Structural and Functional Studies of Ribose-5-Phosphate Isomerase B

Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 332

Structural and Functional Studies of Ribose-5-phosphate isomerase B

ANNETTE K. ROOS

ACTA UNIVERSITATIS UPSALIENSIS ISSN 1651-6214 UPPSALA ISBN 978-91-554-6952-8 2007 urn:nbn:se:uu:diva-8182 ! ""# $%"" & ' & & (' ') *' + , ')

- . /) ""#) & - 010 ' ' ) .

) ) # ) ) 2 3 4#!04$01105410!)

- 10 ' ' 6- 7 & ' 8 9 & ' ' ' '+ +' ' 0 & 10 ' ' 6-1(7 10 ' ') *+ & & ' +' & : - . - ) *' ' - & ' 67 67 & & & +) ' ' - . - +'' ' ' - ' && 050 ' ' ) . ' ' + ' - ' ' ) 2 ' ' & - ) *' & - + +' && ' ' + +' ' -1() *' & ' ' ' ' & - & ' - ) . & - ' + ' ' -1( ' 50 ' ' ) && ' + - - ' ; ) . & - +' ' 44 + ' ' - ' & & ' +'' ' & ' + ' ) 2 ' ' ' ' - - . & ' +'' ' ' ' - .) && ' ' & ' + < - & & & ' +' ' & + 0 )

! ' 0 '' 050 ' ' ' ' '+ -51 =0 '

" # $ % $ % & '()$ $ *+',-. $

> . /) - ""#

2 3 $51$05$ 2 3 4#!04$01105410! % %%% 0!$! 6' %?? );)? @ A % %%% 0!$!7 List of Papers

This thesis is a summary of the results presented in the following publications and manuscript, which will be referred to by their roman numerals:

I Roos A. K., Andersson C. E., Bergfors T., Jacobsson M., Karlen A., Unge T., Jones T. A., Mowbray S. L. (2004) My- cobacterium tuberculosis Ribose-5-phosphate Isomerase has a known fold, but a novel active site. J. Mol. Biol. 335, 799- 809. II Roos A. K., Burgos E., Ericsson D. J., Salmon L., Mowbray S. L. (2005) Competitive Inhibitors of Mycobacterium tuberculosis Ribose-5-phosphate Isomerase B Reveal New Infor- mation About the Reaction Mechanism. J. Biol. Chem. 280, 6416-6422. III Burgos E., Roos A. K., Mowbray S. L., Salmon L. (2005) Synthesis of 5-deoxy-5-phospho-D-ribonohydroxamic acid: a new competitive and selective inhibitor of type B ribose-5- phosphate isomerase from Mycobacterium tuberculosis. Tet- rahedron Lett. 46, 3691-3694. IV Roos A. K., Mariano S., Kowalinski E., Salmon L., Mow- bray S. L. (2007) Ribose-5-phosphate isomerase B from Es- cherichia coli is also a functional allose-6-phosphate isomerase while the Mycobacterium tuberculosis enzyme is not. Manuscript

Articles I-III are printed with permission from the copyright holders.

iii Additional Paper

Nurbo J., Roos A. K., Muthas D., Wahlström E., Ericsson D. J., Lunstedt T., Unge T., Karlen A. (2007) Design, Synthesis and Evaluation of Peptide In- hibitors of Mycobacterium tuberculosis Ribonucleotide Reductase. J. Peptide Science, In press

iv Contents

Introduction...... 1 Conveying ribose across the cell membrane ...... 1 The Pentose Phosphate Pathway ...... 2 Ribose-5-phosphate isomerase – introducing the main character ...... 4 The als operon of E. coli ...... 7 Mycobacterium tuberculosis ...... 10 Sugar transport in M. tuberculosis...... 10 The pentose phosphate pathway of M. tuberculosis...... 11 General considerations in designing new antitubercular drugs...... 12 Rpi essentiality studies in bacteria and fungi...... 12 A human Rpi mutant causing abnormal brain development...... 13 Rpi – a possible drug target? ...... 14 Current investigation...... 15 Aim of thesis ...... 15 Introductory sequence comparisons and genomics searches...... 16 Functional studies of MtRpiB and EcRpiB ...... 18 Assay results of MtRpiB...... 20 Assay results of EcRpiB ...... 25 Allose-6-phosphate isomerase activity (Paper IV) ...... 26 Crystallisation...... 27 Structural results...... 33 Solving the structure of MtRpiB (Paper I)...... 33 The overall structure of MtRpiB...... 35 The active site of MtRpiB...... 37 E. coli RpiB structures...... 42 Comparing the Ec- and MtRpiB structures ...... 48 The proposed reaction mechanism of RpiB ...... 51 The reaction as catalysed by RpiA...... 54 Conclusions and future perspectives...... 56 Summary in Swedish ...... 59 Acknowledgements...... 64 References...... 68

v Abbreviations

Rpi Ribose-5-phosphate isomerase Mt / M. tuberculosis Mycobacterium tuberculosis Ec / E. coli Escherichia coli

PPP Pentose phosphate pathway R5P Ribose 5-phosphate Ru5P Ribulose 5-phosphate All6P Allose 6-phosphate Allu6P Allulose 6-phosphate

4PEA 4-deoxy-4-phospho-D-erythronate 4PEH 4-deoxy-4-phospho-D-erythronohydroxamic acid 4PEAm 4-deoxy-4-phospho-D-erythronamide 4PEHz 4-deoxy-4-phospho-D-erythronhydrazide 4PMEA 4-deoxy-4-phosphonomethyl-D-erythronate 5PRH 5-deoxy-5-phospho-D-ribonohydroxamic acid 5PRA 5-deoxy-5-phospho-D-ribonate Pi Inorganic phosphate F6P Fructose 6-phosphate G6P Glucose 6-phosphate PGI Phosphoglucose isomerase TIM Triose-phosphate isomerase

MESNA 2-mercaptoethanesulfonic acid ME -mercaptoethanol MES 4-morpholineethanesulfonic acid HEPES 4-(2-hydroxyethyl)-1-piperazineethane-sulfonate PDB Protein Data Bank MR molecular replacement CC-F correlation coefficient AU asymmetric unit NCS non-crystallographic symmetry r.m.s. root-mean-square

vi Introduction

All organisms need energy to be able to exist. Some, like plants, get their energy from the sun through the process called photosynthesis. Others, like animals, birds and most bacteria, have to oxidise organic compounds in order to survive. These organic compounds are processed by different enzymes in metabolic pathways. In such pathways, the organic compounds are either rearranged and put together to become larger molecules, or rearranged and disassembled to become smaller building blocks, often releasing energy in the process. This thesis deals with an enzyme involved in one of the central metabolic pathways, the pentose phosphate pathway. The enzyme is called ribose-5-phosphate isomerase and the main metabolite it creates is ribose 5- phosphate. Ribose-5-phosphate isomerase has been studied in two bacterial species, Escherichia coli and Mycobacterium tuberculosis. E. coli was one of the first model organisms employed by biological researchers and many of its main metabolic pathways have now been thoroughly characterised. One such pathway is that needed for utilising ribose.

Conveying ribose across the cell membrane Bacterial cells often use so-called ABC transporters as a means to actively carry small molecules through the hydrophobic cell membrane into the cyto- plasm. The three components of these transporters include a periplasmic binding protein, which has selective affinity to the recognised ligand, a membrane bound protein that is the actual carrier, and an ATP binding protein, which provides the uptake system with energy. For the sugar ribose, E. coli has a high affinity transport system encoded by the genes in the rbsDACBK operon, which has been located to 84 min on the E. coli genetic map (Iida et al., 1984; Lopilato et al., 1984). The actual transport is taken care of by the rbsACB gene products: RbsA a cytoplasmic ATPase, RbsC a membrane permease and RbsB the periplasmic ribose binding protein (Figure 1). RbsD is a mutarotase, that converts the pyranose form of ribose to the furanose form (Ryu et al., 2004). This is thought to help the last gene product of the operon, RbsK or ribokinase, whose job is to phosphorylate ribose. The structure of ribokinase shows that the enzyme binds the -furanose form of the sugar, a form less common in solution

1 compared to the - and -pyranoses (Sigrell et al., 1998). Once ribose is phosphorylated it is trapped within the cell due to the addition of a negative charge, which prevents it from leaking back out across the membrane. Ri- bose 5-phosphate (R5P) now enters the pentose phosphate pathway or is used for building cellular metabolites.

The Pentose Phosphate Pathway The pentose phosphate pathway (PPP), also called the hexose monophos- phate shunt or the phosphogluconate pathway, is needed to generate reduc- tive power in the form of NADPH, which can be used in the biosynthesis of for example fatty acids. This is the PPP’s main role, but it also provides the cell with ribose 5-phosphate, an important precursor of nucleotide synthesis, and a building block for producing the amino acids histidine and tryptophan. In the first part of the PPP, the so-called oxidative part, two molecules of NADPH are generated. The pathway starts with a molecule of glucose 6- phosphate (G6P) that via the action of the enzymes glucose-6-phosphate dehydrogenase (G6PD), 6-phospho-gluconolactonase (PGL) and phosphogluconate dehydrogenase (PGD) is transformed into a molecule of ribulose 5-phosphate (Ru5P) (Figure 1). In the second, non-oxidative part of the pathway, Ru5P is converted into R5P or xylulose 5-phosphate (Xu5P) by the enzymes ribose-5-phosphate isomerase (Rpi) and ribulose-5-phosphate 3- epimerase, (Rpe). These two five-carbon compounds are then the substrates of transketolase (Tkt), which rearranges them into the three-carbon containing glyceraldehyde 3-phosphate (GAP) and sedoheptulose 7-phosphate (S7P, seven carbons). These can then be made into erythrose 4-phosphate (E4P) and fructose 6-phosphate (F6P) by transaldolase (TAL). Further rearrange- ments by transketolase of one molecule of E4P together with another molecule of Xu5P gives a second molecule of F6P and one of GAP. In total, two molecules of Xu5P and one of R5P will generate two molecules of F6P and one GAP, which can enter into the glycolytic pathway. NADPH and R5P are the two most important products of the PPP. How- ever, it is also the only pathway that allows E. coli to break down five- carbon sugars such as ribose, xylose and arabinose (Sprenger, 1995). Xylose and arabinose are both rearranged and phosphorylated and enter the pathway as Xu5P. The other sugar phosphate moieties generated through the non- oxidative branch of the pathway are used as cellular building blocks in various contexts. E4P is a precursor of aromatic amino acids and vitamin B6, and S7P is used for building the cell wall. Depending on the current needs of the cell, the pathway will generate different end products. If a lot of reducing power is needed, F6P can be isomerised into G6P, which can re-enter the oxidative part of the PPP. If R5P is needed for DNA synthesis, F6P and GAP will be diverted from glycolysis and used to generate more R5P via TAL and Tkt.

Figure 1. An overview of ribose metabolism in E. coli focusing on the PPP. The top left corner shows how the gene products of the rbs operon transport D-ribose (simplified pyranose form) into the cell. The bottom corner shows how (a simplified) G6P molecule enters the PPP from the glycolytic pathway. The enzyme ribose- phosphate diphosphonokinase (RPDPK) uses an ATP molecule to produce 5- phosphoribosyl--pyrophosphate (PRPP) the important precursor of nucleotide biosynthesis. The other abbreviations are mentioned in the text.

3 Ribose-5-phosphate isomerase – introducing the main character Ribose-5-phosphate isomerase (EC 5.3.1.6) is an aldose-ketose isomerase that catalyses the conversion of ribulose 5-phosphate to ribose 5-phosphate and vice versa. Although not all organisms have a pentose phosphate pathway (Grochowski et al., 2005) it seems that Rpi is a global enzyme and that all organisms need to have a means of interconverting these two sugar phosphates. In plants, the conversion of R5P to Ru5P is a key reaction as this compound is necessary for carbon dioxide fixation. Ru5P is phosphorylated to form ribulose 1,5-bisphosphate, which in turn is the acceptor of CO2 in the first stage of the Calvin cycle. The Rpi reaction is presented in Figure 2. Briefly, two hydrogen atoms are shifted, resulting in molecules with double bonded oxygens in different positions: the aldose R5P or the ketose Ru5P. The isomerisation is thought to proceed via a cis-enediolate high energy intermediate, which is an unstable negatively charged compound. Calculations have suggested that stabilising this intermediate is the most important role of the enzyme, and the uncata- lysed reaction in water would be extremely slow (Feierberg and Åqvist, 2002).

Figure 2. The isomerisation catalysed by Rpis.

For the isomerisation to occur in the R5P to Ru5P direction, Rpi presumably first has to catalyse the opening of the furanose ring. In solution, R5P exists to a large extent as either - or -ribofuranose with only around 0.1% taking on the linear aldehyde form (Pierce et al., 1985). It would certainly be advantagous for the enzyme if it could assist in this step. This has, for example, been shown to be the case for phosphoglucose isomerase (PGI) that catalyses the interconversion of G6P and F6P (Schray et al., 1973). Once the ring is open, Rpi must provide a basic residue, which abstracts a proton from carbon number 2 of R5P, producing the enediolate. Secondly, the intermediate must be stabilised, for example with a positively charged amino acid. On the other side of the substrate there must further be an acid-

4 base group that can transfer a proton between oxygen 2 and oxygen 1. Fi- nally, the proton abstracted by the initial base will be released to carbon 1 of the product Ru5P. In the opposite direction, the steps are simply reversed. There are two distinct forms of Rpi known: RpiA and RpiB. These share no sequence identity and are presumed to represent an example of conver- gent evolution. RpiA is the most common of the two and can be found in species from all three kingdoms of life. RpiB is mainly found in bacterial genomes with some examples in archaea, protozoa and fungi. Some organisms have genes encoding both enzymes. This was first reported in the 1970s by David and Wiessmeyer whose results with crude bacterial extracts showed that there must be two forms of Rpi in E. coli (David and Wies- meyer, 1970). Skinner and Cooper later found that E. coli had one Rpi that is constitutively expressed (EcRpiA) and one that is formed only after induc- tion with ribose or compounds that can give rise to R5P (EcRpiB) (Skinner and Cooper, 1974). It was also noted that mutants lacking EcRpiA would rapidly revert from being dependant on ribose for survival. It was suggested that rpiB lies under the control of a repressor that easily succumbs to muta- tions, thus making EcRpiB constitutively expressed in such rpiA- mutants (Skinner, 1974). Further characterisation of the two enzymes was performed by Essenberg and Cooper, who determined the molecular weights of the two enzymes: 45 kDa for EcRpiA and around 33 kDa for EcRpiB. They also found that the B form but not the A form could be inhibited with iodoacetate, suggesting the involvement of a cysteine residue in the EcRpiB catalysis (Essenberg and Cooper, 1975). Both genes were later sequenced and mapped to the genome of E. coli (Skinner, 1974; Hove-Jensen and Maigaard, 1993; Sorensen and Hove-Jensen, 1996); from a comparison of the number of bases in each sequence to the observed native molecular weights, it was suggested that both enzymes exist as dimers. In the process, a gene for a repressor of rpiB gene expression, rpiR, was mapped to a locus just downstream of rpiB, at 92.8 min of the E. coli linkage map, but transcribed from the complementary strand. It was also confirmed that ribose induced rpiB gene transcription.

The structures of E. coli RpiA and RpiB The structure of EcRpiA was solved in 2003 to 1.5 Å resolution and revealed a dimer of subunits with an //(/)// fold (Zhang et al., 2003a) illustrated in Figure 3. Each subunit contains a complete active site located in a groove on the surface. In the same publication a structure of EcRpiA in a complex with the inhibitor arabinose 5-phosphate was described and used for assigning roles for the active site residues. A mechanism for the isomerisation was proposed involving residues Asp81, Glu103 and Lys94, where Glu103 is the main catalytic base and Lys94 is involved in stabilising the cis-enediolate intermediate. Of these three residues, the equivalents of Asp81

5 and Lys94 had previously been shown to be important for catalysis in the spinach RpiA (Jung et al., 2000). Upon substrate binding, the overall conformation of each subunit changes slightly to a so-called closed conformation involving a loop that protects the substrate from solvent.

Figure 3. The EcRpiA dimer (PDB code 1KS2 and 1O8B). The A molecule is coloured in light grey and the B molecule in darker grey. The N termini have been marked with the letter N while the C termini are hidden in this view. The inhibitor arabinose 5-phosphate, in black, indicates the location of each active site. This figure was prepared with MacPyMOL (DeLano, 2002) as was Figure 4.

The EcRpiB structure was also solved in 2003 by the same groups that solved EcRpiA (Zhang et al., 2003b). Their construct had an N-terminal histidine tag with a thrombin cleavage site that was not cleaved. Thirteen of the extra residues were seen in the electron density, starting with three of the six histidines. The asymmetric unit (AU) of the crystal contained a tetramer, and this was also the most common species when protein samples were analysed by dynamic light scattering. However, surface accessibility calculations suggested that the tetramer is best viewed as a dimer of dimers, with the long extra N-terminal segment promoting the tetramerisation (Figure 4), although the dimer-dimer interaction also includes residues of two other regions in each molecule (89-94 and 111-115). Each subunit has an fold, with a dimer topology totally different from that of EcRpiA. The dimer has two active sites, each located in a crevice between the two subunits and with contributing residues originating from both monomers. A search in the Protein Data Bank (PDB (Berman et al., 2000)) using the DALI server (Holm and Sander, 1993) revealed no structure with significant similarity to EcRpiB. However, sequence comparisons had previously noted that EcRpiB has a high sequence identity (30-40%) to the LacA and LacB subunits of galactose-6-phosphate isomerase from Staphylococcus mutans, which catalyses the inter-conversion of galactose 6-phosphate and tagatose 6-phosphate (Sorensen, 1996). The LacAB active site residues appear to be different from those of EcRpiB and inspection of these can be used for dis-

6 tinguishing between the two types of sequences (Zhang, 2003b). Interest- ingly, E. coli rpiA- cells that have not been induced with ribose were found to have Rpi activity when transformed with a plasmid containing the gene for S. mutans’ galactose-6-phosphate isomerase (Sorensen, 1996). It would appear as though this enzyme can accept both five- and six-carbon phosphate sugars as substrates. This raised the question as to whether or not the EcR- piB’s main role is to interconvert R5P and Ru5P or if it has a different preferred substrate, perhaps with six carbon atoms.

Figure 4. EcRpiB (PDB code 1NN4). The left panel shows the tetramer observed in the crystallographic AU. Helices are presented as cylinders for clarity, and -strands as arrows. The A, B, C and D molecules are marked with their respective letter and are coloured in different shades of grey. Molecules A and D form one dimer and B and C the other. The dimer-dimer interface of the A and C molecule is highlighted with an arrow. The right panel shows the proposed biological B-C dimer, with the same colouring as the left panel (B molecule in a darker grey) and helices presented as spirals. Both N-terminal ends are marked starting with residue -12 of the histidine tag. The view in the right panel is not the same as in the left.

The als operon of E. coli At the end of the 1990s two papers were published describing the als operon in E. coli (Poulsen et al., 1999; Kim et al., 1997). The gene products of this operon were found to be involved in the growth of E. coli when allose was the main carbon source. The operon, with the gene organisation alsRBACEK, Figure 5, was found to be regulated by alsR, which is identical to rpiR, earlier discussed as the regulator of rpiB gene expression (Sorensen, 1996). The rpiB gene lies just upstream of rpiR but is transcribed in the opposite direction compared to the other genes of the operon. Apart from alsR and rpiB the gene cluster includes alsBAC, encoding an ABC transporter system, consisting of a periplasmic binding protein (AlsB) a transmembrane protein (AlsC) and an ATP binding protein (AlsA). The

7 structure of the allose binding protein was solved in a complex with a molecule of D-allose (Chaudhuri et al., 1999) and the nature of the binding site suggested that this sugar was the optimal ligand. This ABC transport system was, together with AlsE (an epimerase) shown to be necessary for E. coli survival on allose minimal plates. AlsBAC further functions as a low affinity ribose transporter in E. coli cells lacking the rbsDACB genes (Kim, 1997). Cells grew on ribose minimal plates as long as the expression of ribokinase was intact. Iida et al. and Lopilato et al. had earlier suggested the existence of such a low affinity ribose uptake system. Both the all cis-pentose, D- ribose, and the all cis-hexose, D-allose, can induce the als operon (Poulsen, 1999). Finally, the operon encodes an allokinase, AlsK. Both Kim et al. (1997) and Poulsen et al. (1999) found this gene to be non-essential for growth on allose. But in an activity study, allokinase was found to phosphorylate allose, confirming its possible involvement in allose metabolism (Miller and Raines, 2005). The enzyme was also seen to phosphorylate glucose, although less efficiently than allose. Hexokinases are often able to phosphorylate several different sugars (Miller and Raines, 2004), which would explain why allokinase was found to be non-essential. EcRpiB, on the other hand, was deemed essential for growth on allose by Poulsen et al., who subsequently suggested the gene be renamed alsI in reference to its role as an allose-6- phosphate isomerase.

Figure 5. The als operon of E. coli.

The most puzzling aspect of the story is the fact that allose is a rare sugar, generally not thought of in terms of cellular metabolism. It would be expected that the relatively large commitment of a six gene operon out of E. coli’s ~4000 genes should reflect some real need in the organisms survival. Naturally occurring allose was first reported in the 1970s by Perold et al. who found it in two kinds of ester glucosides in leaves of Protea rubropilosa (Perold et al., 1973). Other substances such as flavonoids and iridoids containing D-allose have been isolated from the plants Veronica filiformis (Chari et al., 1981) and Mentzelia species (Jensen et al., 1981). The ketose D- allulose and related alcohol D-allitol have both been isolated from the leaves of plants of the species Itea (Hough and Stacey, 1966). Thus it would seem as though some plants are a natural source of allose, although the concentration is low and the sugar cannot be extracted from the leaves of these plants

8 in a productive manner. Allose used in laboratory studies is instead chemi- cally synthesised or made in vitro with the help of enzymes. One such enzyme is L-rhamnose isomerase from Pseudomonas stutzeri, which can convert D-allulose into D-allose (Menavuvu et al., 2006) in the test tube. This capacity was also noted in RpiB from Clostridium thermocellum (Park et al., 2007), where the authors claim, although without presenting results to support this statement, that the enzyme is more active with D-allose or D- allulose as substrate when compared to ribose 5-phosphate. Nonetheless, this is the first report of an RpiB with a wider substrate specificity range. Mammalian cell cultures have been found to phosphorylate D-allose in small amounts, but the sugar phosphate is not metabolised further or incor- porated into the glycolytic pathway (Ullrey and Kalckar, 1991). On the other hand, certain archea, for example Aerobacter aerogenes, do utilise allose and turn it into fructose 6-phosphate, which can be used in glycolysis (Gibbins and Simpson, 1964) (see Figure 1). The proposed mechanism is via the activity of an allose kinase, an allose-6-phosphate isomerase and an allulose 6- phosphate 3-epimerase, which finally converts allulose 6-phosphate into fructose 6-phosphate. The ability of A. aerogenes to ferment allose was induced by that sugar, thus analogies between this activity in the archaeon and the E. coli als operon can be drawn.

9 Mycobacterium tuberculosis Mycobacterium tuberculosis, is the causative agent of tuberculosis, a disease that The World Health Organization estimates caused 1.6 million deaths in 2005 (WHO, 2007). Each second someone in the world is infected by the bacterium. Of those infected only around 10% develop the disease; especially people with a weakend immune system are susceptible. Together with HIV this makes for a lethal combination and the highest number of deaths caused by tuberculosis are in Africa where HIV also is a large problem. There are four drugs that are generally used to treat tuberculosis and when taken correctly most patients are cured. However, unpleasant side-effects, the long time period for a complete treatment and an unreliable drug supply in certain countries can lead to people interupting their course of medication. This has led to the rise of drug-resistant and multi drug-resistant M. tuberculosis strains, the latter being defined as a strain resistant to the two most efficient tuberculosis medicines rifampicin and isoniazid. The need for new tuberculosis drugs is therefore dire, both to combat the drug-resistant strains that are on the increase and to shorten today’s lengthy course of treatment. An important step in the search for new compounds, which can be used in the fight against tuberculosis, is to learn more about the organism’s methods of survival. As mentioned in the first paragraph, E. coli has been widely studied during the past decades and a great deal is known about its different metabolic pathways. This is not the case for M. tuberculosis. Even though the bacterium was identified by Robert Koch as early as 1882, (Koch, 1882; Kaufmann and Schaible, 2005), it is a difficult bacterium to work with in the laboratory. This is mainly due to the infection risks associated with the mycobacterial cells but the growth rate of M. tuberculosis is also extremely slow, with a generation time of more than 24 hours. This can be compared to 20-40 minutes for laboratory E. coli strains. The sequencing of the M. tuberculosis H37Rv genome (Cole et al., 1998) has increased the level of understanding about the metabolic pathways represented in the bacterium, and structural and functional efforts to characterise the gene products are glob- ally underway (for example the TB Structural Genomics Consortium, http://www.doe-mbi.ucla.edu/TB/). However, many basic questions remain unanswered.

Sugar transport in M. tuberculosis For transport of sugars into the mycobacterial cell, the M. tuberculosis genome has to date been found to encode three different sugar ABC transport systems (Braibant et al., 2000). These are arranged in operons, which is generally the case for microbial ABC transporters. Of the three systems, one is highly similar to the glycerol 3-phosphate importer of E. coli. The genes of this operon are annotated as ugpA, ugpE, ugpB and ugpC (rv2835c-rv2832c)

10 (Cole, 1998) with UgpB as the substrate binding domain, UgpA and E span- ning the membrane and two molecules of UgpC binding ATP. The other two operons have a similar gene organisation, but the substrate specificity of their binding proteins is not known, as they are similar to several bacterial transporters of sugars such as maltose/maltodextrins, lactose, cello- biose/cellotriose and multiple sugars. The two operons comprise the genes rv2041c-rv2038c and sugC, sugB, sugA, lpqY (rv1235-rv1238). In addition to these three operons, a fourth incomplete gene cluster was found, which is lacking a gene for an ATP binding component (rv2316-rv2318, or uspA, uspE, and uspC). This operon is thought to either be inactive or to utilise one of the ATP binding domains of the other operons, as this domain is not involved in substrate specificity. The sugar transport systems identified by Braibant et al. all belong to the ABC transport sub-family 5, involved in carbohydrate transport (Linton and Higgins, 1998). No M. tuberculosis transport system belonging to sub-family 1, which is utilised for monosaccharide transport in E. coli and to which family the rbsDACBK operon belongs, could be detected. Other specific ABC cassettes of the same family that are found in E. coli and are lacking in M. tuberculosis are those used for L-arabinose, D-galactose and D-xylose transport. Either D-ribose and these other monosacharides are exclusively recycled from other metabolites already present within the cell, or one of the above-identified transporters has low substrate specificity and can bind and transport several sugars, or not all transporter systems have been identified in the bacterium.

The pentose phosphate pathway of M. tuberculosis Studies of M. tuberculosis carbohydrate utilisation were undertaken in the 1960s and revealed a functioning glycolysis and a pentose phosphate pathway with all the necessary enzymes present (Ramakrishnan et al., 1972). All these genes have later been annotated in TubercuList† (Cole, 1998). For a few of the steps more than one gene was annotated as encoding the relevant enzyme. M. tuberculosis, for example, has two genes designated as glucose 6-phosphate dehydrogenases (G6PD in Figure 1) (rv1121 and rv1447c) and two as 6-phosphogluconate dehydrogenases (PGD in Figure 1) (rv1844c and rv1122) (Wheeler and Blanchard, 2005). Both are involved in the first few steps of the pentose phosphate pathway where glucose 6-phosphate is converted into ribulose 5-phosphate generating two molecules of NADPH. The gene encoding M. tuberculosis’ ribose-5-phosphate isomerase was annotated as rv2465c and belongs to the B-type of Rpi enzymes.

† http:/genolist.pasteur.fr/TubercuList/

11 General considerations in designing new antitubercular drugs A general rule when searching for proteins to target for designing new medical compounds that will act against a bacterium is to find metabolic pathways, which are essential to the bacteria’s survival. Another desired feature is that the targeted enzyme either is not present in humans or that it is con- siderably different compared to the human form. This makes it possible to design a drug compound that can bind selectively to the bacterial enzyme and not cause any unwanted side effects. For testing whether or not an enzyme is essential to the bacteria, knockout studies are often undertaken. Briefly, the gene coding for the protein of interest is disrupted and the growth of the bacteria is monitored under different growing conditions. The best scenario is one in which the bacteria do not grow at all, even in a rich growth medium containing all possible nutrients. But often when genes involved in different metabolic pathways are disrupted, the growth studies are performed in minimal medium supplemented with various key nutrients.

Rpi essentiality studies in bacteria and fungi For Rpi, two knockout studies have been performed. The first was on E. coli strain K12, where a double mutant lacking both EcRpiA and EcRpiB showed severely impaired growth under all conditions tested (Sorensen, 1996). In the study it was noted that in order for the double mutant to survive it needed ribose as a carbon source but that just ribose was not enough. The bacteria also needed to have the growth medium supplemented with a compound that can be converted to xylulose 5-phosphate, for example glucose or xylose. In these mutant bacteria, ribokinase will phosphorylate ribose to produce R5P, and xylulose 5-phosphate together with R5P can be used in the non-oxidative part of the PPP for generating fructose 6-phosphate for incorporation into glycolysis (Figure 1). Thus, in the absence of an Rpi, the bacterium will find a means to survive, as long as it can obtain the right food, and tolerate a lower growth rate. Saccharomyces cerevisiae (baker’s yeast) has only the gene for RpiA in its genome. When a haploid rpiA deletion was constructed (Miosga and Zimmermann, 1996) the authors found, to their surprise, that the phenotype was lethal when the cells were grown on synthetic mineral-salts medium (0.5% ammonium sulphate, 0.17% Difco yeast nitrogen base and 2% glucose) supplemented with amino acids, adenine and uracil. Addition of a plasmid containing rpiA reverted the phenotype back to wild type. This finding is clearly contradictory to the results from E. coli, considering that the yeast cells had access to glucose to produce Ru5P and adenine and uracil, which ought to be converted to R5P via salvage pathways. As pointed out by

12 the authors, this knockout strain should only have had trouble surviving with just ribose or just xylose as a carbon source. In the E. coli knockout study, it was noted that the double mutant showed impaired growth even in rich medium. If extra ribose was added the growth was even more severely inhibited. It was implied that an accumulation of R5P could be the reason for this inhibition. Methylglyoxal, a toxic and highly reactive metabolic intermediate, has been associated with cell death in E. coli when the ribose transport system has been disturbed (Kim et al., 2004). When the gene for rbsD, the first gene in the operon for ribose transport, was overproduced on a plasmid, growth inhibition was seen shortly after an addition of 0.2% ribose. This effect was only seen when the gene for ribokinase was also overproduced on a second plasmid. As mentioned previously, RbsD is a mutarotase, which converts the pyranose form of ribose to the furanose form (Ryu, 2004), the preferred substrate of ribokinase (Sigrell, 1998) for producing R5P. This would imply that an overproduction of ribose 5-phosphate is harmful for the cells because it causes a build up of methylglyoxal. When studying the als operon in E. coli Poulsen et al. saw that mutant strains lacking the genes for AlsI and AlsE, the isomerase and epimerase in the operon (see Figure 5), were inhibited when allose was added to the growth medium. In this case an accumulation of either or both of allose 6- phosphate and allulose 6-phosphate could be causing the lowering in cell viability.

A human Rpi mutant causing abnormal brain development The human genome contains a gene for RpiA but none for RpiB or any other homologous Rpi. A medical study from 2004 shows that a patient suffering from brain disorders due to damaged myelin sheaths and abnormal cerebral white matter had an inborn mutation in rpiA (Huck et al., 2004). The patient was found to have high concentrations of ribitol and D-arabitol in all body fluids, which caused the authors to investigate the enzymes in the patient’s pentose phosphate pathway. Results demonstrated that Rpi was dysfunc- tional and after sequencing the gene it became clear that there was a single base pair deletion, which caused a frameshift after residue 180 and a truncated protein. Residue 182, the second affected amino acid after the deletion, is the proposed active site base (Glu103 in E. coli), responsible for shuttling a proton between C1 and C2 of the substrate (Zhang, 2003a). A deletion in this region would certainly be detrimental, but just a truncated protein would also be expected to cause problems for normal enzymatic activity. The abnormal cell development seen in the brain was thought to be due to a build- up of pentoses and pentose phosphates, leading to an accumulation of the metabolic end products ribitol and D-arabitol.

13 Rpi – a possible drug target? A lethal knockout, or an essential gene, cannot be rescued by adding end products of the biosynthetic pathway to the medium. It can only be rescued if an active copy of the gene is introduced into the organism. Yeast rpiA ought therefore to be termed as essential, but not the combination rpiAB in E. coli. A knockout study in M. tuberculsis has to date not been performed and it is not possible to make speculations about the gene’s essentiallity in mycobacteria. Since it would seem possible that a bacterium will soon find other means of producing R5P if the enzyme is inactivated, the bacterial Rpi may not be the obvious drug target. Even so, the above studies point toward an important role for Rpi in an organism’s struggle to keep the correct bal- ance of different nutrients in the cell. To target the RpiB found, for example, in M. tuberculosis and find inhibitors specific for this form of the enzyme could be an approach to finding a new medicine for treating tuberculosis. Unpleasant side effects such as that observed with the human mutation certainly ought not to be a problem; as judging from the extensive differences in the active sites of EcRpiA and EcRpiB, it should be possible to design compounds that only bind to RpiB. In the pathogen Trypanosoma cruzei, which only has the RpiB form of the enzyme, it has been suggested that this may be a possible drug target (Stern et al., 2007). In M. tuberculosis a specific transport system for ribose has not been identified. Even though the organism does have a gene for ribokinase (rv2436) and it may be presumed that one of the ABC transporters can take ribose, RpiB must be one of the main routes for obtaining ribose 5- phosphate.

14 Current investigation

Aim of thesis An M. tuberculosis gene for ribose-5-phosphate isomerase (rv2465c) was discovered during the sequencing of the whole genome (Cole, 1998) and was noted to represent an RpiB form of the enzyme. The sequence identity to the EcRpiB is around 30% and many of the conserved residues among other species are also conserved in MtRpiB. However, one remarkable discrepancy was noted. The catalytic base in the E. coli enzyme, Cys66, is not conserved, and what is more noticeable is that it is exchanged for a glycine in the M. tuberculosis sequence. As a glycine cannot act as a base in catalysis the question was raised whether or not this was the Rpi of M. tuberculosis, and if the gene product did have this activity, what was its catalytic mechanism? This thesis work aimed to answer these questions and in the process find out as much as possible about the supposed MtRpiB. EcRpiB has, as was described in the Introduction, been postulated to have a second activity as an allose-6-phosphate isomerase, and would, if so, catalyse the reaction presented in Figure 6. As E. coli generally does not use this protein, it was proposed that this second activity in fact could be the enzyme’s main role in the cell, enabling survival under particular circum- stances. No biochemical evidence for this theory had been put forward, only the observation of the EcRpiB gene’s close proximity to the als operon and the gene’s essentiality when E. coli cells were grown with allose as the only carbon source. The current investigation therefore aimed to consider this second activity of EcRpiB from a structural and biochemical angle, and to determine whether other RpiBs could accept allose 6-phosphate as a substrate.

Figure 6. The isomerisation of allose 6-phosphate and allulose 6-phosphate, proposed to be catalysed by EcRpiB.

15 Introductory sequence comparisons and genomics searches Bioinformatics studies to determine the distribution of RpiA and RpiB in different organisms were performed for Paper I in 2003 and for paper IV in 2007. Many organisms, mainly bacteria, have a gene for both RpiA and RpiB. There are also a significant number with only the rpiB or only the rpiA sequence in their genome, but to date every completely sequenced genome has one or the other. The RpiB sequences can be divided into at least two sub-families, with the MtRpiB as a representative of one group and EcRpiB representing the other. The most striking difference between the two families is the proposed catalytic base, Cys66 of E. coli, which is replaced by glycine in the Mt-like RpiBs. Other residues lining the active site of EcRpiB are conserved over all organisms, for example Asp9, His10, Tyr43, His99, Asn103, Arg133, His134 and Arg137 (Figure 7). In the original (2003) search the MtRpiB sequence uncovered only seven other similar sequences when used in BLAST (Altschul et al., 1997). The E. coli sequence identified many more. In the more recent search (May 2007) the number of sequences related to MtRpiB was 29, which reflects the rapid increase in the number of genomes that have been sequenced in the past four years (777 bacterial, 45 archaeal and 48 eukaryotic species with completed genomes, where protein sequences have been identified could be screened at the time of writing). The sequence alignment seen in Figure 7 is a selection of sequences that resemble MtRpiB and ones that are more similar to EcRpiB. Organisms where the RpiB has been characterised in some way have also been included. The Mt-like RpiB is to date mainly found in the actinobacterial phylum, while the Ec-like enzyme appears scattered through all the other bacterial families. In archaea, Thermofilum pendens is the only representative with a sequenced genome that has an RpiB (Ec-like). All others only have an RpiA. Unfortunately no Rpi sequences for the archaeal species Aerobacter aerogenes that Gibbins and Simpson (1964) found could metabolise D-allose can be located through BLAST searches; it would seem as though not many of this organism’s genes have been sequenced. Most eukaryotes have only an RpiA, although there are some examples of fungal species that also have an EcRpiB. Entamoeba histolytica, Trypano- soma species and Giardia lamblia are eukaryotes that only have an Ec-like RpiB. Some organisms within the actinobacteria together with Magneto- spirillum magnetotacticum, Robiginitalea biformata, Blastopirellula marina have both an Mt- and Ec-like RpiB, but no RpiA. There are only two organisms available that have an MtRpiB together with an RpiA: Rhodococcus sp. RHA1 and Reinekea sp. MED297.

Figure 7. Sequence alignment of a symbolic group of RpiBs, created with ClustalW (Thompson et al., 1994) and BioEdit (Hall, 1999). Identical residues are highlighted in dark grey and similar residues in light grey. The GenBank identities for most sequences are found by Figure 5 in Paper I. Others are: Acidothermus cellulolyticus 11B (gi:117928805), Propionibacterium acnes KPA171202 (gi:50843092), Try- panosoma cruzi (gi:70876518), Clostridium thermocellum ATCC 27405 (gi:125975080), Haemophilus somnus 2336 (gi:46155947), Yersinia intermedia ATCC 29909 (gi:77979015).

17 Functional studies of MtRpiB and EcRpiB

Assays used in this thesis An enzyme can be characterised in many ways. In a structural biology laboratory the method of choice is normally X-ray crystallography to determine what structural features the enzyme of interest has. These features can give many clues to the function of the enzyme but it is also necessary to do studies that link this to the activity of the protein. In this work three different assays have been employed for analysing the activity of RpiB. The main method has been a continuous spectrophotometric assay developed in the 1970s (Wood, 1970). Here the conversion of R5P to Ru5P is monitored directly as a change in absorbance at 290 nm when Ru5P is produced, with the ketose group of the product being responsible for the absorbance change and the substrate R5P having no absorbance at this wavelength. The extinction coefficient for the product Ru5P is 72 cm-1M-1 (Wood, 1970) and so Lambert-Beer’s law can be used for calculating the amount of product. As the reaction catalysed by Rpi is an isomerisation that involves a substrate and product with nearly equal energies, it does not go to completion and so it could be measured in both directions. Due to the difference in cost for the two substrates R5P was chosen and the production of Ru5P monitored instead of the reverse. The assays were performed at room temperature, typically 20-22 degrees, in a 1 mL assay mixture buffered with 50 mM Tris- HCl pH 7.5. For the EcRpiB 5 mM 2-mercaptoethanesulfonic acid (MESNA) was added as a reducing agent for the active site cysteine. These conditions were kept the same for the other assays as well. The spectrophotometer used for the measurements has no software for analysing the initial rate of each reaction automatically. Therefore the curves were inspected both on the spectrophotometer’s computer screen and by plotting the measurements using Microsoft Excel. The initial rates were then estimated by judging by eye how many seconds the slope was linear and calculating the tangent to the line accordingly. The reactions were typically allowed to proceed for five minutes, i.e. beyond the linear region in most cases, to make sure the actual shape of the curve was seen properly. The second assay used was a colorimetric, discontinuous assay adapted from the method of Meredith and Woodard (Dische and Borenfreund, 1951; Meredith and Woodard, 2003). Here the reaction was performed under the same conditions as previously and stopped by adding concentrated acid at different time points. A carbazole-cysteine solution was then added to the reaction mixtures, which together with the product Ru5P forms a purple colour. The colour was allowed to develop for three hours and the absorbance was measured at 540 nm. This assay was adapted to be used in a 96- well format, making it more convenient for screening of potential inhibitory compounds.

18 The third assay was developed by our collaborators in France for investi- gating the proposed allose-6-phosphate isomerase activity of the E. coli and M. tuberculosis RpiBs. Briefly, the method utilized the reaction of thiobarbi- turic acid (TBA) with the product allulose 6-phosphate (Allu6P), giving rise to an intensely yellow compound, which can be measured at 438 nm. The spectrophotometric assay was discontinuous and was stopped by the addition of concentrated acid, after the reaction had been allowed to proceed for 2 minutes; found to be within the initial rate phase of the reaction. This method is explained in more detail in Paper IV.

Interpreting the assay data The assay data were first examined by plotting the initial rates, v, on the y- axis and the corresponding substrate concentration, [S], on the x-axis, i.e. a classical Michaelis-Menten plot. This was mainly as an evaluation of the rates and the different substrate concentrations that were chosen, to see whether any extra points had to be included or remeasured. The kinetic constants were generally not estimated from fitting the Michaelis-Menten equation (1) to the data points.

V maxS v (1) Km S

Instead the data were plotted in a Hanes-Woolf plot with [S]/v against [S], which is considered the best method for visualising the data linearly (Cornish-Bowden, 2004) as all errors will be evenly distributed. In such a plot the slope is 1/Vmax, the y-axis intercept is Km/Vmax and the x-intercept value is –Km.

Inhibition studies Potential inhibitor compounds were generally tested at a concentration of 25- 100 M, if necessary up to 10 mM, to conclude if they had an effect on the activity. If so, more concentrations would be tested to pinpoint the range where the activity of the enzyme was lower but still measurable. For calculating Ki values for inhibitors, a series of measurements with varied compound concentrations was carried out at different substrate concentrations. Apparent Km values were plotted against the different inhibitor concentrations, and the slope of the line, equal to Km/Ki, used for calculating the Ki. Inhibitors were assumed to be competitive with respect to the substrate until activity measurements gave an indication that the case was of otherwise. To make sure that a clear difference in the reaction rate could be detected when compounds were added, the substrate concentration was usually 5 mM. A concentration slightly higher than the Km meant that rates were within the range most easily measured, but that weakly binding inhibitors could still be identified.

19 For compounds that showed a weak inhibitory capacity, only IC50 values were determined. In some cases, an approximate Ki was calculated from

IC50 Ki [S] (2) 1 Km assuming Kd ~ Km and competitive inhibition.

Assay results of MtRpiB Initial measurements of Rpi activity When looking just at the sequence of the proposed MtRpiB it was not clear whether or not the gene product of rv2465c would have the ribose-5- phosphate isomerase activity. However, activity measurements with MtRpiB using the Wood assay soon showed that the enzyme could catalyse the reaction (Paper I). Initial tests were performed under reducing conditions with 5 mM MESNA present, as this was found to be necessary for EcRpiB (Zhang, 2003b), but it could be concluded that MtRpiB did not need this environment for optimal activity. A solution buffered at pH 7.5, a temperature of 37°C and a protein concentration of 46 nM was used for the first characterisation, where kinetic constants were calculated from a Hanes-Woolf plot presented -1 in Figure 8 resulting in a Km of 3.7 mM and a kcat of 120 s for R5P. Fitting the Michaelis-Menten equation (equation 1) to the data (inserted graph of Figure 8) gives a slightly higher Km of 3.8 and the same kcat. Despite the difference in the active site, these values are very similar to those obtained -1 for the E. coli enzyme, where the Km was 1.2 mM and kcat 70 s . EcRpiA, on the other hand, has a strikingly different catalytic activity with a kcat of around 2000 s-1, even though the apparent affinity for the substrate is similar, with a Km of 3 mM. Under most conditions, the slow rate of RpiB will probably not be a serious problem for E. coli, as this enzyme normally is not needed or expressed; RpiA is quite adequate for the task. In M. tuberculosis there is no RpiA, but here the low activity of RpiB most likely will also not cause problems, because the organism is extremely slow growing.

Figure 8. The first enzymatic characterisation of MtRpiB (Paper I). Data points are plotted according to the method of Hanes and Woolf where the linear fit to the data was used to calculate kcat and Km. Inserted are the same data points presented with a curve fitted using the Michaelis-Menten equation.

Searching for inhibitors of the MtRpiB reaction A first search for inhibitors utilised compounds chosen because of their re- semblance to the substrate. These included fructose-6-phosphate, D- erythrose, D-4-erythronic acid and D-3-phosphoglyceric acid. None of these four compounds had a measurable effect on the activity of the M. tuberculosis enzyme at a concentration of 10 mM (D-4-erythronic acid was also tested at 100 mM without effect). Iodoacetate had earlier been found to inhibit EcRpiB (Essenberg, 1975) probably due to a direct modification of Cys66. When MtRpiB was incubated with 10 mM iodoacetate for 10 minutes before the assay was started the enzyme showed no activity (although adding iodoacetate to the reaction mixture and starting the reaction without preincuba- tion had little effect on enzyme activity). Since there is no cysteine present in the active site the compound may react with one of the active site histidines, or perhaps one of the cysteines that are present in the sequence, resulting in an unfolded protein. Clear density for a phosphate molecule was found in the active site of the first structure of MtRpiB. Thus, sodium phosphate was added to the activity assay measurements and an IC50 value of approximately 300 mM was determined. Using equation 2, a Ki of 130 mM was calculated.

21 Tentative inhibitors derived from a structure based virtual screen Through our collaboration with the Medicinal Chemistry group at Uppsala University a structure based virtual screening study of MtRpiB was performed, using the first MtRpiB structure, which has a phosphate molecule in the active site (Paper I), to find potential new inhibitors of the enzyme (un- published results). The protein was allowed no flexibility and no water was included in the docking. A first set of 32 compounds, obtained from the screening results, was purchased from various manufacturers and analysed for their inhibitory capacity using the Wood assay. Compounds were added to the reaction at concentrations ranging between 25 and 100 M. Unfortu- nately, none of these compounds showed any inhibitory effect, although for a large fraction it was difficult to measure a rate at higher concentrations because the background absorbance at 290 nm was too high. Some of the compounds also exhibited solubility problems and interfered with the assay due to precipitation. For this reason the carbazole-cysteine assay was utilised. Compounds could be screened faster and the problem with background absorbance was reduced with a reading at 540 nm. Sixteen new compounds were purchased based on a new virtual screen, that had the MtRpiB-4PEH structure (Paper II) as a starting model, and all 48 compounds were analysed for possible inhibition. Compounds that were not water-soluble were dissolved in DMSO. Of these, nine had to be eliminated from the assay as they precipi- tated upon addition to the reaction mixture. With this method, one compound was found to be a potential inhibitor with a preliminary IC50 value of 0.2 mM. Further measurements with varying substrate concentrations and inhibitor concentrations gave similar apparent Km values whereas differences could be detected in Vmax. These results indicated that the compound was a non-competitive inhibitor of MtRpiB, but it could later be shown that the compound interfered with the colour formation of the assay, and presumably not with the enzymatic activity. With the UV-assay no inhibition was seen, although this compound had a high background absorbance making the results nontrivial to evaluate. Taken together and due to a lack of time, pursuing this compound to a greater extent was considered to be of low priority and no further analysis was performed.

Transition state analog inhibitors A collaboration with Dr Laurent Salmon at the Department of Organic and Inorganic Chemistry at Universite Paris-Sud started in 2003. Compounds designed to resemble the high-energy intermediate of the R5P to Ru5P isomerisation (Figure 2) were synthesised and can be seen in Figure 9. These were evaluated for their ability to function as competitive inhibitors on MtRpiB using the Wood assay (results of Paper II).

Figure 9. Compounds designed as transition state analogs of the R5P to Ru5P isomerisation.

The substance 4-deoxy-4-phospho-D-erythronohydroxamic acid (4PEH) inhibited MtRpiB the best of the five compounds with a Ki of 57 M, see Table 1 for inhibition data. Second best was 4-deoxy-4-phospho-D- erythronate (4PEA) with a Ki of 1.7 mM. Hanes-Woolf plots of the data used for calculating apparent Km values for different inhibitor concentrations and the plots used for estimating the Ki values are seen in Figure 10. Both of these compounds had previously been shown to inhibit spinach RpiA with very similar values, 28 and 29 M respectively for 4PEH and 4PEA (Burgos and Salmon, 2004a). The two compounds were cocrystallised with MtRpiB and the resulting complex structures could explain the difference in binding affinity, which will be discussed on pages 37-38. The other three compounds 4-deoxy-4-phospho-D-erythronamide (4PEAm), 4-deoxy-4-phospho-D-erythronhydrazide (4PEHz) and 4-deoxy-4- phosphonomethyl-D-erythronate (4PMEA) were weaker inhibitors, 4PEAm and 4PEHz so weak that even concentrations of 30 mM did not bring the reaction rate down to half of the original value. For these two, IC50 values were said to be larger than 30 mM and no further measurements were performed. 4PMEA was designed to be a more stable analog of 4PEA due to the oxygen connecting the phosphate group to carbon 4 being exchanged for a methylene group. An IC50 was estimated for this compound and when an approximate Ki was calculated from equation 2, the value, 2 mM, was very close to the Ki of 4PEA.

Table 1. Inhibition data for the transition state analog compounds. MtRpiB Spinach RpiA a Compound Ki (mM) IC50 (mM) Ki (mM) IC50 (mM) 4PEH 0.057 0.04 0.029 0.018 4PEA 1.7 6 0.028 0.01 4PMEA 2b 12 0.074 0.11 4PEAm - >30 2.5 1.8 4PEHz - >30 1.8 1.8 5PRH 0.4 1.5b 6.2 5.0 a. Determined in (Burgos, 2004a; Burgos and Salmon, 2004b) except for 5PRH determined in Paper III. b. Calculated from equation 2.

Figure 10. Inhibition study of MtRpiB with the three inhibitors 4PEH, 4PEA, and 5PRH. To the left are Hanes-Woolf plots of the data points for experiments at different inhibitor concentrations and to the right the apparent Km values plotted against the inhibitor concentration for Ki estimations. A. 4PEH: 0.05 mM 4PEH, Km(app) = 4.1 mM; 0.01 mM 4PEH, Km(app) = 2.8 mM; 0 mM 4PEH, Km = 2.1 mM. B. The slope of the line is equal to Km / Ki. Ki (4PEH) = 57 M. C. 4PEA: 4 mM 4PEA, Km(app) = 5.1 mM; 2 mM 4PEA, Km(app) = 2.9 mM; 0 mM 4PEA, Km = 1.5 mM. D. Ki (4PEA) = 1.7 mM. E. 5PRH: 0.8 mM 5PRH, Km(app) = 5.2 mM 0.4 mM 5PRH, Km(app) = 4.0 mM; 0.2 mM 5PRH, Km(app) = 2.4 mM; 0 mM 5PRH, Km = 1.8 mM. F. Ki (5PRH) = 0.4 mM.

The compound 5-deoxy-5-phospho-D-ribonohydroxamic acid (5PRH) was designed as a mimic of the 1,2-cis-enediolate intermediate of the allose-6- phosphate isomerase reaction, Figure 6, and can be seen in Figure 11. It was evaluated as an inhibitor of the MtRpiB R5P to Ru5P reaction and inhibited the enzyme with a Ki of 0.4 mM (results of Paper III). Data are shown in Figure 10 and Table 1. In the same paper data were presented for 5PRH as a weak inhibitor of the spinach RpiA isomerase reaction using the same assay.

24 For this enzyme 5PRH only had a Ki of 6.2 mM. As such, this was the first reporting of a compound that selectively inhibited the B form of the Rpi enzyme compared to the A form.

Figure 11. Transition state analog inhibitor of the All6P to Allu6P reaction, 5PRH, compared to 4PEH.

Assay results of EcRpiB Mutants EcRpiB was used for teaching purposes in the Swedish University of Agri- cultural Sciences course in Protein Technology. Site-directed mutagenesis was performed by the students in two years, 2004 and 2005. The two mutants used for structural and activity studies were first verified by sequencing. These represented proteins where Cys66 was mutated to a serine (C66S) and where His99 was changed to an aspargine (H99N). The mutant C66S was designed to be inactive, as this cysteine is thought to be the most important catalytic residue (Zhang, 2003b). Activity measurements were initially performed during the course using the Wood assay, with results pointing toward a totally inactive enzyme. These results were confirmed in the present work, thus this cysteine must play a central role in catalysis. In a mutational study of the Trypanosoma cruzei RpiB the equivalent cysteine was mutated to an alanine and no activity could be measured for this mutant either (Stern, 2007). The H99N mutant, on the other hand, was designed to be deficient in opening the ring of the furanose form of R5P. From our structures with MtRpiB (Paper II) we proposed that His102 in MtRpiB (His99 in EcRpiB) could perform this initial step of the reaction, if it is required for the RpiB enzyme. A previous study of spinach RpiA proposed that to maintain a physiological Km for the R5P substrate, determined to 0.63 mM in their case, -1 and with the obtained kcat of 3400 s , RpiA must be able to catalyse ring- opening (Jung, 2000), considering the small fraction of linear aldehyde that is present in solution. The EcRpiA structure with arabinose 5-phosphate, also showed that a five-membered ring easily was accommodated in the active site and residues proposed to be involved in opening the ring were

25 suggested (Zhang, 2003a). Due to the lower activity of the RpiB enzyme it was not clear whether or not this enzyme would need to catalyse this first step. The H99N mutant was assayed in the forward direction of the reaction -1 where R5P is isomerised into Ru5P and was found to have a kcat of only 2 s , -1 compared to 50 s for the wildtype. The Km values for the substrate were on the other hand very similar; 0.8 mM for the H99N mutant compared to 1 mM for the wild type enzyme. In the T. cruzi RpiB, the equivalent histidine was changed to an alanine by site-directed mutagenesis (Stern, 2007). When activity was measured in the forward reaction, the Km was maintained close to the wild-type enzyme’s but the kcat was reduced 10-fold. In the reverse reaction, Ru5P to R5P, both Km and kcat were similar to the wild type values. Since the mutant enzyme functions normally when it has access to the linear substrate Ru5P, but is slowed down when given the ring form R5P, it was concluded that the histidine is important for opening the ring.

Allose-6-phosphate isomerase activity (Paper IV) Both Ec and MtRpiB were analysed for their ability to isomerise All6P to Allu6P using the TBA assay. These experiments were, as mentioned previously, performed by our collaborators in France. Activity measurements using the Wood assay were carried out in parallel, to ensure the enzyme had not become inactive during transportation or storage. The kinetic parameters derived from the activity measurements are presented in Table 2. It could be shown that EcRpiB can catalyse the isomerisation of All6P to Allu6P with a kcat around ten times lower than for the R5P isomerisation. The Km for the substrate was however lower for the six-carbon sugar phosphate, suggesting higher affinity. MtRpiB, on the other hand, gave poor values for both Km and kcat when assayed for the All6P activity. All6P, as well as Allu6P, were instead analysed as inhibitors of MtRpiB when it is catalysing the R5P to Ru5P reaction, with resulting IC50 values of 2 and 6 mM (Table 2). The substrate concentration used in these experiments was 1 mM.

Table 2. Kinetic parameters for EcRpiB and MtRpiB presented in Paper IV. -1 -1 -1 Substrate Km (mM) kcat (s ) kcat/Km (s mM ) IC50 (mM) EcRpiB R5P 1.1 0.2 52 2 47 9 - All6P 0.5 0.2 6 0.3 12 5 nda MtRpiB R5P 1.0 0.4 47 2 50 20 - All6P 16 2 ~0.22 ~0.014 2 0.4b Allu6P nd nd - 6.3 0.4b a. Not determined. b. Inhibitor of the R5P to Ru5P reaction.

26 Crystallisation Crystallisation of MtRpiB In the first 96-well screens set up with MtRpiB crystals appeared only in conditions containing phosphate, not so surprising as the enzyme naturally binds ribose 5-phosphate and ribulose 5-phosphate. Therefore, the first structure that we solved of MtRpiB showed clear density in the active site for a bound phosphate molecule. This crystal was taken from a sitting drop in a 96-well plate with well conditions including 1.26 M NaKPO4 pH 7.5, although at the time optimisation screens had been set up with varying conditions of ammonium phosphate at 1.4-1.8 M. The best crystals tended to appear after about a week in ~1.6 M ammonium phosphate with either 100 mM HEPES, pH 7.5, or MES, pH 6. The protein concentration was for the initial crystallisation experiments, ~20 mg/ml (0.58 mM, Mw = 34.6 kDa) but was later found to be optimal between 7 and 12 mg/ml (0.2-0.35 mM). The two complex structures described in paper II, with 4PEH and 4PEA, were crystallised in the above-mentioned ammonium phosphate conditions, which resulted in a lower occupancy for the 4PEA ligand in the active site. Even though the 4PEA concentration was fairly high (50 mM), its Ki is also high (1.7 mM), and so the 1.65 M of phosphate in the crystallisation drops was sufficient to compete successfully for the phosphate-binding site. To avoid this competition from phosphate, and make it easier to analyse the complexes of weaker inhibitors, new 96-well screens were set up, looking for conditions that would yield crystals without phosphate present. With R5P, 5PRH or All6P, present in the crystallisation experiments crystals hap- pily grew without phosphate (Paper IV). However, it was more difficult to find conditions where good crystals grew without anything present in the active site, which would enable soaking experiments. The general problem with the MtRpiB crystals is that in every condition tested, any crystals are long and thin, like needles or, at best, like rods (Figure 12). With a ligand present it was always possible to find a few crystals that had grown more three-dimensionally, but without anything bound this was a problem. Seed- ing often helped matters slightly, as lower concentrations of both precipitant and protein could be used with crystals still being produced. Lower concentrations decreases the number of crystal nucleation sites, and slows down crystal formation, enabling the crystals that do start growing from a seeded nucleation point to grow larger. Still, not many good quality crystals could be produced without adding a ligand. Therefore, soaking experiments were not easily performed and all ligands had to be cocrystallised to obtain complex structures. In the case of the ligand 5PRH (paper IV), soaking would have been a desirable option. This is explained in more detail later. A caveat is that, in some cases, soaking can be a less reliable technique compared to cocrystallisation, as examples have shown that ligands can bind differently when soaked into the crystal compared to when cocrystallised with the pro-

27 tein. The binding could be an artefact due to unnatural protein contacts or conformations caused by the crystal packing (Danley, 2006). Crystallisation conditions of the different MtRpiB crystals with various ligands can be seen in Table 3.

Table 3. MtRpiB crystal structures. All crystals were of needle- or rod-like morphology and belonged to the space group C2. MtRpiB Crystallisation [Protein], Reso- Density in Paper, ligand condition [ligand] lution active site PDB (mM) (Å) code

Pi 1.26 M NaKPO4 pH 7.5 0.58, 1260 1.9 Pi I, 1USL

4PEH 1.55 M Ammonium 0.49, 3 2.1 4PEH II, phosphate 2BES 0.1 M HEPES pH 7.5

4PEA 1.65 M Ammonium 0.35, 50 2.2 4PEA II, phosphate (+ Pi) 2BET 0.1 M MES pH 6

4PMEA 1.6 M Ammonium 0.52, 25 2.2 Broken unpub- a phosphate density, Pi lished 0.1 M MES pH 6 seen

R5P 20% PEG 3K 0.22, 2 1.65 A mixture IV 0.1 M Tris pH7 of R5P and 0.2 M Ca acetate Ru5P

5PRH 15% PEG 8K, 0.1 M MES 0.35, 25 2.0 Hydrolysed IV (5PRA) pH 6, 5% PEG 1K, 5RPH = 0.2 M Li2SO4 5PRA

All6P 1M tri-Na citrate 0.35, 10 1.85 -pyranose IV 0.1 M Na cacodylate pH 6.5 All6P a. This structure will not be deposited as the density in the active site is too badly defined.

Figure 12. Typical MtRpiB crystals. To the left is a photograph of a crystallisation drop with MtRpiB crystallised without any ligand. The whole drop has a diameter of about 1.5 mm. To the right is a crystal of MtRpiB cocrystallised with All6P mounted in a 0.1 mm nylon loop for crystal testing at the synchrotron (this crystal did not diffract optimally and was never used for data collection).

Crystallisation of EcRpiB The first crystallisation trials of EcRpiB were with the C66S mutant, at a concentration of 8 mg/ml, together with 2 mM R5P. Crystals were easily obtained after just one day in several different conditions and optimisation was not necessary as the crystals in the initial screens gave satisfactory diffraction. These crystals grew as hexagonal plates or with a more three di- mensional, rhombohedral morphology, as seen in Figure 13. Data were collected for a crystal grown in 12% PEG 8K, 0.5 M KCl, 10% glycerol, that were of the space group P3221 and diffracted to 1.7 Å (Table 4). Trials for the H99N mutant with 10 mM R5P resulted in micro-crystals + forming in condition A12 of the JCSG 96-well screen (0.2 M KNO3 pH 6.9 and 20% PEG 3350, Qiagen), and several other conditions with 20% PEG 3350 contained crystalline precipitate. The conditions were optimised to 18- 22% PEG 3350 with 0.1 M MES pH 6 and hexagonal crystals appeared out of the precipitate after 12-24 hours, Figure 13. These were of the space group C2 and diffracted to 2.1 Å. A problem with this morphology was that some crystals grew as thin plates, sometimes stacking on top of each other very closely. During data collection, it was therefore important to examine the images carefully and be aware of the possibility of multiple lattices. Still, only six crystals had to be tested before one with a satisfactory diffraction pattern was found.

Figure 13. EcRpiB mutant crystals. To the left are crystals of the mutant C66S crystallised with 2 mM R5P and to the right are crystals of H99N crystallised with 10 mM R5P.

Wild type EcRpiB proved to be more of a challenge to crystallise. Several different 96-well screens were set up without crystalline material being obtained. Conditions that had been found to produce crystals with the C66S mutant protein were also explored to no avail. Eventually, the Cryo screen (Emerald Biosystems) condition B1 (30% PEG 200, 0.1 M MES pH 6, 5% PEG 1000) yielded large bundles of needle-like crystals; see Figure 14 (A). These conditions included protein, at 8 mg/ml, mixed with 50 mM All6P. Optimisation of these conditions started with a screen of the pH, which resulted in more ordered needles appearing in Na citrate pH 5.4. The drops in the optimisation were seeded from the initial bundles; crystals only appeared at pH 5.4, and in MES pH 6 where the original morphology was reproduced. Higher pHs produced no crystals, and at pHs below 5 only precipitate was formed. The size of the needles could be increased by streak seeding through a series of drops, thus diluting the seeds, and providing fewer nucleation sites. The crystals were also improved when the pH of the protein buffer was changed from 7.5 to 8.3. Crystallisation experiments were consequently performed with 20-50 mM All6P in the protein solution. The needles of EcRpiB did not diffract as well as the MtRpiB needles. Three different data sets were collected to 2.8, 2.6 and 2.5 Å resolution but in all three cases the density in the active site was too disordered for building in the ligand. A new feature was then installed on our laboratory’s crystallisation robot (Oryx 6 from Douglas Instruments) making it possible to seed all drops in a 96-well plate while they are being dispensed (D'Arcy et al., 2007). As a trial project EcRpiB mixed with 30 mM R5P was used together with a seed solution created from crystals of the H99N mutant. A drop containing about five crystals was diluted with 80 l well solution and vortexed together with a 5 mm teflon bead, a so-called seed-bead (Hampton Research, (Luft and DeTitta, 1999)), to crush the crystals. This cross-seeding resulted in a dramatic increase in the success rate. In 17 of the 96 different conditions crystals could be found, some after just a few hours and some after 2-3 days. The

30 crystals were of the same morphology as the H99N crystals (Figure 14 (C)) although in this case, many more of the plates remained very thin. The crystals belonged to the space group P3221, with cell dimensions almost identical to the C66S crystals (Table 4) and diffracted to around 1.8 Å.

Figure 14. Wild type EcRpiB crystal optimisation. A. The first type of crystal produced with wild type protein. Several bundles like this one were present in the drop. B. A pH optimisation resulted in more ordered needles forming in Na citrate pH 5.4. C. Seeding the wild type protein drops with a seed stock solution of H99N crystals produced chunky, hexagonal plates. Quite often several plates would grow on top of each other, an indication of this phenomenon can be seen in the top right crystal.

All data sets collected of EcRpiB crystals resulted in ambiguous electron density in the active site, which is discussed in more detail under the section ‘EcRpiB structures’. This was thought to be correlated with the fact that strong density was frequently seen, apparently covalently attached to the active site base, Cys66. The crystals grew in conditions with a pH of 7, and the protein solution had a pH of 8.3. -mercaptoethanol (ME) was present in the purification buffer and MESNA was regularly added to the crystallisation setup. Thus, it was suspected that the cysteine could react with either reducing agent during the relatively long time periods needed for obtaining crystals. To try to get around this problem, an aliquot of EcRpiB was reduced with 50 mM dithiothreitol for 15 minutes and then the buffer was exchanged to one without any reducing agent, using a PD10 column. The sample was concentrated, divided into three parts and then ligand was added, R5P, All6P and 5PRH, to the three samples. Crystallisation screens were performed with the cross-seeding technique described above. In the screens with 30 mM 5PRH and 30 mM All6P, crystals of the hexagonal morphology appeared readily in several conditions. These were flash frozen in liquid nitrogen after 72 hours. A data set of EcRpiB with 5PRH was collected to 2.15 Å, and with All6P to 2.0 Å; the crystals were once again of the space group P3221.

31 Table 4. The different EcRpiB complexed structures. EcRpiB Crystallisation [Ligand] Reso- Space Mole- type + condition buffer- lution group + cules ligand additive (Å) cell (Å) in AU

C66S 12% PEG 8000 2 mM R5P, 1.8 P3221 2 + 0.5M KCl 10 mM 51.7, 51.7, R5P 10% glycerol ME 185.8

a wt + 30% PEG 200 20 mM 2.8 P21 16 All6P 5% PEG 3000 All6P, 2.6 71.4-72.5, (x3)b 0.1M NaCit pH 5.4 5 mM 2.5 191, 91.2, MESNA =92.5

H99N 20% PEG 3350 10 mM 2.1 C2 6 + 0.1M MES pH 6 R5P, 5 mM 89.7, 51.9, R5P MESNA 207.2, =115

wt 0.1M Hepes pH 7 20 mM 1.8 P3221 2 + 30% Jeffamine R5P, 5 mM 51.2, 51.2, R5P ED-2001 pH 7 MESNA 171.9

wt 0.2 M Na malonate pH 7 20 mM 1.7 P3221 2 + 20% PEG 3350 R5P, 5mM 51.7, 51.7, R5P MESNA 185.7

wt 0.2 M Na malonate 30 mM 2.15 P3221 2 + 0.1 M Bis Tris 5PRH 51.8, 51.8, 5PRH propane pH 6.5 No reduc- 186.1 20% PEG 3350 ing agent

wt 0.2 M Potassium formate 30 mM 2.0 P3221 2 + pH 7.3, 20% PEG 3350 All6P 51.7, 51.7, All6P No reduc- 185.5 ing agent a. Wild type EcRpiB. b. Three data sets were collected on three crystals of the needle like morphology, grown in equivalent conditions.

32 Structural results Solving the structure of MtRpiB (Paper I) For the first MtRpiB structure, data were collected at MAX-laboratory in Lund on beamline I711, from a crystal that diffracted to 1.9 Å resolution and belonged to the space group C2. The structure of MtRpiB was solved by molecular replacement (MR) using AMoRe (Navaza, 1994) via the CCP4 interface and with a dimer of EcRpiB (PDB entry 1NN4.pdb) (Zhang, 2003b) as the search model. Since MtRpiB has a sequence identity of only 32% to EcRpiB and, from calculating the unit cell volume it was clear that at least four molecules had to be present in the AU, it was not surprising that auto AMoRe only gave very low correlation coefficients (CC-F) for the first rotation and translation functions. However, by varying the search radius and resolution range it became clear that two dimer solutions occurred frequently, with a slight drop in CC- F to the third possible solution. The best runs had CC-Fs for the rotation and translation searches of 9.3 and 13.0%, respectively, for a resolution range of 4-20 Å. By fixing and refining the top solution, and performing a new trans- lational search of the rotation solution file, the same solution for the second dimer was obtained, suggesting that these two solutions indeed were correct. The CC-F for the two-dimer solution had increased to 32.7%, while the second and third in the list of possible solutions had CC-Fs of 20.8% and 14.7%. Refined fitting resulted in a substantially improved CC-F of 42.1. A pdb file including the two dimers was created and visualised in O and when symmetry related molecules were generated it could be seen that the solution was consistent with good crystal packing. Potential fifth or sixth molecules could not be located. Although the program ARP/wARP (Perrakis et al., 1997) failed in autobuilding, it was able to produce improved phases when the complete 1.9 Å data set was used. Assuming 4 molecules in the AU, the R-factor dropped from 51.4% to 34.1% over the course of 100 cycles. The improved electron density map gave strong confirmation that the two dimer solutions were correct; -sheets clearly indicated the presence of subunits placed within ~1Å of the positions suggested by molecular replacement, and additional features consistent with the M. tuberculosis sequence began to be apparent.

NCS averaging For convenience in further work, a homology model of MtRpiB was generated using the EcRpiB structure as a template in the program SOD (Kleywegt et al., 2001). The phases resulting from ARP/wARP were then improved by non-crystallographic symmetry (NCS) averaging as imple- mented in the graphics program O (Jones et al., 1991; Jones, 1992). Opera- tors for moving chain A of the homology model onto the other three (mole-

33 cules B, C, and D) were calculated with the command lsq-explicit in the graphics program. A mask covering chain A was then created with MAMA in the RAVE package (Kleywegt and Jones, 1999) using a grid of 170, 128, 90 Å and a radius of 2.5 Å. The operators were improved with IMP from the Uppsala Software Factory (Kleywegt and Jones, 1994). Averaging was performed manually to a resolution of 3 Å by cycling through AVE in the RAVE package, SFALL, SIGMAA and FFT, all in the CCP4 suite (CCP4, 1994). The new map created at each cycle was inspected in O. After 6 cycles of averaging, the map had improved sufficiently to allow the first coarse rebuilding in the A molecule of the homology model, using the grab-build commands in O. The rebuilt A molecule was then used to create new B, C and D molecules using the three improved operators. The resulting subunits were subjected to restrained refinement in REFMAC5 (Murshudov et al., 1997), to 2.5 Å resolution. Then a new mask with a finer grid (228, 170, 120 Å) was made, new operators calculated as previously and the NCS averaging was continued for an additional 10 cycles; the resolution was extended to 2.2 Å after 5 cycles. During the averaging cycles, the R- factor dropped around 2 percentage points per cycle from the start value of 48%, eventually converging at 35% with an R-free of 38%.

Finding the fifth molecule The A chain was rebuilt, once again used to generate the three other chains and the pdb file and phases were refined with NCSref (CCP4i) to the highest resolution of 1.9 Å, but the R-factor would not drop below 35%. The refinement was therefore continued in REFMAC5 without applying non- crystallographic symmetry. It was then possible to detect new electron density in the space between the two dimers (see Figure 15 (B)), of the approximate size of a single subunit, and with a clear -sheet. To check this result, and to place the remaining subunit correctly, AMoRe was re-run using the improved MtRpiB A molecule as the search model. Five clear solutions were found in the molecular replacement (all with correlation coefficients ~9.0%, compared to the next highest solution at 6.4%). When all five molecules had been placed the final correlation coefficient reached 72.4%. The resulting model was then inspected in O. After setting the correct y- origin, the solutions were seen to correspond to the two previous dimers, plus an additional subunit. This last subunit was placed precisely in the “extra” electron density observed after the crystallographic refinement of the four-subunit model. Further, it formed an essentially identical dimer with a subunit that was equivalent by crystallographic symmetry. After placing this last (E) molecule, refinement with REFMAC5 resulted in an immediate and dramatic drop in the R-factor to 27%. C-terminal residues that had previously not been apparent could now be placed and a loop from residue 40 to 45 could be fitted correctly into its density. After 3-4 rounds of REFMAC5

34 water molecules were picked automatically with ARP/wARP and the R- factor dropped to 17.5% with an R-free of 22%.

Figure 15. Sections of the electron density maps of MtRpiB at different stages of refinement, showing how the existence of a fifth molecule was discovered. The electron density is coloured in grey and contoured at 1 sigma. A. The original map with phases improved in ARP/wARP, 1 sigma is 0.17 electrons per Å3. B. The resulting map after refinement of the 4 molecules in REFMAC5 without including NCS restraints, 1 sigma = 0.33 e-/Å3. C. The final electron density map for the deposited model, 1 sigma = 0.61 e-/Å3. The three snap shots are taken from the graphics program O at the same angle, and with the centering on the C-terminal part of the E molecule. The A, B, C and D molecules of MtRpiB are in the density to the right in each picture and the symmetry related molecules are to the left. The E molecule appears in the middle, bottom half of C.

The overall structure of MtRpiB All in all, MtRpiB was crystallised in complex with seven different ligands: Pi, 4PEH, 4PEA, 4PMEA, R5P, 5PRH, and All6P (Table 3). The crystals were consistently of the space group C2 with more or less similar cell dimensions, and diffracted at best to 1.6 Å, ranging down to 2.2 Å. As mentioned in the above section, MtRpiB is found as a dimer in the crystals, which is also seen from size exclusion chromatography during purification and when a protein sample was analysed by dynamic light scattering, although in the latter case tetramerisation was also seen to a small extent. Five molecules are present in the AU of the crystal forming two complete dimers and one dimer formed through a two-fold crystallographic symmetry axis. The molecules can in all cases be traced starting from residue 2 or 3 and ending with residues 158-161 (of the total 172 residues). The N-terminal six- histidine tag (10 residues) was never observed in the electron density. Each MtRpiB subunit forms a Rossman type fold with a five-stranded parallel -sheet surrounded by five -helices, two on one side and three on the other. The sixth -helix, connected to 5 by a long loop, extends out from the main fold and interacts with the second subunit. In MtRpiB the C-

35 terminal end, a long loop after 6, wraps around the second subunit to stabilise the dimer formation. This can be seen in Figure 16, which shows two views of the MtRpiB dimer in (A) and (B).

Figure 16. The overall structure and the active site of MtRpiB. A. The dimer is coloured with the N-terminal of each subunit in blue going through the rainbow to the C-terminal in red. The active site is occupied by an R5P molecule - carbon atoms in grey. B. A different view of the molecule showing how the C-terminal end is involved in the dimerisation. The A molecule is coloured orange and the B molecule blue. C. Residues lining the active site, with white carbon atoms when contributed by the B-molecule. D. EcRpiB aligned to MtRpiB, showing how the catalytic bases terminate in the same position.

36 The dimer interface buries 1990 Å3 of surface area per subunit and involves interactions from five sections of each molecule. Most of the atoms providing the dimer contacts are non-polar, with the polar residues of each subunit supplying 18 direct hydrogen bonds. Superimposing all MtRpiB complexes shows that they are all very similar with root-mean-square (r.m.s.) differences of 0.1-0.36 Å using all C atoms.

The active site of MtRpiB MtRpiB in complex with phosphate (Paper I) The active site of EcRpiB had previously been identified by evaluating the positioning of conserved residues in the structure (Zhang, 2003b). With the first MtRpiB crystal structure a phosphate molecule, from the crystallisation buffer, was seen to bind toward the outer region of this active site cleft and confirmed the residues involved in phosphate binding, two arginines and a histidine (Figure 4 Paper I). A third arginine (Arg113) also interacts with the phosphate group in two of the five molecules of the AU. In the MtRpiB complexes studied, this arginine most frequently adopts the right conformation for phosphate binding. Residues lining the active site originate from both the A and B subunit of the dimer (Figure 16 (C)). Many are conserved within all RpiBs, for example, Asp11 and His12 (phosphate binding) originating from the A molecule of the dimer, and His102, Asn103, Arg137, His138 and Arg141 from the B- molecule (Mt numbering). The most puzzling difference between the Ec- and Mt-like RpiB sequence involved the catalytic base, Cys66 in E. coli, which is a glycine in MtRpiB. From the MtRpiB structure it became clear that the carboxylate moiety of a glutamic acid, six residues away from the glycine, terminated in approximately the same position as the EcRpiB cysteine’s thiol, Figure 16 (D). Both the glycine and glutamic acid are conserved among the Mt-like Rpis, with the glycine being necessary to provide enough space for the glutamate. The second main difference is Ser71 (MtRpiB), which is a threonine in EcRpiB, although among the Ec-like RpiBs a serine is also found in this position (see Figure 7).

Interactions with inhibitors 4PEH and 4PEA (Paper II) Of the two compounds 4PEH and 4PEA, the former binds 30 times more tightly to the active site of MtRpiB. This is explained by the crystal structures, which show that 4PEH has many more specific interactions with the residues of the active site than 4PEA (Figure 17). These interactions could be used to assign possible roles for the active site residues during catalysis, as 4PEH was designed to mimic the transition state compound of the reaction and therefore was expected to bind in a similar way to the substrate.

37 The phosphate moiety of both ligands is placed in the same position as the Pi of the original structure (1USL.pdb). Moving up the ligand, the hydroxyl group of carbon 3 in both structures interact with His102 and a water molecule, which is within hydrogen bonding distance to His138. This oxygen is equivalent to oxygen 4 of the substrate R5P and is that which binds to carbon 1 in the closed furanose form. The next OH group of carbon 2 of 4PEH hydrogen bonds to Asp11 and a backbone nitrogen of residue 70; this OH in 4PEA also interacts with Asp11 but instead of the interaction with the main chain, a hydrogen bond is made to the side chain of Glu75. The top parts of the two ligands are where the main differences in binding occur. Here 4PEA has few interactions with the enzyme; only one hydrogen bond is found between O1 of the top carboxylate and the amide nitrogen of Ser71, and one between the other carboxylate oxygen and the Asn103 side chain, are observed (Figure 17(B)). 4PEH on the other hand has several strong hydrogen bonds to the backbone nitrogens of residues 70, 71 and 74, and the carboxylate oxygens of Glu75 interact with both the NH and the hydroxyl group attached to the nitrogen of the substrate analog (Figure 17). Finally, the Ser71 side chain also interacts with the terminal hydroxyl group. Figure 4 of Paper II gives a more complete visualisation of the hydrogen bond interactions of the two ligands, and shows the electron density of these.

Cocrystallised with R5P (Paper IV) MtRpiB cocrystallised with the substrate R5P resulted in a crystal that diffracted to 1.65 Å. The density found in the active site was very well defined and showed clearly that it was mainly a linear molecule that had bound Figure 18. As the linear R5P aldehyde and the ketose Ru5P only differ in the location of the carbonyl oxygen, and these two molecules would generate similar density it was not possible to determine which of the two had bound. The observed electron density suggested that the most likely scenario is that both molecules are present, however, at 1.65 Å it is not possible to model their relative occupancies with any accuracy. The isomerase reaction favours the production of R5P at a ratio of ~75% over 25% when equilibrium is reached (Wood, 1970). Including both molecules and setting the occupancy of R5P to 0.75 and Ru5P to 0.25 resulted in very low B-factors for Ru5P. With the occupancies set to 0.5 for both molecules the B-factors were more similar (16 for R5P and 13.5 for Ru5P), which seemed more reasonable. The R5P/Ru5P compound was found to bind in the same position as the ligand in the 4PEH structure and all active site residues are oriented in the same way Figure 18. The only small difference is the catalytic base, Glu75, where the carboxylate moiety is seen at slightly different angles. This reflects how the residue is able to adjust its position in relation to the parts of substrate that it will interact with for donating or accepting a proton.

Figure 17. Active site comparison of MtRpiB in complex with 4PEH (pink carbon atoms) and 4PEA (navy carbon atoms). The A-molecule of the dimer is coloured orange and the B-molecule is light blue. A. Overall active site organisation showing the similar binding of the two ligands. The red sphere is a water molecule. B. The top parts of the two ligands showing how 4PEH has more interactions with the enzyme. Hydrogen bonds from 4PEH in orange and from 4PEA in light blue.

Figure 18. Stereo view of the active site of MtRpiB-4PEH (pink carbon atoms) superimposed onto MtRpiB with R5P/Ru5P (green atoms for the residues and R5P and orange atoms for Ru5P). The density of the final refined SIGMAA weighted - 3 2Fo-Fcmap of the R5P/Ru5P ligand in blue mesh, at 1 (0.33 e /Å ).

39 Complexes with the inhibitors 5PRA and All6P (Paper IV) 5PRA Cocrystallisation with the inhibitor 5PRH resulted in clear density for a hy- drolised compound, 5-deoxy-5-phospho-D-ribonate (5PRA) in the active site. The phosphate part of the ligand binds in the same place as in previous structures of MtRpiB, and with this positioning it is obvious that a whole molecule of 5PRH could not be placed in the active site. Before we obtained the structure we had hypothesised that the 6-carbon sugar phosphate, All6P, would bind with the phosphate group positioned further out of the active site and that the long flexible phosphate binding residues, Arg137 and Arg141 merely would adapt to this second positioning (Paper III). If this had been the case, 5PRH, designed to mimic the open chain form of All6P, would have had enough space to dock with its hydroxamic top part interacting in the same way as 4PEH. Instead, one of the oxygens of the carboxylate of carbon 1 and the OH-group of carbon 2 of 5PRA are seen to bind very tightly to Glu75 (Figure 19), and the second oxygen of the top carboxylate is bent down toward carbon 4 with hydrogen bonds to backbone nitrogen atoms of residues 70 and 71, and a water molecule.

Figure 19. The active site of MtRpiB with a molecule of 5PRA bound. A. Fisher projection of 5PRA (compare with Figure 11). B. The electron density of the ligand - 3 in the final refined 2Fo-Fcmap, contoured at 1 (0.35 e /Å ). C. General organisation of the active site, with residues positioned more or less identically to the other MtRpiB structures.

5PRH inhibits the R5P to Ru5P reaction of MtRpiB with a Ki of 0.4 mM (Paper III). This is ~6 times worse than the inhibitory capacity of the one carbon unit shorter 4PEH, which was seen in the crystal structure with no suggestion of hydrolysis. In the activity measurements the initial rate of the reaction is over in roughly one minute and inhibitor stocks were treated carefully, so it would be assumed that it is the non-hydrolysed form of 5PRH that inhibits MtRpiB. In the cocrystallisation experiment, on the other hand, the

40 crystals require at least 2-3 days to form, which would give adequate time for the compound to hydrolyse. Jeffery and co-workers observed a similar phenomenon when crystallising phosphoglucose isomerase (PGI) together with the inhibitor 5-deoxy-5-phospho-D-arabinohydroxamic acid (which is quite analogous to 5PRH, 5-deoxy-5-phospho-D-ribonohydroxamic acid), which in the crystal was found to be the hydrolysed arabinonate (Jeffery et al., 2001). MtRpiB presumably binds the 5PRA molecules selectively over 5PRH as this molecule is easier to fit into the active site. It would be interesting to measure the Ki of the hydrolysed form, as it would be expected to have a better binding affinity compared to the non-hydrolysed form. How- ever, this experiment has not yet been performed. Attempts were made to crystallise MtRpiB without any compound that could bind in the active site, to enable soaking experiments with 5PRH. Crystals would then have been soaked in a drop containing a high concentration of 5PRH and flash frozen directly in liquid nitrogen in order to force the enzyme to bind the non-hydrolysed compound. This strategy was successful in the crystalisation of PGI with the nonhydrolysed form of the above mentioned inhibitor (Arsenieva et al., 2002). Apo-MtRpiB crystals of good enough quality were, however, never obtained.

All6P The MtRpiB-All6P complex resulted in nice density for the -pyranose form of All6P in all five molecules of the AU, Figure 20 (B). Once again the phosphate group binds as in the other MtRpiB structures. With the phosphate in this position it would be difficult for the open form of the sugar to fit into the active site. The residues of the active site are more or less positioned identically as in the other MtRpiB structures. The main differences are Glu75, which was difficult to model due to high B-factors and seems to be mobile, and His102, which is at a slightly different angle when compared to the histidine in the R5P complex (Figure 20 (A)). This results in a loss of an interaction to O5; instead His102 hydrogen bonds to O1 of the -All6P ligand. Glu75 as seen in Figure 20 appears to have been forced away from the ligand due to restricted space; both carboxylate oxygens are within hydrogen bonding distance of O3 of the ligand, which also interacts with the amide nitrogens of Gly70 and Ser71. The bottom part of All6P with the phosphate group followed by C6, C5 2- and O5 lies in the same position as OPO4 , C5, C4 and O4 of R5P. O5 has a hydrogen bond to a water molecule that in turn is close to His138. The hydroxyl group on C4 interacts with Asp11 and the backbone nitrogen of Gly70, and the C2 hydroxyl with the side chains of Ser71 and Asn103 as well as the amide nitrogen of Gly74. Tyr46, which is a conserved residue in all RpiBs, lies with its ring almost parallel to the sugar ring at a distance of around 4 Å, presumably resulting in some stacking interactions, Figure 20

41 (B). This tyrosine may also be involved in positioning Asn103 correctly through an interaction between Tyr-OH and Asn-NH. The ratio of the - and -pyranose of All6P is around 1/3, as measured in D2O solution (Franke et al., 1985). The active site density of MtRpiB-All6P clearly shows that the enzyme has bound only the -pyranose, with no indication of the -form being present. Docking a -pyranose by aligning carbon atoms one to five shows that a hydroxyl group could be accommodated in the -form, although it comes rather close to (3.3 Å) carbon atom C 2 of the ring of Tyr46. The reason why only the -pyranose is observed in the crystal structure is that in this position the hydroxyl group has an additional hydrogen bond to the enzyme via His102. In the -anomer this hydroxyl only has possible interactions with water molecules.

Figure 20. The active site of MtRpiB cocrystallised with All6P. A. Superposition of MtRpiB-All6P (black carbon atoms) onto MtRpiB-R5P (grey carbon atoms) B. The possible stacking interaction of All6P with Tyr46 is shown together with the - All6P density contoured at 0.33 e-/Å3 (1 ).

E. coli RpiB structures First attempts at producing complexed structures E. coli RpiB was crystallised in the presence of the compounds R5P, All6P and 5PRH. The obtained crystals were of several different space groups, see Table 4, and the structures therefore had to be solved by molecular re- placment. Still, the resulting structures could all be aligned with very small r.m.s. differences (0.1-0.3 Å for the dimer). Tetrameric interactions, as seen for the original structure, were still present even though the expressed protein had a shorter N-terminal tag than previously (Zhang, 2003b). The current protein eluted as a dimer during size exclusion chromatography and

42 presumably the tetrameric contacts are formed at the higher concentrations in the crystal. Even so, the fact that several crystal forms were obtained with EcRpiB, but the tetramer remains the same, is an indication of some prefer- ence for this multimeric state. The first EcRpiB structure was of the C66S mutant crystallised together with the substrate R5P. The hope was that as this mutant cannot catalyse the reaction it would be possible to see a molecule of either - or -furanose R5P in the active site. Unfortunately no density for any ligand could be detected. However, some uninterpretable density was found attached to the mutated serine, as if something had bound covalently to the residue. A modification of the protein in this way could have prevented the ligand from binding but a lack of density could also have been due to a too low concentration of ligand added to the crystallisation experiments. The general rule when adding a ligand is that the concentration should be ten times the Kd (estimated in the RpiB case from the value of Km) for 90% of the protein to be occupied. The Km for this mutant could of course not be measured as no activity was seen but as the wild type enzyme has a Km of around 1 mM, 10 mM would have been a better concentration to use in the crystallisation setup instead of the 2 mM used. A higher ligand concentration was used in the following experiments. However, 2 mM was the concentration that gave good density in the active site of MtRpiB crystallised with R5P and this enzyme has a slightly higher Km for the substrate. The first crystals obtained with wild type EcRpiB cocrystallised with the substrate All6P were long, thin needles that generally only diffracted to around 2.8 Å. The space group was P21 and 16 molecules were present in the AU. Density was seen in the active site of most molecules but not very well defined and even with 16-fold averaging and tight NCS restraints it was not possible to place either a linear or a ring form of All6P. Three data sets were collected for this type of needle in the hope that better resolution would be obtained and that slight differences in the crystallisation experiments might make the ligand bind tighter. Cross-seeding from the mutant H99N crystals improved the crystallisation of wild type EcRpiB immensely and data sets could be collected to 1.7- 2 Å. These crystals were of the space group P3221 compared to C2 for the H99N crystals. For wild type EcRpiB cocrystallised with R5P, data were collected from two crystals grown in different conditions (Table 4). For both of these density was seen close to Cys66 in the active site Figure 21 but nothing could be detected close to the phosphate binding residues. In the first case the density on Cys66 suggested a covalently attached molecule and as ME was present during purification and the crystallisation drops were set up in the presence of MESNA it was suspected that one of these two molecules had formed a disulfide bond with the cysteine. A molecule of ME could be placed in the density but it could equally well be a molecule of MESNA. In the second structure the density close to Cys66 was less well

43 defined, but all the same something seemed to be preventing the substrate from binding properly.

Figure 21. Density found attached to C66 in EcRpiB crystallised with the substrate R5P. In A. a molecule of ME could be modeled into the density but only waters could be placed in B. These are shown as small grey spheres. The electron density of the original maps is contoured at 1 sigma, 0.33 and 0.41 e-/Å3 in A and B.

Figure 22. The active site density of wild type EcRpiB cocrystallised with 5PRH in a buffer without a reducing agent. The density is from the original SIGMAA weighted map contoured at 1 (0.31 e-/Å3). A. The main residues of the active site shown together with a manually docked molecule of 5PRH. B. A close up view of C66 and the top part of the ligand.

To try to get around this problem, experiments were set up using a protein solution where the original buffer had been exchanged for one without any reducing agent, as described above. Data were collected on one crystal with 5PRH and one with All6P. Both crystals had two molecules in the AU, i.e. one dimer. This time density could be detected for the phosphate part of the ligand in both active sites of EcRpiB-5PRH but in only one of the active

44 sites of EcRpiB-All6P. The “top” part of 5PRH had some density in one of the active sites but it was not well defined and it was difficult to place the hydroxamic moiety properly, Figure 22. For All6P only the phosphate could be seen.

EcRpiB-H99N in complex with R5P (Paper IV) The only EcRpiB complex with good density for a ligand was the H99N mutant, which crystallised in the space group C2 with 6 molecules in the AU. When the original maps were inspected density suggesting a mixture of - and -furanose together with the linear R5P or Ru5P was seen in each active site. However, in each dimer one of the active sites had less defined density. In fact, closer inspection suggested that different species had bound in the two active sites. In molecules B, C and E where stronger density was observed, both furanoses could be placed with some degree of confidence as well as a linear R5P, see Figure 23 (B). However, in molecules A, D and F the density only accomodated a linear molecule, Figure 23 (A). At 2.1 Å resolution, determination of the most appropriate relative occupancies for a complicated mixture of ligands in two classes of active site is difficult. Therefore the ligands in the two cases were docked into maps that had been obtained from three-fold averaging of the original maps.

Figure 23. The electron density of the averaged map contoured at 0.39 e-/Å3 (which is equivalent to 1 in the unaveraged map) found in the active sites of EcRpiB- H99N crystallised with R5P. A. The A-site of the dimer, which had less defined density, with a docked linear R5P (grey). B. The B-site, which clearly showed the furanose sugar. Residues of the A-molecule have black carbons as does the docked -furanose and in the B-molecule the carbons are grey.

45 The difference between the molecules was found to be correlated with different conformations of His17, located in the first helix. In the B, C and E molecules the histidine interacts with Arg133 of a symmetry related molecule, Figure 24. This arginine is involved in binding the phosphate part of the ligand, and where His17 interacts in this way the active site density is strong for the furanose. His17 of A, D and F does not point toward a neighbouring molecule instead it bends down toward Phe13 of the same chain, perhaps shielding it from water. The symmetry related molecules that pack around the active sites of the A, D and F molecules are farther away, leaving this active site more exposed to solvent, which could be the reason for the lower occupancy of the ligand, Figure 24. The phosphate part of either docked R5P molecule interacts with the equivalent residues of the MtRpiB structure: His10, Arg133, and Arg137. The ring oxygen (O4) of both furanoses, hydrogen bonds to Asn99 and a water molecule, which in turn has a hydrogen bond to His134. O3 interacts with the backbone amide of residue 67 and the carboxylate of Asp9, and O2 has a hydrogen bond to Cys66. The density for the top part of the linear R5P places O1 close to Thr68 and Asn100. In the -furanose this hydroxyl group is close to Tyr43 and has a hydrogen bond to Asn100 and the water molecule, which interacts with His134. The -anomer has its hydroxyl group almost within hydrogen bonding distance to Asn99.

Figure 24. The crystallographic packing of EcRpiB-H99N. A. A close up of the B- type active site showing the interactions with the symmetry related molecules first helix. B. The molecule to the right is a part of the AU while the other two are symmetry related molecules. The A- and B-type active sites are marked, which shows how the A-type site is more exposed.

46 Analysing why EcRpiB crystallises with a more or less empty active site Crystallisation of MtRpiB was relatively straightforward and resulted in several complexed structures; although including a ligand in the experiment was crucial for sucess. The phosphate part of the ligand was found to be involved in the crystal packing between symmetry related dimers, which would explain why such a compound was necessary, see Figure 25. Crystal- lisation of EcRpiB on the other hand was more difficult, especially with the wild type enzyme, and with the aim of studying ligand interactions. In Table 5 all the EcRpiB crystallisation experiments are listed with the resulting active site density. Since the ligand in MtRpiB is involved in the crystal packing two active sites of neighbouring molecules automatically come close together, which makes the environment around the active site fairly closed. In EcRpiB the active site is more exposed to the solvent, and the ligand will not be prevented from escaping in the same way (Figure 25). An exception is the B-site of the EcRpiB-H99N structure and the wild type enzyme crystallised in P3221. In these active sites it is also possible to detect density for a bound ligand. The fact that EcRpiB-H99N seems to have more of the linear substrate in the A-site is probably an artefact of lower occupancy, due to the more open active site. Otherwise this site would have to be more efficient in opening the ring compared to the B-site, which seems strange as both sites lack His99.

Figure 25. Crystal packing observed for three different RpiBs. Maps shown in grey mesh and active site histidines pointed out for help in orientation, which is not the same in the three panels. The placement of R5P is shown in the MtRpiB structure.

Another problem with EcRpiB is of course the density for a mysterious compound attached to Cys66 in many of the structures. Even in the structure with wild type enzyme crystallised with All6P where the buffer had been exchanged to be devoid of any reducing agent, density appears on this cysteine in the A-site. Perhaps the protein sample was not properly reduced with 50 mM DTT and still has a molecule of ME attached. Alternatively, the density comes from some other small molecule present in the experiment. The active site cysteine must be very reactive and as such it could be modi- fied by various compounds. The only other RpiB structure in the PDB is that

47 of the enzyme from Thermotoga maritima (PDB code 1O1X) (Xu et al., 2004). In this RpiB the equivalent cysteine was found to be oxidised and was modeled as a cysteine sulfonic acid. Both EcRpiB mutants were decidedly easier to crystallise compared to the wild type enzyme. An explanation could be their inability to catalyse the reaction efficiently, thus making the active site residues and compounds in it less mobile. It would be interesting to recrystallise C66S with a higher ligand concentration and to crystallise both mutants with the transition state analogs 4PEH and 5PRH.

Table 5. EcRpiB crystal structures with resulting active site density. EcRpiB type + # in Active site density ligand AU C66S + R5P* 2 Unidentified compound bound to S66 wt a + All6P* (x3)b 16 Vague density, improves slightly with averaging H99N + R5P 6 Density accommodates either -, -furanose or linear R5P wt + R5P* 2 ME or MESNA bound to C66 wt + R5P* 2 Unidentified compound bound to C66 wt + 5PRH* 2 Phosphate seen and top of ligand although not well ordered. wt + All6P* 2 Phosphate in B-molecule. In A something on C66 a. Wild type EcRpiB. b. Three data sets were collected on three crystals of the needle like morphology, grown in equivalent conditions. * These structures will not be deposited.

Comparing the Ec- and MtRpiB structures The overall structures MtRpiB has the same overall fold as the original EcRpiB structure (1NN4.pdb) and the two superimpose with an r.m.s difference of 1 Å when 260 C atoms of the MtRpiB AB dimer are aligned to the 1NN4 AD dimer. The carbon atoms were aligned with a cut-off of 2.0 Å. The main difference between the two structures is the larger dimer interface of MtRpiB, 1990 Å2 compared to 1470 for EcRpiB. This dimer interaction is enhanced by the long C-terminal loop, which is absent in EcRpiB, Figure 26. T. maritima RpiB crystallised with only one molecule in the AU, and the dimer is formed through crystallographic symmetry (Xu, 2004). Aligning this one molecule to EcRpiB resulted in an r.m.s difference of 0.8 when 137 C atoms were compared, again at a cut-off of 2.0 Å. With MtRpiB the r.m.s. difference is 0.9 Å for 129 Cs, taking two gaps into account (Figure 7). T. maritima RpiB is a member of the Ec-like family, thus it also has a slightly smaller dimer interface and a shorter C-terminal loop compared to MtRpiB.

Figure 26. EcRpiB (light grey) superimposed onto MtRpiB (black). In A. the C trace is shown together with a molecule of R5P (MtRpiB) and a docked 5PRH (EcRpiB) in one of the two active sites. The view shown in B. highlights the difference in length of the C-terminal loops of the two enzymes.

Active site comparisons At a first glance the active sites of Ec- and MtRpiB look very similar, the main difference being the exchanged catalytic base. Superpositioning the two AB dimers brings most active site residues to seemingly the same positions, especially the residues of the N-terminal part of the helix at the bottom of the active site cleft. Here the side-chain hydroxyl group of Thr68 of EcR- piB aligns beautifully with Ser71 of MtRpiB, as do the C residues of the helix and the preceding loop. The side-chains of Asp9/11 (Ec/Mt) and Tyr43/46 of the A-molecule of the dimer and Arg137/141 of the B-molecule also align perfectly, but a closer inspection of the other residues reveals slight but important differences. The relative positioning of histidines 10/12 (A-molecule) and 99/102, 134/138 and Arg133/137 (B-molecule) are shifted 1 to 2 Å apart in the two species, see Figure 27. In the B-molecule the shift is due to a substitution of Asn94 in MtRpiB, to Tyr91 in EcRpiB. The large tyrosine side chain forces Leu95/98 to have different conformations. In the E. coli enzyme the leucine is pushed down toward His99 (in the figure Asn99 of the mutant is shown), which in turn moves His134 in relation to the equivalent histidines in MtRpiB. These shifted residues enable EcRpiB to bind the phosphate part of the ligand slightly farther out of the active site compared to where it is seen to bind in all the MtRpiB structures. The small differences in positioning of His10 and Arg133 are also linked with the placement of the phosphate. Another significant residue substitution between the two species, found in an outer loop of the active site, is Arg40 in EcRpiB, which is an aspartic acid

49 in the M. tuberculosis enzyme. In the wild type EcRpiB structure crystallised with 5PRH Arg40 has well defined density and is seen to interact with the phosphate of the ligand, see Figure 22 (A). In the EcRpiB-H99N structure this arginine has low occupancy and is difficult to place. In the B-site it is probably interacting with Glu18 in the neighbouring symmetry related molecule as well as with the phosphate, while in the A-site it can be modelled closer to the phosphate. The fact that this residue is better ordered in the complex with the six-carbon containing compound, could be of significance, indicating that this interaction with the phosphate is more important for the longer substrate. The aspartic acid of the M. tuberculosis enzyme found in this position would be repelled by the negative charge of the phosphate and in the MtRpiB structures this residue points out into the surrounding solution. Thus, for the five-carbon substrate this additional phosphate interaction is not necessary.

Figure 27. Stereo view comparing the active sites of EcRpiB-H99N (black carbon atoms) overlayed onto MtRpiB (grey carbons). The main differences found in the two structures are illustrated leaving out the residues that align more or less perfectly for clarity.

50 The proposed reaction mechanism of RpiB In the R5P to Ru5P isomerisation The results presented here allow us to draw conclusions concerning how the reaction catalysed by Mt- and EcRpiB will occur, presented in Figure 28. For RpiA it was clear that ring opening of the ribofuranose 5-phosphate must be the first step catalysed by the enzyme in the R5P to Ru5P direction, but in the slower RpiB the necessity of this step was less obvious. The rate of ring opening in solution should be rapid enough to supply the enzyme with the correct form of substrate. The structure of EcRpiB-H99N, where the active site density clearly indicates the presence of - and -furanose as well as a linear molecule, suggests that the mutant RpiB is deficient in the capacity for opening the ring. This can be contrasted to the active site density in the structure of the wild type MtRpiB, also cocrystallised with R5P, which does not indicate the presence of either furanose; presumably this means that this enzyme already has catalysed the opening step. The kinetic results presented by Stern et al. of the H102A mutant and those of Paper IV further support this conclusion. In Paper II we proposed that the -furanose is the optimal substrate of MtRpiB from a manual docking of this compound in the position of 4PEH. However, from the results of Paper IV it seems likely that both - and - furanoses can be accepted, and opened. The ring opening will involve different combinations of proton donation and acceptance by His99/102 and His134/138 (the latter acting indirectly via a water molecule). In the enzyme PGI, His388 and a water molecule were proposed to be involved in opening the ring of F6P (Lee et al., 2001), and in the case of glucosamine 6- phosphate deaminase, a histidine was also found in a prime position for ring opening (Vincent et al., 2005). In most structures presented here the ring oxygen, O4, is within hydrogen bonding distance of His99/102 and this residue is the most likely candidate for donating a proton. The -furanose in the EcRpiB-H99N model has its O1 close to a water molecule, that is within hydrogen bonding distance of His134, and in the docked model of the -furanose in MtRpiB the hydroxyl group can be placed close to the equivalent His138. This suggests that in the case of the -anomer either the water or the second histidine will accept a proton from O1. In this scenario, the opening of an -furanose is more difficult, as apart from His99/102 there are no residues that could act as a base in the proximity of the hydroxyl group. However, His102 in the MtRpiB-All6P structure is seen to take on a slightly different angle, which is in a better position for exchanging a proton with O1 of the -pyranose, than with the ring oxygen O5. All6P is not this enzyme’s preferred substrate, but the structure still shows that His102 can adopt different conformations. Further, in both this structure and in EcRpiB-H99N, a water molecule is observed close

51 to the ring oxygen. In the case of the -furanose the roles of His99/102 and the water molecule could therefore be switched, with the water initiating ring opening by donating a proton to O4 of R5P.

Figure 28. The proposed reaction mechanism of the MtRpiB catalysed isomerisation starting with the -furanose of R5P.

For the actual isomerisation, the structures of MtRpiB with 4PEH and R5P clearly identify the residues that are involved. Glu75 is well placed for being the catalytic base and transferring a proton between C1 and C2, and Ser71 through its side chain Ser-OH/Ser-O- couple can catalyse the proton transfer between O1 and O2. Structural alignment with EcRpiB shows that Cys66, as had been proposed previously (Zhang, 2003b), will take on the role as base in this enzyme and that Thr68 will transfer the second proton. In the enzymes PGI and triose-phosphate isomerase (TIM), which catalyse

52 analogous isomerisation steps that are also thought to proceed via cis- enediolate intermediates, a glutamic acid is proposed to be the main catalytic base (Davenport et al., 1991; Arsenieva, 2002). Structures of these enzymes together with inhibitors similar to 4PEH (TIM with phosphoglycolohydrox- amic acid, 7TIM.pdb; and PGI with 5-phospho-D-arabinonohydroxamic acid, 1KOJ.pdb), show comparable relative orientations of the base in respect to the hydroxamic moiety of the ligand. On the other side of the ligand in the position of Thr68/Ser71, TIM has a histidine (His95) and in PGI there is a water molecule, which both are considered to be responsible for the proton shuttle between O1 and O2 of the reaction intermediate. In Paper II (Fig- ure 7) the overlap of these catalytic residues can be seen. The anion hole consisting of the backbone amide nitrogens of residues 67-71 in EcRpiB and 70-74 in MtRpiB together with the side chain of Asn100/103 can assist catalysis by stabilising the negative charge of the cis- enediolate high-energy intermediate. In the next step Cys66/Glu75 donate a proton to C1 thus forming the ketose, Ru5P. Finally, His99/102 or the water molecule donates a proton to O4. In the reverse reaction the same steps will occur in the opposite direction. Other conserved residues of the active site are implicated in substrate binding. Arg133/137, Arg137/141 and His10/12 are involved in binding the phosphate moiety. In MtRpiB there is an additional interaction from Arg113 and in EcRpiB Arg40 will at times take on such a role. Asp9/11 is in all structures seen to interact with O3, and will presumably help position the substrate for ideal interactions with the actual catalytic residues. The final conserved active site residue that can be assigned a role in substrate binding is Tyr43/46, which is in a good position for providing some stacking interactions with the sugar ring.

The All6P to Allu6P isomerisation The activity results presented in Paper IV show that EcRpiB can catalyse the isomerisation of All6P to Allu6P, whereas MtRpiB was found to be highly inefficient at this reaction. The structural results reveal why; the phosphate moiety has one mode of binding and in this position it is impossible for the larger 6-carbon sugar phosphate to be accommodated by the enzyme. EcR- piB binds the phosphate farther out and is even equipped with an extra phosphate-binding residue, Arg40. The isomerisation of All6P presumably progresses in the same way as the R5P to Ru5P reaction, with His99 and His134 involved in opening the ring of either - or -anomer, and with Cys66 and Thr68 taking care of the proton transfer. The opposite direction where Allu6P is converted to All6P is harder to visualise, at least for the ring-opening step. The ring form of Allu6P (Figure 6) will probably be placed with the ring oxygen, O5, in the same position as All6P and indeed the ring oxygen of R5P. This leaves a

53 larger group on C2, the position equivalent to C1 in the ribofuranose. Judg- ing by the EcRpiB structures, Allu6P could only be placed with the extra C1-O1H in the -ribofuranose position, otherwise it would clash with Tyr43. Both All6P and Allu6P were tested as inhibitors of MtRpiB, and Allu6P was found to be slightly poorer (Table 2). Even though these compounds will not be binding to MtRpiB in the exact same way as they would to EcRpiB it would be suspected that Allu6P is a less optimal substrate. In the activity study of C. thermocellum RpiB it was noted that the isomerisation was more favourable in the All6P to Allu6P direction than vice versa (Park, 2007). From a biological point of view it is also more difficult to explain the necessity of this reaction in E. coli. If Allu6P is converted to F6P by AlsE (Figure 5) for incorporation into the glycolytic pathway, and AlsE is assumed not to be able to convert All6P into F6P, there would be no reason for E. coli to produce All6P from Allu6P.

The reaction as catalysed by RpiA There are 11 different structures of RpiA in the PDB, which as mentioned in the Introduction has a different overall fold and active site when compared to RpiB. Several of the RpiA structures have been solved together with ligands, and that of Thermus thermofilus RpiA in complex with R5P in particular gives clues to the reaction mechanism in this form of the Rpi enzyme (Hamada et al., 2003). The sugar phosphate has, as in the case of MtRpiB, been found to bind in the linear form, with C1 and C2 clearly identifying the catalytic base, which as in MtRpiB, is a glutamic acid, Glu108. The equivalent residue had previously been identified as the main catalytic residue in spinach RpiA from mutational studies (Jung, 2000) and in the structure of EcRpiA (Zhang, 2003a). For the transfer of a proton between O1 and O2 of the ene-diolate intermediate, T. thermofilus RpiA has no residue in a similar position to Ser71/Thr68 of RpiB or His95 of TIM. However, as in RpiB, the negative charge of the reaction intermediate is thought to be stabilised by an oxyanion hole made up of backbone nitrogen atoms. In addition, there is a highly conserved lysine (Lys99) close to Glu108, which is thought to both stabilise the intermediate and assist the proton transfer from O2 via Glu108 to O1. This lysine could be one of the reasons for why EcRpiA has such a high catalytic rate compared to that of RpiB. For binding the phosphate moiety, RpiAs utilise a serine, a threonine and two lysines, which are all highly conserved. In the study describing the structure of EcRpiA in complex with the ring form of the inhibitor arabinose 5-phosphate, the ring opening step was proposed to be catalysed by Asp81 (Zhang, 2003a). This residue, when mutated to an alanine in the spinach RpiA, had been found to result in a severely diminished kcat when compared to the wild type enzyme. The Km had, on the other hand, remained close to

54 that of the wild type, suggesting a role in catalysis rather than binding (Jung, 2000). In Paper I we describe a virtual screening study of EcRpiA where a docked molecule of -arabinose 5-phosphate was found to bind in the same position as that of the inhibitor found in the structure; the excellent agreement can be seen in Figure 6 (C) of Paper I. In the same study both anomers of the ribofuranose 5-phosphate were docked and surprisingly seen to bind with the ring oxygen flipped by almost 180 degrees compared to how the docked inhibitor was placed. This brought O4 close to another asparagine (Asp84), a residue also proven to be important for catalysis in the spinach RpiA, although the effect of this mutant was not as great as for the Asp equivalent to Asp81 (Jung, 2000). RpiA of T. thermofilus was solved in complex with the linear form of arabinose 5-phosphate, in addition to the complex with R5P. These two linear molecules, which only differ in the conformation of the C2 hydroxyl group (cis in R5P and trans in arabinose 5-phosphate), were found to bind as mirror images of each other. As in the EcRpiA-arabinose 5-phosphate structure, O4 had an interaction with Asp86 (Asp81 in EcRpiA), while the same oxygen of the linear R5P was found to have a hydrogen bond to Asp89 (Asp84 in EcRpiA). These results nicely confirm our docking results. A mechanism for the ring-opening step can thus be visualised to involve Asp84 in the EcRpiA. Either this residue can both donate a proton to O4 and then accept one from O1 or a water molecule will assist in the second step, as was suggested in the original paper (Zhang, 2003a). To conclude, despite the large differences in the contributing residues that make up respective active sites, the mechanisms for the RpiA and RpiB catalysed isomerisations do have some features in common, for example, catalysis of ring opening, and a means for the important stabilisation of the transition state.

55 Conclusions and future perspectives

The only gene for an Rpi found in M. tuberculosis, rv2465c, belongs to the RpiB-type family and was, in this study, found to produce a protein that exhibits the expected isomerase activity. From several structures of MtRpiB together with various inhibitor molecules and with the substrate, a reaction mechanism for the isomerisation of R5P and Ru5P could be proposed, involving Glu75 and Ser71 as the main catalytic residues. Comparisons to the active site of EcRpiB could confirm the residues involved in isomerisation in this enzyme, some of which had been proposed previously (Zhang, 2003b). Mutational studies of the E. coli and T. cruzi RpiBs catalytic cysteine have verified its important role, and such studies would be desirable to perform with MtRpiB in the future. For the RpiBs it has been unclear whether or not the enzyme has to assist the ring-opening step in the isomerisation of R5P to Ru5P. In the MtRpiB structures, His102 was a candidate residue for assisting in the acid-base catalysis required for opening the furanose. The equivalent histidine was mutated in EcRpiB and gave rise to a protein exhibiting catalytic qualities that are in agreement with the proposed function for the residue, and the EcRpiB- H99N crystal structure further supports this theory. The complete ring- opening step is still not fully understood but it is thought to rely mainly on His99/102, and also involve His134/138 and a water molecule. In T. cruzi RpiB an H138A mutant showed little effect on kcat, and was concluded not to be involved in the ring-opening step. This could imply that the water molecule acts as an acid-base on its own. However, the fact that His138 is conserved throughout the RpiBs suggests it does have an important role in the structure. It would be interesting to mutate this residue in the EcRpiB or MtRpiB for cocrystallisation studies with R5P. If only the linear ligand is bound, it would indicate this residue plays no part in ring opening. EcRpiB had previously been proposed to be involved in allose utilisation, and would in such a case have allose-6-phosphate isomerase activity. Kinetic results presented here prove that the enzyme does have this activity, although when compared to the reaction with the shorter substrate, R5P, the kcat/Km is a little lower, suggesting that the latter activity is EcRpiBs primary function. MtRpiB was found to be a very poor catalyst of the All6P to Allu6P isomerisation. The structures of MtRpiB cocrystallised with 5PRA and All6P explain why; the active site simply has no space for a linear 6- carbon compound. Comparisons to the structure of EcRpiB reveals that sub-

56 tle changes in the active site of the two RpiBs enable the E. coli enzyme to bind the substrate lower down, thus creating enough room for the longer substrate to make the necessary contact with the catalytic residues. From sequence searches it could be concluded that all organisms discovered to date, with completed genomes, have at least one form of Rpi. RpiAs frequently have a lot higher activity compared to the RpiBs, so where both genes are found it must be assumed that RpiA plays the main role as the isomerase of R5P and Ru5P in the PPP. In these organisms RpiB, especially the Ec-like type, may have a second function and perform an isomerisation on a different sugar phosphate, for example All6P. Judging how EcRpiB is slightly worse at catalysing the isomerisation with All6P when compared to R5P it would seem as though the gene also is maintained purely as an emer- gency gene. If for some reason RpiA stops functioning, RpiB can step in and take over its activity. In M. tuberculosis and other organisms where only an RpiB is encountered the use of this enzyme must mainly be as a catalyst of the 5-carbon isomerisation. The reason why E. coli has an operon devoted to allose metabolism, remains a mystery, seeing how the sugar is extremely rare in nature. To our knowledge, apart from E. coli, only two other organisms (with completed genome sequences) have the als operon: Yersinia intermedia and Haemophi- lus somnus. Of the two, the operon of Y. intermedia is most similar to that of E. coli, with the same gene organization, and protein sequence identities ranging between 60 and 90%. Both these organisms also have an RpiA. Y. intermedia is a bakterium associated with waste water and can be found in humans with gastrointestinal disorders. H. somnus inhabits mucus mem- branes and can cause disease in humans but is mainly a pathogen of cattle (Challacombe et al., 2007). As E. coli is an intestinal resident, the biological implication of utilising allose for these organisms could be related to mam- mals’ inability to metabolise the sugar. In these bacteria’s surroundings the sugar may to a small extent be building up, which would give the organisms that can use the sugar a slight selective advantage. With the number of fully sequenced genomes rapidly increasing new genomics searches may identify other organisms that have this operon, which perhaps can help to build up a more complete picture of allose utilization. Parallels between the reaction mechanisms of RpiAs and RpiBs can be drawn despite the fact that the active site organisation is clearly different between the two Rpis. Varying inhibition patterns were also obtained in activity measurements of spinach RpiA and MtRpiB. The fact that the compound 4PEA was a several orders of magnitude weaker inhibitor of MtRpiB when compared to the related 4PEH, while both compounds were equally good at inhibiting RpiA, can be explained by Lys99, found close to the catalytic base of this enzyme. The lysine is thought to be involved in stabilising the negative charge of the reaction intermediate. The long and flexible residue will be expected to form an excellent interaction with the carboxylate

57 moiety of 4PEA. In MtRpiB, there is no such charged residue in a good position for creating an interaction; instead the enzyme relies mainly on the anion hole of backbone amides for stabilisation of the transition state, a fact that explains the slow catalytic rate observed for RpiBs. The 4PEA ligand is too short to come into contact with the anion hole, and thus has few interactions with the enzyme. Further inhibition studies of the two different Rpis identified 5PRH as the first compound that inhibits MtRpiB better than the spinach RpiA. The RpiB type active site is apparently better suited for accommodating the larger molecule, even though the MtRpiB in the crystal structure binds the hydrolysed compound (1 unit shorter) 5PRA. An idea for the future is to perform new virtual screens starting from the MtRpiB-5PRA complex in the hope of yielding compounds that can be used as selective and potent inhibitors of the M. tuberculosis enzyme.

58 Summary in Swedish

Strukturella och funktionella studier av ribos-5-fosfat isomeras B

Bakgrund Världshälsoorganisationen har uppskattat att under år 2005 dog runt 1.6 mil- joner människor av sjukdomen tuberkulos, som orsakas av bakterien Myco- bacterium tuberculosis (M. tuberculosis). Till detta kan läggas det skräm- made fakta, att varje sekund smittas någon i världen med bakterien (WHO, 2007). Som tur är utvecklar endast 10% av de smittade sjukdomen, men människor med nedsatt immunförsvar drabbas lättare än andra. Till dessa hör alla HIV bärare, och tuberkulos är därför ett mycket stort problem i länder där HIV epidemin är svår. Något som gör världssituationen än värre är att under de senare åren har resistenta bakteriestammar uppstått och mot dessa bakterier fungerar inte den vanliga tuberkuloskuren. För att få bukt på detta världshälsoproblem behövs nya mediciner, som dels kan ta död på de resistenta bakteriestammarna och dels kan korta ner läkemedelskuren, som i dag uppgår till drygt ett halvår. Ett viktigt steg i vägen mot nya tuberkulosmediciner är att lära sig mer om hur organismen fungerar. Ju mer vi vet om bakterien, dessto lättare blir det att hitta ett nytt sätt att döda den på. I den här avhandlingen presenteras enzymet ribos-5-fosfat isomeras (Rpi), som är inblandad i en av cellens mer centrala metaboliska reaktionsvägar, den så kallade pentosfosfatvägen. Förutom Rpi från M. tuberculosis har även Rpi från Escherichia coli (E. coli) studerats. E. coli var en av de första organismerna, som forskare bör- jade studera, så för denna organism finns en hel del vetskap om dess metaboliska vägar. Men för M. tuberculosis är det fortfarande mycket som är okänt, även om bakterien identifierades redan 1882 av Robert Koch. Den främsta orsaken till detta är för att bakterierna är smittsamma och därför kräver spe- ciella lab för att man ska få arbeta med dem. En annan orsak är att det tar ett dygn för bakteriecellen att föröka sig, vilket kan jämföras med drygt 20 min- uter för en E. coli cell. Reaktionen som Rpi katalyserar är en så kallad isomerisering vilket inne- bär att en molekyl ribos 5-fosfat görs om till en molekyl ribulos 5-fosfat (se figur 1). Ribos 5-fosfat kallas här för substrat och ribulos 5-fosfat för produkt. En isomerisering är en reaktion, som kan gå åt bägge hållen, ribos 5-

59 fosfat kan således både vara substrat eller produkt. Av dessa två molekyler kan ribos 5-fosfat sägas vara den viktigaste då den är en av huvudkomponen- terna i DNA.

Figur 1. Isomeriseringen som katalyseras av Rpi. I rutan visas allos 6-fosfat som är en kolatom längre än ribos 5-fosfat. Siffrorna till vänster om kolen visar namnkon- ventionen.

I E. coli bakterien finns det två former av Rpi: en A och en B-form. RpiA är den form som bakterien använder sig av i vanliga fall och är också den form som vi människor har i vår genuppsättning. RpiB används bara av E. coli i krissituationer till exempel om genen för RpiA har skadats på något sätt. En annan krissituation kan vara om bakterien bara har tillgång till en viss sorts näringsämen, som den vanligen inte har tillgång till. Ett sådant näringsämne är det ovanliga sockret allos. Allos är väldigt likt det mer vanliga sockret ribos. Den enda skillnaden är att allos är uppbyggt av sex stycken kol-atomer medan ribos består av fem, se figur 1. Allos är som sagt ett ovanligt socker som bara påträffats naturligt i vissa sällsynta växtarter. Det är också ett socker som de flesta organismer inte verkar kunna till- godogöra sig, vi människor kan inte få ut någon energi från allos eller an- vända sockret som byggsten i våra celler. Däremot, finns det studier som påvisat att E. coli celler kan leva i en provrörsmiljö där allos är deras enda kol-källa. För att E. coli ska klara av detta har RpiB pekats ut som ett av de enzym som hjälper till att bryta ner allosen genom att isomerisera allos 6- fosfat till allulos 6-fosfat. Ett av målen med det arbete som presenteras här har varit att ta reda på om E. colis RpiB verkligen har denna mer ovanliga funktion. Bakterien M. tuberculosis har endast RpiB i sin genuppsättning. Tuberku- los RpiB genen skiljer sig på ett ställe i gensekvensen markant från E. coli RpiB genen. Skillnaden består egentligen bara av en utbytt aminosyrarest, en cystein i E. coli RpiB som i M. tuberculosis RpiB är en glycin. Denna cystein anses vara den aminosyrarest som gör så att själva katalysen som RpiB utför kan ske. Ett viktigt mål med mitt arbeta har därför varit att visa att

60 RpiB genen i M. tuberculosis verkligen gav upphov till ett funktionsdugligt RpiB enzym.

Använda metoder För att svara på dessa frågor har två skilda metoder använts: struktur- bestämning med hjälp av röntgenkristallografi och enzymkinetik. Enzymi- netik går ut på att man mäter hur pass aktivt ett enzym är när man ger det ett substrat. Mätningen kan ske på flera olika sätt och beror ofta på hur produkten ser ut. I det här fallet har aktiviteten på enzymet RpiB mätts vars produkt ribulos 5-fosfat har egenskapen att den absorberar ljus vid våglängden 290 nm. Substratet ribos 5-fosfat gör inte det. När ribos 5-fosfat omvandlas till ribulos 5-fosfat sker en förändring i ljusabsorbans vilket kan mätas med en spektrofotometer. På så sätt kan man bestämma hur mycket av produkten som enzymet bildar under en viss tid. Detta säger en del om enzymets aktivitet. Vid strukturbestämning av proteiner gör man sig en tredimensionell bild av hur de aminosyror, som bygger upp proteinet förhåller sig till varandra. Tillvägagångssättet för att ta reda på en proteinstruktur börjar med att en stor mängd av det intressanta proteinet produceras för att sedan användas för att göra proteinkristaller. Proteinet framställs med hjälp av en speciellt fram- tagen laboratoriestam av E. coli. Genom att komplettera laboratoriebakteri- ernas DNA med genen för det önskade proteinet, i det här fallet RpiB, kan bakterierna stimuleras att överproducera RpiB. Litervis med bakterier odlas upp. Bakterierna slås sedan sönder vilket frigör alla proteiner som är inuti bakterien. Därefter kan vi med olika metoder rena fram RpiB från alla andra proteiner som finns i laboratoriebakterien normalt.

Figur 2. En schematisk bild över tillvägagångssättet för att producera en proteinkristall.

När en ren RpiB-proteinlösning tagits fram försöker man få protein- molekylerna att bilda kristaller. Detta görs genom att koncentrera proteinet till hög koncentration, och sedan blanda små volymer av protein med många olika lösningar med varierande salthalt, pH och glycerolhalt. När rätt bet- ingelser råder kommer proteinet att packas så att kristaller bildas. En sche-

61 matisk bild över stegen från bakterieodling till färdig proteinkristall kan ses i figur 2. Proteinkristallerna bestrålas sedan med röntgen. Elektronerna runt atomerna som bygger upp proteinmolekylen kommer att böja av röntgen- strålen åt olika håll. De avböjda strålarna fångas upp som prickar på en skärm och utifrån mönstret som bildas på skärmen kan man med matemati- ska formler (sk Fouriertransformer) räkna ut hur elektrontätheten i ett protein ser ut. I elektrontätheten bygger vi sedan in aminosyror och får så en bild av hur proteinet ser ut. Metoden röntgenkristallografi kan urskilja detaljer ner till atomnivå. Oftast syns inte de minsta atomerna, väteatomerna, utan för t.e.x vattenmolekyler (H2O) syns bara syreatomen. Den tredimensionella strukturen på enzymet är viktig för den kan ge ledtrådar i hur enzymet utför sin uppgift i cellen, vilket är starkt förknippat till de aminosyra rester, som sitter i den så kallade aktiva ytan av enzymet. Det är just denna aktiva yta i RpiB som jag har studerat.

Resultat Enzymaktivitetsmätningar av RpiB från M. tuberculosis visade att enzymet visst är ett RpiB och dess katalytiska egenskaper kunde jämföras med dem som tidigare mätts för E. coli RpiB. Tuberkulosenzymet kristalliserades med hjälp av en fosfatbuffert och vi kunde sedan lösa strukturen, som visade sig vara mycket lik strukturen av E. coli RpiB. Nu kunde vi också förklara var- för aminosyrasekvenserna skillde sig mellan de två enzymen. Den viktiga katalytiska aminosyran som saknades i tuberkulos RpiB, bestod av en annan aminosyra, en glutaminsyra, som fanns på en annan plats i sekvensen men som har sin funktionella atom på samma ställe som cysteinen har i E. coli enzymet. Tuberkulos RpiB kristalliserades sedan tillsammans med substratet ribos 5-fosfat. Denna struktur visade hur substratet passar i enzymets aktiva yta, se Figur 3, och när vi såg vilka amiosyrarester som ligger nära substratet kunde vi förstå hur enzymet använder sig av dessa aminosyror för att utföra sin katalys. Viktiga aminosyror i katalysen är den precis omnämnda glutamin- syran som har nummer 75 i aminosyra sekvensen, serin nr 71 och histidin nr 102. Enzymmätningar med E. coli RpiB för att ta reda på om enzymet kunde katalysera en reaktion då allos 6-fosfat är substratet gjordes av våra medar- betare i Frankrike. De kunde visa att E. coli enzymet har denna sekundära aktivitet. Men samma mätningar med tuberkulos RpiB visade att denna form av RpiB inte har den sekundära aktiviteten. Med hjälp av strukturer har jag kunnat visa varför. Tuberkulos RpiB kristalliserades då tillsammans med både allos 6-fosfat och en annan molekyl som är mycket lik allos 6-fosfat, 5PRH. Jag försökte även kristallisera E. coli RpiB med dessa två substanser men de försöken misslyckades. Strukturerna med de två molekylerna som

62 har 6 kol-atomer tydde på att de var för långa för att riktigt få plats i den aktiva ytan hos tuberkulosenzymet. I en jämförelse med E. coli RpiB strukturen blev det tydligt att dennes aktiva yta är något större vilket skulle förk- lara varför den kan katalysera en reaktion med det längre substratet.

Figur 3. Tuberkulos RpiBs aktiva yta med en molekyl av ribos 5-fosfat. Det grå rutmönstret symboliserar elektrontäthetskartan, där aminosyrorna har byggts in. Kolatomer är färgade i svart, likaså fosfatatomen i ribos 5-fosfat. Syre- och kvävea- tomer är grå. Pilarna pekar ut de tre viktigaste aminosyrorna i katalysen.

I enzymaktivitetsstudier kunde jag visa att molekylen 5PRH inhiberar RpiBs aktivitet. Detta jämfördes med aktivitetsstudier av RpiA där det visade sig att 5PRH är en sämre hämmare. Eftersom vi människor har en RpiA och M. tuberculosis har en RpiB och dessa två uppvisade olika inhibi- tionsmönster med 5PRH, skulle denna molekyl kunna vara en bra start för en strukturbaserad läkemedelsdesign. Strukturen med tuberkulos RpiB och 5PRH skulle då användas för att med hjälp av datorprogram skräddarsy nya mer läkemedelslika molekyler, som skulle kunna binda ännu bättre till RpiB och hämma den från att utföra sin uppgift i cellen. Med hjälp av mina resultat har vi lyckats förstå (åtminstone till viss del) hur en liten del av två bakteriers metaboliska reaktionsvägar fungerar. I fallet M. tuberculosis, skulle resultaten i framtiden kunna leda till ett bättre och säkrare läkemedel mot tuberkulos.

63 Acknowledgements

First of all for superb supervision: Thanks Sherry for all the help in the last couple of weeks and years, and for teaching me the snoopy dance (weeeee!), Torsten for positive energy, encouragement with the process of writing, great crystallisation expertise, and of course all the genombrott, Alwyn for the intuitive program O, for looking out for people, employing a biochemist from Arizona, and letting me know each time England has lost a game of rugby, cricket, football, you name it.

A great thanks to collaborators: To Laurant, Emmanuel, and Sandrine thanks for wonderful teamwork with the RpiB projects. I especially want to thank Laurent for last minute answers to my questions despite being on holiday!

The Medicinal Chemistry gang: Anders, Johanna, and Daniel for RNR struggles and a paper well done, Micael for a neverending flow of ideas for RpiB.

Thank you Fredrik, Andrea and Lotta for running the Protein Technology lab with RpiB, and for being so organised!! Thanks to all the students who made the mutants.

A thanks goes up to Umeå and Artur for providing me with the protein from hell, and an extreme learning experience. I would also like to thank Lars, Pelle, and Vladimir for all the help with getting the RNR assay started.

To the RAPID-group: Thank you Angels for great times in the girl’s office. Lena, keeper of big secrets, for understanding the value of colour matching, always being up for a dance and a bit of mischief directed at our Alina, thanks for the adventure spirit and not being shy about the stories you tell. Thanks Nina for excellent help and friendship in Auckland, cat talk and bringing pottery to a new di- mension of book-keeping, Daniel for co-Beer Club royalty and that evil wee grin, Anna-hoppar-som-en-hare uppför väggarna, and fellow shorty, thanks for Couplings and all the skitsnack (I can too you know), Christofer, for swimming stories (brrrr) and of course Glassklubben! I scream for ice cream helped me through this summer. Wojtek thanks for overall cynicism, science discussions and Christmas spirit, Anna for great company at that

64 course so long ago, Henrik for lunchtime chats and your keenness for design, Adrian for trust and trying to teach me salsa, and finally Nisse, master of excel spread sheets, thanks for thesis criticism, heaps of illustration help, and all the encouragement and kicks in the behind these past weeks.

Past RAPID members: Thanks Evalena for black and “hockeybrudar” or was it brud-brudar? For thesis-reading help, stopping me from flaming war, and I guess ultimately for getting me into this sh…t! Eva for being the best exjobbare, and taking cake baking to different dimensions, Erik for maths expertise, teaching me some kinetics and great dancing, Linus for RNR enthusiasm, and cookie competitions with Ulrika thank you both! Patrik klät- trar-som-en-duracell, for climbing energy with synchrotron trips, Martin Hö for riboträsk and great ideas, Jimmy for all kinds of help in the lab and madness on the go-cart track, Emma for perfume and go on the dance floor, and Seved for 37 degree room concerts and steeling leaves off shop plants, I still have one in our bedroom window.

The rest of the xray gang: Magnus Danzig for splatter discussions and sav- ing the Snowman, Pavel for the ghost in the bottle, Lars for asking if I’m stubborn because that’s all it takes to become a scientist – I’m not. Tex for all the crystallization tips; “flush it down the toilet”, Mark for outdoor fun and gear, and the colour of xtrack, Henke for the great beer, Gerard for not killing me at innebandy, and for Bohemian Rapsody (I hope!), Marian for homegrown wine, Fariborz for the axes of evil, Maria for househunting stories, Karin mass producer of wonderful cups, fjordblå is the colour! Thanks for lab tips and crystallisation tricks. Calle thanks for Mustasch ma- nia frenzy, Ragger for G and T (literally) and Sweden Rock, Gooster for never knowing where your climbing gear is and for impossible but fun boul- der problems. Andrea for parties, dinners and trips to Romme, Andreas C for dropping me on my head, Fredrik for Roskilde and protein technology lab enthusiasm, Urszula for cocktails, Martin W for booking the hostel over Hooters and innebandy organisation, Louise for skvaller and finding out everything about everybody, Rosie for some more skvaller, and Malin too! Janos for always asking “How’s life?” The Åqvist group: Johan, Viktor, Martin, Fantomen, Göran (welcome to Xray!), Sinisia, Stefan, Martin, Jens (does “drillborr” qualify for the list?), Fredrik, Isabella, Lars, and Bjørn for Beer Club, lunch and fika discussions, which used to be about whisky and poker but now seem to be more about babies..?

Hasse, Deva, Sherry, Ulla, and Stefan thanks for a great biochemistry course VT1998, which inspired me to want to learn crystallography in the first place. Thanks Talal for never giving up on a climb, Jenny, for synchrotron trips and shoe shopping persuasion, Al, fearless climber, for being time optimist number 2 (I think I may qualify for 3rd place and for 1st I shall men- tion no names), Lotta for Beer Club fun, outdoor revels, great dinners, and

65 making me a little worried about “tvättstugelappar”. Anatoly for stories of vodka and radio-activity and for lending me your cat, Glareh for glada till- rop and for keeping me company at odd hours, Jerry for styrdans, Saeid for suduko marathon, Sanjee for asking how its going, Agata for late night synchrotron chats, cool T-shirts, sweet key rings and holiday organisation tips. Erling and Christer for computer and MAR help, for fika company and old time party anecdotes. Inger, Mats, Anton, Michiel, Kaspars, Marvin, Al- exandra, Nic, Filipe, Erik, Andreas K, Bror, Martin - Ace of Spades- Svenda, Sara, Wimal, Jonas, Lotta, Ulrika, Susanna, Ellinor, Margareta, Linda, David, Tom, Gunnar, Elin, Gunilla, Anke, Remco, Minyan and alla@xray thanks 4 making BMC such an enjoyable place 2 be!

Thank you: Sofia for Julbords organisation encouragement, but sorry you couldn’t make it, Ingrid for keeping track of every important occasion, Sigrid for calm efficiency- thanks for the thesis organisation help! Susanne, Åsa, Christina for answering my stupid questions.

Ted and Heather, thanks for making me feel so at home at SBL. Thanks Vic for crystallography inspiration, Shaun for schmooze tips, Graeme for he- roes, Randall, Moyra, Rochelle, Andrew, Jenny, Dave for Retro Night, Chris, and James (I’m not short!), Ivan, Richard, Harriet, May, Elaine, Matthias, Louise, Tara, Taru, Xiaolin, Nayden, Mel, Jodie, Simon and Simon, Steve, Miriam, Peter, Peter, and Pete thanks for making my stay in Auckland so wonderfully great!

Cathy, Steve, Kade, Rachel, and Tim thanks for super times at 10 Eldon Rd, big fish, little fish, cardboard box! Maria for airplane company and New Zealand fun, Scott and Emma for climbing excursions.

Tack till: Åsa för en finnemang tid på Djäknegatan och ovanliga saker i köksfönstret, Linda för den första lyxmiddagen med tentaplugg, för whisky sour och Simpsons. AnneMarie, Kerstin, Nils, Ulrica och Peter för Harry Potter äventyr och för en glad tid på bioteknologiprogrammet! Tack Jonas, Johanna och Tova för höga berg, det blir nog ingen trad-are av mig men Italien var schysst ändå! Tack Per (jag skulle ha lytt ditt råd) och Ann-Sofie, Pelle och Cynthia för trevliga middagar och sång in på småtimmarna.

Elin och Malin för att vi hängt ihop så länge, för kortspel, långa Knivsta promenader, nära-döden upplevelser, gömma Jäger, medeltidsväx, T-fat, shoppingrundor och fikastunder, Kibibi, Tigi Lady, Öli och Suzi för finfina förfester till ‘Bad Medicine’. Tack till Henrik, Mårten, Hasse med V och T, Lina och Kim för grillkvällar och kräftskivor (med pyramider), allsång och dans till det mesta, Kibibi och Suz för Sweden Rock upptåg.

66 Tack till Familjen Nilsson i Styrsta för att ni är så många och för att ni är så goa och för en otrolig fight över en elefant!

Thanks to The McDonalds for being substitute cousins and for old times in Knivsta, Mamie and Grandpa Gordon, old-aged-power-couple! For inspiration in the methods of growing up, I thank you. 1000 butterfly-kisses to Grandpa. Sven, I’m sorry I never had the sense to get to know you.

Susanne, thanks for pictionary telepathy and keeping me from tantiness.

Thanks Mum and Dad for all your love and encouragement and for being the greatest! Mum, sweety-darling, for answering all my English questions and Dad for inspiring me, at the age of 10, to think that my dream job was to become a scientist…

And finally a huge thanks to Nisse my darling Dr Nice! And since you know best, why should I reinvent what has already been said so well – for love and fun and (adventures, 5 babies, a house with a garage, and a garden, a katt, a dog? and lots of) belay to come.

That’s all folks!

Annette

67 References

Altschul, S. F., Madden, T. L., Schaffer, A. A., Zhang, J., Zhang, Z., Miller, W. and Lipman, D. J. (1997). Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 25 (17): 3389-3402. Arsenieva, D., Hardré, R., Salmon, L. and Jeffery, C. J. (2002). The crystal structure of rabbit phosphoglucose isomerase complexed with 5- phospho-D-arabinonohydroxamic acid. Proc Natl Acad Sci U S A 99 (9): 5872-5877. Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T. N., Weissig, H., Shindyalov, I. N. and Bourne, P. E. (2000). The Protein Data Bank. Nucleic Acids Res. 28 (1): 235-242. Braibant, M., Gilot, P. and Content, J. (2000). The ATP binding cassette (ABC) transport systems of Mycobacterium tuberculosis. FEMS Mi- crobiol Rev 24 (4): 449-467. Burgos, E. and Salmon, L. (2004a). Synthesis and evaluation of new 4- phospho-D-erythronic acid derivatives as competitive inhibitors of spinach ribose-5-phosphate isomerase. Tetrahedron Letters 45 (4): 753-756. Burgos, E. and Salmon, L. (2004b). Synthesis and kinetic evaluation of 4- deoxy-4-phosphonomethyl-D-erythronate, the first hydrolytically stable and potent competitive inhibitor of ribose-5-phosphate isomerase. Tetrahedron Letters 45 (17): 3465-3469. CCP4 - Collaborative Computational Project Number 4. (1994). The CCP4 Suite: Programs for Protein Crystallography. Acta Crystallogr. D50: 760-763. Challacombe, J. F., Duncan, A. J., Brettin, T. S., Bruce, D., Chertkov, O., Detter, J. C., Han, C. S., Misra, M., Richardson, P., Tapia, R., Thayer, N., Xie, G. and Inzana, T. J. (2007). Complete genome sequence of Haemophilus somnus (Histophilus somni) strain 129Pt and comparison to Haemophilus ducreyi 35000HP and Haemophilus influenzae Rd. J Bacteriol 189 (5): 1890-1898. Chari, V. M., Grayer-Barkmeijer, J., Harborne, J. B. and Oesterdahl, B. G. (1981). An acylated allose-containing 8-hydroxyflavone glycoside from Veronica filiformis. Phytochemistry 20: 1977-1979. Chaudhuri, B. N., Ko, J., Park, C., Jones, T. A. and Mowbray, S. L. (1999). Structure of a D-allose binding protein from Escherichia coli bound to allose at 1.8 Å resolution. J. Mol. Biol. 286: 1519-1531. Cole, S. T., Brosch, R., Parkhill, J., Garnier, T., Churcher, C., Harris, D., Gordon, S. V., Eiglmeier, K., Gas, S., Barry, C. E., 3rd, Tekaia, F.,

68 Badcock, K., Basham, D., Brown, D., Chillingworth, T., Connor, R., Davies, R., Devlin, K., Feltwell, T., Gentles, S., Hamlin, N., Hol- royd, S., Hornsby, T., Jagels, K., Barrell, B. G. and et al. (1998). Deciphering the biology of Mycobacterium tuberculosis from the complete genome sequence. Nature 393 (6685): 537-544. Cornish-Bowden, A. (2004). Fundamentals of Enzyme Kinetics. London, UK, Portland Press Ltd. D'Arcy, A., Villard, F. and Marsh, M. (2007). An automated microseed matrix-screening method for protein crystallization. Acta Crystallogr D Biol Crystallogr 63 (Pt 4): 550-554. Danley, D. E. (2006). Crystallization to obtain protein-ligand complexes for structure-aided drug design. Acta Crystallogr D Biol Crystallogr 62 (Pt 6): 569-575. Davenport, R. C., Bash, P. A., Seaton, B. A., Karplus, M., Petsko, G. A. and Ringe, D. (1991). Structure of the triosephosphate isomerase- phosphoglycolohydroxamate complex: an analogue of the intermediate on the reaction pathway. Biochemistry 30 (24): 5821-5826. David, J. and Wiesmeyer, H. (1970). Regulation of ribose metabolism in Escherichia coli. II. Evidence for two ribose-5-phosphate isomerase activities. Biochim Biophys Acta 208: 56-67. DeLano, W. L. (2002). The PyMOL Molecular Graphics System. http://www.pymol.org Dische, Z. and Borenfreund, E. (1951). A new spectrophotometric method for the detection and determination of keto sugars and trioses. J Biol Chem 192 (2): 583-587. Essenberg, M. K. and Cooper, R. A. (1975). Two ribose-5-phosphate isom- erases from Escherichia coli K12: partial characterisation of the enzymes and consideration of their possible physiological roles. Eur J Biochem 55: 323-332. Feierberg, I. and Åqvist, J. (2002). Computational modeling of enzymatic keot-enol isomerization reactions. Theor Chem. Acc 108: 71-84. Franke, F. P., Kapuscinsky, M. and Lyndon, P. (1985). Synthesis and N.M.R. characterization of intermediates in the L-type pentose phosphate cycle. Carbohydrydrate Research 143: 69-76. Gibbins, L. N. and Simpson, F. J. (1964). The incorporation of D-allose into the glycolytic pathway by Aerobacter aerogenes. Can. J. Microbiol. 10: 829-836. Grochowski, L. L., Xu, H. and White, R. H. (2005). Ribose-5-phosphate biosynthesis in Methanocaldococcus jannaschii occurs in the absence of a pentose-phosphate pathway. J Bacteriol 187 (21): 7382- 7389. Hall, T. A. (1999). BioEdit: a user-friendly biological sequence alignmnet editor and analysis program for Windows 95?98/NT. Nucl. Acids Symp. Ser. 41: 95-98. Hamada, K., Ago, H., Sugahara, M., Nodake, Y., Kuramitsu, S. and Miyano, M. (2003). Oxyanion hole-stabilized stereospecific isomerization in ribose-5-phosphate isomerase (Rpi). J Biol Chem.

69 Holm, L. and Sander, C. (1993). Protein structure comparison by alignment of distance matrices. J. Mol. Biol. 233: 123-138. Hough, L. and Stacey, B. E. (1966). Variation in the allitol content of Itea plants during photosynthesis. Phytochemistry 5: 171-175. Hove-Jensen, B. and Maigaard, M. (1993). Escherichia coli rpiA gene encoding ribose phosphate isomerase A. J. Bacteriol. 175: 5628-5635. Huck, J. H., Verhoeven, N. M., Struys, E. A., Salomons, G. S., Jakobs, C. and van der Knaap, M. S. (2004). Ribose-5-phosphate isomerase de- ficiency: new inborn error in the pentose phosphate pathway associated with a slowly progressive leukoencephalopathy. Am J Hum Genet 74 (4): 745-751. Iida, A., Harayama, S., Iino, T. and Hazelbauer, G. L. (1984). Molecular cloning and characterization of genes required for ribose transport and utilization in Escherichia coli K-12. J. Bacteriol. 158 (#2): 674- 682. Jeffery, C. J., Hardre, R. and Salmon, L. (2001). Crystal structure of rabbit phosphoglucose isomerase complexed with 5-phospho-D- arabinonate identifies the role of Glu357 in catalysis. Biochemistry 40 (6): 1560-1566. Jensen, S. R., Mikkelsen, C. B. and Nielsen, B. J. (1981). Iridoid mono- and di-glycosides in Mentzelia. Phytochemistry 20: 71-83. Jones, T. A. A, yaap, asap, @#*? A set of averaging programs. (1992). Mo- lecular Replacement. E. J. Dodson, S. Glover and S. Wolf. Warring- ton, UK, Darebury Laboratory. Jones, T. A., Zou, J.-Y., Cowan, S. W. and Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A47: 110- 119. Jung, C. H., Hartman, F. C., Lu, T. Y. and Larimer, F. W. (2000). D-Ribose- 5-phosphate isomerase from spinach: heterologous overexpression, purification, characterization, and site-directed mutagenesis of the recombinant enzyme. Arch Biochem Biophys 373 (2): 409-417. Kaufmann, S. H. and Schaible, U. E. (2005). 100th anniversary of Robert Koch's Nobel Prize for the discovery of the tubercle bacillus. Trends Microbiol 13 (10): 469-475. Kim, C., Song, S. and Park, C. (1997). The D-allose operon of Escherichia coli K-12. J. Bacteriol. 179: 7631-7637. Kim, I., Kim, E., Yoo, S., Shin, D., Min, B., Song, J. and Park, C. (2004). Ribose utilization with an excess of mutarotase causes cell death due to accumulation of methylglyoxal. J Bacteriol 186 (21): 7229-7235. Kleywegt, G. J. and Jones, T. A. Halloween... Masks and Bones. (1994). From First Map to Final Model. S. Bailey, R. Hubbard and D. Waller. Warrington, UK, SERC Daresbury Laboratory: 59-66. Kleywegt, G. J. and Jones, T. A. (1999). Software for handling macromolecular envelopes. Acta Crystallogr D Biol Crystallogr 55 (Pt 4): 941-944.

70 Kleywegt, G. J., Zou, J. Y., Kjeldgaard, M. and Jones, T. A. Around O. (2001). International Tables for Crystallography: Crystallography of Biological Macromolecules. M. G. Rossmann and E. Arnold. Dordrecht, Netherlands, Kluwer Academic Publishers. F: 353-356. Koch, R. (1882). Die Aetiologie der Tuberculose (Nach einem in der physi- ologischen Gesellschaft zu Berlin am 24.März cr. gehaltenem Vor- trage). Berliner klin. Wochenschr. 19: 221-230. Lee, J. H., Chang, K. Z., Patel, V. and Jeffery, C. J. (2001). Crystal structure of rabbit phosphoglucose isomerase complexed with its substrate D- fructose 6-phosphate. Biochemistry 40 (26): 7799-7805. Linton, K. J. and Higgins, C. F. (1998). The Escherichia coli ATP-binding cassette (ABC) proteins. Mol Microbiol 28 (1): 5-13. Lopilato, J. E., Garwin, J. L., Emr, S. D., Silhavy, T. J. and Beckwith, J. R. (1984). D-Ribose metabolism in Escherichia coli K-12: genetics, regulation, and transport. J. Bacteriol. 158 (#2 of May): 665-673. Luft, J. R. and DeTitta, G. T. (1999). A method to produce microseed stock for use in the crystallization of biological macromolecules. Acta Crystallogr D Biol Crystallogr 55 (Pt 5): 988-993. Menavuvu, B. T., Poonperm, W., Leang, K., Noguchi, N., Okada, H., Mori- moto, K., Granstrom, T. B., Takada, G. and Izumori, K. (2006). Ef- ficient biosynthesis of D-allose from D-psicose by cross-linked recombinant L-rhamnose isomerase: separation of product by ethanol crystallization. J Biosci Bioeng 101 (4): 340-345. Meredith, T. C. and Woodard, R. W. (2003). Escherichia coli YrbH is a D- arabinose 5-phosphate isomerase. J Biol Chem 278 (35): 32771- 32777. Miller, B. G. and Raines, R. T. (2004). Identifying Latent Enzyme Activi- ties: Substrate Ambiguity within Modern Bacterial Sugar Kinases. Biochemistry 43 (21): 6387-6392. Miller, B. G. and Raines, R. T. (2005). Reconstitution of a defunct glycolytic pathway via recruitment of ambiguous sugar kinases. Biochemistry 44 (32): 10776-10783. Miosga, T. and Zimmermann, F. K. (1996). Cloning and characterization of the first two genes of the non-oxidative part of the Saccharomyces cerevisiae pentose-phosphate pathway. Curr Genet 30: 404-409. Murshudov, G. N., Vagin, A. A. and Dodson, E. J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr. D53: 240-255. Navaza, J. (1994). AMoRe: an automated package for molecular replacement. Acta Crystallogr. A50: 157-163. Park, C. S., Yeom, S. J., Kim, H. J., Lee, S. H., Lee, J. K., Kim, S. W. and Oh, D. K. (2007). Characterization of ribose-5-phosphate isomerase of Clostridium thermocellum producing D: -allose from D: -psicose. Biotechnol Lett. Perold, G. W., Beylis, P. and Howard, A. S. (1973). Metabolites of pro- teaceae. 8. The occurrence of (+)-D-allose in nature: rubropilosin

71 and pilorubrosin from Protea rubropilosa Beard. J Chem Soc [Perkin 1] 6: 643-649. Perrakis, A., Sixma, T. K., Wilson, K. S. and Lamzin, V. S. (1997). wARP: improvement and extension of crystallogaphic phases by weighted averaging of multiple-refined dummy atomic models. Acta Crystal- logr. D53: 448-455. Pierce, J., Serianni, A. S. and Barker, R. (1985). Anomerization of furanose sugars and sugar phosphates. J. Amer. Chem. Soc. 107: 2448-2456. Poulsen, T. S., Chang, Y. Y. and Hove-Jensen, B. (1999). D-Allose catabolism of Escherichia coli: involvement of alsI and regulation of als regulon expression by allose and ribose. J Bacteriol 181 (22): 7126- 7130. Ramakrishnan, T., Murthy, P. S. and Gopinathan, K. P. (1972). Intermediary metabolism of mycobacteria. Bacteriol Rev 36 (1): 65-108. Ryu, K. S., Kim, C., Kim, I., Yoo, S., Choi, B. S. and Park, C. (2004). NMR application probes a novel and ubiquitous family of enzymes that al- ter monosaccharide configuration. J Biol Chem 279 (24): 25544- 25548. Schray, K. J., Benkovic, S. J., Benkovic, P. A. and Rose, I. A. (1973). Cata- lytic reactions of phosphoglucose isomerase with cyclic forms of glucose 6-phosphate and fructose 6-phosphate. J Biol Chem 248 (6): 2219-2224. Sigrell, J. A., Cameron, A. D., Jones, T. A. and Mowbray, S. L. (1998). Structure of Escherichia coli ribokinase in complex with ribose and dinucleotide determined to 1.8 Å resolution: insights into a new family of kinase structures. Structure 6: 183-193. Skinner, A. J. and Cooper, R. A. (1974). Genetic studies on ribose-5- phosphate isomerase mutants of Escherichia coli K-12. J. Bacteriol. 118: 1183-1185. Sorensen, K. I. and Hove-Jensen, B. (1996). Ribose catabolism of Es- cherichia coli: characterization of the rpiB gene encoding ribose phosphate isomerase B and of the rpiR gene, which is involved in regulation of rpiB expression. J. Bacteriol. 178: 1003-1011. Sprenger, G. A. (1995). Genetics of pentose-phosphate pathway enzymes of Escherichia coli K-12. Arch Microbiol 164: 324-330. Stern, A. L., Burgos, E., Salmon, L. and Cazzulo, J. J. (2007). Ribose-5- phosphate isomerase type B from Trypanosoma cruzi: kinetic prop- erties and site-directed mutagenesis reveal information about the reaction mechanism. Biochem J 401 (1): 279-285. Thompson, J. D., Higgins, D. G. and Gibson, T. J. (1994). CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, positions-specific gap penalties and weight matrix choice. Nucleic Acids Res. 22: 4673-4680. Ullrey, D. B. and Kalckar, H. M. (1991). Search for cellular phosphorylation products of D-allose. Proc. Natl. Acad. Sci. USA 88: 1504-1505.

72 Vincent, F., Davies, G. J. and Brannigan, J. A. (2005). Structure and kinetics of a monomeric glucosamine 6-phosphate deaminase: missing link of the NagB superfamily? J Biol Chem 280 (20): 19649-19655. Wheeler, P. R. and Blanchard, J. S. General Metabolism and Biochemical Pathways of Tubercle Bacilli. (2005). Tuberculosis and the Tubercle Bacillus. S. T. Cole, K. D. Eisenach, D. N. McMurray and W. R. Ja- cobs. Washington, DC, ASM Press: 309-339. WHO (2007). Tuberculosis, Fact Sheet No 104, Revised March 2007. http://www.who.int Wood, T. (1970). Spectrophotometric assay for D-ribose-5-phosphateketol- isomerase and for D-ribulose-5-phosphate 3-epimerase. Anal Bio- chem 33 (2): 297-306. Xu, Q., Schwarzenbacher, R., McMullan, D., von Delft, F., Brinen, L. S., Canaves, J. M., Dai, X., Deacon, A. M., Elsliger, M. A., Eshagi, S., Floyd, R., Godzik, A., Grittini, C., Grzechnik, S. K., Jaroszewski, L., Karlak, C., Klock, H. E., Koesema, E., Kovarik, J. S., Kreusch, A., Kuhn, P., Lesley, S. A., Levin, I., McPhillips, T. M., Miller, M. D., Morse, A., Moy, K., Ouyang, J., Page, R., Quijano, K., Robb, A., Spraggon, G., Stevens, R. C., van den Bedem, H., Velasquez, J., Vincent, J., Wang, X., West, B., Wolf, G., Hodgson, K. O., Wooley, J. and Wilson, I. A. (2004). Crystal structure of a ribose-5-phosphate isomerase RpiB (TM1080) from Thermotoga maritima at 1.90 A resolution. Proteins 56 (1): 171-175. Zhang, R., Andersson, C. E., Savchenko, A., Skarina, T., Evdokimova, E., Beasley, S., Arrowsmith, C. H., Edwards, A. M., Joachimiak, A. and Mowbray, S. L. (2003a). Structure of Escherichia coli ribose-5- phosphate isomerase: a ubiquitous enzyme of the pentose phosphate pathway and the Calvin cycle. Structure (Camb) 11 (1): 31-42. Zhang, R., Andersson, C. E., Skarina, T., Evdokimova, E., Edwards, A. M., Joachimiak, A., Savchenko, A. and Mowbray, S. L. (2003b). The 2.2 Å resolution structure of RpiB/AlsB from Escherichia coli illustrates a new approach to the ribose-5-phosphate isomerase reaction. J. Mol. Biol. 332 (5): 1083-1094.

73 Acta Universitatis Upsaliensis Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 332

Editor: The Dean of the Faculty of Science and Technology

A doctoral dissertation from the Faculty of Science and Technology, Uppsala University, is usually a summary of a number of papers. A few copies of the complete dissertation are kept at major Swedish research libraries, while the summary alone is distributed internationally through the series Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology. (Prior to January, 2005, the series was published under the title “Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology”.)

ACTA UNIVERSITATIS UPSALIENSIS Distribution: publications.uu.se UPPSALA urn:nbn:se:uu:diva-8182 2007