WO2009019484A1

WO2009019484A1 - Crystal structure of ampk and uses thereof

Info

Publication number: WO2009019484A1
Application number: PCT/GB2008/002688
Authority: WO
Inventors: Steve Gamblin; David Carling; Bing Xiao
Original assignee: Medical Research Council
Current assignee: Medical Research Council
Priority date: 2007-08-06
Filing date: 2008-08-06
Publication date: 2009-02-12
Anticipated expiration: 2010-02-06
Also published as: GB0715284D0

Abstract

The invention provides the crystal structure of the gamma subunit of an AMPK protein molecule. The structure is set out in Table 1 and Table 2. The structure may be used in to model the interaction of compounds such as pharmaceuticals with this protein, and to determine the structure of related protein molecules.

Description

CRYSTAL STRUCTURE OF AMPK AND USES THEREOF Field of the Invention.

The present invention relates to crystals and co-crystals of adenosine monophosphate- activated protein kinase (AMPK) and their 3-dimensional structures, and uses thereof.

Background to the Invention.

AMPK

AMP-activated protein kinase (AMPK) regulates cellular metabolism in response to the availability of energy. The enzyme is therefore a target for type Il diabetes treatment¹. It senses changes in the ratio of AMP/ATP by binding both species in a competitive manner². Thus increases in the concentration of AMP, activate AMPK resulting in the phosphorylation and differential regulation of a series of downstream targets that control anabolic and catabolic pathways^1"3.

AMPK has been described as a cellular fuel gauge⁴. As early as 1980, the phosphorylation and inactivation of acetyl-CoA carboxylase was shown to depend on the relative levels of ATP and AMP⁵, although at that time it was not known that the kinase responsible was AMPK. It is now becoming clear that AMPK plays a central role in controlling whole body metabolism in response to nutrients and hormonal signals¹. AMPK is involved in the regulation of lipid metabolism ⁶, feeding and body weight⁷, glucose homeostasis⁸, and mitochondrial biogenesis⁹. In addition to metabolic stress, AMPK is activated by hormones, including leptin¹⁰, adiponectin¹¹ and ciliary neurotrophic factor¹². AMPK is also activated by metformin¹³, one of the most widely used drugs for the treatment of type 2 diabetes, prescribed to over 120 million people worldwide. Moreover, a recent study suggests that AMPK activation is essential for the therapeutic effects of this drug in the liver¹⁴. In mammalian cells AMPK is activated by increases in intracellular AMP by an allosteric mechanism and by regulating the level of AMPK phosphorylation by inhibiting the dephosphorylation of Thr-172 in the activation loop of the kinase domain¹⁵. AMPK is a heterotrimeric complex, consisting of an alpha catalytic subunit (2 isoforms) with beta (2 isoforms) and gamma (3 isoforms) regulatory subunits³. The beta subunit acts as a scaffold binding both the alpha and gamma subunits¹⁶ and the C-terminal 85 residues of beta is sufficient to form a functional AMP-dependent heterotrimer¹⁷. Recently, the structure of the AMPK homologue from S. pombe has revealed the architecture of the heterotrimer subunit interactions¹⁸ but appears to be fundamentally different from the mammalian enzyme in the way that it senses AMP/ATP. The yeast structure showed AMP binding to only one site.

The three isoforms of the gamma subunit are highly homologous in the C-terminal region. The gamma-2 and gamma-3 isoforms have N-terminal extensions which are different from each other and not present in the gamma-1 subunit. The gamma-1 subunit is distributed throughout the body whereas the distribution of the other two subunits varies. The homologous C-terminal regions of all three isoforms of the gamma subunit are composed of four cystathionine β- synthase domains, termed CBS domains. These are numbered, in the N- to C- terminal direction as CBS-1 to CBS-4 respectively. These domains are paired to form two Bateman domains (CBS1+2 and CBS3+4), and it is believed that these Bateman domains form AMP binding pockets.

The gamma subunit of AMPK is known to bind AMP and ATP. However, it is still not known in detail how AMP and ATP interact with this subunit of AMPK to provide regulation of the enzyme.

In another recent study, a Bateman fragment of the human AMPK gamma subunit was crystallised with AMP and the structure of this fragment resolved, to show binding of the fragment with a single AMP molecule.

WO2003/03945 describes an AMPK inhibitor screening assay. The publication proposes that tandem pairs of CBS domains from AMPK form allosteric binding sites for two molecules of AMP or ATP.

Despite the above studies, key questions remain as to the mechanism and regulation of the mammalian enzyme.

Background to Crystallisation It is well-known in the art of protein chemistry, that crystallising a protein is a chancy and difficult process without any clear expectation of success. It is now evident that protein crystallization is the main hurdle in protein structure determination. For this reason, protein crystallization has become a research subject in and of itself, and is not simply an extension of the protein crystailographer's laboratory. There are many references which describe the difficulties associated with growing protein crystals. For example, Kierzek, A.M. and

Zielenkiewicz, P., (2001), Biophysical Chemistry, 91, 1-20, Models of protein crystal growth, and Wiencek, J. M. (1999) Annu. Rev. Biomed. Eng., 1, 505-534, New Strategies for crystal growth.

It is commonly held that crystallization of protein molecules from solution is the major obstacle in the process of determining protein structures. The reasons for this are many; proteins are complex molecules, and the delicate balance involving specific and non-specific interactions with other protein molecules and small molecules in solution, is difficult to predict.

Each protein crystallizes under a unique set of conditions, which cannot be predicted in advance. Simply supersaturating the protein to bring it out of solution may not work, the result would, in most cases, be an amorphous precipitate. Many precipitating agents are used, common ones are different salts, and polyethylene glycols, but others are known. In addition, additives such as metals and detergents can be added to modulate the behaviour of the protein in solution. Many kits are available (e.g. from Hampton Research), which attempt to cover as many parameters in crystallization space as possible, but in many cases these are just a starting point to optimise crystalline precipitates and crystals which are unsuitable for diffraction analysis. Successful crystallization is aided by a knowledge of the proteins behaviour in terms of solubility, dependence on metal ions for correct folding or activity, interactions with other molecules and any other information that is available. Even so, crystallization of proteins is often regarded as a time-consuming process, whereby subsequent experiments build on observations of past trials.

In cases where protein crystals are obtained, these are not necessarily always suitable for diffraction analysis; they may be limited in resolution, and it may subsequently be difficult to improve them to the point at which they will diffract to the resolution required for analysis. Limited resolution in a crystal can be due to several things. It may be due to intrinsic mobility of the protein within the crystal, which can be difficult to overcome, even with other crystal forms. It may be due to high solvent content within the crystal, which consequently results in weak scattering. Alternatively, it could be due to defects within the crystal lattice which mean that the diffracted x-rays will not be completely in phase from unit to unit within the lattice. Any one of these or a combination of these could mean that the crystals are not suitable for structure determination.

Some proteins never crystallize, and after a reasonable attempt it is necessary to examine the protein itself and consider whether it is possible to make individual domains, different N or C- terminal truncations, or point mutations. It is often hard to predict how a protein could be re- engineered in such a manner as to improve crystallisability. Our understanding of crystallisation mechanisms are still incomplete and the factors of protein structure which are involved in crystallisation are poorly understood.

Disclosure of the Invention.

We have now crystallised a full length gamma subunit of mammalian AMPK. We have found that AMPK contains not, as commonly believed, two nucleotide binding sites, but three AMP binding sites. Surprisingly, one of these sites (which we refer to as "AMP3") binds AMP very tightly whereas the other two sites appear ("AMP1" and "AMP2") to be capable of exchanging AMP for ATP.

Thus, unlike the previously published structures mentioned above, we have now found that three-binding sites are present and these sites have different characteristics from each other to explain how the enzyme is regulated.

We also propose a model to explain how relatively small changes in AMP concentrations can cause activation or inactivation of AMPK.

These new insights into the structure and function of AMPK open new avenues to the development of new compounds which are activators or inhibitors of AMPK. One such avenue is in the use of the new crystal structures disclosed herein for modelling new chemical structures or remodelling existing structures to bind AMPK. Another avenue is in the provision of assays which utilise a fluorescently labelled nucleotide to identify potential modulators of AMPK. In one aspect, the present invention provides a three dimensional structure of the gamma subunit set out in Table 1 or Table 2, and uses, described further herein below of the three dimensional structure of the gamma subunit of AMPK set out in Table 1 or Table 2. Thus there is provided a computer-based method for the analysis of the interaction of a molecular structure with an AMPK gamma subunit structure, which comprises: providing an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof; providing a molecular structure to be fitted to said AMPK structure or selected coordinates thereof; and fitting the molecular structure to said AMPK structure.

In general aspects, the present invention is concerned with the provision of an AMPK structure and its use in modelling the interaction of molecular structures, e.g. potential and existing pharmaceutical compounds, including inhibitors or activators, with this AMPK structure. These and other aspects and embodiments of the present invention are discussed below.

In further aspects, the invention is defined by the accompanying claims.

Brief Description of the Tables

Table 1a (Figure 1) sets out the coordinate data of the gamma subunit of AMPK, derived from a co-crystai of AMPK and three AMP molecules.

Table 1b (Figure 1) is a further refinement of the structure of Table 1a.

Table 2a (Figure 2) sets out data of the gamma subunit of AMPK, derived from a co-crystal of AMPK and one AMP with two ATP molecules.

Table 2b (Figure 2) is a further refinement of the structure of Table 2a.

Table 3 sets out binding site residues of the AMP1 pocket which bind AMP and ATP.

Table 4 sets out binding site residues of the AMP2 pocket which bind AMP and ATP

Table 5 sets out binding site residues of the AMP3 pocket which bind AMP.

Brief Description of the Drawings

Figure 1 sets out Tables 1a and 1b

Figure 2 sets out Tables 2a and 2b.

Figure 3 sets out an alignment of rat (SEQ ID NO:1) and human (SEQ ID NO:2) AMPK gamma 1 sequences.

Figure 4 illustrates the location of AMP binding in the AMP-1 binding pocket of AMPK.

Figure 5 illustrates the location of ATP binding in the AMP-1 binding pocket of AMPK.

Figure 6 illustrates the location of AMP binding in the AMP-2 binding pocket of AMPK.

Figure 7 illustrates the location of ATP binding in the AMP-2 binding pocket of AMPK. Figure 8 illustrates the location of AMP binding in the AMP-3 binding pocket of AMPK. Detailed Description of the Invention

The present invention is based upon the use, e.g. in crystals and assay methods, of a mammalian gamma subunit comprising both Bateman domains, together with, in some aspects, at least a C-terminal domain of a mammalian beta subunit and/or at least a C-terminal domain of a mammalian alpha subunit. These proteins are readily obtainable in the prior art and the production of the protein complex that has been crystallized in the accompanying examples was obtained as described by Neumann et a/ (Ref: 31 ). To provide further guidance to those of skill in the art the proteins for use in the invention may be further characterised by the following definitions.

Sequences referred to below by their database accession number are hereby incorporated by reference in their entirety in the form as entered onto the database at the filing date of the present application.

In the description below, where reference is made to Table 1 , a structure of Table 1 , or the like, then it will be understood that unless explicitly stated to the contrary, this may be read as reference to either Table 1a or Table 1b. Likewise, where reference is made to Table 2, a structure of Table 2, or the like, then it will be understood that unless explicitly stated to the contrary, this may be read as reference to either Table 2a or Table 2b.

In one aspect of the invention, reference to Table 1 or a structure of Table 1 , or the like, may be read as a reference to Table 1a or a structure of Table 1a. In another aspect of the invention, reference to Table 1 or a structure of Table 1, or the like, may be read as a reference to Table 1 b or a structure of Table 1 b.

In one aspect of the invention, reference to Table 2 or a structure of Table 2, or the like, may be read as a reference to Table 2a or a structure of Table 2a. In another aspect of the invention, reference to Table 2 or a structure of Table 2, or the like, may be read as a reference to Table 2b or a structure of Table 2b.

A mammalian AMPK gamma subunit comprising both Bateman domains

This may be any isoform of any mammalian AMPK gamma subunit. In the accompanying examples, we have used the rat gamma isoform whose sequence is shown as SEQ ID NO:1 (Genbank X95578). However, other mammalian forms have a high degree of homology and may equally be used. For example, the human gamma 1 isoform of SEQ ID NO:2 (Genbank NP_002724) may also be used. As can be seen from the alignment of Figure 3, the sequences are identical apart from eight, mainly conserved, changes in the N-terminal region. Other mammalian gamma-1 subunits are of a similar level of homology, including bovine or murine forms.

Thus in one embodiment the gamma subunit may be that of SEQ ID NO:1 or SEQ ID NO:2 (or a fragment thereof as defined in the following paragraph), or a protein in which from 1 to 10, for example from 1 to 8, e.g. from 1 to 5, such as 1 or 2 amino acid substitutions, deletions or insertions are present. (In the case of deletions or insertions, the numbers recited herein relate to the numbers of amino acids deleted or substituted). Combinations of substitutions, deletions or insertions may be present, as illustrated by the alignment of Figure 3.

Further, although it is preferred that the full-length gamma 1 subunit is used, small N- and/or C- terminal truncations are not excluded from the scope of the invention provided that the protein retains both Bateman domains which include the CBS 1+2 pair and the CBS 3+4 pair.

Alternatively, the gamma subunit may be a gamma-2 or gamma-3 subunit C-terminal fragment corresponding to the gamma-1 region. These C-terminal fragments are highly conserved and also contain four CBS domains folded into two Bateman domains. Thus the C-terminal 330 amino acid region of any mammalian AMPK gamma-2 or AMPK gamma-3 protein may be used in place of the AMPK gamma-1 protein, or a fragment thereof retaining both Bateman domains.

Suitable gamma-2 proteins include the human gamma-2 (long form) (Genbank AJ249976 ), the human (short form) (Genbank NP_077747), the rat (short form) (Genbank (NP_908940) and the mouse (long form) (Genbank Q91WG5).

Suitable gamma-3 proteins include the human gamma-3 (Genbank AJ249977) or the corresponding murine (Genbank NP_714966), bovine (Genbank Q2LL38) or porcine (Genbank Q9MYP4) forms.

Gamma-2 and gamma-3 C-terminal regions also include variants of wild-type sequences in which from 1 to 10, for example from 1 to 8, e.g. from 1 to 5, such as 1 or 2 amino acid substitutions, deletions or insertions are present. (In the case of deletions or insertions, the numbers recited herein relate to the numbers of amino acids deleted or substituted). Combinations of substitutions, deletions or insertions may be present.

C-terminal domain of a mammalian beta subunit The beta subunit of AMPK exists in two mammalian isoforms. In the accompanying examples the use of the beta-2 isoform is illustrated, though the beta-1 isoform may also be used.

The following beta-1 and beta-2 isoforms are available by reference to the following Genbank accession numbers:

Beta-1 Beta-2

Species Accession No. Species Accession No.

Human AJ224525 Human AJ224538

Rat X95577 Rat Q9QZH4

Bovine Q5BIS9 Mouse Q6PAM0

Mouse Q9R078

The C-terminal domain of the beta isoform contains an alpha/gamma binding domain (in yeast, the equivalent region is termed the ASC domain). Thus the C-terminal domain will usually include about 100 or fewer amino acids of a beta subunit, e.g. such as a fragment having residues 187-272 of the human beta 2 domain or the corresponding regions of any other mammalian beta domain. The corresponding region of such beta domains may be determined by conventional alignment of protein sequences.

Fragments of this portion comprising at least 60, for example at least 70, e.g. at least 80 amino acids may be used.

Beta-1 and Beta-2 C-terminal domains also include variants of wild-type sequences in which from 1 to 10, for example from 1 to 8, e.g. from 1 to 5, such as 1 or 2 amino acid substitutions, deletions or insertions are present. (In the case of deletions or insertions, the numbers recited herein relate to the numbers of amino acids deleted or substituted). Combinations of substitutions, deletions or insertions may be present.

The C-terminal domains, fragments and variant thereof as defined above will retain the ability to form a complex with a C-terminal domain of an AMPK subunit comprising both Bateman domains.

C-terminal domain of a mammalian alpha subunit The alpha subunit of AMPK exists in two mammalian isoforms. In the accompanying examples the use of the alpha -1 isoform is illustrated, though the alpha -2 isoform may also be used.

The following alpha -1 and alpha -2 isoforms are available by reference to the following Genbank accession numbers:

Alpha-1 Alpha-2

Species Accession No. Species Accession No.

Human (long form) NP_006242 Human NP_006243

Human (short form) AAH37303 Rat Q09137

Mouse (short form) NP_001013385

Rat (short form) U40819

The C-terminal domain of the alpha isoform contains a beta binding domain and accordingly the C-terminal domain will usually include about 200 or fewer amino acids of a alpha subunit, e.g. such as a fragment having residues 396-548 of the rat alpha 1 domain or the corresponding regions of any other mammalian alpha domain, or such as a fragment having residues 396-544 of the rat alpha 1 domain or the corresponding region of any other mammalian alpha domain. The corresponding region of such alpha domains may be determined by conventional alignment of protein sequences. Fragments of this portion comprising at least 60, for example at least 70, e.g. at least 80 amino acids may be used.

Alpha-1 and alpha-2 C-terminal domains also include variants of wild-type sequences in which from 1 to 10, for example from 1 to 8, e.g. from 1 to 5, such as 1 or 2 amino acid substitutions, deletions or insertions are present. (In the case of deletions or insertions, the numbers recited herein relate to the numbers of amino acids deleted or substituted). Combinations of substitutions, deletions or insertions may be present. A variant of the rat alpha 1 subunit was used for crystallisation as described in the accompanying examples.

Protein Crystals. The present invention provides a co-crystal of AMPK complex of a mammalian gamma subunit comprising both Bateman domains bound to three nucleotides, C-terminal domains of a mammalian beta and alpha subunits. In one aspect, the co-crystal have a space group P2i2i2 with cell dimensions of about a=49, b=120, c=128 A, 90°, 90°, 90°. Unit cell variability of 5% may be observed in all dimensions.

Two specific examples of a crystal of the invention are the co-crystal comprising three AMP nucleotides having a=48.52, b=119.39, c=129.31 , and a co-crystal comprising two ATP nucleotides and an AMP nucleotide having a=48.79, b=120.69, c=127.07.

Other nucleotides which may be co-crystallized or obtained by soaking, used, analysed or modified either in silico or chemically as a result of the use of methods of the invention include adenosine diphosphate (ADP) and synthetic adenosine nucleotide analogues including 5- Aminoimidazole-4-carboxamide ribonucleoside (AICAR) or ZMP, the monophosphorylated derivative of AICAR as well as adenosine analogues including thiophosphate or phosphoramidate analogues.

Alternatively the co-crystal could comprise a compound whose interaction with AMPK is unknown.

Such crystals may be obtained using the methods described in the accompanying examples. Generally, such methods include co-crystallization and soaking. Since AMPK has three separate nucleotide binding pockets, co-crystals may comprise a mixture of two or three different nucleotides or other compounds located in such pockets.

The present invention is believed to be the first time that a complete gamma domain of a mammalian AMPK domain has been crystallized and its structure resolved. The successful crystallization of the gamma domain is believed to have been achieved through the use of the presence of the C-terminal regions of mammalian alpha and beta subunits to stabilize the folding of the gamma subunit.

This methodology is generally applicable to the provision of other crystals of mammalian AMPK gamma subunits. Thus crystals of the invention may be formed from the following components: The methodology used to provide a crystal illustrated herein may be used generally to provide an AMPK crystal resolvable at a resolution of at least about 2.4 A.

In a further aspect, the invention provides a method for making a protein crystal, particularly of an AMPK protein comprising a mammalian gamma subunit comprising both Bateman domains in association with a C-terminal domain of a mammalian AMPK beta subunit and a C-terminal domain of a mammalian AMPK alpha subunit, which method comprises growing a crystal by vapor diffusion. Protein stock solution comprising 5-30mg/ml, e.g. about 15mg/ml in 5OmM Tris, pH 7.0, 10OmM NaCI and 1mM TCEP (Tris(2-carboxyethyl)phosphine hydrochloride) is mixed with an excess of nucleotide from 1.1 to 5 times e.g. 3-fold excess. Vapour diffusion is done by mixing protein stock solution with a reservoir buffer in the ratio 0.3 to 3.0 to one, e.g. 1 to 1 , where the reservoir solution contains from 6-22% isopropanol e.g. about 15%. Crystals are prepared for data collection by replacing the crystallisation mixture with crystallisation mixture with 25-35% ethylene glycol e.g. about 30%.

Crystals may be prepared in the presence of a nucleotide, particularly AMP, ADP or ATP, or a mixture of two or more of these nucleotides.

Crystal Coordinates. Overview

In one aspect, the invention utilizes the novel crystal coordinates set out in Table 1 and Table 2. The uses of the coordinates are described further herein below and include various methods relating to the computer modelling of ligands that may interact with AMPK, the delivery of the coordinates through computer systems for use by those of skill in the art, and the use of the coordinates in constructing homology models of related proteins or in interpreting X-ray or nmr data of other proteins in order provide a structure of such other proteins. These uses are not exhaustive and other uses are set out herein below.

Format of Table 1 and Table 2 Table 1 and Table 2 give atomic coordinate data for the structures of AMPK. In these Tables the first column is the identifier "ATOM", the second column a contiguous number to uniquely refer to an atom of a particular row, the third column denotes the atom and its location within an amino acid residue, the fourth the residue type, the fifth the chain identification, the sixth the residue number (the atom numbering is with respect to the full length wild type protein), the seventh, eighth and ninth columns are the X, Y, Z coordinates respectively of the atom in question, the tenth column the occupancy of the atom, and the eleventh the temperature factor of the atom, the twelfth the atom type.

Table 1 and Table 2 are set out in an internally consistent format. For example the coordinates of the atoms of each amino acid residue are listed such that the backbone nitrogen atom is first, followed by the C-alpha backbone carbon atom, designated CA, followed by side chain residues (designated according to one standard convention) and finally the carbon and oxygen of the protein backbone. Alternative file formats (e.g. such as a format consistent with that of the EBI Macromolecular Structure Database (Hinxton, UK)) which may include a different ordering of these atoms, or a different designation of the side-chain residues, ligand or water, may be used or preferred by others of skill in the art. However it will be apparent that the use of a different file format to present or manipulate the coordinates of the Table is within the scope of the present invention.

Thus the coordinates of Table 1 and Table 2 provide a measure of atomic location in Angstroms, given to 3 decimal places. The coordinates are a relative set of positions that define a shape in three dimensions, but the skilled person would understand that an entirely different set of coordinates having a different origin and/or axes could define a similar or identical shape. Thus for the purposes described herein as being aspects of the present invention, it is within the scope of the invention if: the coordinates of any of Table 1 and Table 2 are transposed to a different origin and/or axes.

AMPK gamma subunit structure

Tables 1 and 2 contain not only the structure of the AMPK gamma subunit, but also include the coordinate data the C-terminal AMPK alpha and AMPK beta domains, the nucleotide molecules present in the gamma subunit, water molecules, and in the case of Table 2, Magnesium ions. Thus in Table 1a, the gamma subunit is identified as chain "E", in the rows numbered 1444- 3884 of the Table. In Table 2a, the gamma subunit is identified as chain "C", in the rows numbered 1438-3878. In Table 1b and Table 2b, the gamma subunit is identified as chain "E", in the rows numbered 1458-3898 of the Table.

In the present invention, reference to the "AMPK gamma subunit structure of Table 1 or Table 2", or the like, is to be taken as reference to those coordinates in the Tables which relate to this subunit alone. However, unless explicitly set out to the contrary, or otherwise clear from the context, reference throughout the present specification to the use of all or selected coordinates of or from Table 1 or Table 2 does not exclude the use of additional coordinates, particularly some or all of the coordinate of alpha or beta subunit domains, the nucleotides, water molecules or magnesium ions.

RMSD

Protein structure similarity is routinely expressed and measured by the root mean square deviation (rmsd), which measures the difference in positioning in space between two sets of atoms, based on the square root of the arithmetic mean of the squares of the deviations from the mean. The rmsd measures distance between equivalent atoms after their optimal superposition. The rmsd can be calculated over all atoms, over residue backbone atoms (i.e. the nitrogen-carbon-carbon backbone atoms of the protein amino acid residues), main chain atoms only (i.e. the nitrogen-carbon-oxygen-carbon backbone atoms of the protein amino acid residues), side chain atoms only or more usually over C-alpha atoms only. For the purposes of this invention, the rmsd can be calculated over any of these, using any of the methods outlined below. In one embodiment, rmsd is measured by reference to the C-alpha atoms. For the purposes of the present invention, the skilled person would understand that varying the relative atomic positions of the atoms of the structure so that the root mean square deviation of the C-alpha atoms is less than 1.5 A, preferably less than 1.0 A, more preferably less than 0.5 A, more preferably less than 0.3 A, such as less than 0.25 A, or less than 0.2 A, and most preferably less than 0.1 A, when superimposed on the coordinates of AMPK gamma subunit structure of Table 1 or Table 2 for the residue backbone atoms, will generally result in a structure which is substantially the same as the structure of Table 1 and Table 2 in terms of both its structural characteristics and usefulness for structure-based analysis of -interactivity molecular structures.

Reference herein to the AMPK gamma subunit structure of Table 1 or Table 2 and the like thus includes the AMPK gamma subunit coordinate data of these Tables in which one or more individual values of either Table are varied within an rmsd of less than 1.5 A or any of the narrower values identified above.

Methods of comparing protein structures are discussed in Methods of Enzymology, vol 115, pg 397-420. The necessary least-squares algebra to calculate rmsd has been given by Rossman and Argos (J. Biol. Chem. , vol 250, pp7525 (1975)) although faster methods have been described by Kabsch (Acta Crystallogr., Section A, A92, 922 (1976)); Acta Cryst. A34, 827-828 (1978)), Hendrickson (Acta Crystallogr., Section A, A35, 158 (1979)); McLachan (J. MoI. Biol., vol 128, pp49 (1979)) and Kearsley (Acta Crystallogr., Section A, A45, 208 (1989)). Some algorithms use an iterative procedure in which the one molecule is moved relative to the other, such as that described by Ferro and Hermans (Ferro and Hermans, Acta Crystallographic, A33, 345-347 (1977)). Other methods e.g. Kabsch's algorithm locate the best fit directly.

Programs for determining rmsd include MNYFIT (part of a collection of programs called COMPOSER, Sutcliffe, M. J., Haneef, I., Carney, D. and Blundell, T.L. (1987) Protein Engineering, 1 , 377-384), MAPS (Lu, G. An Approach for Multiple Alignment of Protein Structures (1998, in manuscript and on http://bioinfo1.mbfys.lu.se/TOP/maps.html)).

It is usual to consider C-alpha atoms and the rmsd can then be calculated using programs such as LSQKAB (Collaborative Computational Project 4. The CCP4 Suite: Programs for Protein Crystallography, Acta Crystallographica, D50, (1994), 760-763), QUANTA (Jones et al., Acta Crystallography A47 (1991), 110-119 and commercially available from Accelerys, San Diego, CA), Insight (commercially available from Accelerys, San Diego, CA)₁ Sybyl® (commercially available from Tripos, Inc., St Louis), O (Jones et al., Acta Crystallographica, A47, (1991), 110- 119), and other coordinate fitting programs.

In, for example the programs LSQKAB and O, the user can define the residues in the two proteins that are to be paired for the purpose of the calculation. Alternatively, the pairing of residues can be determined by generating a sequence alignment of the two proteins. Programs for sequence alignment are well known in the art, e.g. the BLAST package of alignment programs. The atomic coordinates can then be superimposed according to this alignment and an rmsd value calculated. The program Sequoia (CM. Bruns, I. Hubatsch, M. Ridderstrδm, B. Mannervik, and J.A. Tainer (1999) Human Glutathione Transferase A4-4 Crystal Structures and Mutagenesis Reveal the Basis of High Catalytic Efficiency with Toxic Lipid Peroxidation Products, Journal of Molecular Biology 288(3): 427-439) performs the alignment of homologous protein sequences, and the superposition of homologous protein atomic coordinates. Once aligned, the rmsd can be calculated using programs detailed above. For sequence identical, or highly identical, the structural alignment of proteins can be done manually or automatically as outlined above. Another approach would be to generate a superposition of protein atomic coordinates without considering the sequence.

Selected coordinates

Those of skill in the art will appreciate that in many applications of the invention, it is not necessary to utilise all the coordinates of Table 1 and Table 2, but merely a portion of them. For example, as described below, in methods of modelling candidate compounds, selected coordinates of the AMPK gamma subunit structure of Table 1 or Table 2 may be used.

By "selected coordinates" it is meant for example at least 5, preferably at least 10, more preferably at least 50 and even more preferably at least 100, for example at least 500 or at least 1000 atoms of the AMPK gamma subunit structure of Table 1 or Table 2. Likewise, the other applications of the invention described herein, including homology modelling and structure solution, and data storage and computer assisted manipulation of the coordinates, may also utilise all or a portion of the coordinates (i.e. selected coordinates) of Table 1 and Table 2.

The structures of Tables 1 and 2 have allowed us to identify particular residues of AMPK which are involved in the interaction with AMP or ATP at the AMP-1 , 2 and 3 regions of the protein. Figures 4, 6 and 8 illustrate the binding of AMP to the AMP-1 , 2 and 3 sites respectively, Figures 5 and 7 illustrate the binding of ATP to the AMP-1 and 2 sites respectively. Thus the residues at these regions are thus of particular interest, and are as set out in Tables 3 to 5 below:

Table 3: "AMP-1" region:

Table 5: "AMP-3" region

In some aspects of the invention, the selected coordinates may include one or more coordinates of an atom from an amino acid set out in any one of Tables 3 to 5.

Preferably the selected coordinates include one or more, such as at least 5, preferably at least 10, more preferably at least 50 and even more preferably at least 100, coordinates of atoms of the amino acids set out in any one of Tables 3 to 5. More preferably the selected coordinates are of atoms of residues of Table 3 or Table 4. When the selected coordinates are of atoms of residues of either of Tables 3 or 4, the coordinates are desirably derived from at least 2, more preferably at least 4, such as at least 10 different amino acids. Likewise, when the selected coordinates are of atoms of residues of Table 5, the coordinates are desirably derived from at least 2, more preferably at least 4, such as at least 10 different amino acids.

In one embodiment the invention uses at least 10 selected coordinates from at least 4 different amino acids in any one of Tables 3 or 4. In another embodiment the invention uses at least 50 selected coordinates from at least 10 different amino acids in any one of Tables 3 or 4.

In one embodiment the invention uses at least 10 selected coordinates from at least 4 different amino acids in Table 5. In another embodiment the invention uses at least 50 selected coordinates from at least 10 different amino acids in Table 5.

In another embodiment the invention uses at least 10 selected coordinates from at least 4 different amino acids in Table 3 and at least 10 selected coordinates from at least 4 different amino acids in Table 4. In this embodiment, optionally at least a further 10 selected coordinates from at least 4 different amino acids in Table 5 may also be used.

In another embodiment the invention uses at least 50 selected coordinates from at least 10 different amino acids in Table 3 and at least 50 selected coordinates from at least 10 different amino acids in Table 4. In this embodiment, optionally at least a further 10 selected coordinates from at least 4 different amino acids in Table 5 may also be used.

In embodiments of the invention in which some or all of the coordinates of atoms of the amino acids in Table 5 are used, the coordinates of some (e.g. at least 5, such as at least 10) or all of the AMP atoms of Table 1 or 2 may also be used.

Residues in the above tables whose mutation is implicated in Wolff-Parkinson-White syndrome are Arg 69, His 150, Thr 167 and Arg 298. The equivalent residues in γ2 are: Arg-69≡Arg-302, His-150sHis-383, Thr-167≡Thr-400, Arg-29&≡Arg-531. Atoms from these resides, when the relevant tables are being used, may be included in the methods of the present invention.

Description of Structure. The overall structure of the enzyme can be regarded as being made up of two components: an α+β module at the bottom interacting with one shoulder of the Y domain module at the top. Whilst the α subunit is composed of both helices and strands, and has a recognisable hydrophobic core, the β chain consists of beta structure and generates most of its hydrophobic contacts via interactions with the α subunit.

However, the interface between the α+β and y modules is mediated by the last two strands of the β subunit forming an interdomain beta sheet with a beta-strand from y. Thus it appears that although the C-terminal domain of β relies for its folding on the structure of α, it is responsible for mediating complex formation. The y subunit, made up of four cystathionine β-synthase (CBS) domains, is shaped like a flattened disk about 60 A across and 30 A deep. These CBS domains, packing together to generate two Bateman domains, obey approximate 2,2,2 point group symmetry. Each Bateman domain (CBS1+2 and CBS3+4) arises from the packing of a pair of anti-parallel strands from each CBS domain according to a local dyad axis. It is this interaction that generates the adenyl binding sites. The two Bateman domains pack around a second dyad axis through an interface dominated by a pair of α-helices from each CBS domain. Passing through the centre of the y domain there is a small solvent accessible channel.

Comparison of the structure of mammalian AMPK with its homologue from yeast reveals three striking differences. Firstly, the mammalian enzyme, both in the crystal and in solution, is a monomer whereas the yeast enzyme has been reported to be a dimer. Secondly, there is hinging movement between the α+β and Y modules such that there is about a 12° difference in their relative orientations between the two species. Thirdly, mammalian AMPK binds three AMP molecules; one does not exchange but the other two compete for binding with Mg.ATP and are responsible for the adenyl sensing properties of the mammalian enzyme. In the case of the yeast homologue the structure shows one adenyl-binding site, equivalent to the non- exchangeable site in our structure, which seems to bind either AMP or non Mg²⁺-bound ATP¹⁸.

The symmetry relating the CBS domains of y means that there are four potential adenyl binding sites. There is well-defined electron density for AMP in three of these sites but not the fourth. In these three sites, the nucleotide binds in a surface pocket at the interface of the two CBS domains, within each Bateman domain. The adenine moiety sits in a hydrophobic pocket making hydrogen bonds with main chain groups from two different strands. The phosphate group interacts with the basic side chains of a number of different residues (including Arg-69, His-150, Arg-151 , Lys-169, His-297 & Arg-298) and with the hydroxyl groups of either serine or threonine residues.

Importantly, there are some interactions of the bound AMP molecules with protein residues from the alternate Bateman to which the binding site is predominantly generated (for example Lys-169 and His-150). The 2' and 3' hydroxyl groups of the ribose group make a bi-dentate interaction with an aspartic acid residue located on the first turn of a α-helix adjacent to the site (Asp-244 with AMP-1 , Asp-89 with AMP-2 & Asp-316 with AMP-3). In the fourth potential adenyl binding site there is an arginine residue, at position 170, instead of this aspartic acid that likely accounts for the fact that AMP does not bind at this fourth site.

Only small crystals were obtained from co-crystallisation of AMPK with ATP so we soaked ATP into our AMPK/AMP₃ crystals. To test the results of our solution studies we carried out soaking experiments with Mg.ATP and with Mg²⁺-free ATP. In both cases the resulting electron density maps revealed that ATP replaced two out of the three molecules of AMP and that the metal- bound and metal-free forms of the ATP adopted very similar conformations. Consistent with the results of our binding studies, that showed no difference in affinity between the two forms, the magnesium ions complexed between the β and y phosphate oxygens of the ATP do not interact with the protein. Two views of the structure of the Y subunit with bound nucleotide from both the AMPK/AMP₃ and AMPK/AMP₁ (MgATP)₂ complexes were compared. The two molecules of ATP bind in a similar manner to the AMP molecules that they replace. The β- and γ-phosphates adopt an extended conformation within the binding channel necessitating the rearrangement of the side-chains of a few basic residues; otherwise there are no significant differences in the two different complexes.

Comparison of the AMP and ATP structures suggests that AMP-3, which does not exchange for ATP, represents the tightly bound adenyl moiety that co-purifies with the protein and, as noted above, AMP-3 occupies the equivalent site to the single molecule of AMP seen bound to the yeast homologue¹⁸. All three adenyl binding sites are structurally related but comparison suggests that the positioning of two serine residues (225 & 315), that interact with the phosphate group, and a threonine residue (199), that interacts with the ribose, which does not occur in the other two sites, may account for the enhanced affinity of this site. Interestingly, the only one of these residues not conserved in the S. pombe sequence is Ser-315 which is an alanine residue in that sequence (residues 307 of the C chain in PDB file 2OOX). At least for mammalian AMPK our results show that AMP binding at this non-exchangeable site is not involved in AMP/ATP sensing. Clearly mutagenic and biophysical/biochemical studies will be required to elucidate the role of this tightly bound adenyl species.

A glycogen storage cardiomyopathy associated with an electrical conductance disorder in humans (Refs 26-28) has been attributed to naturally occurring mutations in the γ2 isoform of AMPK. These mutations generally interfere with the normal activation of AMPK by AMP. The corresponding residues mapped onto the γ1 structure reveals that eight out of the ten mutations that have been characterised involve amino acids whose side-chains are in proximity to the adenyl binding sites (Table 6). The remaining two (Tyr-254 & Asn-255) are located towards the periphery of the gamma domain and may affect protein stability/aggregation rather than enzymatic activity directly.

Six of the remaining eight are seen to be contacting the phosphate groups of the AMP/ATP. The structure also shows how small changes in the side-chain conformations of these residues accommodate either AMP or ATP in the binding site. All the mutations tested for adenyl binding have been shown to reduce the enzyme's affinity for both AMP and ATP, in some cases quite substantially. All but one of these mutations is identified here as contacting the adenyl phosphate groups and the remaining one (a leucine insertion) is close to this site and may well affect binding also. Once the molecular basis for signal propagation by AMP binding to AMPK is further elucidated additional insights into the actions of these mutations may be revealed.

Table 6:

Homology Modelling.

The invention also provides a means for homology modelling of other proteins (referred to below as target proteins). By "homology modelling", it is meant the prediction of related structures based either on X-ray crystallographic data or computer-assisted de novo prediction of structure, based upon manipulation of the coordinate data derivable from any one of Table 1 and Table 2 or selected portions thereof.

"Homology modelling" extends to target proteins which are analogues or homologues of the AMPK gamma subunit structure that has been determined herein. It also extends to protein mutants of AMPK gamma subunit protein itself.

The term "homologous regions" describes amino acid residues in two sequences that are identical or have similar (e.g. aliphatic, aromatic, polar, negatively charged, or positively charged) side-chain chemical groups. Identical and similar residues in homologous regions are sometimes described as being respectively "invariant" and "conserved" by those skilled in the art. In general, the method involves comparing the amino acid sequences of the AMPK gamma subunit of SEQ ID NO:1 with a target protein by aligning the amino acid sequences. Amino acids in the sequences are then compared and groups of amino acids that are homologous (conveniently referred to as "corresponding regions") are grouped together. This method detects conserved regions of the polypeptides and accounts for amino acid insertions or deletions.

Homology between amino acid sequences can be determined using commercially available algorithms. The programs BLAST, gapped BLAST, BLASTN, PSI-BLAST and BLAST 2 (provided by the National Center for Biotechnology Information) are widely used in the art for this purpose, and can align homologous regions of two amino acid sequences. These may be used with default parameters to determine the degree of homology between the amino acid sequence of the SEQ ID NO:1 protein and other target proteins which are to be modelled.

Analogues are defined as proteins with similar three-dimensional structures and/or functions with little evidence of a common ancestor at a sequence level.

Homologues are defined as proteins with evidence of a common ancestor, i.e. likely to be the result of evolutionary divergence and are divided into remote, medium and close sub-divisions based on the degree (usually expressed as a percentage) of sequence identity.

A homologue is defined here as a protein with at least 15%, preferably at least 40%, more preferably at least 70%, sequence identity or which has at least one functional domain, which is characteristic of the AMPK gamma subunit.

The homlogues could also be polymorphic forms of the AMPK gamma subunit such as alleles or mutants.

Once the amino acid sequences of the polypeptides with known and unknown structures are aligned, the structures of the conserved amino acids in a computer representation of the polypeptide with known structure are transferred to the corresponding amino acids of the polypeptide whose structure is unknown. For example, a tyrosine in the amino acid sequence of known structure may be replaced by a phenylalanine, the corresponding homologous amino acid in the amino acid sequence of unknown structure.

The structures of amino acids located in non-conserved regions may be assigned manually by using standard peptide geometries or by molecular simulation techniques, such as molecular dynamics. The final step in the process is accomplished by refining the entire structure using molecular dynamics and/or energy minimization.

Homology modelling as such is a technique that is well known to those skilled in the art (see e.g. Greer, Science, Vol. 228, (1985), 1055, and Blundell et al., Eur. J. Biochem, Vol. 172, (1988), 513). The techniques described in these references, as well as other homology modelling techniques, generally available in the art, may be used in performing the present invention.

Thus the invention provides a method of homology modelling comprising the steps of: (a) aligning a representation of an amino acid sequence of a target protein of unknown three-dimensional structure with the amino acid sequence of SEQ ID NO:1 to match homologous regions of the amino acid sequences;

(b) modelling the structure of the matched homologous regions of said target of unknown structure on the corresponding regions of the AMPK gamma subunit structure of Table 1 or Table 2, or selected coordinates thereof; and

(c) determining a conformation (e.g. so that favourable interactions are formed within the target of unknown structure and/or so that a low energy conformation is formed) for said target of unknown structure which substantially preserves the structure of said matched homologous regions.

Preferably one or all of steps (a) to (c) are performed by computer modelling.

The co-ordinate data of Table 1 and Table 2 or selected coordinates thereof, will be particularly advantageous for homology modelling of other mammalian proteins, in particular the C-terminal Bateman domains of AMPK gamma-2 or AMPK gamma-3, including the human forms thereof.

The aspects of the invention described herein which utilise the structure in silico may be equally applied to homologue models of obtained by the above aspect of the invention, and this application forms a further aspect of the present invention. Thus having determined a conformation of a by the method described above, such a conformation may be used in a computer-based method of rational drug design as described herein.

Structure Solution

The atomic coordinate data of an AMPK gamma subunit structure of Table 1 or Table 2 can also be used to solve the crystal structure of other target proteins including other crystal forms of AMPK, mutants, co-complexes of AMPK gamma, where X-ray diffraction data or NMR spectroscopic data of these target proteins has been generated and requires interpretation in order to provide a structure.

In the case of AMPK gamma, this protein may crystallize in more than one crystal form. The data of Table 1 and Table 2, or portions thereof, as provided by this invention, are particularly useful to solve the structure of those other crystal forms of AMPK. It may also be used to solve the structure of AMPK mutants, AMPK co-complexes, or of the crystalline form of any other protein with significant amino acid sequence homology to any functional domain of AMPK.

In the case of other target proteins, particularly the human proteins referred to above, the present invention allows the structures of such targets to be obtained more readily where raw X-ray diffraction data is generated. Thus, where X-ray crystallographic or NMR spectroscopic data is provided for a target of unknown three-dimensional structure, the atomic coordinate data derived from any one of AMPK gamma subunit structure of Table 1 or Table 2, may be used to interpret that data to provide a likely structure for the other by techniques which are well known in the art, e.g. phasing in the case of X-ray crystallography and assisting peak assignments in NMR spectra.

One method that may be employed for these purposes is molecular replacement. This method will provide an accurate structural form for the unknown crystal more quickly and efficiently than attempting to determine such information ab initio. Examples of computer programs known in the art for performing molecular replacement are CNX (Brunger AT.; Adams P.D.; Rice L.M., Current Opinion in Structural Biology, Volume 8, Issue 5, October 1998, Pages 606-611 (also commercially available from Accelrys San Diego, CA), MOLREP (A.Vagin, A.Teplyakov, MOLREP: an automated program for molecular replacement, J. Appl. Cryst. (1997) 30, 1022- 1025, part of the CCP4 suite) or AMoRe (Navaza, J. (1994). AMoRe: an automated package for molecular replacement. Acta Cryst. A50, 157-163).

Thus, in a further aspect of the invention provides a method for determining the structure of a protein, which method comprises; providing an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof, and positioning the coordinates in the crystal unit cell of said protein so as to provide a structure for said protein.

The invention may also be used to assign peaks of NMR spectra of such proteins, by manipulation of the data of any one of Table 1 and Table 2.

In a preferred aspect of this invention the co-ordinates are used to solve the structure of target AMPK gamma subunits, particularly the gamma-2 and gamma-3 subunits.

Computer Systems

In another aspect, the present invention provides systems, particularly a computer system, intended to generate structures and/or perform optimisation of compounds which interact with an AMPK gamma subunit, the system containing computer-readable data comprising one or more of:

(a) AMPK co-ordinate data of an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof, said data defining the three-dimensional structure of the gamma subunit of AMPK or said selected coordinates thereof; (b) atomic coordinate data of a target protein generated by homology modelling of the target based on the coordinate data of an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof; or (c) atomic coordinate data of a target protein generated by interpreting X-ray crystallographic data or NMR data by reference to the co-ordinate data of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof.

For example the computer system may comprise: (i) a computer-readable data storage medium comprising data storage material encoded with the computer-readable data; (ii) a working memory for storing instructions for processing said computer-readable data; and (iii) a central- processing unit coupled to said working memory and to said computer-readable data storage medium for processing said computer-readable data and thereby generating structures and/or performing rational drug design. The computer system may further comprise a display coupled to said central-processing unit for displaying said structures.

The instructions for processing said computer-readable data may include instructions for constructing a model of the AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof, for constructing a model of an AMPK gamma subunit structure based on homology modelling or structure solution using the data derived from said Tables, or performing the computer-based modelling methods of the invention described herein.

The display may be used to provide an output in which an AMPK gamma subunit structure of the invention, optionally including one or more molecular structures fitted said AMPK gamma subunit structure is displayed in the form of a model. Such a model may be (a) a wire-frame model; (b) a chicken-wire model; (c) a ball-and-stick model; (d) a space-filling model; (e) a stick-model; (f) a ribbon model; (g) a snake model; (h) an arrow and cylinder model; (i) an electron density map; (j) a molecular surface model.

The invention also provides such systems containing atomic coordinate data of target proteins wherein such data has been generated according to the methods of the invention described herein based on the starting data provided the data of Table 1 or Table 2 or selected coordinates thereof.

Such data is useful for a number of purposes, including the generation of structures to analyse the mechanisms of action of, and/or to perform rational drug design of compounds, which interact with, AMPK and are potential activators or inhibitors of the enzyme.

In a further aspect, the present invention provides a computer-readable storage medium, comprising a data storage material encoded with computer readable data, wherein the data are defined by all or a portion of the structure coordinates of the AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof.

As used herein, "computer readable media" refers to any medium or media, which can be read and accessed directly by a computer. Such media include, but are not limited to: magnetic storage media such as floppy discs, hard disc storage medium and magnetic tape; optical storage media such as optical discs or CD-ROM; electrical storage media such as RAM and ROM; and hybrids of these categories such as magnetic/optical storage media.

By providing such computer readable media, the atomic coordinate data of the invention can be routinely accessed to model s or selected coordinates thereof. For example, RASMOL (Sayle et al., TIBS, Vol. 20, (1995), 374) is a publicly available computer software package, which allows access and analysis of atomic coordinate data for structure determination and/or rational drug design.

As used herein, "a computer system" refers to the hardware means, software means and data storage means used to analyse the atomic coordinate data of the invention. The minimum hardware means of the computer-based systems of the present invention comprises a central processing unit (CPU), input means, output means and data storage means. Desirably a monitor is provided to visualize structure data. The data storage means may be RAM or means for accessing computer readable media of the invention. Examples of such systems are microcomputer workstations available from Silicon Graphics Incorporated and Sun Microsystems running Unix based, Windows NT or IBM OS/2 operating systems.

The invention also provides a computer-readable data storage medium comprising a data storage material encoded with a first set of computer-readable data comprising the AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof, which, when combined with a second set of machine readable data comprising an X-ray diffraction pattern of a molecule or molecular complex of unknown structure, using a machine programmed with the instructions for using said first set of data and said second set of data, can determine at least a portion of the electron density corresponding to the second set of machine readable data.

A further aspect of the invention provides a method of providing data for generating structures and/or performing optimisation of compounds which interact with an AMPK gamma subunit the method comprising: (i) establishing communication with a remote device containing

(a) AMPK co-ordinate data of an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof, said data defining the three-dimensional structure of the gamma subunit of AMPK or said selected coordinates thereof;

(b) atomic coordinate data of a target protein generated by homology modelling of the target based on the coordinate data of an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof; or

(c) atomic coordinate data of a target protein generated by interpreting X-ray crystallographic data or NMR data by reference to the co-ordinate data of an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof; and (ii) receiving said computer-readable data from said remote device.

When such data is received it may be utilised in the various methods of the invention described herein, including computer-based methods of the analysis of the interaction of a molecular structure with an AMPK gamma subunit structure. The atomic coordinate data may include coordinates of amino acids set out in Table 3, Table 4 and/or Table 5.

Thus the remote device may comprise e.g. a computer system or computer readable media of one of the previous aspects of the invention. The device may be in a different country or jurisdiction from where the computer-readable data is received.

The communication may be via the internet, intranet, e-mail etc, transmitted through wires or by wireless means such as by terrestrial radio or by satellite. Typically the communication will be electronic in nature, but some or all of the communication pathway may be optical, for example, over optical fibres.

The data received from said remote device may be used to perform the methods of the invention described herein, including but not limited to the computer based methods for the analysis of the interaction of a molecular structure with an AMPK gamma subunit structure.

Uses of the Structures of the Invention.

The crystal structures obtained according to the present invention as well as the structures of target proteins obtained in accordance with the methods described herein), may be used in several ways for drug design.

In particular, the finding that AMPK is activated by binding of AMP into two of the three AMP binding sites of AMPK allows the rational drug design of potential activator compounds specifically directed to one or other of these sites. Similarly, potential inhibitors of AMPK that binds at the "AMP-1" and "AMP-2" but not the AMP-3 sites allows the development of compounds which interact with AMPK in novel ways. Alternatively, the high-affinity AMP-3 provides a further novel target for drug development.

(i) Obtaining and analysing crystal complexes.

In one approach, the structure of a compound bound to a may be determined by experiment. This will provide a starting point in the analysis of the compound bound to , thus providing those of skill in the art with a detailed insight as to how that particular compound interacts with AMPK.

Many of the techniques and approaches to structure-based drug design described above rely at some stage on X-ray analysis to identify the binding position of a ligand in a ligand-protein complex. A common way of doing this is to perform X-ray crystallography on the complex, produce a difference Fourier electron density map, and associate a particular pattern of electron density with the ligand. However, in order to produce the map (as explained e.g. by Blundell et al., in Protein Crystallography, Academic Press, New York, London and San Francisco, (1976)), it is necessary to know beforehand the protein 3D structure (or at least the protein structure factors). Therefore, determination of the structure also allows difference Fourier electron density maps of -compound complexes to be produced, determination of the binding position of the drug and hence may greatly assist the process of rational drug design.

Accordingly, the invention provides a method for determining the structure of a compound bound to an AMPK gamma subunit, said method comprising: providing a crystal of an AMPK gamma subunit according to the invention; soaking the crystal with said compounds; and determining the structure of said compound complex by employing the coordinate data of any one of Table 1 and Table 2 or selected coordinates thereof.

Alternatively, an AMPK gamma subunit and compound may be co-crystallized. Thus the invention provides a method for determining the structure of a compound bound to an AMPK gamma subunit, said method comprising; mixing the protein with the compound(s), crystallizing the protein-compound(s) complex; and determining the structure of said -compound(s) complex by reference to the coordinate data of Table 1 or Table 2, or selected coordinates thereof.

The analysis of such structures may employ (i) X-ray crystallographic diffraction data from the complex and (ii) a three-dimensional structure of an AMPK gamma subunit, or at least selected coordinates thereof, to generate a difference Fourier electron density map of the complex, the three-dimensional structure being defined by atomic coordinate data of Table 1 or Table 2, or selected coordinates thereof. The difference Fourier electron density map may then be analysed.

Therefore, such complexes can be crystallized and analysed using X-ray diffraction methods, e.g. according to the approach described by Greer et al., J. of Medicinal Chemistry, Vol. 37, (1994), 1035-1054, and difference Fourier electron density maps can be calculated based on X-ray diffraction patterns of soaked or co-crystallized and the solved structure of uncomplexed. These maps can then be analysed e.g. to determine whether and where a particular compound binds to an AMPK gamma subunit and/or changes the conformation of an AMPK gamma subunit.

Electron density maps can be calculated using programs such as those from the CCP4 computing package (Collaborative Computational Project 4. The CCP4 Suite: Programs for Protein Crystallography, Acta Crystallographica, D50, (1994), 760-763.). For map visualization and model building programs such as "O" (Jones et al., Acta Crystallographica, A47, (1991), 110-119) can be used.

In addition, in accordance with this invention, an AMPK gamma subunit variants or isoforms may be crystallized in co-complex with known an AMPK gamma subunit substrates or inhibitors or novel compounds. The crystal structures of a series of such complexes may then be solved by molecular replacement and compared with that of the an AMPK gamma subunit structure of Table 1 or Table 2, or selected coordinates thereof. Potential sites for modification within the - various binding sites of the enzyme may thus be identified. This information provides an additional too! for determining the most efficient binding interactions, for example, increased hydrophobic interactions, between an AMPK gamma subunit and a chemical entity or compound.

(H) In silico analysis and design

Although the invention will facilitate the determination of actual crystal structures comprising a and a compound, which interacts with an AMPK gamma subunit, current computational techniques provide a powerful alternative to the need to generate such crystals and generate and analyse diffraction date. Accordingly, a particularly preferred aspect of the invention relates to "in silico" methods directed to the analysis and development of compounds which interact with structures of the present invention.

Determination of the three-dimensional structure of an AMPK gamma subunit provides important information about the binding sites of this protein. This information may then be used for rational design and modification of AMPK ligands including activators or inhibitors, e.g. by computational techniques which identify possible binding ligands for the binding sites, by enabling linked-fragment approaches to drug design, and by enabling the identification and location of bound ligands (e.g. including those ligands mentioned herein above) using X-ray crystallographic analysis. These techniques are discussed in more detail below.

Thus as a result of the determination of the three-dimensional structures provided by the present invention, more purely computational techniques for rational drug design may also be used to design structures whose interaction with AMPK is better understood (for an overview of these techniques see e.g. Walters et al (Drug Discovery Today, Vol.3, No.4, (1998), 160-178; Abagyan, R.; Totrov, M. Curr. Opin. Chem. Biol. 2001, 5, 375-382). For example, automated ligand-receptor docking programs (discussed e.g. by Jones et al. in Current Opinion in Biotechnology, Voi.6, (1995), 652-656 and Halperin, I.; Ma, B.; Wolfson, H.; Nussinov, R. Proteins 2002, 47, 409-443), which require accurate information on the atomic coordinates of target receptors may be used.

The aspects of the invention described herein which utilize the structures of the present invention in silico may be equally applied to both the an AMPK gamma subunit structure of Table 1 or Table 2, or selected coordinates thereof, and the models of target proteins obtained by other aspects of the invention. Thus having determined a conformation of an AMPK gamma subunit by the method described above, such a conformation may be used in a pomputer- based method of rational drug design as described herein. In addition the availability of the structure of an AMPK gamma subunit will allow the generation of highly predictive pharmacophore models for virtual library screening or compound design.

Accordingly, the invention provides a computer-based method for the analysis of the interaction of a molecular structure with an AMPK gamma subunit structure, which comprises: providing an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A₁ or selected coordinates thereof; providing a molecular structure to be fitted to said AMPK structure or selected coordinates thereof; and fitting the molecular structure to said AMPK structure.

By "fitting", it is meant determining at least one interaction between an atom of said AMPK structure and said molecular structure.

Interactions may be determined by automatic, semi-automatic or manual means. Computer programs can be employed to estimate interactions including the attraction, repulsion, and steric hindrance of the two binding partners (i.e. the AMPK structure and a molecular structure). Various computer-based methods for fitting are described further herein. Generally the process involves calculating the extent to which such an interaction is present or stable.

In practice, it will be desirable to model a sufficient number of atoms of the as defined by the coordinates of any one of Table 1 and Table 2, or selected coordinates thereof, which represent a binding pocket. Thus preferred numbers and/or locations of selected coordinates as defined above may be used in the present invention.

In order to provide a three-dimensional structure of compounds to be fitted to a structure of the invention, the compound structure may be modelled in three dimensions using commercially available software for this purpose or, if its crystal structure of a compound is available, the coordinates of the structure may be used to provide a representation of the compound for fitting to a structure of the invention.

More specifically, the interaction of a compound or compounds with can be examined through the use of computer modelling using a docking program such as GOLD (Jones et al., J. MoI. Biol., 245, 43-53 (1995), Jones et al., J. MoI. Biol., 267, 727-748 (1997)), GRAMM (Vakser, I.A., Proteins , Suppl., 1:226-230 (1997)), DOCK (Kuntz et al, J.Mol.Biol. 1982 , 161, 269-288, Makino et al, J.Comput.Chem. 1997, 18, 1812-1825), AUTODOCK (Goodsell et al, Proteins 1990, 8, 195-202, Morris et al, J.Comput.Chem. 1998, 19, 1639-1662.), FlexX, (Rarey et al, J.Mol.Biol. 1996, 261, 470-489) or ICM (Abagyan et al, J.Comput.Chem. 1994, 15, 488-506). This procedure can include computer fitting of compounds to an AMPK gamma subunit structure of the invention to ascertain how well the shape and the chemical structure of the compound will bind to AMPK.

In one aspect, the invention provides a method of identifying a potential ligand of an AMPK gamma subunit protein comprising a nucleotide binding pocket, said method comprising the steps of:

(a) using the atomic coordinates of Table 1 or Table 2 to generate a three-dimensional model; (b) identifying said binding pocket residues which comprises at least one or more of the residues set out in any one of Tables 3 to 5, and using said residues to generate a specific three-dimensional (3-D) target;

(c) employing said 3-D target of (b) to design or select said potential ligand; (d) synthesizing said potential ligand; and

(e) contacting said potential ligand with said AMPK gamma subunit protein in vitro to determine the ability of said potential inhibitor to interact with said AMPK gamma subunit protein; whereby the ability to interact is an indication that said potential ligand of said AMPK gamma subunit protein is identified.

The above method may use any of the AMPK subunit proteins described herein, for example an AMPK gamma subunit protein selected from the group consisting of a gamma 1 subunit or fragment thereof comprising both Bateman domains which include the CBS 1+2 pair and the CBS 3+4 pair; a gamma-2 subunit or C-terminal fragment thereof retaining both Bateman domains; and a gamma-3 subunit retaining both Bateman domains. Suitable subunit proteins include those of SEQ ID NO:1 and SEQ ID NO:2.

The invention further provides a method of designing a compound which binds to a nucleotide binding pocket of an AMPK gamma subunit protein, wherein said protein is characterized by: (i) the atomic coordinates of amino acids 23-326 of SEQ ID No: 1 shown in Table 1 , optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A; (ii) the atomic coordinates of amino acids 23-326 of SEQ ID No: 1 shown in Table 2, optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A; and/or (iii) the atomic coordinates of a nucleotide binding pocket defined by amino acids of any one of Tables 3, 4 and 5, optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A; said method comprising the steps of:

(a) using the atomic coordinates of Table 1 or Table 2 to build a 3-D computer model of a compound interaction region of said protein comprising at least one of (i), (ii) and (iii);

(b) assessing the stereochemical complementarity between a compound and said interaction region;

(c) optimizing said stereochemical complementarity in an iterative approach by observing changes in the protein or compound that affect the protein/compound associations; and

(d) designing a compound which optimizes said protein/compound stereochemical complementarity.

Also computer-assisted, manual examination of the active site structure of may be performed. The use of programs such as GRID (Goodford, J. Med. Chem., 28, (1985), 849-857) - a program that determines probable interaction sites between molecules with various functional groups and an enzyme surface - may also be used to analyse the active site to predict, for example, the types of modifications which will alter the rate of metabolism of a compound. Since AMPK has been found to contain three nucleotide binding pockets a plurality of molecular structures, e.g. two or three, may be provided in the modelling methods of the invention and the interactions between each of said structures and an AMPK gamma subunit structure of the invention may be determined. In addition, the interactions between two or more of such molecular structures may be examined.

It is also envisaged that two or more such structures may be assembled to form a single larger structure. For example, the binding of one or more molecular fragments can be determined in the protein binding pocket by X-ray crystallography. Molecular fragments are typically compounds with a molecular weight between 100 and 200 Da. This can then provide a starting point for medicinal chemistry to optimise the interactions using a structure-based approach. The fragments can be combined onto a template or used as the starting point for 'growing out' an inhibitor into other pockets of the protein (Blundell et al, 2002, Nat Rev Drug Discov. Jan;1(1):45-54). The fragments can be positioned in the binding pocket of the and then 'grown' to fill the space available, exploring the electrostatic, van der Waals or hydrogen-bonding interactions that are involved in molecular recognition. The potency of the original weakly binding fragment thus can be rapidly improved using iterative structure-based chemical synthesis.

At one or more stages in the fragment growing approach, the compound may be synthesized and tested in a biological system for its activity. This can be used to guide the further growing out of the fragment.

As indicated above, where two fragment-binding regions are identified, a linked fragment approach may be based upon attempting to link the two fragments directly, or growing one or both fragments in the manner described above in order to obtain a larger, linked structure, which may have the desired properties.

Where the binding site of two or more ligands are determined they may be connected to form a potential lead compound that can be further refined using e.g. the iterative technique of Greer et al. For a virtual linked-fragment approach see Verlinde et al., J. of Computer-Aided Molecular Design, 6, (1992), 131-147, and for NMR and X-ray approaches see Shuker et al., Science, 274, (1996), 1531-1534 and Stout et al., Structure, 6, (1998), 839-848. The use of these approaches to design inhibitors is made possible by the determination of the structure.

The various computer-based methods of analysis described herein may be performed using computer systems such as those described in the preceding section. Generally, the computer systems used will be configured to display or transmit a model of the structure of Table 1 or 2, or selected coordinates thereof and a molecular structure so as to indicate one or more interactions between the two. A variety of formats of display are known in the art and may be selected by a person of ordinary skill in the art dependent upon a variety of factors including, for example, the nature of the interactions being determined. By using the present invention, detailed structural information can be obtained about the binding of compounds to AMPK, and in the light of this information adjustments can be made to the structure or functionality of the compound, e.g. to alter its interaction with AMPK. The above steps may be repeated and re-repeated as necessary. For example, Greer et al. ( J. of Medicinal Chemistry, Vol. 37, (1994), 1035-1054) describes an iterative approach to ligand design based on repeated sequences of computer modelling, protein-ligand complex formation and X-ray crystallographic or NMR spectroscopic analysis. Thus novel thymidylate synthase inhibitor series were designed de novo by Greer et al., and ligands may also be designed or modified in the this way. More specifically, using e.g. GRID on the solved structure of AMPK, a ligand may be designed that complements the functionalities of the protein's binding sites. Alternatively a ligand may be modified such that it complements the functionalities of protein's binding sites better or less well. The ligand can then be synthesised, formed into a complex with an AMPK gamma subunit, and the complex then analysed by X-ray crystallography to identify the actual position of the bound ligand. The structure and/or functional groups of the ligand can then be adjusted, if necessary, in view of the results of the X-ray analysis, and the synthesis and analysis sequence repeated until an optimised ligand is obtained. Related approaches to structure-based drug design are also discussed in Bohacek et al., Medicinal Research Reviews, Vol.16, (1996), 3-50. Design of a compound with alternative properties using structure based drug design may also take into account the requirements for high affinity to a second, target protein. Gschwend et al., (Bioorganic & Medicinal Chemistry Letters, VoI 9, (1999), 307-312) and Bayley etal., (Proteins: Structure, Function and Genetics, VoI 29, (1997) 29-67) describe approaches where structure based drug design is used to reduce affinity to one protein whilst maintaining affinity for a target protein.

Modification will be those conventional in the art known to the skilled medicinal chemist, and will include, for example, substitutions or removal of groups containing residues which interact with the amino acid side chain groups of a structure of the invention. For example, the replacements may include the addition or removal of groups in order to decrease or increase the charge of a group in a test compound, the replacement of a group to increase or decrease the size of the group in a test compound, the replacement of a charge group with a group of the opposite charge, or the replacement of a hydrophobic group with a hydrophilic group or vice versa. It will be understood that these are only examples of the type of substitutions considered by medicinal chemists in the development of new pharmaceutical compounds and other modifications may be made, depending upon the nature of the starting compound and its activity.

As indicated above, molecular structures, which may be fitted to an AMPK structure of the invention, include compounds under development as potential pharmaceutical agents. The agents may be fitted in order to determine how they bind to AMPK and to provide, for example, a basis for modelling further candidate modulators of AMPK activity, particularly activators.

Molecular structures, which may be used in the present invention, will usually be compounds under development for pharmaceutical use. Generally such compounds will be organic molecules, which are typically from about 100 to 2000 Da, more preferably from about 100 to 1000 Da in molecular weight. Such compounds include nucleotide analogues such as AICAR and ZMP. In principle, any compound under development in the field of pharmacy can be used in the present invention in order to facilitate its development or to allow further rational drug design to improve its properties.

Where a potential modified compound has been developed by fitting a starting compound to the structure of the invention and predicting from this a modified compound with an altered activity, the invention further includes the step of synthesizing the modified compound and testing it in a in vivo or in vitro biological system in order to determine its activity.

The above-described processes of the invention may be iterated in that the modified compound may itself be the basis for further compound design.

Assay Methods In one aspect of the invention, there is provided an assay method for a modulator of AMPK, which method comprises: providing a candidate AMPK modulator compound; bringing said compound into contact with a mammalian AMPK gamma subunit or fragment thereof having two Bateman domains; and determining the binding of said candidate compound for said gamma subunit or fragment thereof in competition with an adenosine nucleotide; wherein said adenosine nucleotide is labelled with a fluorescent label.

Although the exact assay format may be varied according to the preferences of any individual practitioner of ordinary skill in the art, the method can conveniently be performed in a multi-well format, allowing different candidate modulators and/or different concentrations of such candidate modulators to be assayed simultaneously. Such a format will also permit automation of the method using standard robotic technology.

The candidate modulator compound may be any compound available to those of skill in the art. For example, commercially available combinatorial chemical libraries may be used. Candidate compounds which are nucleotide analogues may be used.

In performing the assay of the invention, the mammalian AMPK gamma subunit or fragment thereof having two Bateman domains may be used alone or in combination with either or both of an AMPK gamma alpha subunit or C-terminal fragment thereof, and/or an AMPK beta subunit or C-terminal fragment thereof. Any mammalian alpha, beta and gamma subunit may be used, given the high degree of conservation between these subunits. This is evidenced by the present examples in which a rat gamma subunit is co-expressed and forms an active AMPK with a human beta subunit and rat alpha subunit. In view of this, it is believed that other mammalian subunits may be used interchangeably in the present invention without loss of assay function. The mammalian AMPK gamma subunit or fragment thereof is preferably a gamma-1 subunit, though the use of a gamma-2 or gamma-3 subunit is also contemplated.

Any method for measuring the binding of the candidate compound in competition an adenosine nucleotide, particularly AMP or ATP, may be used. ADP may also be used. Generally, this will involve labelling one or other, or both, of the candidate compound or the adenosine nucleotide with a detectable label, such that the amount of label that can be detected either bound to or free of the AMPK gamma subunit is altered when or if the candidate compound binds to the subunit in place of ATP/AMP, or vice versa.

In one currently preferred aspect the fluorescent label is N-methylanthraniloyl ("mant").

The concentrations of the protein, candidate compound and adenosine nucleotide may be varied according to the particular format of the assay. Suitably, 300 μl of from 1 to 100 μm mant-AXP (where "AXP" may be AMP, ADP or ATP), e.g. 10 μM mant-AXP, titrated up to 5-200 μM AMPK, e.g. about 15 μm AMPK, in the presence of a fixed concentration of the candidate compound which might be in the range of 1OnM up to 1mM.

Compounds of the invention Where a potential compound has been developed by the computer based methods or assay methods described above, the invention provides such a compound, which may be manufactured and/or used in the preparation, i.e. manufacture or formulation, of a composition such as a medicament, pharmaceutical composition or drug. These formulations or compositions may be administered to individuals, e.g. to an individual with a metabolic syndrome. Such a condition includes a disease selected from the group of diabetes, obesity, insulin-resistance, hypertension, cardiovascular disease, and dyslipidaemia.

Thus, the present invention extends in various aspects not only to a compound as provided by the invention, but also a pharmaceutical composition, medicament, drug or other composition comprising such a compound. The compositions may be used, for treatment (which may include preventative treatment) of disease such as those mentioned above. Such a treatment may comprise administration of such a composition to a patient, e.g. for treatment of disease; the use of such an inhibitor in the manufacture of a composition for administration, e.g. for treatment of disease; and a method of making a pharmaceutical composition comprising admixing such an inhibitor with a pharmaceutically acceptable excipient, vehicle or carrier, and optionally other ingredients.

Thus a further aspect of the present invention provides a method for preparing a medicament, pharmaceutical composition or drug, the method comprising: (a) identifying or modifying a compound by a method of any one of the other aspects of the invention disclosed herein; (b) optimising the structure of the molecule; and (c) preparing a medicament, pharmaceutical composition or drug containing the optimised compound. The above-described processes of the invention may be iterated in that the modified compound may itself be the basis for further compound design.

By "optimising the structure" we mean e.g. adding molecular scaffolding, adding or varying functional groups, or connecting the molecule with other molecules (e.g. using a fragment linking approach) such that the chemical structure of the modulator molecule is changed while its original modulating functionality is maintained or enhanced. Such optimisation is regularly undertaken during drug development programmes to e.g. enhance potency, promote pharmacological acceptability, increase chemical stability etc. of lead compounds.

Modification will be those conventional in the art known to the skilled medicinal chemist, and will include, for example, substitutions or removal of groups containing residues which interact with the amino acid side chain groups of a structure of the invention. For example, the replacements may include the addition or removal of groups in order to decrease or increase the charge of a group in a test compound, the replacement of a charge group with a group of the opposite charge, or the replacement of a hydrophobic group with a hydrophilic group or vice versa. It will be understood that these are only examples of the type of substitutions considered by medicinal chemists in the development of new pharmaceutical compounds and other modifications may be made, depending upon the nature of the starting compound and its activity.

Compositions may be formulated for any suitable route and means of administration. Pharmaceutically acceptable carriers or diluents include those used in formulations suitable for oral, rectal, nasal, topical (including buccal and sublingual), vaginal or parenteral (including subcutaneous, intramuscular, intravenous, intradermal, intrathecal and epidural) administration. The formulations may conveniently be presented in unit dosage form and may be prepared by any of the methods well known in the art of pharmacy.

For solid compositions, conventional non-toxic solid carriers include, for example, pharmaceutical grades of mannitol, lactose, cellulose, cellulose derivatives, starch, magnesium stearate, sodium saccharin, talcum, glucose, sucrose, magnesium carbonate, and the like may be used. Liquid pharmaceutically administrable compositions can, for example, be prepared by dissolving, dispersing, etc, an active compound as defined above and optional pharmaceutical adjuvants in a carrier, such as, for example, water, saline aqueous dextrose, glycerol, ethanol, and the like, to thereby form a solution or suspension. If desired, the pharmaceutical composition to be administered may also contain minor amounts of non-toxic auxiliary substances such as wetting or emulsifying agents, pH buffering agents and the like, for example, sodium acetate, sorbitan monolaurate, triethanolamine sodium acetate, sorbitan monolaurate, triethanolamine oleate, etc. Actual methods of preparing such dosage forms are known, or will be apparent, to those skilled in this art; for example, see Remington's

Pharmaceutical Sciences, Mack Publishing Company, Easton, Pennsylvania, 15th Edition, 1975. Examples

The invention is illustrated by the following examples. In summary, the examples show that truncated AMPK was cloned into a tricistronic vector and expressed in E.coli. Proteins were initially purified by Nickel affinity chromatography. Crystals were grown by vapour diffusion in hanging drops using iospropanol as precipitant and were cryo-protected with 28% ethylene glycol. Crystals were soaked in ATP or MgATP made up in reservoir solution. Diffraction data were collected at 100K and processed using Denzo and Scalepack. The structures were solved by molecular replacement and refined using Refmacδ with manual model building using O. Bound nucleotide analysis was carried out by denaturation of samples, removal of protein and analysis by hplc. Λ/-methylanthraniloyl (mant) nucleotides were prepared according to published methods. Uncorrected fluorescence emission spectra of the mant nucleotides and their complexes with AMPK were recorded at 20° C using a SPEX FluoroMax fluorimeter. Binding of mant-AXPs was monitored by titrating 10μM of the nucleotides with AMPK. Dissociation constants for AMP and ATP were determined using competition assays.

Expression and crystallization of AMPK

Using a tri-cistronic expression system we were able to express and purify full-length mammalian AMPK. Truncated AMPK, γ1_(ral)β2_(human)(187-272)His-α1 ₍^(396-548), was cloned into a tricistronic vector and subsequently expressed in E.coli BL21 (BL21-CodonPlus-RIL, stratagene)³¹. Proteins were purified using a Nickel affinity chromatography (His-Trap, GE Healthcare), anion exchange (Mono Q, GE Healthcare), and gel filtration (Superdex 200, GE Healthcare).

As a result of cloning the sequence of the alpha subunit fragment into the vector, residues 545- 548 (ILAQ) of the alpha chain were replaced by NSCT and two additional residues, V549 and N550, were incorporated.

We measured diffraction data extending to 2.1 A Bragg spacing from crystals obtained from the trimeric complex of the C-terminal domains of the α fragment and β (187-272) with full-length v.

Protein complex stock solution was prepared at 15 mg/ml in 50 mM Tris, pH 7.0, 100 mM NaCI and 1 mM TCEP, and then mixed with a three-fold molar excess of AMP. Crystals were grown by vapour diffusion technique at 18⁰C in hanging drops. Drops were prepared by mixing equal volumes of protein complex with reservoir solution containing 15% isopropanol. Crystals were first transferred into mother liquor with an additional 25% ethylene glycol, prior to plunging into liquid nitrogen. For crystal soaking experiments 1mM ATP/0.1 mM EDTA and 0.5mM ATP/0.6 mM MgCI₂ were made up in reservoir buffer, crystals were soaked for 1-5 hours prior to flash cooling. Diffraction data were collected on an in-house MicroMax 007HF rotating anode coupled to a RaxislV⁺⁺ detector. Data were integrated using Denzo and scaled with Scalepack³². The structure was solved by molecular replacement using Amore³³ using

2OOX.pdb as the search model, because of the differences in domain orientation between the yeast and mammalian enzyme the calculations were first performed with the gamma domain, which was then fixed, and the search repeated with the combined alpha and beta domains. Standard refinement was carried out with Refmacδ³⁴ and CNS³⁵ together with manual model building with O³⁶. Figures were created with Pymol (http://pymol.soiarceforge.net/) and Grasp, The structure was solved by molecular replacement and relevant crystallographic statistics are given in Table 7.

Table 7:

AMPK/AMP3 AMPK/(ATP.Mg)₂AMP AMPK/ATP₂AMP complex complex complex

Data collection

Space group P2i2i2, P2₁2₁2₁ P2₁2₁2₁

Cell dimensions a, b, c (A) 48.5, 119.4, 129.2 48.8, 120.8, 126.8 49.0, 121.4,

127.1 a, A r C) 90, 90, 90

Resolution (A) 20-2.1 (2.18-2.1)* 25-2.6 (2.7-2.6) 20-2.3 (2.4-2.3)

"sym Or /Emerge 5.9 (41.6) 9.9 (48.0) 9.8 (44.3)

IhI 18.7 (2.2) 12.0 (1.7) 16.5 (1.9)

Completeness (%) 99.0 (98.4) 95.8 (87.9) 99.1 (95.2)

Redundancy 5.6 (5.1) 6.0 (2.1 ) 7.3 (2.5)

Refinement

Resolution (A) 20-2.1 25-2.6 20-2.3

No. reflections 42054 21913 31118

°work/ "free 21.4/24.6 24.8/27.1 23.4/28.4

No. atoms

Protein 3884 3884 3884

Ligand/ion 69 87 85

Water 442 - 176

B-factors

Protein 44.6 56.9 53.1

Ligand/ion 35.0 37.6 45.7

Water 62.1 - 59.2

R.m.s deviations

Bond lengths (A) 0.009 0.010 0.010

Bond angles (°) 1.44 1.52 1.47

One crystal was used for each of the two datasets. "Highest resolution shell is shown in parenthesis. Further refinement

The AMPK/AMP₃ structure of Table 1a and AMPK/(ATP.Mg)₂AMP structure of Table 2a were subject to further refinement to produce the structures of Tables 1b and 2b respectively.

The structures of Table 1a and Table 2a were determined as for a complex including the native sequence of the rat alpha subunit fragment 396-548. The structures of Table 1b and Table 2b were determined as for a complex including the rat alpha subunit fragment 396-548 in which wild type residues 545 to 548 were replaced by NSCT and with an additional two C terminal residues VN as residues 549 and 550. As noted above, the fragment used for crystallisation consisted of residues 396 to 544 of the rat α1 subunit, fused at the C terminus to NSCTVN.

The co-ordinates in Table 1b and Table 2b also reflect the rebuilding of some loop residues as a further refinement of the structure of Table 1a and Table 2a respectively.

Binding Studies To understand better the regulation of AMPK by changes in AMP/ATP ratio we carried out a series of binding studies.

Λ/-methylanthraniloyl (mant) nucleotides were synthesised, purified and characterised as described elsewhere³⁷. Uncorrected fluorescence emission spectra of the mant nucleotides and their complexes with AMPK were recorded at 20° C in 25mM Tris, 1mM TCEP (pH 7.5) using a SPEX FluoroMax fluorimeter with lambda ex =366 or 380 nm (bandwidth 1.7 nm) and emission scanned from 400 to 550 nm (bandwidth 5 nm). Binding of mant-AXPs was monitored by titrating 10μM of the nucleotides with AMPK. Dissociation constants for AMP and ATP were determined using two different types of competition assay.

The uncomplexed mant-nucleotides have an emission maximum at 446 nm. In the presence of saturating AMPK the emission maximum is blue-shifted and the fluorescence intensity is increased. The emission maxima and intensity ratios (Complexed/Free) for total integrated fluorescence (400-550 nm) are: 429 nm and 5.25 for mant-AMP; 436 nm and 5.35 for mant- ADP; 437 nm and 5.55 for mant-ATP. These fluorescence changes were used to determine apparent dissociation constants by titrating ~ 10 μM mant-nuceotides with AMPK in 25 mM Tris (pH 8) at 2O⁰C. Titrations of mant-ATP with the full-length protein and with the crystallization fragment clearly indicated that the two proteins bind mant-ATP with similar affinity and show that binding cannot be described using a simple 1:1 binding model (the dashed line is the best fit attainable with such a model). The K_ds were therefore determined using standard methods with the assumptions that there are two identical, non-interacting sites for nucleotide binding and that the binding at each site generates the same change in the fluorescence spectrum of the mant-nucleotide. The observation that the intensity ratios are closely similar for all three nucleotides is consistent with the view that the same number of nucleotides is bound in all cases. Binding data for mant-AMP were therefore analysed with the same two-site model.

Dissociation constants were determined to be 4 ± 0.5 μM for mant-ATP and 10 + 2 μM for mant-AMP. Increasing the salt concentration to 100 mM increased the Ka by a factor of ~ 4 for both nucleotides, to 14 ± 3 μM for mant-ATP and to 38 ± 7 μM for mant-AMP. Increasing the magnesium concentration to 125 μM slightly increased the Kd for mant-ATP (to 5.5 ± 0.8 μM) but had no effect on the affinity for mant-AMP. Much higher concentrations of magnesium led to significant decreases in the affinity for mant-ATP but had relatively little effect on the affinity for mant-AMP. Since the effect of, for example, 1.25 mM MgCI₂ is significantly greater than the effect of 100 mM NaCI the magnesium effect is not related to ionic strength changes. In the low salt buffer used in these measurements the K_d for mant-ATP binding to Mg²⁺ was determined using ITC to be 5.3 ± 1 μM (cf. 15 ± 2 μM for ATP itself) in 25 mM Tris (pH 8) at 20⁰C. Therefore, the mant-ATP is largely Mg²⁺ saturated at 125 μM Mg²⁺ and the observed effects cannot be attributed to changes in the saturation of the nucleotide by Mg²⁺. We do not know whether this phenomenon is a result of a weak-binding site for the metal on the enzyme, or on the nucleotide itself, but we conclude that the effect is not physiologically relevant to AMPK regulation.

Dissociation constants for unlabelled nucleotides were determined by using the mant- nucleotides as reporters for the binding of unlabelled nucleotides. Competition assays were performed in which the mant-nucleotide plus unlabelled nucleotide (with and without magnesium) was titrated with AMPK. These data were then analyzed with the previously determined K_d for the mant-nucleotide held constant in the analysis using methods described elsewhere (Ref 41 ).

From the titration studies described in the next paragraph we observed two AMP binding sites on AMPK, apparently in contradiction to the three molecules seen in our crystal structure. We therefore analysed our purified, over-expressed protein by high-pressure-liquid- chromatography for bound nucleotides and discovered that for both the crystallisation fragment, and full-length protein, there was approximately one adenyl molecule present for every molecule of AMPK. Most of the adenyl species (85%) was present as AMP with a small amount of ADP (15%). As yet we have not managed to displace this co-purifying adenyl moiety and are therefore unable to determine its binding affinity. However, the fact that it persists through the various stages of protein purification, including gel filtration, suggests that the binding is very tight.

We synthesised /V-methylanthraniloyl (mant) labelled analogues of AMP and ATP to use as probes for determining the affinity of the unlabelled nucleotides. AMPK binds two mant- nucleotides with dissociation constants of 4 ± 0.5 μM (mant-ATP) and 10 ± 2 μM (mant-AMP). Dissociation constants for the unlabelled nucleotides were determined using competition assays in which a mant-nucleotide plus unlabelled nucleotide (either AMP or ATP) was titrated with AMPK. The values determined were 12.5 + 4 μM for ATP and 24.5 ± 4.5 μM for AMP. We also obtained very similar results from titrations carried out with full-length AMPK. Given the recent description of the structure of the AMPK homologue from S. pombe in complex with Mg²⁺-free ATP we investigated the effect of the metal ion on nucleotide binding with mammalian AMPK. Initial experiments indicated that Mg²⁺ reduced the apparent affinity of AMPK for ATP, however, this effect is only observed at very high concentrations of the metal ion that are not physiologically relevant. Titrations of ATP carried out in the presence of 125 μM Mg²⁺ (which will saturate most of the ATP) are the same as those done in the absence of the metal ion. Therefore, at least for mammalian AMPK, the enzyme binds ATP or Mg.ATP with the same affinity. This means that under physiological conditions (where nearly all ATP is complexed with Mg²⁺ and there is about 0.4mM free Mg²⁺²⁰) the majority of the enzyme will be in complex with two Mg.ATP moieties. Mammalian AMPK is therefore different to proteins, like kinases and small GTPases, whose affinity for the nucleotide is strongly dependent on Mg²⁺ binding but it is also different to its homolgue from S. pombe where it is inferred that ATP- bound metal counter ions must be stripped before binding to AMPK. Our results are consistent with earlier studies of adenyl binding to AMPK v domain²¹, which concluded that the affinity of the protein for ATP was independent of Mg²⁺.

Taken together, our studies indicate that under physiological conditions the majority of AMPK will contain Mg.ATP and only a small proportion of the enzyme will have AMP bound at the two regulatory binding sites.

Modelling AMPK activity

AMPK has substantial basal activity once phosphorylated that can be modestly enhanced by the allosteric effect of AMP binding. However, AMP binding also significantly decreases the rate of dephosphorylation of activated AMPK. Any molecular model of AMPK action must account for both of these phenomena. We think the lack of any significant conformational change in the Y subunit between the AMP and ATP bound forms implies that the activation signal, initiated by AMP binding, is not propagated via the limited rearrangements of some basic side-chains at the binding site. Instead, we hypothesise that AMP binding triggers the formation of inter-subunit interactions that are not possible when ATP is bound. Such changes in the quaternary structure of the heterotrimer then influence the activity of the kinase domain and its susceptibility to dephosphorylation. Given the structure and charge associated with the ATP binding sites, a plausible binding partner for AMP would be a phosphorylated residue from either the α and/or β subunits (at least eight such phosphorylation sites have been identified). The fact that our structure shows there are two exchangeable AMP/ATP sites, on opposite faces of the y subunit, suggests the possibility that two distinct inter-subunit interactions with y occur, perhaps providing for separate mechanisms for the allosteric activation and dephosphorylation protection activities of AMP.

To describe how changes in the cellular ATP/AMP ratio affect AMPK activity we set up a simple model to simulate the enzyme's activity using the binding constants determined here together with data from the literature concerning AMPK activation and adenyl concentrations. Key assumptions of the model are that AMP binding to the enzyme causes allosteric activation as well as inhibiting de-phosphorylation of the enzyme's T-loop and that the non-phosphorylated form has negligible activity. The model shows that 2-3 fold increases in the low micromolar concentrations of AMP bring about a similar fold enhancement in the proportion of AMP-bound enzyme in the presence of a much higher, and essentially unchanging, concentration of Mg.ATP. Importantly, this change in AMP-bound AMPK accounts for the 2-3 fold enhancement of muscle AMPK activity observed following moderate exercise²²'²³. A key prediction of this model is that only a small proportion of AMPK is active at any one time. This notion, of a large pool of essentially inactive AMPK, is supported by studies using prolonged exercise²⁴ or non- physiologic methods to activate AMPK, e.g. respiratory uncouplers, that generate a large increase in AMP and lead to increases in enzyme activity at least 5-10 fold greater than that stimulated by moderate exercise²⁵. The model is set forth as follows:

1. Under physiological conditions the majority of AMPK will contain Mg.ATP at the two regulatory sites and only a very small proportion (<1%) will have AMP bound at these sites because the nucleotides bind with similar affinity and Mg.ATP is in large excess over AMP.

2. How then do changes in the cellular ATP/AMP ratio affect AMPK activity?

3. To address this issue we set up the following highly simplified model. In this model the asterisk indicates that the enzyme is phosphorylated, the K_d (and K_d*) values are for binding of AMP and ATP to the non-phosphorylated (and phosphorylated) states, and the kP (and kdP) values are rates of phosphorylation (and de-phosphorylation) for the ATP-bound, AMP-bound, and apo (nucleotide free) forms of the enzyme.

K_dT K_dM

AMPK-ATP AMPK AMPK-AMP

kP(T) kdP(T) kP() kdP() kP(M) kdP(M)

AMPK*-ATP AMPK* AMPK*-AMP κ_dτ* K_dM*

4. Two key features of the model are that all the non-phosphorylated forms have negligible activity and that AMP binding to the enzyme causes allosteric activation, with AMPK*- AMP being 2-5 times more active than AMPK*-ATP, depending on the isoform composition of the complex (Ref 42).

5. The experimental finding is that a doubling of the low micromolar AMP concentration in the presence of say 5 mM Mg.ATP produces a 2-3 fold enhancement of muscle AMPK activity following exercise (Ref 43). If phosphorylation does not affect the affinity of

AMPK for the nucleotides (see below) then doubling the amount of AMP will effectively double the concentration of enzyme with AMP bound (Note: This will still only result in a small proportion of the total enzyme being in the AMP-bound form).

6. At low [AMP] the 'activity' will be:

A_AMPK--ATP[AMPK*-ATP] + A_AMPK*-_AMP[AMPK*-AMP]

where A_AMPK*-ATP and A_AMPK*-_AMP are the activities of the ATP and AMP bound forms. If we assume that doubling the AMP concentration doubles the concentration of AMPK*-AMP but leaves the concentration of AMPK*-ATP unchanged then the 'activity' at high AMP will be:

A_AMPK._-ATP[AMPK*-ATP] + 2 x A_AMP,<._-AMP[AMPK*-AMP]

The activity ratio (activity at high [AMP] divided by activity at low [AMP] will then be:

A_AMPκ*-_ATp[AMPK*-ATP] + 2 X A_AMPK*-AMP[AMPK*-AMP]

A_AMPK*-Aτp[AMPK*-ATP] + A_AMPK*-_AMP[AMPK*-AMP]

Let A_AMPK*-_Aτp = 0.33 x A_AMPK*-AMP, divide top and bottom by [AMPK*-ATP] and let β = [AMPK*- AMP]/[AMPK*-ATP] gives the activity ratio as:

0.33 X A_AMPK*-AMP + 2β X A_AMPK*-AMP 0.33 * A_AMPK*-AMP + β X A_AMPK*-AMP

7. This relationship clearly shows that a doubling in activity can only be observed if β (i.e., [AMPK*-AMP] divided by [AMPK*-ATP]) is close to or greater than one. However, if nucleotide binding is not affected by phosphorylation then R would be < 0.002 (see (1)) under physiological conditions.

8. Because most of the AMPK will have Mg.ATP bound at the regulatory sites (see (1)) the ratio β can only approach 1 if the concentration of AMPK*-ATP is very much smaller than the concentration of AMPK-ATP. If it is not then [AMPK*-ATP] would always remain much greater than [AMPK*-AMP] (assuming that the rates of phosphorylation and de-phosphorylation are unaffected by nucleotide content - see below). Thus, a key prediction of this model is that only a very small proportion of the total AMPK is active at any one time.

9. The third key assumption of this model is then that the binding of AMP inhibits de- phosphorylation of the enzyme's T-loop (i.e. kdP(M) is much smaller than kdP(T) or kdP()). With this assumption, a large fraction of the enzyme with AMP bound could be in the phosphorylated (active) state whereas almost all of the enzyme with Mg.ATP bound would be in the non-phosphorylated (inactive) state, and the concentrations of AMPK*- ATP and AMPK*-AMP would be similar (β close to 1).

10. Another factor that could contribute to maintaining β close to 1 would be if phosphorylation reduced the affinity of the enzyme for ATP (i.e. K_d(T*) > Kd(T)). 11. Although many much more complex models are possible (we have not included cooperativity in nucleotide binding, or the possibility that mixed nucleotide species might have different activity etc.) this model provides a working explanation as to the activation of AMPK and is consistent with the presence of the three nucleotide biding sites described herein.

All publications and patents mentioned in the above specification are herein incorporated by reference. Various modifications and variations of the described invention will be apparent to those of skill in the art without departing from the scope and spirit of the invention. Although the invention has been described in connection with specific preferred embodiments, it should be understood that the invention as claimed should not be unduly limited to such specific embodiments.

Sequences:

SEQ ID NO:1 - Rat AMPK Gamma Subunit (Genbank X95578)

MΞSVAAESAP APENEHSQET PESNSSVYTT FMKSHRCYDL IPTSSKLVVF DTSLQVKKAF FALVTNGVRA APLWDSKKQS FVGMLTITDF INILHRYYKS ALVQIYELEE HKIETWREVY

LQDSFKPLVC ISPNASLFDA VSSLIRNKIH RLPVIDPESG NTLYILTHKR ILKFLKLFIT

EFPKPEFMSK SLEELQIGTY ANIAMVRTTT PVYVALGIFV QHRVSALPW DEKGRVVDIY

SKFDVINLAA EKTYNNLDVS VTKALQHRSH YFEGVLKCYL HETLEAIINR LVEAEVHRLV VVDEHDVVKG IVSLSDILQA LVLTGGEKKP

SEQ ID NO:2 - Human AMPK Gamma Subunit (Genbank NP_002724)

METVISSDSS PAVENEHPQE TPESNNSVYT SFMKSHRCYD LIPTSSKLVV FDTSLQVKKA FFALVTNGVR AAPLWDSKKQ SFVGMLTITD FINILHRYYK SALVQIYELE EHKIETWREV YLQDSFKPLV CISPNASLFD AVSSLIRNKI HRLPVIDPES GNTLYILTHK RILKFLKLFI TEFPKPEFMS KSLEELQIGT YANIAMVRTT TPVYVALGIF VQHRVSALPV VDEKGRWDI YSKFDVINLA AEKTYNNLDV SVTKALQHRS HYFEGVLKCY LHETLETIIN RLVEAEVHRL VVVDENDVVK GIVSLSDILQ ALVLTGGEKK P

References:

1. Kahn, B. B., Alquier, T., Carling, D. & Hardie, D. G. AMP-activated protein kinase: Ancient energy gauge provides clues to modern understanding of metabolism. Cell Metab. 1, 15-25 (2005).

2. Hardie, D. G., Carling, D. & Carlson, M. The AMP-activated/SNF1 protein kinase subfamily: metabolic sensors of the eukaryotic cell? Annu. Rev. Biochem. 67, 821-855 (1998).

3. Carling, D. The AMP-activated protein kinase cascade-a unifying system for energy control. Trends Biochem Sci 29, 18-24 (2004).

4. Hardie, D. G. & Carling, D. The AMP-activated protein kinase: Fuel gauge of the mammalian cell. Eur. J. Biochem. 246, 259-273 (1997).

5. Yeh, L., Lee, K. & Kim, K. Regulation of rat liver acetyl-CoA carboxylase. Regulation of phosphorylation and inactivation of acetyl-CoA carboxylase by adenylate energy charge. J. Biol. Chem. 255, 2308-2134 (1980).

6. Hardie, D. G., Carling, D. & Sim, A. T. R. The AMP-activated protein kinase-a multisubstrate regulator of lipid metabolism. Trends Biochem. Sci. 14, 20-23 (1989).

7. Minokoshi, Y. et al. AMP-kinase regulates food intake by responding to hormonal and nutrient signals in the hypothalamus. Nature 428, 569-574 (2004).

8. Viollet, B. et al. The AMP-activated protein kinase a2 catalytic subunit controls whole body insulin sensitivity. J. Clin. Invest. 111 , 91-98 (2003).

9. Zong, H. et al. AMP kinase is required for mitochondrial biogenesis in skeletal muscle in response to chronic energy deprivation. Proc Natl Acad Sci U S A 99, 15983-7 (2002).

10. Minokoshi, Y. et al. Leptin stimulates fatty-acid oxidation by activating AMP-activated protein kinase. Nature 415, 339-43 (2002). 11. Yamauchi, T. et al. Adiponectin stimulates glucose utilization and fatty-acid oxidation by activating AMP-activated protein kinase. Nature Med. 8, 1288-1295 (2002).

12. Watt, M. J., et al. CNTF reverses obesity-induced insulin resistance by activating skeletal muscle AMPK. Nat Med 12, 541-8 (2006).

13. Zhou, G. et al. Role of AMP-activated protein kinase in mechanism of metformin action. J Clin Invest 108, 1167-74 (2001).

14. Shaw, R. J. et al. The kinase LKB1 mediates glucose homeostasis in liver and therapeutic effects of metformin. Science 310, 1642-6 (2005).

15. Sanders, M. J., Grondin, P. O., Hegarty, B. D., Snowden, M. A. & Carling, D. Investigating the mechanism for AMP activation of the AMP-activated protein kinase cascade. Biochem. J. in press (2006).

16. Woods, A. et al. Characterization of AMP-activated protein kinase b subunit and g subunit - assembly of the heterotrimeric complex in vitro. J. Biol. Chem. 271 , 10282-10290 (1996).

17. Iseli, T. J., Walter, M., van Denderen, B.J., Katsis, F., Witters, L.A., Kemp, B.E., Michell, BJ. , Stapleton, D. AMP-activated protein kinase beta subunit tethers alpha and gamma subunits via its C-terminal sequence (186-270). J Biol Chem 280, 13395-400 (2005).

18. Townley, R. & Shapiro, L. Crystal structures of the adenylate sensor from fission yeast AMP-activated protein kinase. Science 315, 1726-9 (2007).

19. Rudolph, M. J., Amodeo, G.A., Bai, Y., Tong, L Crystal structure of the protein kinase domain of yeast AMP-activated protein kinase Snfl . Biochem Biophys Res Commun. 337, 1224-8 (2005).

20. Corkey, B. E., Duszynski, J., Rich, T.L., Matschinsky, B., Williamson, J. R. Regulation of free and bound magnesium in rat hepatocytes and isolated mitochondria. J. Biol. Chem. 261 , 2567-2574 (1986).

21. Scott, J. W. et al. CBS domains form energy-sensing modules whose binding of adenosine ligands is disrupted by disease mutations. J Clin Invest 113, 274-84 (2004).

22. Chen, Z. et al. AMPK signaling in contracting human skeletal muscle: acetyl-CoA carboxylase and NO synthase phosphorylation. Am J Physiol Endocrinol Metab. 279, 1202- 1206 (2000).

23. Fujii, N. et al. Exercise induces isoform-specific increase in 5'AMP-activated protein kinase activity in human skeletal muscle. Biochem. Biophys. Res. Commun. 273, 1150-5 (2000).

24. McConell, G. K., Lee-Young, R.S., Chen, Z.P., Stepto, N.K., Huynh, N.N., Stephens, TJ., Canny, BJ. , Kemp, B. E. Short-term exercise training in humans reduces AMPK signalling during prolonged exercise independent of muscle glycogen. J Physiol. 568, 665-76 (2005).

25. Fryer, L. G., Parbu-Patel, A. & Carling, D. The Anti-diabetic drugs rosiglitazone and metformin stimulate AMP-activated protein kinase through distinct signaling pathways. J Biol Chem 277, 25226-32 (2002).

26. Gollob, M. H. et al. Identification of a gene responsible for familial Wolff-Parkinson- White syndrome. New Eng. J. Med. 344, 1823-1831 (2001).

27. Arad, M. et al. Constitutively active AMP kinase mutations cause glycogen storage disease mimicking hypertrophic cardiomyopathy. J. Clin. Invest. 109, 357-362 (2002). 28. Blair, E., Redwood, C, Ashrafian, H., Ostman-Smith, I. & Watkins, H. Mutations in the g2 subunit of AMP-activated protein kinase cause familial hypertrophic cardiomyopathy: evidence for the central role of energy compromise in disease pathogenesis. Hum. MoI. Genet. 10, 1215-1220 (2001).

29. Daniel, T. & Carling, D. Functional analysis of mutations in the gamma2 Subunit of AMP-activated protein kinase associated with cardiac hypertrophy and Wolff-Parkinson-White Syndrome. J. Biol. Chem. 277, 51017-51024 (2003).

30. Adams, J. et al. lntrasteric control of AMPK via the gammal subunit AMP allosteric regulatory site. Protein Sci 13, 155-65 (2004).

31. Neumann, D., Woods, A., Carling, D., Wallimann, T. & Schlattner, U. Mammalian AMP- activated protein kinase: functional, heterotrimeric complexes by co-expression of subunits in Escherichia coli. Protein Expr Purif 30, 230-7 (2003).

32. Otwinowski, Z. & Minor, W. in Data Collection and Processing (eds. Sawyer, L., Isaacs, N. & Bailey, S.) 556-562 (SERC Daresbury Laboratory, Warrington, 1993).

33. Navaza, J. AMoRe: an Automated Package for Molecular Replacement. Acta Crystallogr A 50, 157-163 (1994).

34. CCP4. The CCP4 suite: programs for protein crystallography. Acta Crystallogr D 50, 760-763 (1994).

35. Brunger, A. T. et al. Crystallography & NMR system: A new software suite for macromolecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905-21. (1998).

36. Jones, T. A., Zhou, J. Y., Cowan, S. W. & Kjeldgaard, M. Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47, 110-119 (1991).

37. Jameson, D. M. & Eccleston, J. F. Fluorescent nucleotide analogs: synthesis and applications. Methods Enzymol. 278, 363-390 (1997).

38. Arad, M. et al. Transgenic Mice Overexpressing Mutant PRKAG2 Define the Cause of Wolff-Parkinson-White Syndrome in Glycogen Storage Cardiomyopathy. Circulation 107, 2850- 2856 (2003).

39. Daniel, T. & Carling, D. Functional analysis of mutations in the gamma2 Subunit of AMP-activated protein kinase associated with cardiac hypertrophy and Wolff-Parkinson-White Syndrome. J. Biol. Chem. 277, 51017-51024 (2003).

40. Scott, J. W. et al. CBS domains form energy-sensing modules whose binding of adenosine ligands is disrupted by disease mutations. J Clin Invest 113, 274-84 (2004).

41. Martin, S. R., Bayley, P.M. Regulatory implications of a novel mode of interaction of calmodulin with a double IQ-motif target sequence from murine dilute myosin V. Protein Sci. 11 , 2909-23 (2002).

42. Cheung, P. C. F., Salt, I. P., Davies, S. P., Hardie, D. G. & Carling, D. Characterization of AMP-activated protein kinase g-subunit isoforms and their role in AMP binding. Biochem. J. 346, 659-669 (2000).

43. Stephens, T. J. et al. Progressive increase in human skeletal muscle AMPKalpha2 activity and ACC phosphorylation during exercise. Am J Physiol Endocrinol Metab 282, E688- 94 (2002).

Claims

Claims:

1. A computer-based method for the analysis of the interaction of a molecular structure with an AMPK gamma subunit structure, which comprises: providing an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof; providing a molecular structure to be fitted to said AMPK structure or selected coordinates thereof; and fitting the molecular structure to said AMPK structure.

2. The method of claim 1 wherein said selected coordinates include atoms from one or more of the residues of Tables 3, 4 or 5.

3 The method of claim 1 or 2 wherein the selected coordinates are of at least 5, 10, 50, 100, 500 or 1000 atoms.

4. The method of any one of the claims 1 to 3 which further comprises the steps of:

(a) obtaining or synthesising a compound which has said molecular structure; and

(b) contacting said compound with an AMPK gamma subunit protein to determine the ability of said compound to interact with said protein.

5. The method of any one of claims 1 to 3 which further comprises the steps of:

(a) obtaining or synthesising a compound which has said molecular structure;

(b) forming a complex of an AMPK gamma subunit protein and said compound; and

(c) analysing said complex by X-ray crystallography.

6. The method of any one of claims 1 to 3 which further comprises the steps of:

(b) determining or predicting how said compound is bound to an AMPK gamma subunit; and

(c) modifying the compound structure so as to alter the interaction between it and an AMPK gamma subunit.

7. The method of claim 6 wherein the compound is modified to alter its interaction with one or more atoms of the residues of any one of Tables 3, 4 or 5.

8. The method of any one of claims 1 to 7 wherein a plurality of molecular structures are fitted to said AMPK gamma subunit structure are assembled into a single molecule to form a candidate modulator molecule.

9. A method of identifying a potential ligand of an AMPK gamma subunit protein comprising a nucleotide binding pocket, said method comprising the steps of: (a) using the atomic coordinates of Table 1 or Table 2 to generate a three-dimensional model;

(b) identifying said binding pocket residues which comprises at least one or more of the residues set out in any one of Tables 3 to 5, and using said residues to generate a specific three-dimensional (3-D) target;

(c) employing said 3-D target of (b) to design or select said potential ligand;

(d) synthesizing said potential ligand; and

10. The method of claim 9 wherein said AMPK gamma subunit protein is selected from the group consisting of a gamma 1 subunit or fragment thereof comprising both Bateman domains which include the CBS 1+2 pair and the CBS 3+4 pair; a gamma-2 subunit or C-terminal fragment thereof retaining both Bateman domains; and a gamma-3 subunit retaining both Bateman domains.

11. The method of claim 9 wherein said AMPK gamma subunit protein is selected from the group of SEQ ID NO:1 and SEQ ID NO:2.

12. A method of designing a compound which binds to a nucleotide binding pocket of an AMPK gamma subunit protein, wherein said protein is characterized by:

(i) the atomic coordinates of amino acids 23-326 of SEQ ID No: 1 shown in Table 1, optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A;

(ii) the atomic coordinates of amino acids 23-326 of SEQ ID No: 1 shown in Table 2, optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A; and/or

(iii) the atomic coordinates of a nucleotide binding pocket defined by amino acids of any one of Tables 3, 4 and 5, optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A; said method comprising the steps of:

(b) assessing the stereochemical complementarity between a compound and said interaction region; (c) optimizing said stereochemical complementarity in an iterative approach by observing changes in the protein or compound that affect the protein/compound associations; and

13. A method of obtaining a structure of an AMPK gamma subunit protein, the method comprises the steps of: providing a crystal of said protein; obtaining an X-ray diffraction pattern of said crystal, calculating a three-dimensional atomic coordinate structure of said protein, by modelling its structure on the AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof.

14. The method of claim 9 wherein said target protein is a mammalian AMPK gamma-2 or a mammalian AMPK gamma-3 protein.

15. A method for determining the structure of a compound bound to an AMPK gamma subunit protein, said method comprising:

(a) providing a crystal of said protein;

(b) soaking the crystal with the compound to form a complex; and

(c) determining the structure of the complex by employing the data of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof.

16. A method for determining the structure of a compound bound to an AMPK gamma subunit protein, said method comprising:

(a) mixing protein with the compound;

(b) crystallizing a protein-compound complex; and

17. A method of obtaining a representation of the three dimensional structure of a crystal of the AMPK gamma subunit, which method comprises providing the data of an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof, and constructing a three- dimensional structure representing said coordinates.

18. A computer system, intended to generate structures and/or perform optimisation of compounds which interact with an AMPK gamma subunit, the system containing computer- readable data comprising one or more of: (a) AMPK co-ordinate data of an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-σ atoms of less than 1.5 A, or selected coordinates thereof, said data defining the three-dimensional structure of the gamma subunit of AMPK or said selected coordinates thereof;

(c) atomic coordinate data of a target protein generated by interpreting X-ray crystallographic data or NMR data by reference to the co-ordinate data of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof.

19. A computer system according to claim 18, wherein said atomic coordinate data is for at least one of the atoms of any one of Tables 3, 4 or 5.

20. A computer system according to any one of claims 18 or 19 comprising: a computer-readable data storage medium comprising data storage material encoded with said computer-readable data; a working memory for storing instructions for processing said computer-readable data; and a central-processing unit coupled to said working memory and to said computer- readable data storage medium for processing said computer-readable data and thereby generating structures and/or performing rational drug design.

21. A computer system according to claim 20 further comprising a display coupled to said central-processing unit for displaying said structures.

22. A method of providing data for generating structures and/or performing optimisation of compounds which interact with an AMPK gamma subunit the method comprising:

(i) establishing communication with a remote device containing

(c) atomic coordinate data of a target protein generated by interpreting X-ray crystallographic data or NMR data by reference to the co-ordinate data of an AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof; and

(ii) receiving said computer-readable data from said remote device.

23. The method of claim 22 which further comprises use of the computer readable data to perform the method of any one of claims 1 to 17.

24 A computer-readable storage medium, comprising a data storage material encoded with computer readable data, wherein the data are defined by all or a portion of the structure coordinates of the AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof.

25. Use of a computer-readable storage medium, comprising a data storage material encoded with computer readable data, wherein the data are defined by the structure coordinates of the AMPK gamma subunit structure of Table 1 or Table 2 optionally varied by a root mean square deviation of residue C-α atoms of less than 1.5 A, or selected coordinates thereof in a method of performing a computer-based method of any one of claims 1 to 17.

26 A co-crystal of AMPK gamma subunit protein and a ligand, wherein said crystal further comprises a C-terminal portion of an AMPK alpha subunit protein and a C-terminal portion of an AMPK beta subunit protein.

27. The crystal of claim 26 with cell dimensions a=49, b=120, c=128 A, 90°, 90°, 90°.with a unit cell variability of 5% in all dimensions.

28. An assay method for a modulator of AMPK activity, which method comprises: providing a candidate AMPK modulator compound; bringing said compound into contact with a mammalian AMPK gamma subunit or fragment thereof having two Bateman domains; and determining the binding of said candidate compound for said gamma subunit or fragment thereof in competition with an adenosine nucleotide; wherein said adenosine nucleotide is labelled with a fluorescent label.

29. The method of claim 28 wherein the fluorescent label is N-methylanthraniloyl.

30. The method of claim 28 wherein said adenosine nucleotide is mant-AMP or mant-ATP.