al., 2006). This is quite close to the 256 predicted in 1996, however one should take into account that the growth conditions are of importance. Surprisingly one of the largest groups of essential genes was found to be genes with unknown function. M. genitalium was a good candidate for these experiments due to low genomic redundancy, bacteria with larger genomes contains higher degree of genomic redundancy. However, using M. genitalium, a parasite and human pathogen as the base for the minimal cell can be doubtful, since it has genes that are of importance for the pathogenicity and interaction with the host cell, e.g. the bacterium needs to send signals to the host cell to obtain nutrients. These functions are essential

2

to M. genitalium, but would not be of importance to other bacteria. The largest groups of essential genes all consists of genes encoding proteins involved in replication, transcription, translation, and repair, the latter is however only a minor group of genes in Mollicutes sequenced until today.

Mycoplasmas are known to have the ability to fuse with, invade the cell or live in close contact with the host cell, being obligate parasites (Baseman et al., 1995; Dimitrov et al., 1993). They are able to incorporate essential components such as fatty acids and amino acids from the host cell to save energy (Baseman and Tully, 1997). It is also known that some mycoplasmas incorporate host lipids into the membrane (Uemura et al., 1988), which would mimic the host cell and prevent them being discovered by the host immune system. These bacteria thus have a large fraction of specific transporters essential for a parasitic life, which would not be essential for other organisms. Even if the bacterium is small, it has functions that a larger bacterium does not need, which makes estimations of the minimal genome difficult.

Mutagenesis experiments, making knockout mutants in M. genitalium have shown that as mentioned, 382 out of 482 genes are needed for survival. Within these, the genes encoding three glycosyltransferases (GT) could not be disrupted by the transposones (Glass et al., 2006), and two of these are included in the minimal genome set estimated in 1996 (Mushegian and Koonin, 1996). M. pneumoniae is almost an exact copy of M. genitalium with 207 additional ORFs that are unique (Himmelreich et al., 1997). When considering the high amino acid identity between M. genitalium and M. pneumoniae orthologous, the protein functions are believed to be similar in the two species and thus be essential in both species. Why would a glycolipid synthase be essential to the obligate human parasite, M. pneumoniae and why can’t some lipids be replaced by others? The goal in this thesis was to identify the glycosyltransferases used in the glycolipid synthesis in M. pneumoniae. Only three GTs had been proposed after sequencing the genome, however the function of these could not be predicted. The closest homologues were all LPS enzymes from Gram- positive bacteria, a structure not found in Mollicutes, however lipoligosaccharides have been found (Smith, 1992). Glycolipids are only a minor component of the M. pneumoniae membrane, but the variation is large. How are glycolipids synthesised and how do the enzymes producing them work? These are the questions that will be addressed in this thesis.

3

The biological membrane

Life would not exist without the biological membrane, the shield that separates the cell from its surrounding making it possible to have a different environment within. The membrane consists of both a large variety of different lipids forming a bilayer and membrane proteins embedded in the lipid layers. The lipid bilayer is formed due to hydrophobic and hydrophilic properties, with the hydrophilic head groups shielding the hydrophobic hydrocarbon chains from the water surrounding it. The physical properties and the shape of the lipids will influence the function of the membrane (van den Brink-van der Laan et al., 2004). The biological membrane will also work as a selective filter, determining all communication with the world outside. Only a very limited number of molecule species can pass the bilayer without the assistance of a transporter or channel.

The cellular lipidome consists of thousands of different lipid molecules and the lipid content and amounts vary greatly between different bacterial and eukaryotic species (van Meer, 2005). The total lipid content of the cell is only about 5 % of the dry weight and 40-50% of the membrane with the rest being proteins (Smith, 1992). The major functions of lipids are the same, but the way to perform them are different. The major lipids in for example Echerichia coli are phosphatidylethanolamine (PE) 75% phosphatidlyglycerol (PG) 20% and cardiolipin (CL) 5% (all mol %) (Cronan, 2003). Acholeplasma laidlawii, a well studied Mollicute, contains 15-25% PG and large amounts of glycolipids, 15-50% GlcDAG and GlcGlcDAG (Rilfors et al., 1993). The thylakoid membrane, the most common bio-membrane on Earth, is unusual with its low phospholipid content, PG 10% but large amounts of glycolipids, 50% GalDAG and 30% GalGalDAG (all mole %) (Lee, 2000). The large differences in lipid composition of different organisms reflect the different properties of lipids and the differences in growth. In the next section, the properties of lipids will be discussed.

4

Lipid properties Why are there so many kinds of lipids? Lipids differ in head group, back bone and acyl chain length and chain unsaturation (Dowhan and Bogdanov, 2002). These differences will have effects on the property and function of the lipid. The physical state of membrane lipids has been found to strongly affect membrane associated processes (Dowhan, 1997), such as insertion of membrane proteins, membrane protein function, and binding of membrane associated proteins. The membrane flexibility and ability to adapt to different environmental changes is only possible with a large variety of lipids, since the membrane needs to be fluid and flexible at all times for the cell to function.

There are many different forces that control the behaviour of the lipids, for example the attraction or repulsion of the head groups, the molecular structure and ratio between the volumes of the head groups, the chain length and saturation, and the hydration forces (Boggs, 1987). Within these, the attraction or repulsion of charged head groups is one of the most important properties, which causes the lipids to become more or less densely packed. This also affects their tendency to form hydrogen bonds. The intramolecular bonds are important for the stability of the membrane, charged or neutral lipids with hydrogen-donating and accepting groups will form these interactions. It is important to remember that the membrane is in a fluid state and the interhydrogen bonds will constantly break and reform between different lipid molecules (Boggs, 1987).

Lipid shape The lipid shape is another important property that will affect the bilayer. There are three major classes of lipid structures, inverted cone, cylindrical, and cone shaped (Fig. 1). The first type is most common for lysophospholipids, having only one acyl chain, phospholipids with short akyl chains, and detergents. This will cause the lipids to form micells due to a large head group and small hydrophobic domain, with the alkyl chains facing inwards (Reviewed in) (Dowhan and Bogdanov, 2002). Cylindrically shaped lipids such as phosphatidylcholine (PC), sphingomyelin, cardiolipin, phospahtidic acid (PA), and digalactosyldiacylglycerol, (GalGalDAG) will pack into bilayers, either as

5

ordered gel phase (Lβ) or liquid crystalline phase (Lα). This allows for tight packing of acyl chains and will efficiently exclude water. These lipids have head groups and hydrophobic domains of similar size (Dowhan and Bogdanov, 2002) and will thus have a cylindrical shape (se Fig. 1). The last group of lipids have smaller head and larger hydrophobic domains, causing them to become cone shaped. PE and monogalactosyldiacylglycerol (GalDAG) together with, PA, and CL are examples of lipids with this shape, the latter, however only in the presence of divalent cations. The ion reduces the repulsion of the negatively charged head groups, making CL cone shaped instead of cylindrical. This shape will tend to form a hexagonal (HII) phase or cubic phase, with the acyl chains facing outwards and the head groups inwards towards the aqueous core in long tubes (see Fig. 1) (Dowhan and Bogdanov, 2002). The inverted cone and cone shaped lipids are considered to be nonbilayer-prone and when mixed with bilayer forming lipids, the cone shape will change the physical properties of the bilayer and cause stress or strain to the bilayer structure. The chain saturation can also affect the shape of the lipid, when increasing the unsaturation the acyl chains will become larger the shape will change from cylindrical to conical, this will also happen with increased temperature (Dowhan and Bogdanov, 2002). Figure 1. Phases and molecular shapes of different lipids. The physical shape and the phase they adopt is shown for the three different shapes, inverted cone, cylindrical and cone types (Dowhan and Bogdanov, 2002).

6

The liquid crystalline phase, formed by cylindrical lipids has been found to be essential for the viability of the cell and the lipid content is therefore adjusted to maintain this property. The strain caused by the insertion of non-bilayer forming lipids, results in the curvature stress that exist in the membrane. Forcing a non-bilayer prone lipid into a cone shape in a membrane will result in the exposure of the hydrophobic core to the aqueous surrounding (Dowhan and Bogdanov, 2002). This is probably driving the formation of micro domains, separating the different types of lipids that will decrease the tension in the membrane. The existence of rafts, specific lipid domains in eukaryotes, which constitutes micro environments has been under debate for decades. Rafts are believed to be important in many cellular functions, such as cellular signal transduction and trafficking (Hanzal-Bayer and Hancock, 2007). Lipid domains in bacteria would produce a unique milieu for different enzymes and could potentially regulate their functions. Different lipids could therefore regulate different enzymes in both eukaryotes and prokaryotes (Matsumoto et al., 2006).

The importance of the existence of different types of lipids can be illustrated with the mutant AD93 (Cronan, 2003; DeChavigny et al., 1991). This mutant lacks the non-bilayer prone PE and grows very poorly. Several membrane proteins have been found to have an altered topology in this mutant e.g. lactose permease and gamma-aminobutyric acid permease (Xie et al., 2006; Zhang et al., 2005). The properties of PE have been found to be important for the insertion of these proteins in the correct transmembrane orientation. The non-bilayer prone lipid GlcDAG was found to be able to replace PE and restore both the orientation of LacY and the growth rate of the mutant (Xie et al., 2006). This shows how important the structure of the lipids is, and how different enzymes use the different lipids and probably the bilayer stress that non-bilayer forming lipids causes to insert correctly into the membrane. Charged lipids are of great functional importance to the cell, such as the anionic lipid PG, but is only a minor component (20- 30%) in E.coli and most other membrane. The charged head group is utilised in the membrane association of many cytoplasmic proteins (Dowhan, 1997). It is believed for example that the negative charges of PG are not equally distributed in the membrane. Charge islands would be more attractive binding sites than a random distribution (Dowhan, 1997). Other lipids in bacteria are also thought to exist in micro domains such as CL and PE (reviewed in (Matsumoto et al., 2006)) offering specific environments for certain proteins and enzymes. Micro domains in the bacterial membrane, were also believed to be formed by the differences between PE, nonbilayer-prone and CL, bilayer-prone. The preference for certain lipids by enzymes is known, for exampel MurG a membrane bound glycosyltransferase involved

7

in the cell wall synthesis in E. coli binds preferentially to CL (van den Brink-van der Laan et al., 2003).

Glycolipids As described above, the glycolipid content of various membranes varies greatly. One of the organisms that has been very well studied and from which most information of the importance of glycolipids originates is A. laidlawii. Studies have shown that unsaturated GlcDAG is more nonbilayer-prone then saturated GlcDAG, with the unsaturated being more common in this bacterium. To overcome the tension created by unsaturated GlcDAG, the ratio of GlcDAG to GlcGlcDAG (bilayer-prone) decreases when the bacterium changes the properties of the membrane by altering the glycolipid content, instead of the unsaturation of the fatty acid chains (Wieslander et al., 1978). The latter is more common in other bacteria e.g. E. coli. Why would a bacterium use this system to make new glycolipids instead of just changing the saturation of e.g. PG or PE? The glycolipid content in bacteria lacking a cell wall, e.g. Mycoplasmas is found to be higher than in bacteria having a this structure. Glycolipids have the ability to form hydrogen bonds between the sugar molecules and would therefore stabilise membrane without the need of a cell wall, (Boggs, 1987). However, the glycolipid content has been found to vary greatly in this bacterial group. This could however be explained by the presence of a lipoglycan layer in some spieces, replacing the glycolipids (Smith, 1992).

Glycolipids have also been found to be of major importance for photosynthesis. The transition temperature is higher for glycolipids than for phospholipids, increasing the potential to form clusters in the membrane (Curatolo, 1987b). The glycolipid GalDAG is also found to drive the stacking of the membrane to form grana. This stacking allows for efficient packing of the thylakoid membrane within the chloroplast and leads to the separation of photosystem I and II (Lee, 2000). This glycolipid is a nonbilayer-prone lipid, due to its smaller head group and bulky highly unsaturated chains. GalDAG has been found to be highly associated with the light harvesting complex II, with 80 molecules per protein complex (Simidjiev et al., 2000). The binding to the membrane proteins forces the lipids to become more bilayer prone in structure. The strained nonbilayer properties of GalDAG and the energy stored in this tension is believed to change or modulate membrane protein packing properties and also the function (Lee, 2000). This phenomenon has also been seen in cytochrome c oxidase that forces CL into a bilayer prone state (Rietveld et al., 1987). The

8

ratio between GalDAG and GalGalDAG in the thylakoid membrane has been found be similar to the PE/PC ratio found in most plasma membranes, hence the proportion of bilayer prone and nonbilayer-prone is the same in the two membranes. It is believed that the reason for using glycolipids in the thylakoid is to make it more stable due to the intra hydrogen bonds (Curatolo, 1987a). The membrane stability is of major importance due to the transfer of charge and where the trans-bilayer ionic environment must be keept constant.

9

Mollicutes and Mycoplasmas

Most of the work in this thesis has been done on or with enzymes from M. pneumoniae a bacterium classified as a Mollicute. This group of bacteria is the smallest capable of autonomous growth and is characterised by the small genomes (0.58 kb to 2.2 Mb), their low G+C % content (23-40%) (Johansson and Pettersson, 2002), and their lack of cell wall. Until today 200 different species have been identified. Mollicutes belong to the phylum Firmicutes to which also other Gram-positive bacteria with low G+C % content belong, such as Bacilli and Clostridia (see Fig 2). Studies of the 16S rRNA show that the Mollicutes diverged from the Streptococcus branch of the Gram positive bacteria with low G+C content about 605 million years ago. By genome reduction different branches of the Mollicutes were developed during 400 million years, giving the different families of Mollicutes found today (Manillof, 2002). Figure 2 Schematic representation of relationships within the Firmicutes (Wolf et al., 2004)

In Mollicutes, the majority of the lipids are located in the cytoplasmic membrane and constitute between 25-35% of the dry weight (Hayashi and Wu, 1990). Lipid synthesis is diverse and sequence analysis has revealed that the identified phospholipid synthesis (see Fig. 3), must differ from other bacteria with known metabolic pathways. Analogous enzymes/genes from E. coli and other bacteria with known lipid synthesis have not been found in the

10

sequenced Mycoplasma genomes. One of the two acyltransferases, used in the phosphatidic acid synthesis is for example missing, in M. pneumoniae, M. genitalium, and Ureaplasma urealyticum and phosphatidylglycerol phosphate phosphatase in M. pulmonis (Chambaud et al., 2001; Fraser et al., 1995; Glass et al., 2006; Himmelreich et al., 1996). However, the one found might be adding both acyl chains on the glycerol in these species.

Most of the studies on the lipid content and lipid synthesis in Mollicutes were performed in the late sixties and not much work has been done since then. The lipid content in approximately 20 different Mycoplasma species have been characterised (Smith, 1992). Later studies have mainly been focusing on A. laidlawii, M. pneumoniae, M. fermentans. Mollicutes have no ability to synthesise fatty acids, nor change the fatty acid chain length or saturation with a few exceptions, e.g. Acholeplasmas are capable of synthesising saturated fatty acids (Smith, 1992) and also elongate exogenous chains (McElhaney, 1992). Mollicutes depend on the environment for the supply of cholesterol, an essential component for the structure and properties of the membrane. Acholeplasma is an exception, where cholesterol is not essential. These components together with some phospholipids, e.g. PC and PE are known to be incorporated into the membrane from the surroundings. These lipids can also be further metabolised or the saturation of the fatty acids in PC changed (Rottem et al., 1986). The lack of cell wall peptidoglycan layer in mollicutes makes them unique in the bacterial kingdom. This property places them in a direct contact with the surrounding, making it easier to take up essential components. A lipoglycan layer has been found in some species, for example in Acholeplasma, Anaeroplasma, Ureaplasma and a few Mycoplasma. This layer is firmly attached to the membrane and constitutes about 10% of the dry weight (Smith, 1992). The absence of a cell wall has made Mollicutes an ideal organism for studying the biomembrane, making it easy to remove and add membrane components that are otherwise difficult.

Lipid content and synthesis The membrane of Mollicutes contains mainly three classes of lipids, phospholipids, glycolipids, and neutral lipids (Smith, 1992). The major lipids in the membrane are phospholipids and especially PG. The lipid synthesis pathways are believed to be very similar in most Mollicutes, however the pathways have been mostly studied in A. laidlawii. PA is the key lipid in the biosynthesis of both glycolipids and phospholipids in A. laidlawii (McElhaney, 1992) and believed to be in most Mollicutes (Fig. 3). PA is synthesised by acylation of sn-glycerol-3-phosphate with two fatty acids. PA

11

can be dephosphorylated to a diacylglycerol, DAG by a PA phosphatase (Berg and Wieslander, 1997). DAG can be glucosylated in two steps to form GlcDAG and GlcGlcDAG (Dahlqvist et al., 1992). The glycolipids can be further acylated by a third acyl chain to form MAGlcDAG and MAGlcGlcDAG (Hauksson et al., 1994a, b). Phosphoglycolipids, GPGlcGlcDAG and MABGPGlcGlcDAG from GlcGlcDAG are also synthesised by A. laidlawii however by an unknown process (Hauksson et al., 1994b). PA is also a precursor in the phospholipid synthesis, where cytidine diphosphate diglyceride is formed from CTP and PA. Phosphatidylglycerol-3-phosphate is formed by CDP-diacylglycerol and glycerol phosphate and further metabolised of a phosphate group to form PG. Two PG molecules can form a diphosphatidylglycerol or cardiolipin (Smith, 1992) (Fig. 3).

Figure 3. Metabolic pathways for the synthesis of glycolipids and phospholipids in Mollicutes. GP, glycerol-phosphate; PA, phosphatidic acid; CPD-DAG, Cytidine diphosphate diacylglycerol; DAG, diacylglycerol; MH, monohexose; DH, dihexose; TH, trihexose; PG, phosphatidylglycerol; APG, acylphosphatidylglycerol; CL, cardiolipin; and Cer, ceramide. All steps between PA and PG are not shown.

Cholesterol

Cholesterol is as stated previously, an essential component in Mollicutes and is inserted in between the long hydrophobic fatty acid chains in the membrane and account for between 4 to 20% of the total lipid content (Smith, 1992). Cholesterol enhances the formation of a liquid ordered phase (Simons and Vaz, 2004) and decreases the surface charge caused by PG and Cardiolipin in the membrane, by modifying the intermolecular interaction between these polar head groups (Tarshis et al., 1993). However, cholesterol

12

has been found to have a preference for sphingolipids, a component found in many Mycoplasma species, due to the hydrogen bonding ability and the saturation of the chains that is naturally occurring in sphingolipids (Li et al., 2001). All sphingolipids have both hydrogen bond acceptors and donors (the carbonyl group and the amide group), whereas DAG based lipids such as PC only have the hydrogen bond acceptor, the carbonyl group (Li et al., 2001). By incorporating cholesterol into the membrane, the bilayer can consist of lipids in a physical intermediate state between gel and liquid–crystalline states (Ladbrooke et al., 1968; Shah and Schulman, 1967), forming membrane domains. The cholesterol content has also been found to be very important in the fusion of Mycoplasma cells, e.g. M. capricolium with the eukaryotic host cell (Franzoso et al., 1992; O'Toole and Lowdell 1990), but is believed to be important for other cholesterol containing Mollicutes such as M. pneumoniae. This is probably due to the Mycoplasma membrane becoming more like the host cell making it easier to fuse the membranes.

Immune response in a Mycoplasma infection

M. pneumoniae is a common human pathogen, and is the bacterium behind pneumonia in about 20% of all the cases, and up to 50% of all pneumonia in the age group of 20-30 years. It is often referred to as walking pneumonia, due to rather week symptoms and long incubation time. As most other Mycoplasmas it is a mucosal pathogen, living in close association with epithelial cells (Waites and Talkington, 2004). The progression of an M. pneumoniae infection is slow and when the bacterium reaches the lower respiratory tract, it meets the host defence. Macrophages become activated and start to phagocytose the bacteria. The macrophages, also called antigen presenting cell, (APC) present mainly peptides from digested surface associated proteins and LPS on the surface of the cell through MHC I and MHC II molecules. One of the most important peptide antigens for the immune system in mycoplasma infections are lipoproteins, due to their surface exposure in the bacterium (Chambaud et al., 1999).

Recently it was discovered that a new class of molecules, named CD1, present lipids instead of peptides on the surface of the antigen presenting cells (Porcelli and Modlin, 1999). The activation of invariant nature killer cells (iNKT) cells through these molecules has been found to be an important factor in the defense of bacteria that lack a cell wall (Kinjo et al., 2006). Kinjo and co-workers showed that the glycolipids, α-GalDAG from the human pathogen Borrelia burgdorferi is presented by the CD1d molecule on APC (Kinjo et al., 2006). Activation of iNKT cells by CD1d and through the T cell receptor will also activate B-cells to proliferate and

13

produce immunoglobulins (Galli et al., 2003). The sugar preference and orientation for CD1d binding is still not clearly understood. Both α-GalCer and β- GalCer (Parekh et al., 2004) as well α-GlcCer (Brigl et al., 2006) can bind to CD1d and activate iNTK cells. When analysing the structure of a CD1d molecule with the bound α-GalCer substrate, it was however discovered that the molecule have structural preferences for the alpha configuration. The structure shows that a beta linkage between the ceramide and the galactose would disrupt the interaction between the CD1d molecule and the T-cell receptor (Borg et al., 2007). The activation of the immune system through a MHC unrestricted pathway has previously been reported after a Mycoplasma infection (Ruuth and Praz, 1989). These results indicate that glycolipids from M. pneumoniae and other cell wall lacking bacteria could be processed in similar fashion and be presented by CD1d or by a similar molecule. The activation of iNKT stimulates the immune system and activates the production of antibodies.

Autoimmunity

M. pneumoniae is known to rarely, in 0.1% of the cases cause Guillain-Barré syndrome (GBS) (Tsiodras et al., 2005). This autoimmune disease breaks down the myelin sheets surrounding nerve cells, which causes limb weakness and loss of nerve reflexes. The auto antibodies recognise glycolipids, the major component in the myelin sheets, indicating that the bacteria infecting the patient prior to GBS have these epitopes (Kumada et al., 1997). The GM1 lipid sugar structure, the most common epitope for antibodies in GBS patients, has been found in some strains of Campylobacter jejuni, indicating molecular mimicry as the cause of GBS (Gilbert et al., 2000). In GBS patients where the disease is preceded by an M. pneumoniae infection, anti-GalCer antibodies are most common, and anti-GA1 (Kusunoki et al., 1995; Kusunoki et al., 2001) and GM1 have also been found (Susuki et al., 2004). Susuki and co-workers showed that the GM1 sugar structure is probably a part of the membrane of M. pneumoniae (Susuki et al., 2004). Antibodies directed against this lipid react to total lipid extracts from M. pneumoniae. Cross reactivity between anti-GM1 and anti-GalCer has however been seen. The bacterial GTs, found to be essential by knock out studies, must have the ability to use the eukaryotic ceramide as base in the sugar lipid synthesis. Molecular mimicry described above is a way to hide from the immune system, however once the bacterium has been discovered, the production of autoantibodies can become lethal to the patient.

14

Glycosyltransferases

The turn-over of sugars is one of the most widespread enzymatic functions on this planet. Approximately 1-2% of the coding regions of all genomes are glycosyltransferase genes (Davies and Henrissat, 1995). However, a massive gene loss has been observed in bacteria with a parasitic or symbiotic life style (Davies and Henrissat, 1995). For example, M. pneumoniae has only 0.4% predicted GTs, or three genes encoded in its genome, instead of the expected 10. However, these GTs are all essential indicating their importance for supporting life, probably being the minimal number of GTs bacteria need.

The energy stored in glycogen and starch is one of the most important metabolic energy sources. Sugar structures are not only used as energy storage, but also e.g. building blocks. The production of lipopolysaccharide (LPS) is of major importance for the survival of free living bacteria, protecting the cell from osmotic lysis and making a first barrier for aggressive macromolecules and the host defence. Glycolipids, discussed above, are other important products made by GTs and are essential to photosynthesis and for membrane stability in many bacteria. The galactoceramide molecule is another example, being an essential component of nerve cells, isolating the cell from its surrounding, making it possible to rapidly transfer electric impulses along the axon (Ruvolo, 2003). However, only the function of glycosyltransferases, transferring an activated sugar to an acceptor will be discussed here.

Glycosyltransferase classification

Glycosyltransferases have been classified by Henrissat and colleagues in a system called CAZy (Coutinho and Henrissat, 1999). The list of GT families is growing monthly and the data base consists of over 20 000 amino acid sequences, with an established or predicted function. The GTs are divided into 90 families according to function, with 21 of these having at least one GT with an established structure (CAZy GT family 1, 2, 5–9, 13, 15, 20, 27, 28, 35, 42–44, 63, 64, 72, 78, and 80) (Martinez-Fleites et al., 2006). The classification is amino acid sequence-based, a GT is grouped to a specific

15

CAZy family if it has a BLAST probability value lower than 10-3 over a stretch of 100 amino acids and a hydrophobic cluster pattern that is similar to other members in that group (Campbell et al., 1997). The origin of the sequences varies greatly, with some families containing enzymes from all kingdoms, e.g. CAZy GT 2 and GT 4 whereas others contain only one taxonomic group. These two families, CAZy GT 2 being inverting and CAZY GT 4 retaining are probably the two ancestor families to all others (Martinez-Fleites et al., 2006). The fact that they also contain all kingdoms and that Archea GTs mainly are classified into these two families make this a good assumption (Martinez-Fleites et al., 2006).

Glycosyltransferases use an activated sugar donor, where an α

configuration between the sugar and the nucleoside diphosphate, nucleoside monophosphate or lipid phosphosugar is the most common (Tarbouriech et al., 2001) phosphate, UDP and TDP, is believed to be used by around 60% of all predicted GTs (Hu and Walker, 2002). The linkage of the sugar nucleotide is either retained or inverted between the sugar and the acceptor, hence giving the naming retaining or inverting GTs (Fig. 5).

Figure 5. The stereochemical outcome. An α-D-linked activated sugar donor can give rise to a product with the anomeric configuration either retained or inverted.

The sequence similarity within each CAZy family can be very low, and where not all members share sequence identity or similarity, eg CAZy GT 2 and GT 4. This is due to the large variation of the origin of the GTs. The fold is however believed to be conserved within each family, and there is mostly no or very low sequence similarity between different families (Campbell et al., 1997; Davies and Henrissat, 1995; Henrissat and Davies, 1997). However, only one GT in family 2 (Charnock and Davies, 1999) and three in CAZy GT 4 (Guerin et al., 2005; Martinez-Fleites et al., 2006), have been crystallised, making this generalisation uncertain for these two large families. The substrate diversity within some large families varies greatly, for example in the large families 1, 2, 4, although they still share substantial sequence similarities (Henrissat and Davies, 2000). Other families consist of GTs utilising the same substrate, e.g. CAZy GT 3. There are also examples of almost identical enzymes using different substrates. The blood group A

16

and B transferases from CAZy GT 6 differ in only four amino acids, but still use different sugar donors (Yamamoto et al., 1990). Only two of these four amino acids are found in the active site in the crystal structure, involved in the binding of the different sugars, UDP-Galactose and UDP-Acetylgalactoseamine (Patenaude et al., 2002). The opposite has also been recorded, enzymes having very low or no sequence similarity can utilise the same substrates (Breton et al., 2001), indicating convergent evolution. The CAZy GT families are believed to be classified according to both structure and mechanism, however this may not be completely correct for all families. Distant similarities have been found between families with different mechanism, retaining and inverting, and some families have been spliced into new families, when the mechanism has been characterised for some family members (Breton et al., 2006). The prediction of substrate specificity and mechanism of a newly discovered GT are therefore almost impossible.

Glycosyltransferase fold types

Within the 21 families with an establish fold, only two fold types have been characterised until today, named GT-A and GT-B (Ang et al., 2000; Bourne and Henrissat, 2001; Coutinho et al., 2003; Davies et al., 2005) (Fig. 4). Both fold types are globular proteins and with partial Rossman folds (Bourne and Henrissat, 2001), with GT-A having anti parallel beta sheets and GT-B parallel. The Rossman fold was first identified in proteins binding diphosphate containing co-factors such as NADH (Buehner et al., 1973). Even though GT-B consists of two Rossmann fold domains and GT-A of one, no sequence similarities have been found between the two different fold families, indicating that they have evolved separately (Liu and Mushegian, 2003). The idea of a gene duplication of a GT-A fold enzyme is otherwise easy to assume. The topology with parallel beta-sheets found in the GT-B family is quite common in the SCOP database. However, when analysing the GT-A family, it is evident that this family have anti-parallel beta-sheets that seem to be unique in the fold database (Liu and Mushegian, 2003). The fold is otherwise very similar between the two families.

Most glycosyltransferases use activated sugar donors where UDP and other sugar nucleotides are the most common. A third glycosyltransferase group (GT-C), (see Fig. 4) has been discovered by iterative BLAST searches followed by structural comparisons (Liu and Mushegian, 2003). However, the latter families are only predicted and no crystal structures have yet been solved. Proteins with this fold are all integral membrane proteins with the active site in the loops, and with the transmembrane helix number varying between 8 and 13. This family can also be found with Hidden Markov

17

method searches within the GT families (Liu and Mushegian, 2003). This method have also identified a fourth family, unique to eukaryotes, named GT-D, consisting of only 3 CAZy GT families (Kikuchi et al., 2003). There are however, no fold models for this enzyme group. The structural diversity is believed to be very small for GTs, however the opposite is seen for the glycosylhydrolases, breaking sugar linkages, where the structural diversity is large (Davies et al., 2005).

Figure 4. The different fold families discovered until today within the GT enzyme groups (Charnock and Davies, 1999; Ha et al., 2000).

GT-A The GT-A fold family is the largest with 25 CAZy families classified into

this group, compared to 12 for the GT-B fold family. The fold consists of a Rossman fold, with two dissimilar domains, with the nucleotide sugar binding domain N-terminally and the acceptor binding domain in the C-terminal. Almost all enzymes with a GT-A fold type have a DxD motif in the N-domain. This motif is found to interact with the phosphate groups of the nucleotide donor through a divalent cation, most commonly Mn2+. There are differences in the function of the DxD motif when comparing inverting and retaining GTs of the GT-A fold. The two aspartate residues, interacting with the ion in the retaining GTs, but only the last aspartate interacting with the ion in inverting GTs (Breton et al., 2006). The variable amino acid, most often a polar or small and non cyclic is usually involved in binding the ribose. For example, in the inverting enzyme SpsA (CAZy GT 2) the motif is xDD where the first aspartate is known to interact with the ribose group of the nucleotide and the second aspartate interacts with the ion (Tarbouriech et al., 2001). The DxD motif is not fully conserved in the GT-A family with for example SpsA having, xDD and DxH has been found in the retaining family 27 (Fritz et al., 2004). The inverting C2GnT-L (CAZy GT 14) completely

18

lacks this motif. The function of the metal ion is replaced by the two basic amino acids arginine and lysine (Pak et al., 2006). This active site arrangement has also been seen in the GT-B fold family discussed below, indicating that the reaction mechanism is a junction between the two fold families GT-A and GT-B.

Cst II from Campylobacter jejeni believed by some to have a GT-A like

fold, also lacks the DxD motif. It is known to be further activated by Mg2+ and Mn2+, even though the ions are not known to be essential (Chiu et al., 2004). Due to the absence of the DxD motif, an ion in the active site, as well as a catalytic base, the reaction mechanism is believed to be similar to the retaining GTs discussed below, the product is although inverted (Chiu et al., 2004). This structure is however not a clear GT-A fold. The flexible loops found in most GT-A structures are here containing short helixes that will fold over the active site (Chiu et al., 2004). The secondary structure arrangement in the cst II enzyme is also different from the GT-A family and has been designate to a fold family of its own by SCOP (Murzin et al., 1995; Pak et al., 2006).

Inverting reaction mechanism The inverting mechanism is the mostly widely studied, and it has been

easier to “capture” intermediates during structure analysis. The mechanism is believed to be a SN2 type, with the bound divalent cation being the acid catalyst and involved in the coordination of the substrate and the donor (Breton et al., 2006). The inverting mechanism uses a base to deprotonate the reactive hydroxyl of the sugar acceptor to further activate it (Charnock et al., 2001). This base is aspartate-191 in SpsA, (CAZy GT 2) (Charnock and Davies, 1999), conserved in CAZy GT families 2, 7 (Gastinel et al., 1999) and 13 (Gordon et al., 2006; Unligil et al., 2000) and replaced by a glutamate in CAZy GT families 14 (Pak et al., 2006), and 43 (Pedersen et al., 2000). In the later families the sugar acceptor is known to be hydrogen bonded to the glutamate in the position for a nucleophilic attack on the UDP-sugar donor (Pak et al., 2006; Pedersen et al., 2000). The function of the conserved aspartate has also been verified by mutational studies (Garinot-Schneider et al., 2000; Keenleyside et al., 2001). These enzymes use three aspartates, (Asp 39, 98, 99 in SpsA) to coordinate the nucleotide in the donor sugar and a fourth aspartate, Asp 191 or glutamate acts as the catalytic base (Charnock et al., 2001). Aspartate 99 is also known to coordinate the divalent cation (Tarbouriech et al., 2001).

19

Figure 6. Coordination of the UDP-sugar in the Gal-T1 in CAZy GT 7 and the conserved amino acids in other GT-A enzymes with an established fold (Tarbouriech et al., 2001).

All GT-A structures solved have a flexible loop that is believed to be

involved in the active site and the binding of the substrate (Boix et al., 2001; Ramakrishnan et al., 2002b; Ramakrishnan et al., 2006), but this has only been functionally characterised in the inverting Gal-T1 (CAZy GT 7) enzyme (Ramakrishnan et al., 2006) and partially in LgtC (CAZy GT 8) (Persson et al., 2001). The Gal-T1 enzyme has been very extensively studied and was first crystallised in 1999, and has since then been crystallised with various substrates and at different resolutions (Persson et al., 2001; Ramakrishnan and Qasba, 2001; Ramakrishnan et al., 2002a; Ramakrishnan et al., 2002b; Ramakrishnan et al., 2006). The mechanism established until today involves first the binding of Mn2+ followed by UDP-Gal and the acceptor substrate. The binding of the ion has been found to be essential for the binding of the donor (Ramakrishnan et al., 2006). This is probably true for most inverting GT-A enzymes. The formation of the acceptor binding site has also been seen in the GnT-I enzyme (CAZy GT 13) (Gordon et al., 2006). The metal ion is bound in an octahedral coordination, where it interacts with one or two residues in the DxD motif, two oxygen atoms in the UDP molecule and finally with three water molecules (Qasba et al., 2005). The binding of the ion and the donor induces a conformational change in the flexible loops described above, leading to a change in the coordination of the water molecules around the ion. This change creates an active ground state in the enzyme that can assist catalysis and makes the acceptor binding site accessible to the solvent (Fig. 7) (Qasba et al., 2005). This conformational change and the change in the coordination of different water molecules also assist the rotation of the phosphate group in the donor sugar nucleotide, preventing the reaction to revert (Ramakrishnan et al., 2006). The water molecules found to be coordinated with the ion have been found to be conserved in the structures solved where the reaction is dependent on a

20

divalent ion (Ramakrishnan et al., 2006). The ion is also believed to neutralise the nucleotide once the sugar has been cleaved off, helping it to leave the enzyme (Qasba et al., 2005). This function has been replaced by two basic residues in the CAZy GT 14 family and in the CstII enzyme that has a novel fold. These enzymes are also believed to be able to bind the acceptor in the absence of the donor (Pak et al., 2006).

Figure 7. Schematic drawing of the conformational changes during catalysis observed in GNT1. i) The binding of the ion makes it possible for the ii) sugar donor to bind. iii) The loop folds in to the structure, iv) this conformational change creates a binding site for the acceptor and the reaction can take place. The donor and the product leaves the active site and the cycle start once again (Qasba et al., 2005).

Retaining reaction mechanism

The retaining reaction mechanism is not as well understood, but is believed by many to involve a double displacement reaction where the donor sugar is first transferred to an amino acid in the protein and then transferred to the acceptor, as seen in glycosylhydrolases (Qasba et al., 2005). To support the double displacement mechanism, a base is needed in the active site (Tvaroska, 2004). However, this has not been identified in the only retaining enzyme crystallised in the presence of donor or substrate, the LgtC enzyme (CAZy GT 8) (Persson et al., 2001). Molecular modelling has shown that the most likely reaction mechanism occurs in one step, where an oxygen in the acceptor attacks a carbon in the donor, breaking the bond between the sugar and the nucleotide (Tvaroska, 2004). This is in agreement with the proposed mechanism from crystal structure data from LgtC (Persson et al., 2001). In this suggested mechanism, there is no need for a catalytic acid or base in the protein. No intermediates, expected to be present in a double displacement mechanism, have been captured in the structures

21

solved until today, further indicating the mechanism described above. The active site is only providing the right environment for the reaction to take place and is not actively taking part in the reaction. This would also explain why there are no conserved conformational signatures in this enzyme family, except for the binding of the sugar donor and the acceptor (Flint et al., 2005).

In the structure of LgtC, the metal ion is coordinated with the nucleotide

and three amino acids in the protein, histidine 244, aspartate 103 and aspartate 105, the DxD motif as in the inverting Gal-T1. The three amino acids are conserved within the CAZy GT families (Persson et al., 2001). As in the inverting mechanism the sugar donor binds first to the enzyme, followed by movement of the flexible loops, forming the active site, making it possible for the acceptor to bind (Persson et al., 2001). The metal ion is believed to bind early in the reaction, and as in the inverting mechanism acting as an acid catalyst (Persson et al., 2001). In the structures of the inverting Gal-T1 enzyme, water has been found to be important in the reaction mechanism, but is excluded once the acceptor binds (Ramakrishnan et al., 2006; Snajdrova et al., 2004). The retaining enzyme group seems more diverse than the inverting. In the latter, flexible loops are of importance in the donor and acceptor binding, however in the retaining Kre2p/Mnt1p enzyme (CAZy GT 15), this has not been observed. The movement of the protein domains or loops during binding of donor and acceptor is minor in this protein. On the other hand, a few side chains rotate upon binding of the sugar donor. This small movement makes it possible once again for the acceptor to bind (Lobsanov et al., 2004), but no large conformational changes have been seen as for inverting GT-A.

GT-B

The GT-B fold family group of enzymes are the most diverse in sugar acceptor usage when compared to GT-A. Vancomycin, LPS, glycolipids, and proteins, and DNA are some of the products that this group of enzymes is involved in producing (Hu and Walker, 2002). The GT-B structure is better suited for large acceptor molecules compared to GT-A, due to the large cleft in between the two domains (Fig. 4) (Breton et al., 2006). The GT-A fold with its compact structure would have difficulties in binding large acceptor molecules. The GT-B fold consists of two separate Rossmann fold domains with a linker region and the catalytic site in between the two domains. Most GT-B structures have a kink in the last C- terminal helix making it cross over into the N-domain (Gibson et al., 2004; Hu and Walker, 2002). The C-terminal of proteins with this structure is well conserved and the N-terminal

22

domain is diverse, reflecting the diversity in used sugar acceptors used and the conservation of used sugar nucleotides (Breton et al., 2006).

A sequence motif encoding a α/β/α subdomain has been identified in the

GT-B family (Hu and Walker, 2002), (Fig. 8) that consists of Cα4, Cβ5, and Cα5 but with few other invariant residues. It has been found to be involved in binding of the sugar donor (Hu and Walker, 2002), with the first helix being involved in the sugar binding and the second in binding the pyranose. The first helix that is enriched in positively charged amino acids is believed to stabilise the binding of the negatively charged oxygen in the nucleotide (Hu and Walker, 2002). The GT-B family lacks the otherwise common DxD motif, but it is known that most GT-B enzymes are further activated by divalent cations. No ions have however been found in the active site in the crystal structures solved (Breton et al., 2006; Gibson et al., 2004; Hu et al., 2003; Mulichak et al., 2001). The α/β/α subdomain described above, is believed to have the same function as the cation in GT-A (Hu and Walker, 2002). In for exampel OtsA (CAZy GT 20), the phosphate groups of UDP are in direct contact with arginine 262 and leucin 267 in the C-domain and a conserved Gly,Gly,Leu motif in the N-domain instead of through an ion (Gibson et al., 2004).

Inverting mechanism The inverting mechanism is the most studied in the GT-B fold family as

in GT-A, but the exact reaction mechanism and conformational changes are not as well established as for the GT-A group. The inverting reaction mechanism is believed to go through an SN2 mechanism (Breton et al., 2006). As in the GT-A fold family the enzymes undergo conformational changes after the sugar donor has bound, however the GT-B fold type has shorter loops involved in the donor and acceptor binding than the GT-A. When the sugar donor binds to the N-domain in MurG, (CAZy GT 28) (Hu et al., 2003), and GtfA (Mulichak et al., 2003) and GtfD (CAZy 1) (Mulichak et al., 2004) a loop in this domain, the GGS loop undergoes notable changes. It moves towards the α/β/α subdomain located in C-domain of the protein and mediates contact between residues in both the N and the C-domain. This conformational change has been found to be essential for activity (Hu et al., 2003; Mulichak et al., 2003). The movement of the N- and C-domain has also been observed in WaaG (CAZy GT 4) (Martinez-Fleites et al., 2006). The movement of the GGS loop is believed to be important for substrate binding. The conformational change moves an aromatic amino acid, conserved in many GT-B proteins and works as a lid in the binding pocket by stacking over the uracil ring of the donor sugar (Hu et

23

al., 2003; Mulichak et al., 2003; Mulichak et al., 2004). Many GT-B enzymes have a conserved arginine or histidine that is part of the R/H/Ex7E motif found in the GT-B family. These residues are located on the first helix of the α/β/α described above and are believed to stabilises the leaving group through electrostatic interactions or protonation (Ha et al., 2001; Mulichak et al., 2003). The conserved glutamate in the R/H/EX7E motif is involved in donor binding and not catalysis like the first residues. Mutational studies in AceA (CAZy GT 4), a protein not yet crystallised have shown that the glutamate residue in the EX7E motif is essential for activity (Fig. 8) (Abdian et al., 2000).

A B

C NmMurG --CVEFITDMVSAYRDADLVICRAGALTIAELTAAGLGALLVPYPHAVDDHQTAN- EcMurG HKVTEFIDDMAAAYAWADVVVCRSGALTVSEIAAAGLPALFVPFQHK-DRQQYWNA AoGtfD -FAIDEVN-FQALFRRVAAVIHHGSAGTEHVATRAGVPQLVIPRNTD---QPYFA- AoGtfB -FAIGEVN-HQVLFGRVAAVIHHGGAGTTHVAARAGAPQILLPQMAD---QPYYA- EcWaaG RNDVSELMAAADLLLHPAYQEAAGIVLLEAITAGLPVLTTAVCGYAHYIADANCGTVI SpMGS IAPSETALYYKAADFFISASTSETQGLTYLESLASGTPVIAHGNPYLNNLISDKMFG--- BbMGS IPWEEIYYYYKISDIFASLSKSEVYPMTVIEALTAGIPAILINDYIYKDVIKEGIN----AlDGS VDGAVIKGAFSGADCVFFPSYEETEGIVVLEGLASKTPVVLRDIPVYYDWLFHK------ Figure 8. A; The structure of MurG, with the α/β/α subdomain in blue. B; The α/β/α subdomain, with the conserved H/R/EX7E/V motif in green and the donor substrate in green. C. An alignment of the α/β/α sub-domain. The protein names in italic are retaining enzymes and the other four are inverting. The conserved R/H/E residue, marked in red, is believed to stabilise the positive charge in the transition state and the E/V residue, marked in red, involved in binding the hydroxyls on the ribose sugar of the UDP group. (The figure A and B were made in PDB viewer and PovRay)

The exact catalytic residues have not definitely been established by

capturing reaction intermediates for inverting GT-B enzymes. The only residues at the right position from the donor sugar and the acceptor are serine 10 and aspartate 13 in GtfA that can form hydrogen bonds to the reactive

24

hydroxyl groups of the sugar donor (Mulichak et al., 2003). However, mutation of the serine and aspartate in GtfA did not abolish activity (Mulichak et al., 2001). The corresponding residues in WaaC (CAZy GT 9) are aspartate 13 and 261, histidine 22 and aspartate 121 in UGT71G1 (CAZy GT 1). The importance of these residues was confirmed by mutagenesis, after which the activity of the enzymes was almost abolished (Grizot et al., 2006; Shao et al., 2005). Aspartate 261 (WaaC) is conserved in several GT-B inverting enzymes, but not the histidine required in the proposed mechanism.

Retaining mechanism

The retaining reaction mechanism is believed to be similar to the GT-A retaining group, being Substitution Nucleophilic internal like (SNi). It is however not entirely understood due to the lack of both sugar donor and acceptor in the solved crystal structures. As in the retaining family of GT-A, no conserved amino acids that could act as a nucleophile have been found (Martinez-Fleites et al., 2006). A conformational change occurring when the acceptor and donor are bound has been observed in OtsA. As stated above most interactions between the enzyme and the donor is in the C-domain of the protein, the conformational change occurring in a loop is believed to provide a cross talk between the two domains (Gibson et al., 2004). It is however not clear if the sugar donor needs to bind to enzyme for the acceptor to be able to bind, as in the GT-A fold family (Gibson et al., 2004). In both OtsA and WaaG most interactions between the sugar in the donor and the enzyme are with main chain amide groups and not side chains, except for the caroxylate moiety recognising the sugar that is donated by either a glutamate 281 (WaaG) or an aspartate (OtsA). This carboxylate is found in many retaining GT-B enzymes and is believed to stabilise the positive charge in the transition state (Martinez-Fleites et al., 2006). The extensive use of backbone amides instead of side chains is consistent with the flexibility reported for the donor (Gibson et al., 2002; Gibson et al., 2004). OtsA homologues have been reported to be able to use different sugar nucleotides. The donor sugar binding is expected to vary depending on the specificity of the enzyme and the usage of different nucleotides.

25

Membrane binding

Many glycosyltransferases act on membrane bound substrates and to be able to interact with the substrate, the enzyme needs to be membrane associated. This is solved in mainly three ways; peripheral association with the membrane by an amphipathic helix, that will snorkel into the membrane to hide the hydrophobic amino acids; integral membrane proteins with one or several transmembrane (TM) helices that will anchor the protein to the membrane with globular domains outside the membrane, and finally transmembrane proteins with several TM segments and that have the active site in the loops connecting the TM segments. All three classes are found among glycosyltransferases, with the first being most common in the GT-B family (Leipold et al., 2007; van den Brink-van der Laan et al., 2003), and with a few anchoring TM segments being most common in the GT-A fold family. There are however exceptions with TM segments in the GT-B family and amphipathic helices in the GT-A. The last class of membrane bound enzymes, containing mainly TM segments, has been found only in the third structural GT, family GT-C. However, only the membrane association through amphipathic helices will be discussed here.

The binding of a protein to the membrane is dependent on the nature of

binding peptide segments and the lipid environment. Most studies have been done on peptides, but in nature, the peptide is usually a part of a full length protein. The binding domains of membrane associated proteins are often amphipathic helices that are enriched in cationic amino acids (Ladokhin and White, 2001). This interaction yields more or less strong associations with the membrane, and the helix is often infiltrated into the membrane for a more stable interaction (Seelig, 2004). The binding of the protein to the membrane is dependent on electrostatic forces, van der Waals interactions, and hydrogen bonds. The depth to which the helix penetrates into the membrane varies between different proteins, and depends on hydrophobic/hydrophilic equilibrium between the molecular groups, and forces involved (Seelig, 2004). However, most amphipathic helices are inserted to the glycerol backbone level of the membrane lipids. Not all membrane proteins interact with the membrane in this way. There are also proteins with specific lipid binding domains usually found in loops, binding to specific single lipids (Cornell and Taneva, 2006), however the nature of these will not be discussed here.

26

Enzymes acting on membrane bound substrates often use amphitropism,

to regulate the activity by changing the binding strength to the membrane. Most amphitropic enzymes are active when membrane bound and inactive when soluble. The activity is altered by binding soluble ligands or by an alteration in the membrane lipid composition that will change the membrane affinity (Cornell and Northwood, 2000). For example CTP:phosphocholine cytidylyltransferase that is involved in the synthesis of PC is regulated by the lipid environment and by phosphorylation of the protein (Cornell and Northwood, 2000). The amphipathic helix that is involved in the regulation has mostly acidic residues at the polar face, the interfacial zone consists of mostly basic residues, and the non-polar face mostly of hydrophobic residues that drive the membrane binding. The hydrophobic residues are believed to be buried in the structure when the protein is soluble. The many basic residues, in the interface region of the helix are believed to mediate the binding to a negatively charged membrane, followed by the hydrophobic interactions between the amino acid and the lipid (Cornell and Northwood, 2000). Hence, changes in the lipid content of the membrane will change the strength of the binding. There are also membrane proteins that sense the curvature of the membrane. These proteins have amphipathic helices that are enriched in serine and threonine residues on the polar face of the helix (Drin et al., 2007). These types of regulation are also believed to be used by GTs, especially by GT enzymes using lipids as substrates. A. laidlawii MGS, synthesising GlcDAG is activated by negatively charged lipids such as PG (Karlsson et al., 1994), however the enzyme activity is not affected by the curvature (Karlsson et al., 1997). The DGS enzyme from the same species, synthesising GlcGlcDAG is regulated by both PG and sensing the curvature of the membrane (Vikström et al., 1999). These two enzymes will work together to keep the membrane properties constant, MGS is only active at high PG levels in the membrane, making it less charged, and DGS will regulate the production of nonbilayer-prone or bilayer-prone lipids, hence GlcDAG or GlcGlcDAG. MGS is known to bind to the membrane through an amphipathic helix (Lind et al., 2007).

27

Sequence analysis and protein classification

Traditionally the methods that have been used to compare and classify proteins are mainly based on sequence alignments, where only amino acid identity or similarity in the same position is accounted for (e.g. BLAST). However, a small number of gaps are allowed. To classify and predict the function and fold of new proteins with low sequence similarity to proteins of known function other methods are needed. It is believed that protein sequences behave in manners far from random, and the amino acid sequence should therefor be organised to reflect the corresponding function and structure (Rackovsky, 1998). Hence, the property patterns are believed to be the same in proteins with the same (or similar) structure or function and the property patterns are thought to be conserved even at low sequence similarities. These thoughts are especially clear for proteins with repeated motifs, e.g. TIM barrels (Rackovsky, 1998; Wold and Sjöström, 1998), but should be valid for other protein families as well. Proteins with the same function, but with different folds are still believed to have certain local property patterns in the sequence that determine the function. With a multivariate sequence analysis method, based on amino acid properties, proteins with no sequence similarity can be grouped together and conserved property patterns can be visualised within the proteins.

Multivariate data analysis To be able to use a multivariate approach, the protein sequence needs to

be translated into numbers, variables. This is done by translating the amino acid sequence into Z-scores describing different amino acid properties. The scores used in this study are a combination of 29 different physico-chemical variables calculated by principal component analysis, PCA (Hellberg et al., 1987). The scores are experimentally determined e.g. NMR, HPLC-retention, pKa, and logP and theoretical values such as Mw and van der Waals volume. The three Z-scores used, named z1 to z3 describes hydrophilicity, side chain bulk volume and polarisability/charge respectively.

28

The periodicity of the amino acid properties in a protein is described by calculating the auto-covariance and cross-covariance (ACC). The covariance is analysed with different “window” sizes, i.e. lags where the amino acid properties at different positions from each other, up to the decided lag/window are calculated and an average of all positions along the sequence is obtained. The difference between the two methods is that auto-covariance is using the same property along the sequence and cross-covariance a combination of different properties (Fig. 9). This method provides an analyses of the relationship between close and more distant neighbouring amino acids and not only amino acid identity and similarity. The obtained data matrix is then analysed by a multivariate method described below.

Figure 9. The amino acid sequence is translated into Z-values that are used to calculate the auto covariance and the cross covariance variables up to the decided lag.

Previous studies

The above method has been used to successfully classify ABC transporters in M. pneumoniae and other genomes, and signal peptides in E. coli, Mycoplasma and other Gram-positive bacteria (Edman et al., 1999; Edman, 2001), predict the localisation of proteins in Synechocystis (Rajalahti et al., 2007), and to classify and identify sequence difference between GPCRs (Lapinsh et al., 2002).

29

Multivariate projection methods

Principal component analysis

Multivariate data analysis can be used to analyse the ACC data. The method has been developed to analyse the numerous amounts of research data, obtained from many different techniques used today. The multivariate data analysis programs is based on a so called projection method where a swarm of points in a K-dimensional space is used, where K can be any number greater the two. One dimension for each variable and one point for each observation is projected down in one plane, a two dimensional space. The coordinates of the points in the new hyperplane contain a compressed representation of the observation in the original data table. The resulting vectors, when compressing the original matrix, are scores (t-vectors) that describe the relationship between the objects, and loadings that describe the contribution of the variables to the score-vectors. The compression is done to find underlying structures in the data set using a set of new latent variables, principal components. The first principal component t1,p1 describes most of the variation in the data set and the second (t2,p2) the second most and so on. The two first principal components are orthologous to each other, meaning that they are independent of each other. When reducing the data set, it becomes easier to obtain an overview of the variation, and outliers in the group can be identified. This compression is possible when the variables are correlated to each other. When using this technique to separate and classify protein groups it is believed that proteins with similar function or structure have similar sequence periodicity as described above (Wold and Sjöström, 1998). To analyse the results, the t and p scores are plotted, where the t scores are plotted to visualise objects with similar properties and the loadings to search for the important variables. The loadings are later used to find the relationship between different objects. However, it should be noted that the new principal components are a combinations of all variables and can usually not be assigned to a specific one.

To simplify the projection method, one can think about it as a looking at

a shadow of a 3D object. When holding a pair of scissors in front of a light,

30

holding it like in Fig. 10, A the shadow will describe the object well, however a 45º rotation will make it harder and turning it 45º further will make it impossible to identify the object. Hence, the direction in space that is chosen when projecting the 3D object into a 2D plane is of importance. The direction that has the largest variation should be chosen. In this example, 3 variables (3-dimensions) are projected down to new principal components. The first two dimensions/variables, describe the object best and the width is not needed.

Figure 10. Projection of a K dimensional object onto a 2D plane, here a pair of scissors (a 3D object) is projected on to a sheet of paper by the help of a light source. The image, the shadow, represents two dimensions of the 3D object. The object is rotated 45° in two steps to generate different shadows of the same object. The same method is used with multivariate analysis programs. A projection of a 3 dimensional problem that is projected down into two dimensions and the resulting t vectors are plotted as a score plot.

Partial least squares projection to latent structures

Partial least squares projection to latent structures (PLS), is used to find relationships between X-matrix observations and a response matrix Y. The Y matrix contains additional descriptions of the samples in X. PLS-discriminator analysis, PLS-DA is a special case of PLS, used to find class characteristics within objects in a X matrix. In this special case, the Y-matrix is composed of dummy variables instead of e.g. biological activity. Each member of a group has the value 1 and non-members 0. PLS-DA uses the fact that objects belonging to the same class are expected to have common features and should behave similarly in the score plot. The method also gives the opportunity to predict new unclassified sequences. To evaluate the complexity of a PLS model, cross validation is preformed, objects are randomly taken out one by one during the calculation, and are predicted into

31

the model and the Y-values calculated (Wold 1978). This is an internal validation method, where objects that create the model are used to validate it. The validation is repeated until all objects have been left out once. The cross validation is done to calculate the Q2 value. Hence, the ability of the model to predict the object back to the same place as it had in the plot before it was taken out, is calculated. The Q2 value describes how much of the Y matrix can be predicted by the model, hence if all objects that are taken out and predicted back again are predicted to the same position a Q2 score of 1 is obtained. A Q2 larger than 0.1 corresponds to 95% significance. As in PCA a hyper plane is calculated from the original X matrix that gives t-scores, and the Y variables u-scores. These scores can be plotted to analyse the data and w*c plots are made to evaluate the correlation between the X and the Y variables. Hence, these plots are used to identify variables that are important for the separation of the groups. With PLS-DA it is possible to predict whether a object belongs to a class or not, by looking at the predicted class value. A predicted value greater then 1 suggests that the specific object belongs to the class and a value closer to 1 that it doesn’t. By using PLS-DA for protein family classification, a protein with an unknown function can be predicted by using the described method to investigate if it has the same property patterns as other members of the group.

32

Summary of papers

Paper I The aim of this paper was to use a new approach based on multivariate

data analyses and ACC to separate and classify GTs. With this method, not only similarity in specific positions is accounted for, but also property patterns along the sequence. This method would in theory find relationship between sequences that share no similarity according to BLAST, but that share the same fold or function. This would be possible since it is believed that sequences behave far from random and that the protein sequence reflects both the structure and the function. A data set consisting of GTs classified by CAZy, with an established function and from families with an established fold was constructed. The data set was used to divide fold families, and reaction mechanism, and to predict the function of proteins with unknown function.

We found that it was possible to separate the three different fold groups

from each other with this method with high scores, a Q(cum)2 value of 0.730 was obtained. It was also possible to separate the inverting from retaining reaction mechanism within the GT-A and GT-B fold, with a score of Q(cum)2 values of 0.702 and 0.562 respectively. The GT-C family was not included in these experiments due the occurrence of only one reaction mechanism within the family. The fold classification showed that the occurrence of TM segments within the GTs was easily recognised. GT-A and GT-B proteins with TMs moved closer to the GT-C family, consisting of several TM segments and small cytoplasmic domains. Divisions with only two families were also done, yielding high scores.

The MVD method also gives information about the separating

parameters in the data set. The parameters are visualised on a w*c plots. Only the separating parameters for the reaction mechanism division within each fold family were plotted. The important variables for the GT-A family were (in rank) z(2)1z(3)13 (e.g. size in the first position and polarisability 13 amino acids downstream), z(2)1z(2)5, z(1)1z(1)18, z(2)1z(2)10, z(2)1z(1)4, and z(1)1z(1)14. The shorter distances variables were size (z(2)) and hydrophilicity/hydrophobicity (z(1)) patterns, e.g. positions 1 to 4 and 1 to 5.

33

This most likely corresponds to amino acids on the same side of a helix. Alignments established by pair wise aligning sequences with an established structure in the following CAZy families, for the retaining enzymes GT6, GT 8, and for the inverting group GT7, GT43, GT2, and GT13 were achieved. A stretch of 84 amino acids in the retaining group could successfully be aligned. The top variables were searched for in the sequences using the alignment, only the z(1)1z(1)18 variable could easily be visualised. However, a comparison between the two reaction mechanisms could not be done since no corresponding alignment for the inverting group could be established.

When analysing the GT-B fold group the most important variables were

z(2)1z(2)5, z(1)1z(1)16, z(2)1z(2)8, z(1)1z(3)14, z(3)1z(1)11, and z(2)1z(2)2 for the separation of the retaining and inverting sequences. The GT-B fold group was also analysed by the same method as described above. Here, sequences from the GT1 and GT28 families were aligned. The variable z(1)1z(3)16 was marked out at the sequence level by its z products and should be positively correlated to the inverting enzyme group. In the alignment 18 positive and 7 negative pairs were found within the inverting group, and 3 positive and 7 negative pairs were found for the retaining families. This confirms the importance of the z(1)1z(3)16 variable. The alignment was found to coincide with the UDP-binding site (se fig 11), and also a structural correlation could be made.

Figure 11 Alignment (ClusalW) of selected GTs with the GT-B fold. The z(1)1z(3)16 variable pairs indicated in the alignment. p, first position in positive variable pairs, (+) the second position; n, first position in a negative variable pair; (-) the second position. Conserved amino acids are marked with * and amino acids with the same properties, (:) and similar properties (.) as marked in Clustalw.

The method could also be used to predict the fold and reaction

mechanism of GTs from E. coli and Synechocystis. The proteins used were all taken from the CAZy database, however the fold was only known for a few of the included CAZy families, hence a few proteins. To evaluate the predictions, multiple fold predictions, made by the MetaServer (bioinfo.pl/meta) were used. When predicting the fold, with all three fold

34

groups in the reference set, 82% of the E. coli GTs were predicted to have the right fold, and excluding the GT-C group increased the prediction correctness to 89%. The reaction mechanism predictions also gave very good results, with 100% of the E. coli and 86% of the Synechocystis set within the GT-A fold group, and 60% versus 80% for the GT-B group, being correctly predicted.

In conclusion, the multivariate data analysis approach could successfully

be used to separate different fold groups from each other and reaction mechanism. The separating parameters were potentially found in the protein sequences, however further mutagenic studies need to be made. The method could also be used predict the fold and function of GTs from E. coli and Synechocystis and M. pneumoniae.

Paper II Enzymes acting on membrane bound substrates are known to be

regulated by factors such as lipid charge and the curvature stress. This indicates that the enzymes must be attached to the membrane in order to sense these changes, however the binding is not well characterised among glycosyltransferases of the GT-B fold. When comparing the fold of membrane bound GTs from the GT-B family with soluble enzymes, structural differences are difficult to detect. The aim of this paper was to identify the binding region of these GTs and characterise it.

We used as in the previous paper a multivariate data analysis approach

to detect differences between soluble and membrane bound GTs with the GT-B fold. The score plot, from the separation of the two different groups reveales two well separated clusters, and the model has 3 significant PLS components and a Q(cum)2 value for the model of 0.52. The variables were in order of importance, pI of N-domain, pI of full enzyme, z21*z39, z31*z39, z31*z13 and z31*z33. The two first variables were substantially stronger than all the following ones. To summarise these variables it can be said that the differences between the two groups are seen in the charge and amphipathic patterns. Membrane bound GTs have higher pI in the N-terminal domain and amphipathic patterns that we believe are localised in helices on the surface. These characteristics are believed to mediate the binding of the protein to the membrane surface, with the positively charged surface interacting with the negatively charged membrane lipids. The hydrophobic amino acids in the amphipathic helix will be buried into the lipid bilayer. Using in silico mutated GTs, where whole domains or just

35

helices with amphipathic characteristics were swapped, showed that the N-domains had a large impact on the localisation of the protein in its group, but also the helix. This indicates that the pI of the domain and the amphipathic helix only about 20 aa long, absent in soluble GTs, are important for the membrane binding (se fig 12 for localisation within the proteins).

A B

Figure 12. A; MurG showing the amphipathic helix, involved in membrane binding in green, B; Fold model of AlMGS with the peptide in green. The figure was made in PDB wiever and PovRay.

The features of the amphipathic helix found in alMGS, a well studied

GT from A. laidlawii, and that we believe mediates the binding of the proteins to the bilayer were further investigated by CD, Biacore, and NMR studies. The experimental study showed that the helix, S65-L87 adopted a α-helix conformation in several membrane mimicking environments, however not in water. The ability to bind to a lipid surface, investigated by Biacore studies revealed that the peptide had the same binding affinities as the full size protein. The usage of spin-labelled phospholipids showed that the N-terminal part of the peptide is interacting with the lipids in the bilayer, with zwitterionic bicells influencing the binding more than partly negatively charged bicells. The C-terminus of the peptide was not affected. This indicates that the charged C-terminus is situated outside the membrane, while the N-terminus is penetrating into the bilayer. These studies show that the peptide probably undergoes conformational changes when the full size protein reaches the membrane, due to the fact that lipids induce the helical fold. The N-terminal part of the helix will when reaching the membrane fold and go in into the lipid bilayer and anchoring the protein to the surface. The cluster of charges found in the C-terminal will further stabilise the binding by interacting with the charged lipids.

36

Paper III M. pneumoniae is one of the smallest organisms capable of autonomous

growth. It has only 688 ORF with 25% of the ORFs having an unknown function. Its ability to synthesise lipids is surprisingly large considering the down sized-genome. In this paper, we analysed the ability to make glycolipids and cloned three potential GTs.

We found that the ability to synthesise glycolipids is greater than previously known. In vivo labelling studies showed that this bacterium can make at least three glycolipids, five phosphoglycolipids, and six phospholipids, using glucose and phosphate as the labelled source. In vitro studies using solubilsed M. pneumoniae cells showed that the bacterium has a large ability to use a variety of different acceptors as well as both UDP-glucose and UDP-galactose as sugar donors, although the latter was preferred. The opposite was seen in vivo, which is due to the epimerase that epimerise UDP-glucose to UDP-galactose, which is preferred by the GTs. The bacterium has the ability to use diacylglycerol, αGalDAG, and βGalDAG as substrates with UDP-glucose as the donor and diacylglycerol, αGalDAG, and βGalDAG as well as αGlcDAG, αGlcαGlcDAG, βGlcDAG with UDP-galactose as donor. Up to three galactose residues could be attached to a diacylglycerol, and one to two residues to the glycolipids. Most reactions were stimulated by PG, however one sugar could be added to a diacylglycerol, but the elongation stopped in the absence of PG. Ceramide, was used in the assays to investigate the ability to use eukaryotic lipids as substrates. A similar pattern as for diacylglycerol-based lipids was seen, and once again, PG stimulated the reaction. When combining these results, we suspected that all these glycolipids could not be produced by three GTs. To investigate if the bacteria had more GTs than the three proposed GTs, GTs with an established function from Gram positive bacteria were used as search probes against the M.pnumoniae genome. However, no further GTs genes where found.

The three proposed GTs were cloned and solubilised E. coli expressing

the M. pneumoniae GTs were used in assays using the same substrates as with solubilsed M. pneumoniae cells. Activity were only seen with MPN483 expressing cells. The cloned MPN483 enzyme showed the largest activity with β-GalDAG as the acceptor and UDP-galactose as the donor in vitro. This lipid was used as a reference to compare the activity with other substrates. Substituting UDP-galactose for UDP-glucose gave only 30% of the activity. Other substrates yielding a high activity were DAG, βGalDAG, αGalβGalDAG, βGalCer with UDP-galactose, and DAG, and βGalDAG with UDP-glucose, respectively. These results indicate that the enzyme

37

prefers a terminal galactose as the acceptor sugar to be able to extend the sugar chain. PG was once again needed as an activator in all reactions. A lipid product was also purified from E.coli cells expressing MPN483 and structurally determined. The product were found to be a βGlcβGalDAG. The phospholipid could be replaced by cardiolipin, that also is anionic, but not by the zwitterionic PC or PE. The charge thus seems important. The lipid environment thus has effects on the GT enzyme, with both the head group elongation and the total amount being affected.

Glycolipids are known to be important in the infection of M. pneumoniae in humans and the immune system is known to react to these. We used both purified lipids from M. pneumoniae and pure glycolipids with established structures, to investigate the immunological responses towards glycolipids during a M. pneumoniae infection. ELISA studies of purified M. pneumoniae lipids showed a highly increased response with an IgM fraction from patient sera, compared to the IgM negative control for the three different glycolipids. The di-hexoseDAG variant both purified from M. pneumoniae and E.coli (synthesised by MPN483) gave the highest response. The MPN483 enzyme was also used to synthesise different glycolipids to be used in the ELISA test. These studies showed that the MPN483 produced βGalCer and especially βGalDAG were reactive towards the positive IgM serum. None of the above mentioned lipids gave any negative results. Recently it has been shown that bacterial lipids, such as αGalDAG from B.burgdorferi as well as αGalCer and βGalCer bind to the CD1d receptor, presented on antigen-presenting cells (Kinjo et al., 2006). This binding stimulates iNKT through TCR and further activate factors in the host defence against bacteria lacking lipopolysaccharide (LPS) (Porcelli and Modlin, 1999).

In conclusion, glycolipids do not constitute the major membrane lipid

fraction in M. pneumoniae, but their important characteristics are essential to the cell. The glycolipids are also important in the recognition of the bacteria by the host immune system. The promiscuous and processive character of the MPN483-enzyme probably yields bacterial-human lipid hybrids that might be important for an auto-immune process.

Paper IV

Many membrane associated mechanisms are belived to be regulated by the curvature stress in the membrane. The proportion of bilayer/nonbilayer-prone lipids in the membrane regulates this. The E.coli AD93 strain, lacking the nonbilayer-prone lipid PE, was used to investigate the ability of different

38

glycolipids to restore cell growth, cell division, stress response, and membrane permeability malfunctioning here. The membrane in wild type E.coli normally contains 75 mole% PE, and the depletion of this lipid has a large effect on many vital cell functions. The membrane of AD93 cells have a very high level, about 85 mole % of the negatively charged PG and the lipid mutant needs divalent cation like Mg2+ to survive, by neutrally a negatively charged membrane, which is critical for the cell to survive. The neutral glycolipid, αGlcDAG also nonbilayer-prone, (NB-prone) can restore the cell growth of AD93, however the membrane permeability was not fully restored (Wikström et al., 2004). This lipid, constituting about 50% of the membrane lipids, is believed to dilute the high surface charge and restore the environment of many membrane processes. However, the shape of this lipid is also similar to PE.

To further investigate what properties can restore cellular processes in the

PE mutant, surface charge or lipid shape, four new GTs were introduced into this E.coli strain. These were DGS from A. laidlawii making αGlcαGlcDAG, two Arabidopsis thaliana GTs making βGalDAG, αGalβGalDAG, and a processive GT, MPN483 from M. pneumoniae making GalDAG, βGlcβGalDAG, and GalGalGalDAG. The disugar lipids are bilayer-prone in nature due to their larger uncharged head groups and as the mono variants also uncharged. The glycolipid content in these strains reached up to 60 mol%, however the M. pneumoniae GT could only produce up to 28 mol% glycolipid in total. The βGlcβGalDAG lipid was the major glycolipid in this strain. These results reflect the maximum levels of glycolipid produced by the different organisms. The strain producing the bilayer-prone αGlcαGlcDAG lipid was here found to behave like wt E. coli in most aspects, and αGlcDAG and βGalDAG performed almost equally well. However, the βGalβGalDAG and MPN483 strains hardly survived.

Analysis of the acyl chain composition revealed that all strains had an

increased level of unsaturated chains, with the GalGalDAG and MPN483 glycolipid mix having the highest levels. The average chain length was increased in all but the GlcDAG strains. This is probably a way to compensate for loss of NB-prone lipids, increasing the curvature. All glycolipid strains except MPN483, have a major fraction of noncharged glycolipids, and reduced anionic ones compared to the AD93 host. The wt, GlcDAG and GalDAG strains have high levels of NB-prone lipids, however the AD93, GlcGlcDAG and the GalGalDAG strains lack these totally. The MPN483 clone has low levels of NB-prone lipids, since the GalDAG lipids is a very minor species, and the bilayer-forming GlcGalDAG and tri-sugar-DAG dominate the glycolipids. All glycolipid strains except MPN483, have a major fraction of noncharged glycolipids. The increased acyl chain unsaturation and length, increase the curvature stress and potentially

39

compensate for properties lost, and not fully restored by the foreign lipids. The saturation and chain length, thus seem to be the major factor in vivo for E. coli to control the curvature stress of the lipid bilayer. The levels of the NB-prone CL is increased in all glycolipid strains, and was found to be highest in the bilayer-prone GlcGlcDAG strain and lowest for the monosugar-DAGs. The observed increased levels of CL is probably due to compensation for the loss of nonbilayer-prone PE, hence to regulate the membrane curvature by other NB-prone lipids. The βGalβGalDAG strain also requires Mg2+ to grow. Hence, curvature may increase by a large membrane fraction of CL plus Mg2+, that is making CL NB-prone.

The lateral pressure between the acyl-chains was measured in all glycolipid strain to especially investigate the difference observed between the GlcGlcDAG and GalGalDAG lipids. It was seen that GalGalDAG has a larger lateral headgroup area, that changes the physical state of the membrane towards less nonbilayer-prone (less curvature) compared to GlcGlcDAG. This change seems to be compensate for E. coli by increased CL levels. This decrease of lateral pressure in the membrane decreases and prevents activities of several membrane functions. Smaller headgroups of lipids, most likely in combination with dilution of high surface concentration of anionic lipid charges, seems to bring advantages to a number of membrane associated proteins in E. coli. This was tested by using the transmembrane transporter LacY, known to have an altered topology in the absence of PE, as a marker. The passive downhill transport in this channel was restored in the presence of both GlcDAG (Wikström et al., 2004) and GlcGlcDAG, however only active transport was supported in the GlcDAG clone (Wikström et al., 2004). The topology of LacY was correct in both clones, still the uphill transport was absent in the GlcGlcDAG clone. The anionic phospholipid and non-charged glycolipid levels are similar in both clones, but the GlcGlcDAG one is essentially lacking nonbilayer-prone lipids species and has a substantially lower chain ordering due to its large headgroup. This will substantially lower the curvature stress and thus has an affect on the function of the LacY. Accordingly, LacY needs a certain curvature stress to function correctly.

40

Concluding remarks

The lipid bilayer has several important functions as described earlier. The membrane protects and shields the organism from its surrounding and provides the right properties for many enzymes and membrane proteins. It has been found that several enzymes are regulated by the membrane curvature stress or charge, and lipids are also closely associated with many membrane proteins e.g. bacteriorhodopsin (Luecke et al., 1999). The membrane is also an immediate contact layer, e.g. in bacteria lacking LPS, where the membrane is the first barrier against the host immune system.

I have in this thesis shown that a multivariate method can be used to

classify the different glycosyltransferase folds, reaction mechanisms within each fold family (paper I), and soluble and membrane bound GTs of the GT-B type (paper II). The multivariate sequence analysis method used here (paper I), could conveniently separate the three different fold types from each other, and divide the two different reaction mechanisms within the A and B structural fold groups. The method could also be used to predict the fold and reaction mechanism within E.coli and Synechocystis GTs in the CAZy database. This indicates that ORFs with unknown function and that share no sequence homology with known glycosyltransferases could potentially be identified with this method. When analysing the MPN483 enzyme with the above method, it was classified as a GT-A enzyme performing an inverting reaction mechanism. This method is easy to use and a powerful bioinformatic tool.

Amphipathic helices was found to be important for the group separation

of membrane-bound and soluble GTs (paper II), analysed by swapping these helixes from alMGS and E.coli MurG to a corresponding pair of soluble enzymes. The pI of the N-domain was also found to be important. In this work we have also shown that an amphipathic helix, here represented by a synthetic peptide, probably is unfolded until it reaches the membrane where it undergoes conformational changes. The helix will also regulate the enzymatic activity by binding to charged lipids, hence as the lipid content of the membrane changes it will affect the binding and also the activity. It was previously known that the MGS activity is regulated by charged lipids. This is believed to be mediated through the cluster of charges found in the C-terminal of the peptide (in enzyme the N-domain). These properties are most

41

likely found in other types of membrane acting enzymes, such as both GT-A and GT-B glycosyltransferases. The activity of the processive MPN483 is regulated by the PG and CL content of the micelle (paper III), indicating that the enzyme activity is regulated by a similar bilayer properties. It also has predicted amphipathic helixes, however this needs to be further investigated.

In paper III, I have shown that the MPN483 enzyme from M. pneumoniae

is both processive and promiscuous. The enzyme has the ability to use both ceramide, an eukaryotic lipid and DAG as precursors in the synthesis of glycolipids. This ability can be a strategy to make itself look more like the human host cell that the bacterium is attached to, or just a way to be flexible in the usage of different substrates (to take what one gets), which is probably beneficial for a very small cell. The use of both DAG and ceramide would normally be handled by having several enzymes, however having a promiscuous enzymes will solve this problem and also be energy saving. During evolution this bacterium has lost many enzymes, e.g. several GTs and this loss have been covered by developing of enzymes like MPN483. This GT has also been found to be essential, emphasising the importance of processive enzymes and glycolipids. The ability of neutral glycolipids to replace the neutral PE in the PE-depleted AD93 E. coli strain was analysed in paper IV. In this work, it was shown that GlcDAG and GalDAG could replace this lipid and restore many cellular functions, however di-sugar variants were less suited for this. The mono-sugar variants are all nonbilayer-prone due to its smaller sugar headgroups and the larger DAG backbone, this is a similar structural feature to the depleted PE, and substantially affects the curvature stress, compared to the AD93 strain. Disugar lipid molecules are instead bilayer-prone, making them less suitable for this. This was clearly demonstrated with LacY, where the topology of the protein were correct with both GlcDAG and GlcGlcDAG, but the activity much reduced in the latter strain. This channel thus needs a certain curvature stress to function correctly. These lipid molecules can not replace PE in its cellular function, showing the importance of different lipid molecules. The MPN483 enzyme preferentially making GlcGalDAG or GalGalDAG is essential in M pneumoniae showing that the variety of different sugar molecules are important. The produced glycolipids in M. pneumoniae are however a target for the immune system and the lipids are recognised by (early) antibodies during an infection. The exact mechanism for this is unclear today. Further studies needs to be made to analyse the mechanism by which this occur. The exact function of glycolipids in cellular processes also needs to be analysed, as well as mutational studies to investigate the importance of the separating parameters found in the different multivariate studies.

42

Varför fett? Sammanfattning på svenska

Alla bakterier, djur och växter består av celler, och alla celler har minst ett membran. Detta är cellens väggar och det gör det möjligt för cellen att ha olika miljö på dess ut- och insida, vilket är en mycket viktig funktion för att möjliggöra uppkomsten av liv. Väggarna kring cellen består av ca 50% fett (lipider) och 50% protein, sk äggviteämnen, där fettet är själva byggmaterialet och proteinerna cellens dörrar och fönster som används för att transportera in och ut ämnen som cellen behöver. Fett består av ett vattenälskande huvud (hydrofil) och icke vattenälskande svansar (hydrofoba), dessa organiserar sig så att svansarna lägger sig mot varandra och med huvudena utåt (se figuren), för att skyla svansarna från vatten. Det är därför fett lägger sig som ringar på vatten och inte löser sig.

Det finns många olika typer av lipider, laddade, oladdade, phospholipider

och sockerlipider, koniska och raka, där alla har en speciell funktion. Jag har i denna avhandling bla sett att formen på lipiden påverkar funktionen av proteinerna som är insatta i membranet. Om membranet innehåller fel typ av lipid fungerar inte proteinet som det ska. I papper 4 har vi sett att i ett membran med för lite koniska lipider blir för hårt packat och vissa proteiner förlorar sin funktion. Det är alltså viktigt att membranet innehåller rätt sorts fett. Det är inte bara proteiner som är insatta i membranet som påverkas av lipidinnehållet, utan även proteiner som binder mot ytan påverkas. I papper 2 har jag undersökt hur dessa proteiner binder mot ytan och hur denna bindning påverkas av olika lipider.

Fett har tidigare bara ansett som ett byggmaterial för cellen som utan

någon större funktion, men betydelsen av olika fettstrukturer börjar nu

43

klarna. Forskar värden ansåg tidigare att endast proteiner kändes igen av kroppens försvar, men även här har detta visats vara fel. Även vissa typer av lipider känns igen av kroppens försvar. Detta har visat sig vara av stor betydelse vid autoimmuna sjukdomar, där kroppen börjar brytas ned, istället för en inkräktare. Detta tros bero på att både bakterier, inkräktare och kroppsegna celler består av liknande fett molekyler och där försvaret förvirras av dess likheter. Det finns även bakterier som tros använda detta som ett tillvägagångssätt att gömma sig i kroppen. De använder fett som är så pass likt mänskliga varianter att försvaret inte ska kunna se någon skillnad. När inkräktaren väl upptäcks bryts även kroppens egna celler ner. Mycoplasma pneumoniae, är en av de minsta bakterierna som hittats på vår jord och som är en vanlig orsak till lunginflammation tros använda detta tillväga gångs sätt för att dels kunna undvika att bli upptäckt, men också för att spara energi. Denna bakterie (papper 3) har utvecklat en förmåga att använda fettet som finns i sin omgivning och sedan modifiera dessa, för att själv behöva tillverka fettmolekyler. Detta får till följd att den lättare kan interagera med mänskliga celler, men det kan även orsaka att den egna kroppen bryts ner efter en infektion.

Fett är sammanfattningsvis väldigt viktigt för oss, utan fett med rätt form

och innehåll fungerar inte våra proteinmaskiner.

44

Acknowledgment

I would like to express my gratitude to all the people that is making DBB such a great place to work at. Especially I would like to thank:

Åke, min handledare, tack för att din entusiasm för forskning och att du gett mig så mycket frihet under alla dess år. Tack för att du överraskade mig med att flytta till Stockholm när jag trodde att jag skulle få tillbringa ytterligare många år framöver i Umeå.

Mikael Sjöström, för att du hjälpt mig att förstå ACC, PLS och PLS-DA. Utan din hjälp hade papper I aldrig blivit till. Stefan, för att din dörr alltid är öppen och för att du tar dig tid. Du är bra boss för oss doktorander. Anki, Kicki, Ann, Charlotta och Sofia för att ni alltid med ett leende hjälpt mig med allt det praktiska. Malin W, för att du fått timmarna på labb att gå så mycket fortare. Vi har lärt känna varandra väldigt väl under alla dessa år, och gått igenom graviditeter och småbarns år tillsammans och mycket annat. Efter vår tid på plan 5 kan alla där upp allt om spädbarn och vet vilka underbarn vi har!!! Tuulia, för att du är en så fantastik person! För att du sprider glädje omkring dig och förlängt mitt liv med många år av alla goda skratt. Linda, som helt plötsligt klev in på kontoret och frågade om du inte kunde få göra D-kurs projekt hos mig. Utan din hjälp hade jag aldrig blivit klar med M.p lipid pappret i tid. Alex, for all good laughs at lunch time and for interesting conversations in the office. Amélie, for helping me with difficult clones. I value your deep knowledge very high, you always had an answer for everything.

45

Maria E, för att du introducerade fler dimensioner i mitt liv. Stefan B, och Patrik för tuffa innebandy matcher. Lu, for all great discussions in the lab. Hanna, för allt morgonkaffe du kokat det sista året, trevliga (korta) lunch diskussioner och gott samarbete i doktorandrådet. Kajsa, för den värme du sprider på plan 5. Det gamla lunch och fika gänget på plan 4, ni förgyllde tillvaron med många goda skratt och för att ni alltid lyssnat och peppat. Elin, för att du är en sådan fantastiks vän. Det finns ingen som känner mig så väl, och alltid förstår vad jag menar. För att du alltid finns där. För alla trevliga middagar på ”Haga deli”. Vi fixade det! Jag hoppas att vi hinner träffas oftare nu när avhandlingarna är klara (nästan). Karin, Lovisa och Erika, mitt glada tjejgäng. Vi är spridda från norr till söder, öst till väst men vänskapen består. Jag tänker ofta på de fantastiska studentåren vi tillbringade tillsammans i Umeå. Tack för att ni alltid finns där, även om det ibland går lång tid mellan de tillfällen då vi ses. Mormor och Morfar, för alla trevliga år i Umeå och underbara somrar. Morfar för att du introducerat jakt för mig, vilket gett oss många fantastiska timmar i skogen. Mormor för all god mat och fantastiska bullar. Anna, min underbara syster, för att du alltid finns där och peppar mig och för att du får mig att tänka på annat än jobb och familj. Det ska bli skönt att få hem dig igen. Gunnar för ditt intresse för forskning, och för att du som samhällsvetare försöker förstå lipider. Mamma och Pappa, utan allt ert stöd hade denna avhandling aldrig blivit skriven. Tack för att ni alltid trott på mig, uppmuntrat mig och för att ni alltid finns där. Vad vore jag utan er? Mina fantastiska tjejer Emmy och Thyra. Emmy för dina oerhörda värme och den glädje du ger mig, att resonera om varför saker är som det är en underbar utmaning. Thyra för att du alltid är så glad. Ni har gett livet en ny mening!!! Göran, du är mitt allt!!! Jag älskar dig innerligt. Nu ska vi äntligen få mer tid för varandra!!!

46

References

. Abdian, P.L., Lellouch, A.C., Gautier, C., Ielpi, L., and Geremia, R.A.

(2000) Identification of essential amino acids in the bacterial alpha -mannosyltransferase aceA. J Biol Chem 275: 40568-40575.

Ang, C.W., Jacobs, B.C., Brandenburg, A.H., Laman, J.D., van_der_Meche, F.G., Osterhaus, A.D., and van_Doorn, P.A. (2000) Cross-reactive antibodies against GM2 and CMV-infected fibroblasts in Guillain-Barré syndrome. Neurology 54: 1453-1458.

Baseman, J.B., Lange, M., Criscimagna, N.L., Giron, J.A., and Thomas, C. (1995) Interplay between mycoplasmas and host target cells. Microb Pathog 19: 105-116.

Baseman, J.B., and Tully, J.G. (1997) Mycoplasmas: sophisticated, reemerging, and burdened by their notoriety. Emerg Infect Dis 3: 21-32.

Berg, S., and Wieslander, Å. (1997) Purification of a phosphatase which hydrolyzes phosphatidic acid, a key intermediate in glucolipid synthesis in Acholeplasma laidlawii A membranes. Biochim Biophys Acta 1330: 225-232.

Boggs, J.M. (1987) Lipid intermolecular hydrogen bonding: influence on structural organization and membrane function. Biochim Biophys Acta 906: 353-404.

Boix, E., Swaminathan, G.J., Zhang, Y., Natesh, R., Brew, K., and Acharya, K.R. (2001) Structure of UDP complex of UDP-galactose:beta-galactoside-alpha -1,3-galactosyltransferase at 1.53-Å resolution reveals a conformational change in the catalytically important C terminus. J Biol Chem 276: 48608-48614.

Borg, N.A., Wun, K.S., Kjer-Nielsen, L., Wilce, M.C., Pellicci, D.G., Koh, R., Besra, G.S., Bharadwaj, M., Godfrey, D.I., McCluskey, J., and Rossjohn, J. (2007) CD1d-lipid-antigen recognition by the semi-invariant NKT T-cell receptor. Nature 448: 44-49.

Bourne, Y., and Henrissat, B. (2001) Glycoside hydrolases and glycosyltransferases: families and functional modules. Curr Opin Struct Biol 11: 593-600.

Breton, C., Mucha, J., and Jeanneau, C. (2001) Structural and functional features of glycosyltransferases. Biochimie 83: 713-718.

47

Breton, C., Snajdrova, L., Jeanneau, C., Koca, J., and Imberty, A. (2006) Structures and mechanisms of glycosyltransferases. Glycobiology 16: 29R-37R.

Brigl, M., van den Elzen, P., Chen, X., Meyers, J.H., Wu, D., Wong, C.H., Reddington, F., Illarianov, P.A., Besra, G.S., Brenner, M.B., and Gumperz, J.E. (2006) Conserved and heterogeneous lipid antigen specificities of CD1d-restricted NKT cell receptors. J Immunol 176: 3625-3634.

Buehner, M., Ford, G.C., Moras, D., Olsen, K.W., and Rossman, M.G. (1973) D-glyceraldehyde-3-phosphate dehydrogenase: three-dimensional structure and evolutionary significance. Proc Natl Acad Sci U S A 70: 3052-3054.

Campbell, J.A., Davies, G.J., Bulone, V., and Henrissat, B. (1997) A classification of nucleotide-diphospho-sugar glycosyltransferases based on amino acid sequence similarities. Biochem. J. 326: 929-939.

Chambaud, I., Wroblewski, H., and Blanchard, A. (1999) Interactions between mycoplasma lipoproteins and the host immune system. Trend Microbiol 7: 493-499.

Chambaud, I., Heilig, R., Ferris, S., Barbe, V., Samson, D., Galisson, F., Moszer, I., Dybvig, K., Wroblewski, H., Viari, A., Rocha, E.P., and Blanchard, A. (2001) The complete genome sequence of the murine respiratory pathogen Mycoplasma pulmonis. Nucl Acid Res 29: 2145-2153.

Charnock, S.J., and Davies, G.J. (1999) Structure of the nucleotide-diphospho-sugar transferase, SpsA from Bacillus subtilis, in native and nucleotide-complexed forms. Biochemistry 38: 6380-6385.

Charnock, S.J., Henrissat, B., and Davies, G.J. (2001) Three-dimensional structures of UDP-sugar glycosyltransferases illuminate the biosynthesis of plant polysaccharides. Plant Physiol 125: 527-531.

Chiu, C.P., Watts, A.G., Lairson, L.L., Gilbert, M., Lim, D., Wakarchuk, W.W., Withers, S.G., and Strynadka, N.C. (2004) Structural analysis of the sialyltransferase CstII from Campylobacter jejuni in complex with a substrate analog. Nat Struct Mol Biol 11: 163-170.

Cornell, R.B., and Northwood, I.C. (2000) Regulation of CTP:phosphocholine cytidylyltransferase by amphitropism and relocalization. Trends Biochem Sci 25: 441-447.

Cornell, R.B., and Taneva, S.G. (2006) Amphipathic helices as mediators of the membrane interaction of amphitropic proteins, and as modulators of bilayer physical properties. Curr Protein Pept Sci 7: 539-552.

Coutinho, P.M., and Henrissat, B. (1999) Carbohydrate-active enzymes: an integrated database approach. In Recent Advances in Carbohydrate Bioengineering. Gilbert, H.J., Davies, G., Henrissat, B. and Svensson, B. (eds). Cambridge: The Royal Society of Chemistry, pp. 3-12.

48

Coutinho, P.M., Deleury, E., Davies, G.J., and Henrissat, B. (2003) An evolving hierarchical family classification for glycosyltransferases. J Mol Biol 328: 307-317.

Cronan, J.E. (2003) Bacterial membrane lipids: where do we stand? Annu Rev Microbiol 57: 203-224.

Curatolo, W. (1987a) Glycolipid function. Biochim Biophys Acta 906: 137-160.

Curatolo, W. (1987b) The physical properties of glycolipids. Biochim Biophys Acta 906: 111-136.

Dahlqvist, A., Andersson, S., and Wieslander, Å. (1992) The enzymatic synthesis of membrane glucolipids in Acholeplasma laidlawii. Biochim. Biophys. Acta. 23: 131-140.

Davies, G., and Henrissat, B. (1995) Structures and mechanisms of glycosyl hydrolases. Structure 3: 853-859.

Davies, G.J., Gloster, T.M., and Henrissat, B. (2005) Recent structural insights into the expanding world of carbohydrate-active enzymes. Curr Opin Struct Biol 15: 637-645.

DeChavigny, A., Heacock, P.N., and Dowhan, W. (1991) Sequence and inactivation of the pss gene of Escherichia coli. Phosphatidylethanolamine may not be essential for cell viability. J Biol Chem 266: 5323-5332.

Dimitrov, D.S., Franzoso, G., Salman, M., Blumenthal, R., Tarshis, M., Barile, M.F., and Rottem, S. (1993) Mycoplasma fermentans (incognitus strain) cells are able to fuse with T lymphocytes. Clin Infect Dis 17 Suppl 1: S305-308.

Dowhan, W. (1997) Molecular basis for membrane phospholipid diversity: why are there so many lipids? Annu Rev Biochem 66: 199-232.

Dowhan, W., and Bogdanov, M. (2002) Functional roles of lipids in membranes. In Biochemistry of lipids, lipoproteins and membranes. Vance, D.E.V.a.J.E. (ed). Amsterdam: Elsevier, pp. 1-35.

Drin, G., Casella, J.F., Gautier, R., Boehmer, T., Schwartz, T.U., and Antonny, B. (2007) A general amphipathic alpha-helical motif for sensing membrane curvature. Nat Struct Mol Biol 14: 138-146.

Edman, M., Jarhede, T., Sjöström, M., and Wieslander, A. (1999) Different sequence patterns in signal peptides from mycoplasmas, other gram-positive bacteria, and Escherichia coli: a multivariate data analysis. Proteins 35: 195-205.

Edman, M. (2001) Detection of sequence patterns in membrane proteins : a story of many dimensions. Umeå: Univ.

Flint, J., Taylor, E., Yang, M., Bolam, D.N., Tailford, L.E., Martinez-Fleites, C., Dodson, E.J., Davis, B.G., Gilbert, H.J., and Davies, G.J. (2005) Structural dissection and high-throughput screening of mannosylglycerate synthase. Nat Struct Mol Biol 12: 608-614.

Franzoso, G., Dimitrov, D.S., Blumenthal, R., Barile, M.F., and Rottem, S. (1992) Fusion of Mycoplasma fermentans strain incognitus with T-lymphocytes. FEBS Letters 303: 251-254.

49

Fraser, C.M., Gocayne, J.D., White, O., Adams, M.D., Clayton, R.A., Fleischmann, R.D., Bult, C.J., Kerlavage, A.R., Sutton, G., Kelley, J.M., and et al. (1995) The minimal gene complement of Mycoplasma genitalium. Science 270: 397-403.

Fritz, T.A., Hurley, J.H., Trinh, L.B., Shiloach, J., and Tabak, L.A. (2004) The beginnings of mucin biosynthesis: the crystal structure of UDP-GalNAc:polypeptide alpha-N-acetylgalactosaminyltransferase-T1. Proc Natl Acad Sci U S A 101: 15307-15312.

Galli, G., Nuti, S., Tavarini, S., Galli-Stampino, L., De Lalla, C., Casorati, G., Dellabona, P., and Abrignani, S. (2003) Innate immune responses support adaptive immunity: NKT cells induce B cell activation. Vaccine 21 Suppl 2: S48-54.

Garinot-Schneider, C., Lellouch, A.C., and Geremia, R.A. (2000) Identification of essential amino acid residues in the Sinorhizobium meliloti glucosyltransferase ExoM. J Biol Chem 275: 31407-31413.

Gastinel, L.N., Cambillau, C., and Bourne, Y. (1999) Crystal structures of the bovine beta4galactosyltransferase catalytic domain and its complex with uridine diphosphogalactose. EMBO 18: 3546-3557.

Gibson, R.P., Turkenburg, J.P., Charnock, S.J., Lloyd, R., and Davies, G.J. (2002) Insights into trehalose synthesis provided by the structure of the retaining glucosyltransferase OtsA. Chem Biol 9: 1337-1346.

Gibson, R.P., Tarling, C.A., Roberts, S., Withers, S.G., and Davies, G.J. (2004) The donor subsite of trehalose-6-phosphate synthase: binary complexes with UDP-glucose and UDP-2-deoxy-2-fluoro-glucose at 2 A resolution. J Biol Chem 279: 1950-1955.

Gilbert, M., Brisson, J., Karwaski, M.F., Michniewicz, J., Cunningham, A.M., Wu, Y., Young, N.M., and Wakarchuk, W.W. (2000) Biosynthesis of ganglioside mimics in Campylobacter jejuni OH4384. Identification of the glycosyltransferase genes, enzymatic synthesis of model compounds, and characterization of nanomole amounts by 600-mhz (1)h and (13)c NMR analysis. J Biol Chem 275: 3896-3906.

Glass, J.I., Assad-Garcia, N., Alperovich, N., Yooseph, S., Lewis, M.R., Maruf, M., Hutchison, C.A., 3rd, Smith, H.O., and Venter, J.C. (2006) Essential genes of a minimal bacterium. Proc Natl Acad Sci U S A 103: 425-430.

Gordon, R.D., Sivarajah, P., Satkunarajah, M., Ma, D., Tarling, C.A., Vizitiu, D., Withers, S.G., and Rini, J.M. (2006) X-ray crystal structures of rabbit N-acetylglucosaminyltransferase I (GnT I) in complex with donor substrate analogues. J Mol Biol 360: 67-79.

Grizot, S., Salem, M., Vongsouthi, V., Durand, L., Moreau, F., Dohi, H., Vincent, S., Escaich, S., and Ducruix, A. (2006) Structure of the Escherichia coli heptosyltransferase WaaC: binary complexes with ADP and ADP-2-deoxy-2-fluoro heptose. J Mol Biol 363: 383-394.

Guerin, M.E., Buschiazzo, A., Kordulakova, J., Jackson, M., and Alzari, P.M. (2005) Crystallization and preliminary crystallographic analysis of PimA, an essential mannosyltransferase from

50

Mycobacterium smegmatis. Acta Crystallograph Sect F Struct Biol Cryst Commun 61: 518-520.

Ha, S., Walker, D., Shi, Y., and Walker, S. (2000) The 1.9 Å crystal structure of Escherichia coli MurG, a membrane-associated glycosyltransferase involved in peptidoglycan biosynthesis. Protein Sci 9: 1045-1052.

Ha, S., Gross, B., and Walker, S. (2001) E. coli MurG: a paradigm for a superfamily of glycosyltransferases. Curr Drug Targets Infect Disord 1: 201-213.

Hanzal-Bayer, M.F., and Hancock, J.F. (2007) Lipid rafts and membrane traffic. FEBS Lett 581: 2098-2104.

Hauksson, J.B., Lindblom, G., and Rilfors, L. (1994a) Structures of glucolipids from the membrane of Acholeplasma laidlawii strain A-EF22. II. Monoacylmonoglucosyldiacylglycerol. Biochim. Biophys. Acta 1215: 341-345.

Hauksson, J.B., Lindblom, G., and Rilfors, L. (1994b) Structures of glucolipids from the membrane of Acholeplasma laidlawii strain A-EF22. I. Glycerophosphoryldiglucosyldiacylglycerol and monoacylbisglycerophosphoryldiglucosyldiacylglycerol. Biochim. Biophys. Acta. 1214: 124-130.

Hayashi, S., and Wu, H.C. (1990) Lipoproteins in bacteria. J Bioenerg Biomembr 22: 451-471.

Hellberg, S., Sjöström, M., Skagerberg, B., and Wold, S. (1987) Peptide quantitative structure-activity relationships, a multivariate approach. J Med Chem 30: 1126-1135.

Henrissat, B., and Davies, G. (1997) Structural and sequence-based classification of glycoside hydrolases. Curr Opin Struct Biol 7: 637-644.

Henrissat, B., and Davies, G.J. (2000) Glycoside hydrolases and glycosyltransferases. Families, modules, and implications for genomics. Plant Physiol 124: 1515-1519.

Himmelreich, R., Hilbert, H., Plagens, H., Pirkl, E., Li, B.C., and Herrmann, R. (1996) Complete sequence analysis of the genome of the bacterium Mycoplasma pneumoniae. Nucl Acid Res 24: 4420-4449.

Himmelreich, R., Plagens, H., Hilbert, H., Reiner, B., and Herrmann, R. (1997) Comparative analysis of the genomes of the bacteria Mycoplasma pneumoniae and Mycoplasma genitalium. Nucl Acid Res 25: 701-712.

Hu, Y., and Walker, S. (2002) Remarkable structural similarities between diverse glycosyltransferases. Chem Biol 9: 1287-1296.

Hu, Y., Chen, L., Ha, S., Gross, B., Falcone, B., Walker, D., Mokhtarzadeh, M., and Walker, S. (2003) Crystal structure of the MurG:UDP-GlcNAc complex reveals common structural principles of a superfamily of glycosyltransferases. Proc Natl Acad Sci U S A 100: 845-849.

Johansson, E.K., and Pettersson, B. (2002) Taxonomy of Mollicutes. In Molecular biology and phatogenicity of Mycoplasmas. Razin,

51

S.a.H.R. (ed). New York: Kluwer Academic/Plenum Puplishers, pp. 1-30.

Karlsson, O.P., Dahlqvist, A., and Wieslander, Å. (1994) Activation of the membrane glucolipid synthesis in Acholeplasma laidlawii by phosphatidylglycerol and other anionic lipids. J Biol Chem 269: 23484-23490.

Karlsson, O.P., Dahlqvist, A., Vikstrom, S., and Wieslander, Å. (1997) Lipid dependence and basic kinetics of the purified 1,2-diacylglycerol 3-glucosyltransferase from membranes of Acholeplasma laidlawii. J Biol Chem 272: 929-936.

Keenleyside, W.J., Clarke, A.J., and Whitfield, C. (2001) Identification of residues involved in catalytic activity of the inverting glycosyl transferase WbbE from Salmonella enterica serovar borreze. J Bacteriol 183: 77-85.

Kikuchi, N., Kwon, Y.D., Gotoh, M., and Narimatsu, H. (2003) Comparison of glycosyltransferase families using the profile hidden Markov model. Biochem Biophys Res Commun 310: 574-579.

Kinjo, Y., Tupin, E., Wu, D., Fujio, M., Garcia-Navarro, R., Benhnia, M.R., Zajonc, D.M., Ben-Menachem, G., Ainge, G.D., Painter, G.F., Khurana, A., Hoebe, K., Behar, S.M., Beutler, B., Wilson, I.A., Tsuji, M., Sellati, T.J., Wong, C.H., and Kronenberg, M. (2006) Natural killer T cells recognize diacylglycerol antigens from pathogenic bacteria. Nat Immunol 7: 978-986.

Kumada, S., Kusaka, H., Okaniwa, M., Kobayashi, O., and Kusunoki, S. (1997) Encephalomyelitis subsequent to mycoplasma infection with elevated serum. Pediatr Neurol 16: 241-244.

Kusunoki, S., Chiba, A., Hitoshi, S., Takizawa, H., and Kanazawa, I. (1995) Anti-Gal-C antibody in autoimmune neuropathies subsequent to mycoplasma infection. Muscle. Nerve 18: 409-413.

Kusunoki, S., Shiina, M., and Kanazawa, I. (2001) Anti-Gal-C antibodies in GBS subsequent to mycoplasma infection: evidence of molecular mimicry. Neurology 57: 736-738.

Ladbrooke, B.D., Williams, R.M., and Chapman, D. (1968) Studies on lecithin-cholesterol-water interactions by differential scanning calorimetry and X-ray diffraction. Biochim. Biophys. Acta. 150: 333-340.

Ladokhin, A.S., and White, S.H. (2001) Protein chemistry at membrane interfaces: non-additivity of electrostatic and hydrophobic interactions. J Mol Biol 309: 543-552.

Lapinsh, M., Gutcaits, A., Prusis, P., Post, C., Lundstedt, T., and Wikberg, J.E. (2002) Classification of G-protein coupled receptors by alignment-independent extraction of principal chemical properties of primary amino acid sequences. Protein Sci 11: 795-805.

Lee, A.G. (2000) Membrane lipids: it's only a phase. Curr Biol 10: R377-380.

Leipold, M.D., Kaniuk, N.A., and Whitfield, C. (2007) The C-terminal Domain of the Escherichia coli WaaJ glycosyltransferase is

52

important for catalytic activity and membrane association. J Biol Chem 282: 1257-1264.

Li, X.M., Momsen, M.M., Smaby, J.M., Brockman, H.L., and Brown, R.E. (2001) Cholesterol decreases the interfacial elasticity and detergent solubility of sphingomyelins. Biochemistry 40: 5954-5963.

Liu, J., and Mushegian, A. (2003) Three monophyletic superfamilies account for the majority of the known glycosyltransferases. Protein Sci 12: 1418-1431.

Lobsanov, Y.D., Romero, P.A., Sleno, B., Yu, B., Yip, P., Herscovics, A., and Howell, P.L. (2004) Structure of Kre2p/Mnt1p: a yeast alpha1,2-mannosyltransferase involved in mannoprotein biosynthesis. J Biol Chem 279: 17921-17931.

Luecke, H., Schobert, B., Richter, H.T., Cartailler, J.P., and Lanyi, J.K. (1999) Structure of bacteriorhodopsin at 1.55 A resolution. J Mol Biol 291: 899-911.

Manillof, J. (2002) Phylogeny and Evolution. In Molecular biology and phatogenicity of Mycoplasmas. Razin, S.a.H.R. (ed). New York: Kluwer Academic/Plenum Puplishers, pp. 31-44.

Martinez-Fleites, C., Proctor, M., Roberts, S., Bolam, D.N., Gilbert, H.J., and Davies, G.J. (2006) Insights into the synthesis of lipopolysaccharide and antibiotics through the structures of two retaining glycosyltransferases from family GT4. Chem Biol 13: 1143-1152.

Matsumoto, K., Kusaka, J., Nishibori, A., and Hara, H. (2006) Lipid domains in bacterial membranes. Mol Microbiol 61: 1110-1117.

McElhaney (1992) Membrane structure. In Mycoplasma; Molecular biology and pathogenesis. Maniloff, J., McElhaney, R.N., Finch, L.R. and Baseman , J.B. (eds). Washington D.C.: ASM, pp. 113-157.

Mulichak, A.M., Losey, H.C., Walsh, C.T., and Garavito, R.M. (2001) Structure of the UDP-glucosyltransferase Gtfb that modifies the heptapeptide aglycone in the biosynthesis of vancomycin group antibiotics. Structure 9: 547-557.

Mulichak, A.M., Losey, H.C., Lu, W., Wawrzak, Z., Walsh, C.T., and Garavito, R.M. (2003) Structure of the TDP-epi-vancosaminyltransferase GtfA from the chloroeremomycin biosynthetic pathway. Proc Natl Acad Sci U S A 100: 9238-9243.

Mulichak, A.M., Lu, W., Losey, H.C., Walsh, C.T., and Garavito, R.M. (2004) Crystal structure of vancosaminyltransferase GtfD from the vancomycin biosynthetic pathway: interactions with acceptor and nucleotide ligands. Biochemistry 43: 5170-5180.

Murzin, A.G., Brenner, S.E., Hubbard, T., and Chothia, C. (1995) SCOP: a structural classification of proteins database for the investigation of sequences and structures. J Mol Biol 247: 536-540.

Mushegian, A.R., and Koonin, E.V. (1996) A minimal gene set for cellular life derived by comparison of complete bacterial genomes. Proc Natl Acad Sci U S A 93: 10268-10273.

53

O'Toole, C., and Lowdell , M. (1990) Infection of human T-cells with mycoplasma , inhibition of CD4 expression and HIV-1 gp 120 glycoprotein binding and infectivity. Lancet 336: 1067.

Pak, J.E., Arnoux, P., Zhou, S., Sivarajah, P., Satkunarajah, M., Xing, X., and Rini, J.M. (2006) X-ray crystal structure of leukocyte type core 2 beta1,6-N-acetylglucosaminyltransferase. Evidence for a convergence of metal ion-independent glycosyltransferase mechanism. J Biol Chem 281: 26693-26701.

Parekh, V.V., Singh, A.K., Wilson, M.T., Olivares-Villagomez, D., Bezbradica, J.S., Inazawa, H., Ehara, H., Sakai, T., Serizawa, I., Wu, L., Wang, C.R., Joyce, S., and Van Kaer, L. (2004) Quantitative and qualitative differences in the in vivo response of NKT cells to distinct alpha- and beta-anomeric glycolipids. J Immunol 173: 3693-3706.

Patenaude, S.I., Seto, N.O., Borisova, S.N., Szpacenko, A., Marcus, S.L., Palcic, M.M., and Evans, S.V. (2002) The structural basis for specificity in human ABO(H) blood group biosynthesis. Nat Struct Biol 9: 685-690.

Pedersen, L.C., Tsuchida, K., Kitagawa, H., Sugahara, K., Darden, T.A., and Negishi, M. (2000) Heparan/chondroitin sulfate biosynthesis. Structure and mechanism of human glucuronyltransferase I. J Biol Chem 275: 34580-34585.

Persson, K., Ly, H.D., Dieckelmann, M., Wakarchuk, W.W., Withers, S.G., and Strynadka, N.C. (2001) Crystal structure of the retaining galactosyltransferase LgtC from Neisseria meningitidis in complex with donor and acceptor sugar analogs. Nat Struct Biol 8: 166-175.

Porcelli, S.A., and Modlin, R.L. (1999) The CD1 system: antigen-presenting molecules for T cell recognition of lipids and glycolipids. Annu Rev Immunol 17: 297-329.

Qasba, P.K., Ramakrishnan, B., and Boeggeman, E. (2005) Substrate-induced conformational changes in glycosyltransferases. Trends Biochem Sci 30: 53-62.

Rackovsky, S. (1998) "Hidden" sequence periodicities and protein architecture. Proc Natl Acad Sci U S A 95: 8580-8584.

Ramakrishnan, B., and Qasba, P.K. (2001) Crystal structure of lactose synthase reveals a large conformational change in its catalytic component, the beta1,4-galactosyltransferase-I. J Mol Biol 310: 205-218.

Ramakrishnan, B., Balaji, P.V., and Qasba, P.K. (2002a) Crystal structure of beta1,4-galactosyltransferase complex with UDP-Gal reveals an oligosaccharide acceptor binding site. J Mol Biol 318: 491-502.

Ramakrishnan, B., Boeggeman, E., and Qasba, P.K. (2002b) Beta-1,4-galactosyltransferase and lactose synthase: molecular mechanical devices. Biochem Biophys Res Commun 291: 1113-1118.

Ramakrishnan, B., Ramasamy, V., and Qasba, P.K. (2006) Structural snapshots of beta-1,4-galactosyltransferase-I along the kinetic pathway. J Mol Biol 357: 1619-1633.

54

Rietveld, A., van Kemenade, T.J., Hak, T., Verkleij, A.J., and de Kruijff, B. (1987) The effect of cytochrome c oxidase on lipid polymorphism of model membranes containing cardiolipin. Eur J Biochem 164: 137-140.

Rilfors, L., Wieslander, Å., and Lindblom, G. (1993) Regulation and physicochemical properties of the polar lipids in Acholeplasma laidlawii. Subcell Biochem 20: 109-166.

Rottem, S., Adar, L., Gross, Z., Ne'eman, Z., and Davis, P.J. (1986) Incorporation and modification of exogenous Phosphatidylcholines by mycoplasmas. J Bacteriol 167: 299-304.

Ruuth, E., and Praz, F. (1989) Interactionbetween mycoplamsa and the immune system. Immun Rev 112: 133-160.

Ruvolo, P.P. (2003) Intracellular signal transduction pathways activated by ceramide and its metabolites. Pharmacol Res 47: 383-392.

Seelig, J. (2004) Thermodynamics of lipid-peptide interactions. Biochim Biophys Acta 1666: 40-50.

Shah, D.O., and Schulman, J.H. (1967) Influence of calcium, cholesterol, and unsaturation on lecithin monolayers. J Lipid Res 8: 215-226.

Shao, H., He, X., Achnine, L., Blount, J.W., Dixon, R.A., and Wang, X. (2005) Crystal structures of a multifunctional triterpene/flavonoid glycosyltransferase from Medicago truncatula. Plant Cell 17: 3141-3154.

Simidjiev, I., Stoylova, S., Amenitsch, H., Javorfi, T., Mustardy, L., Laggner, P., Holzenburg, A., and Garab, G. (2000) Self-assembly of large, ordered lamellae from non-bilayer lipids and integral membrane proteins in vitro. Proc Natl Acad Sci U S A 97: 1473-1476.

Simons, K., and Vaz, W.L. (2004) Model systems, lipid rafts, and cell membranes. Annu Rev Biophys Biomol Struct 33: 269-295.

Smith, P.F. (1992) Membrane Lipid and Lipopolysaccharide Structures. In Mycoplasma; Molecular biology and pathogenesis. Maniloff, J., McElhaney, R.N., Finch, L.R. and Baseman , J.B. (eds). Washington D.C.: ASM, pp. 79-93.

Snajdrova, L., Kulhanek, P., Imberty, A., and Koca, J. (2004) Molecular dynamics simulations of glycosyltransferase LgtC. Carbohydr Res 339: 995-1006.

Susuki, K., Odaka, M., Mori, M., Hirata, K., and Yuki, N. (2004) Acute motor axonal neuropathy after Mycoplasma infection: Evidence of molecular mimicry. Neurology 62: 949-956.

Tarbouriech, N., Qasba, P.K., and Davies, G.J. (2001) Three-dimensional structures of the Mn and Mg dTDP complexes of the family GT-2 glycosyltransferase SpsA: a comparison with related NDP-sugar glycosyltransferases. J Mol Biol 314: 655-661.

Tarshis, M., Salman, M., and Rottem, S. (1993) Cholesterol is required for the fusion of single unilamellar vesicles with Mycoplasma capricolum. Biophys J 64: 709-715.

55

Tsiodras, S., Kelesidis, I., Kelesidis, T., Stamboulis, E., and Giamarellou, H. (2005) Central nervous system manifestations of Mycoplasma pneumoniae infections. J Infect 51: 343-354.

Tvaroska, I. (2004) Molecular modeling insights into the catalytic mechanism of the retaining galactosyltransferase LgtC. Carbohydr Res 339: 1007-1014.

Uemura, K., Loomes, L.M., Childs, R.A., and Feizi, T. (1988) Evidence for occurrence of passively adsorbed I antigen activity on a cultured strain of Mycoplasma pneumoniae. Infect. Immun. 56: 3015-3020.

Unligil, U.M., Zhou, S., Yuwaraj, S., Sarkar, M., Schachter, H., and Rini, J.M. (2000) X-ray crystal structure of rabbit N-acetylglucosaminyltransferase I: catalytic mechanism and a new protein superfamily. EMBO J. 19: 5269-5280.

Waites, K.B., and Talkington, D.F. (2004) Mycoplasma pneumoniae and its role as a human pathogen. Clin Microbiol Rev 17: 697-728

van den Brink-van der Laan, E., Boots, J.W., Spelbrink, R.E., Kool, G.M., Breukink, E., Killian, J.A., and de Kruijff, B. (2003) Membrane interaction of the glycosyltransferase MurG: a special role for cardiolipin. J Bacteriol 185: 3773-3779.

van den Brink-van der Laan, E., Killian, J.A., and de Kruijff, B. (2004) Nonbilayer lipids affect peripheral and integral membrane proteins via changes in the lateral pressure profile. Biochim Biophys Acta 1666: 275-288.

van Meer, G. (2005) Cellular lipidomics. Embo J 24: 3159-3165. Wieslander, Å., Ulmius, J., Lindblom, G., and Fontell, K. (1978) Water

binding and phase structures for different Acholeplasma laidlawii membrane lipids studied by deuteron nuclear magnetic resonance and x-ray diffraction. Biochim Biophys Acta 512: 241-253.

Wikström, M., Xie, J., Bogdanov, M., Mileykovskaya, E., Heacock, P., Wieslander, A., and Dowhan, W. (2004) Monoglucosyldiacylglycerol, a foreign lipid, can substitute for phosphatidylethanolamine in essential membrane-associated functions in Escherichia coli. J Biol Chem 279: 10484-10493.

Vikström, S., Li, L., Karlsson, O.P., and Wieslander, Å. (1999) A Key role of the diglucosyldiacylglycerol synthase for the nonbilayer-bilayer lipid balance of Acholeplasma laidlawii membranes. Biochemistry 38: 5511-5520.

Wold, S., and Sjöström, M. (1998) Chemometrics and its roots in physical organic chemistry. Acta Chemi Scand 52: 517-523.

Wolf, M., Muller, T., Dandekar, T., and Pollack, J.D. (2004) Phylogeny of Firmicutes with special reference to Mycoplasma (Mollicutes) as inferred from phosphoglycerate kinase amino acid sequence data

Int J Syst Evol Microbiol 54: 871-875. Xie, J., Bogdanov, M., Heacock, P., and Dowhan, W. (2006)

Phosphatidylethanolamine and monoglucosyldiacylglycerol are interchangeable in supporting topogenesis and function of the

56

polytopic membrane protein lactose permease. J Biol Chem 281: 19172-19178.

Yamamoto, F., Clausen, H., White, T., Marken, J., and Hakomori, S. (1990) Molecular genetic basis of the histo-blood group ABO system. Nature 345: 229-233.

Zhang, W., Campbell, H.A., King, S.C., and Dowhan, W. (2005) Phospholipids as determinants of membrane protein topology. Phosphatidylethanolamine is required for the proper topological organization of the gamma-aminobutyric acid permease (GabP) of Escherichia coli. J Biol Chem 280: 26032-26038.