Michael Dunn explains how analysing the total protein content of organisms is helping chemists to decipher the mechanisms controlling gene expression.
Scientists published the first complete genome sequence � of the bacterium Haemophilus influenzae � less than three years ago. Since that time, researchers have completed the genome sequences of 11 more microorganisms and many others are in progress (Table 1 and http://www.tigr.org/tdb/mdb/mdb.html). The! y have completed the sequence of only one eukaryotic organism to date � the yeast Saccharomyces cerevisiae � but are making significant progress for several other species, with a current target date for completing the human genome by 2005. This wealth of information represents an invaluable resource in terms of understanding how an organism functions and its evolutionary relationships with other forms of life.
Table 1. Some sequenced and partially sequenced genomes Organism Size/million base &nbs! p; &nbs p; pairs of DNA ORFs
Year completed Microorganisms
Mycoplasma genitalium 0.58 470
1995 Ureaplasma urealyticum &n! bsp; 0.75 640
Mycoplasma pneumoniae 0.81 679
1996 Treponema pallidum  ! ; 1.14 1000
Borrelia burgdorferi 1.44 843
1997 Aquifex aeolicus 1.50 ! 1512
1998 Helicobacter pylori 1.66 1590
1997 Methanococcus jannaschii 1.66 &nbs! p; 1738
1996 M. thermoautotrophicum 1.75 1855
1997 Haemophilus influenzae 1.83 1743
1995 &! nbsp; Streptococcus pyogenes 1.98 1900
Archaeoglobus fulgidis 2.18 2436
1997 Nisseria gonorrhoreae &n! bsp; &n bsp; 2.20 2100
Pyrobaculum aerophilum 2.22 1900
Synechocystis PCC6803 ! 3.57 3168
1996 Bacillus subtilis 4.20 4100
1997 Escherichia coli 4.60 &n! bsp; &n bsp; 4288
Saccharomyces cerevisiae (yeast) 13.0 5885
1996 Dictyostelium doscoideum (slime mould) 70 &n! bsp; 12 500
Arabidopsis thalania (cress) 70 14 000
Caenorhabditis elegans (nematode worm) 80 &n! bsp; &n bsp; 17 800
Drosophila melanogaster (fruit fly) 170 30 000
Homo sapiens (human) 2900 ! 50 000
However, it is becoming clear from these genome programmes that it is usually impossible to attribute even putative functions to as many as 40 per cent of the structural genes within a particular organism. In addition, as many as 30 per cent of the open reading frames (ORFs) are assigned functions only on the basis of homology to genes encoding proteins of known function. Although the genomics approach provides information on all of the possible ways that an organism may express its genes, it does not provide insights into the ways in which an organism may modify its pattern of gene expression in response to particular conditions.
! Direct answers
&nb sp; Scientists can solve these problems only by investigating gene expression directly, by studies at either the messenger RNA (mRNA) or protein level. They have also developed powerful techniques such as DNA micro-arrays and serial analysis of gene expression (Sage), which make it possible to undertake mass screening of mRNA expression and obtain information on which mRNAs are expressed in an organism at any particular time. However, it is becoming increasingly apparent that there is often a poor correlation between mRNA abundance and the amount of the corresponding functional molecule � the protein � present in the cell. The other major problem is tha! t studies at the mRNA level provide no information on processes of co- and post-translational modification (Fig 1) that result in polypeptides being modified by the addition of other groups. These modifications include phosphorylation, sulphation, glycosylation, hydroxylation, N-methylation, carboxymethylation, acylation, prenylation, and N-myristoylation. Such modifications usually have a profound influence on the functional properties of proteins, so that knowledge of these processes is fundamental to understanding gene expression.
Fig 1. Transcription, translation and post translational &! nbsp; processe s resulting in the production of a mature protein product from a particular gene.
These problems can be resolved by protein biochemistry techniques, and this interface between protein biochemistry and molecular biology has become known as 'proteome analysis'. The term 'proteome' was first coined by a collaborative team at Macquarie and Sydney Universities, Australia, in 1995, and they defined it as the protein complement of the genome of an organism. Currently, researchers are undertaking proteome analysis for simple organisms such as the mycoplasmas (the smallest free living organisms known), bacteri! a and yeast. Characterising the proteome of eukaryotic organisms will obviously require an enormous effort because of the size of the genome and the co- and post-translational modifications, which result in protein diversity far exceeding the complexity of the genome. The complexity of eukaryotic proteomes means that we can use proteomics in a narrower context, to define patterns of quantitative gene expression in particular cells and tissues. Researchers can then exploit this information to characterise biological processes, such as those involved in development, during the cell cycle, during cell death, in disease, and in response to p! harmaceutical intervention, extracellular stimuli and toxic agents. U ltimately, our goal is to decipher the mechanisms controlling gene expression.
The first requirement for proteome analysis is that we must be able to separate the complex mixture of proteins obtained from whole cells, tissues or organisms. Currently, the best method for separating complex protein mixtures simultaneously is two-dimensional polyacrylamide gel electrophoresis (2-DE), in which sample proteins are separated according to different properties in each dimension. The first 2-DE experiments reported in 1956 used a combination of paper and starch gel electrophoresis for separating serum proteins, and since ! that time researchers have described a variety of improved methods. However, the most commonly used approach for 2-DE is the combination of a first dimension separation by isoelectric focusing (IEF) under denaturing conditions with a second dimension separation by sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-Page). The first separation is according to charge (different proteins are focused at their respective isoelectric points, pI - the pH at which the net charge is zero); this is followed by a size-based separation (molecular weight, Mr). This orthogonal combination of two separations carried out at right angles to e! ach other results in the sample proteins being distributed across the two-dimensional (2-D) gel profile (Fig 2).
Fig 2. A 2-DE separation of 100 ug bovine heart proteins using a non-linear pH 3.5 to 10 IPG IEF gel in the first dimension and 12 per cent SDS-Page in the second dimension.
For effective use in proteome analysis, 2-DE must be capable of reproducible high resolution protein separation. In the early days of 2-DE this proved to be a problem largely because of! the nature of the synthetic carrier ampholytes (SCA) used to generate the pH gradients required for IEF. The electroendosmotic flow of water (migration of H3O+ towards the cathode during electrophoresis) occurring during IEF results in migration of these small SCA molecules towards the cathode. This process, known as cathodic drift, results in pH gradient instability and loss of the more basic proteins from the gel. This problem was overcome, at least in theory, when Amersham Pharmacia Biotech developed the Immobiline reagents for generating immobilised pH gradients (IPG) for IEF in the early 1980s. The Immobiline reagents are a series ! of eight acrylamide derivatives with the structure CH2=CH-CO-NH-R, & nbsp; where R contains either a carboxyl or tertiary amino group, giving a series of buffers with different pK values distributed throughout the pH 3 to 10 range. One can add the appropriate IPG reagents, calculated according to published recipes, to the mixture used for gel polymerisation. During polymerisation, the buffering groups that will form the pH gradient are covalently attached via vinyl bonds to the polyacrylamide backbone. IPG generated in this way are immune to the effects of electroendosmosis, so that they give us the opportunity to carry out IEF separations that are extremely stable, allowing the true equilibrium state to be attain! ed.
Scientists encountered several problems during initial attempts to implement the IPG technology to 2-D separations, but these were solved largely by Angelika G�rg and her colleagues at the Technical University in Munich, and we now prefer to use IPG IEF for the first dimension separation of 2-DE. Researchers perform IPG IEF in individual IPG gel strips, 3 to 5mm wide, cast on a plastic support. After steady-state IEF, strips are equilibrated and the separated proteins then applied either to the surface of a horizontal or to the top of a vertical SDS-containing gel. Inter-laboratory studies of heart, barley and yeast proteins show the &n! bsp;excellent reproducibility of both protein spot position and quant itation attained by this 2-DE method.
The separation capacity of 2-DE is critically dependent on gel size. The current standard combination of 18cm IPG strips for IEF with 20cm long SDS-Page gels is capable of routinely separating 2000 proteins from tissue and cell extracts. Very large format gels of about 30cm in each dimension are capable of separating up to 10,000 proteins, but this is achieved at the expense of the ease of gel handling and processing. Only a few hundred proteins can be separated using mini-gel formats, but these are quick to run and are useful for rapid screening &! nbsp;purposes. One can also choose the range of the pH gradient for IPG IEF to maximise protein separation. Wide pH gradients covering pH 3-10 allow us to analyse the protein diversity in a particular sample, while narrower pH gradients improve the resolution in particular regions of the protein profile. Very basic pH gradients, up to pH 12, are now available for 2-DE and we can use these to separate very basic proteins, such as nuclear and ribosomal proteins. A further and very important advantage of IPG IEF is that it has a very high capacity for micro-preparative 2-DE protein separations, particularly using a recently described method in which dry IPG strips are reswollen dir! ectly in a solution containing up to several mg of the protein sample to be analysed.
The next critical issue in proteome analysis is to detect the separated proteins at high sensitivity. Traditional methods of protein staining following gel electrophoresis based on the use of the dye Coomassie brilliant blue have limited sensitivity. Researchers can achieve higher detection sensitivity � 0.1ng protein per spot � by silver staining, but there can be problems using this method as a quantitative procedure because it is known to be non-stoichiometric and prone to saturation and negative staining effects, where regions of very high protein concentra! tion do not stain and appear as 'holes' in the pattern of stained spots. Detection methods based on using fluorescent compounds, which scientists are developing for use with 2-DE separations, promise to overcome these problems because of their excellent linearity and high dynamic range (ie they can be used over a wide range of protein concentrations). We can also achieve very sensitive protein detection in situations where metabolic labelling of proteins with a radiolabelled amino acid can be used before their 2-DE separation, for example in cell culture systems.
We need automated computer analysis systems for the rigorous qualitative and quantitative &n! bsp; analysis of the complex patterns of protein expressi on visualised by 2-DE. The first step in this process is to obtain a digitised image of the 2-D protein separation. A flat-bed scanning laser densitometer providing high resolution � down to 50�m � combined with a high dynamic range � up to 4OD (optical density) � is the best option for stained gels. The current generation of desktop scanners can also achieve high resolution (600dpi is equivalent to 42�m) and a high dynamic range (12bits), but care must be taken to ensure the linearity of such devices. We can also prepare film images of radiolabelled 2-D separations using such devices, but accurate quantitation is complicated both by the limited dynamic range and the &nb! sp; non-linearity of film response.
We can overcome this problem by using phosphorimaging screens, the surface of which contain a thin layer of BaFbr:Eu2+ in a plastic support. We place the dried 2-D gel in contact with the screen; during this exposure step the b-particles emitted by the radiolabelled proteins pass through the layer, converting Eu2+ to Eu3+. After a suitable exposure time, we transfer the screen to a 'phosphorimaging' scanner where light from a high intensity HeNe laser (633nm) is absorbed, causing the excited state Eu3+ ions to decay to the Eu2+ ground state by the emission of blue (390nm) luminescence proportional to the amount of! radiation incident on the s creen. The major advantages of this approach compared with conventional autoradiography is that relatively short exposure times are required, it has a high dynamic range and good linearity of response is achieved. The only disadvantage is the high capital cost of the phosphorimaging screens and the dedicated imaging device that is required. Similarly, a dedicated densitometer is required for the imaging of 2-D gel profiles of fluorescently labelled proteins
In the early days of 2-DE scientists needed large, dedicated computer systems for analysing 2-DE protein separation profiles. The rapid progress made in microcomputer technolog! y has changed this situation, so that we can run the current generation of commercial 2-DE analysis software systems � Melanie II, Bio-Rad Laboratories; BioImage, BioImage Systems; Phoretix, Phoretix International; Kepler, LSB � on desktop workstations, such as Unix, PC or Mac. These systems mean that we can derive qualitative and quantitative information from individual 2-D gels, to match the protein separation profiles from large numbers of 2-D gels and construct comprehensive databases of quantitative protein expression for cells, tissues and whole organisms.
Next, we need! to be able to identify and characterise the separated proteins. The 2-DE separation provides us with information on the apparent Mr, pI and relative abundance of the proteins, but gives us no direct clues to their identities or functions. We can search the sequence databases for proteins of matching Mr and pI, for example using the TagIdent tool in Expasy ( http://www.expasy.ch/www/guess-prot.html) but this approach is unlikely on its own to result in unequivocal protein identification. This problem is exacerbated by the uncertainty of mass values � around �10 per cent � derived from protein mobility during SDS-Page. However, we can now determine directly the abs! olute mass of proteins by mass spectrometry (Fig 3). The best approach is to transfer the separated proteins by Western blotting onto the surface of a nitrocellulose or polyvinylidene difluoride membrane. We then treat the blot with the matrix required for MS, cut out the protein of interest, and mount it directly into a matrix-assisted laser desorption (Maldi)-MS and measure the mass of the intact protein. This method is very accurate, usually within 1 per cent of the predicted mass, but requires a Maldi-MS instrument fitted with an infrared laser and works best if the proteins are not stained. This approach can also be extended to two-dimensional 'scanning' of the sample targ! et, thereby generating mass contour images.
Fig 3. Methods for identifying and characterising proteins separated by 2-DE.
Until recently we only had available laborious and time-consuming methods, such as specific staining methods, co-electrophoresis with purified proteins, cellular subfractionation and over-expression of genes introduced via plasmid or viral constructs, to identify proteins from 2-D gels. The first major advance was developing methods for Western blotting, which allowed us to probe gel-separated proteins with a variety of ligands, particularly poly- and monoclonal antibodies, after their transfer to the surface ! of an inert support such as nitrocellulose. Biochemists use this approach extensively for identifying proteins separated by 2-DE, but it is still a time-consuming method that depends on the availability of a suitable panel of specific antibodies reactive with the denatured proteins in 2-DE gels.
Amino acid sequence, even if this is only a few residues in length, is the most specific method of identifying proteins. The preferred method for chemically sequencing proteins is still based on the Edman degradation method. Edman degradation was first described in 1949, although it did not become a practical method until the first automated protein sequenator was &nbs! p; developed in 1967. A commercial version became available in 1971, which had a sensitivity limit of 10 nanomole (equivalent to 500�g for a 50kDa protein). Since 1971, progress in sequenator technology and the optimisation of Edman degradation chemistry has resulted in the current generation of gas-liquid and solid-phase sequenators, which we can use to determine N-terminal sequences from low picomole quantities of protein (5 picomole is equivalent to 0.25�g for a 50kDa protein). This level of sensitivity is compatible with the amount of protein present in many of the spots present in micro-preparative 2-D gels, and this is our preferred method if extended runs of N-terminal protein sequences are required.
&n! bsp; A major problem with N-terminal sequencing by Edman degradation is that many proteins are 'blocked', that is they lack a free a-amino group. This results from processes of co- and post-translational modification, such as those involving the addition of formyl, acetyl or acyl groups. In the case of eukaryotic organisms, we usually find that 50 per cent of all cellular proteins are modified in this way. The best approach to this problem is to cleave the blocked protein with either a chemical reagent such as CNBr or an enzyme such as trypsin, to generate shorter, 'internal' peptides that can be isolated by HPLC and sequenced. We can carry out the cleavage s! tep either in situ within the 2-D gel or after Western blot transfer of the protein onto a nitrocellulose or PVDF membrane. While this procedure is effective, it requires considerably more protein than direct N-terminal sequencing, so that normally we have to use multiple spots collected from a series of replicate 2-D gels.
An alternative approach to sequence determination of peptides is to use mass spectrometry. This can be effected either by peptide fragmentation within the spectrometer or by a technique known as ladder sequencing. In the latter method, we use either chemical (Edman degradation) or enzymic (aminopeptidase, carboxypeptidase) degradation under limiting conditions to generate a series of ! truncated peptides that differ in size according to the number of amino acids that have been removed from their N- or C-terminus. We can then measure the masses of these peptides, usually by Maldi-MS, to deduce the sequence. We require very high mass accuracy to generate unambiguous sequence and cannot distinguish between the amino acids leucine and isoleucine because these have an identical mass.
Although chemical protein sequencing is a sensitive and highly discriminating method of protein identification, sample throughput using an automated sequenator is low, t! ypically one or two samples per day. We therefore need techniques that are capable of high-throughput sensitive screening of proteins separated by 2-DE, so that only those proteins that cannot be identified unequivocally or appear to be novel need further characterisation by protein sequencing.
Current methods for the HPLC analysis of amino acid derivatives are capable of very high (sub-picomole) sensitivity. We can apply this method directly to proteins separated by 2-DE and have found it to be an excellent method for their rapid identification. This approach depends on individual proteins having more or less unique amino acid compositions. We use the experiment! al amino acid composition determined for the protein of interest to interrogate databases of amino acid compositions derived in silico from sequences of known proteins or predicted from translated nucleotide sequences. We have access to a number of search algorithms via the Internet and can filter the search data by including Mr and pI search windows, or species specificity of the target protein.
A major drawback to this approach is that the end-point is a list of protein identities ranked in order of probability, but the 'correct' protein does not necessarily occur as the first ranked entry. While scientists find score patterns a useful way of increasing confidence of identi! ty, we generally prefer to a dopt an orthogonal approach and combine amino acid compositional analysis with another rapid method of protein identification, such as peptide mass profiling. Recently, scientists have developed a method in which Edman degradation is used to create a three or four amino acid N-terminal 'sequence tag', following which the proteins are subjected to amino acid compositional analysis. The combined amino acid composition and 'sequence tag' data are then used for protein identification.
A major breakthrough in rapid protein identification came when we realised that the set of peptide masses obtained by MS analysis of a protein digest provides a characteristic &! nbsp; fingerprint of that protein (Fig 4a). We then use this information to interrogate databases of peptide masses derived from sequences of known proteins or predicted from nucleotide sequences and a number of algorithms have been implemented to facilitate this process. As in the case of amino acid compositional analysis, this technique of peptide mass profiling or fingerprinting is a statistical method with putative protein identities being ranked in order of probability (Fig 4b). We can improve the reliability of this approach by combining peptide mass profiling data from two separate digests � eg with trypsin and Lys-C � or by adopting an orthogonal approach in! combination with amino acid compositional analysis.
; Fig 4. Maldi-MS spectrum of the tryptic digest of the protein spot; the peptide mass database search indicates that the protein is the M-chain of creatine kinase.
Scientists have developed methods for peptide mass profiling based on digests generated by enzymatic cleavage either while the protein spot is in situ within the gel matrix or after transfer by electroblotting to a suitable membrane. After recovery, we can analyse the unfractionated peptides by Maldi-MS (Fig 4). Alternatively, we fractionate the peptide mixture by reverse-phase HPLC. Systematic screening of HPLC fractions can be done either by &! nbsp; Maldi-MS or by electrospray ionisation ESI-MS using a quadropole or ion-trap instrument. We can also couple ESI-MS on-line with the HPLC separation by splitting the column effluent, allowing simultaneous mass measurement and fraction collection. The peptides present in the fractions may be useful for other identification strategies such as protein sequencing and carbohydrate analysis.
Partial peptide sequence data is an extremely powerful adjunct to identifying proteins by peptide mass profiling. While we can generate this sequence data by automated Edman degradation or by MS ladder sequencing, there are also two MS-based approaches tha! t we can use to identify pro teins. These take advantage of the ability of two-stage mass spectrometers, either Maldi-MS with post-source decay (PSD) or ESI-MS/MS triple-quadropole or ion-trap instruments, to induce fragmentation of peptide bonds. The first of these methods, termed peptide sequence tagging, is based on interpreting a portion of the ESI-MS/MS or PSD-Maldi-MS fragmentation data to generate a short partial sequence or 'tag', which is used in combination with the mass of the intact parent peptide ion, and provides a significant amount of additional information for the homology search. An elegant extension of this approach is a nano-electrospray ion source that allows spraying times of more than! 30min from ca 1�l of sample. Using this method, we can sequence multiple peptides from a digest mixture without the need for prior separation by HPLC. Biochemists have found this method to be sensitive in the low femtomole range and have successfully analysed silver stained 2-DE protein spots containing as little as 5ng protein.
The second method is based on the automated interpretation of ESI-MS/MS fragmentation data, which is used to search sequence databases directly. In this method, the program first identifies all those peptides that can be generated from proteins in the sequence database &n! bsp; and whose masses match those of the measured peptide ion. In the second step, the program predicts the fragment ions expected for each of the candidate peptides if they were fragmented under the experimental conditions used. The experimentally determined MS/MS spectrum is then compared with the predicted spectra using cluster analysis algorithms. Each of the comparisons is allocated a score and the highest scoring peptide sequences are reported. The method has been adapted for automated identification of peptide digests analysed by ESI-MS/MS where the ions subjected to fragmentation are automatically selected during the run and the data automatically analysed. Proteins present in mixtures can be readily &nbs! p; identified with a 30-fold difference in molar quantity and sensitivity is at the low femtomole level.
Our final requirement for proteomic technology is that we must be able to store all the data that are generated in databases that we can interrogate effectively in the laboratory and also, where possible, make it available to other scientists worldwide. Our best approach to this at present is to use the World Wide Web (WWW). In order to provide optimal interconnectivity between these 2-D gel protein databases and other databases of related information available via the WWW, it has been suggested that 2-D gel databases are constructed &nbs! p; according to a set of fundamental rules. Databa ses conforming to these rules are said to be 'federated 2-D databases', while many of the other databases conform to at least some of the rules. You can view a list of these 2-D protein databases at World2Dpage on the WWW ( http://www.expasy.ch/ch2d/2d-index.html).
Proteomics interfaces with and complements genomics to provide information on quantitative protein expression in biological systems. Such information will provide new insights into complex cellular processes and improve our understanding of cellular responses to external stimu! li. Proteomics also promises to yield information on the ways in which cells respond to disease processes, leading to an understanding of disease at the molecular level and providing new opportunities for developing diagnostics and therapeutics. For these reasons, proteomics is generating considerable excitement in the biotechnology and pharmaceutical industry. Moreover, scientists in this sector believe that proteomics can also be used to validate new therapeutic agents by providing information about their effects on protein expression, thereby accelerating the process of drug discovery.
Michael Dunn is a reader in biochemistry in the department of cardiothoracic surgery, ! National Heart and Lung Insti tute, Imperial College School of Medicine, Heart Science Centre, Harefield Hospital, Harefield, Middlesex UB9 6JH.
A. G�rg, G. Boguth, C. Obermaier, A. Posch, W. Weiss, Electrophoresis, 1995, 16, 1079. I. Humphery-Smith, S. J. Cordwell, W. P. Blackstock, Electrophoresis, 1997, 18, 1217. A. I. Lamond, M. Mann, Trends Cell Biol., 1997, 7, 139. S. D. Patterson, R. Aebersold, Electrophoresis, 1995, 16, 1791. S. R. Pennington, M. R. Wilkins, D. F. Hochstrasser, M. J. Dunn, Trends Cell Biol., &! nbsp; 1997, 7, 168. M. R. Wilkins, K. L. Williams, R. D. Appel, D. F. Hochstrasser (eds), Proteome research: new frontiers in functional genomics. Berlin: Springer, 1997. J. R. Yates, J. Mass Spectrom., 1998, 33, 1.
Densitometer an instrument that measures the optical transmission or reflecting properties of a material, particularly the optical density (OD) of exposed and processed photographic images. Eukaryotes organisms � eg anima! ls, plants, fungi, whose cells have a membrane-bound nucleus and organelles. Genome All of the DNA sequences in an organism. Mycoplasma the smallest free-living organism known. N-terminal sequencing determination of the sequence of a protein using Edman degradation, resulting in the sequential release of amino acids from the N-terminus of the protein. Open reading frame A DNA sequence with the potential to encode a protein. Prokaryotes single-! celled organisms � eg bacteria � that have no defined nucleus and whose genetic material is usually a circular duplex of DNA. Reverse phase HPLC separation of peptides by differences in polarity of amino acid side chains. Reversed phase systems are so-called because the mobile phase (usually a mixture of water and an organic modifier) is less polar than the silica-based stationary phase. Transcription process by which RNA polymerase enzyme produces complementary single-stranded messenger RNA. Translation &nbs! p; process occurring in cell organelles called ribosomes, to decipher the code in mRNA in order to synthesise a specific polypeptide. Western blotting the transfer, usually by electrophoresis, of separated proteins from a gel onto the surface of a thin support such as nitrocellulose or polyvinylidene difluoride. The immobilised proteins can readily interact with antibodies � a process termed immunoprobing.
03/13/06 03/14/06 03/15/06 03/17/06 03/19/06 03/20/06 03/21/06 03/25/06 03/26/06 03/28/06 03/29/06 03/30/06 03/31/06 04/01/06 04/05/06 05/04/06 05/07/06 05/08/06 05/10/06 05/11/06 05/12/06 05/13/06 05/15/06 05/16/06 05/17/06 05/19/06 05/20/06 05/23/06 05/24/06 05/26/06 05/27/06 05/29/06 05/30/06 05/31/06 06/03/06 06/04/06 06/05/06 06/07/06 06/08/06 06/09/06 06/10/06 06/12/06 06/14/06 06/17/06