 |
Online ISSN: 1097-0134 Print ISSN: 0887-3585 Copyright © 2003 Wiley-Liss, Inc. |
|
| Strategies for structural proteomics of prokaryotes: Quantifying the advantages of studying orthologous proteins and of using both NMR and X-ray crystallography approaches |
| Alexei Savchenko 1, Adelinda Yee 1, Anna Khachatryan 1, Tatiana Skarina 1, Elena Evdokimova 1, Marina Pavlova 1, Anthony Semesi 1, Julian Northey 1, Steven Beasley 1, Ning Lan 2, Rajdeep Das 2, Mark Gerstein 2, Cheryl H. Arrowmith 1 4, Aled M. Edwards 1 3 4 * |
| 1Ontario Center for Structural Proteomics, University Health Network, Toronto, Ontario, Canada 2Molecular Biophysics and Biochemistry Department, Yale University, New Haven, Connecticut 3Banting and Best Department of Medical Research, Toronto, Ontario, Canada 4Ontario Cancer Institute and Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada |
| email: Aled M. Edwards (aled.edwards@utoronto.ca) |
*Correspondence to Aled M. Edwards, Banting and Best Department of Medical Research, 112 College St., Toronto, Ontario, Canada
| Keywords |
| protein expression • structural proteomics • nuclear magnetic resonance spectroscopy • X-ray crystallography |
| Abstract |
 |  | | | | |
| Only about half of non-membrane-bound proteins encoded by either bacterial or archaeal genomes are soluble when expressed in Escherichia coli (Yee et al., [Proc Natl Acad Sci USA 2002;99:1825-1830]; Christendat et al., [Prog Biophys Mol Biol 200;73:339-345)]. This property limits genome-scale functional and structural proteomics studies, which depend on having a recombinant, soluble version of each protein. An emerging strategy to increase the probability of deriving a soluble derivative of a protein is to study different sequence homologues of the same protein, including representatives from thermophilic organisms, based on the assumption that the stability of these proteins will facilitate structural analysis. To estimate the relative merits of this strategy, we compared the recombinant expression, solubility, and suitability for structural analysis by NMR and/or X-ray crystallography for 68 pairs of homologous proteins from E. coli and Thermotoga maritima. A sample suitable for structural studies was obtained for 62 of the 68 pairs of homologs under standardized growth and purification procedures. Fourteen (eight E. coli and six T. maritima proteins) samples generated NMR spectra of a quality suitable for structure determination and 30 (14 E. coli and 16 T. maritima proteins) samples formed crystals. Only three (one E. coli and two T. maritima proteins) samples both crystallized and had excellent NMR properties. The conclusions from this work are: (1) The inclusion of even a single ortholog of a target protein increases the number of samples for structural studies almost twofold; (2) there was no clear advantage to the use of thermophilic proteins to generate samples for structural studies; and (3) for the small proteins analyzed here, the use of both NMR and crystallography approaches almost doubled the number of samples for structural studies. Proteins 2003;50:392-399. © 2003 Wiley-Liss, Inc. |
| Digital Object Identifier (DOI) |
| Article Text |
 | | | | | |
The field of proteomics seeks to determine the biochemical and cellular functions and structure of proteins on a genome-wide scale. Proteomics includes, among other things, studies of protein-protein interactions,[3-6] protein structure,[7-9] protein abundance,[10][11] post-translational modification,[12-14] and protein fossils.[15] The biochemical analysis of purified proteins is referred to as functional proteomics. Structural proteomics aims to derive the three-dimensional structures for all proteins. The consensus strategy for structural proteomics is to determine the experimental structures for enough proteins such that remaining structures can be predicted accurately using computational approaches.[16]
One of the most significant challenges facing experimentally based functional or structural proteomics is the reality that a large proportion of proteins are insoluble when expressed in heterologous systems or when purified and concentrated to the levels required for structural techniques.[1][17] Although there continue to be improvements in the techniques for generating soluble versions of an individual protein, such as variation of fusion tags,[18] genetic screens,[19] or using computational methods[17] to engineer soluble variants, these strategies are time-consuming and the success rate remains poor.
An alternative approach, and one that is afforded by the wealth of genome sequence information, is to express and purify a series of orthologs from various species for a given protein of interest. The underlying hypothesis is that subtle differences in the surface properties and/or stability of a protein, which might arise from the sequence variation among orthologs, may contribute to altered solubility and perhaps more suitable properties for NMR or X-ray structure determination. There is anecdotal evidence that this strategy may improve the probability of crystallization or NMR structure determination for individual proteins, but the extent to which this strategy improves the structure determination process remains unclear.
There is also a prevailing belief that orthologous proteins from thermophilic organisms are more amenable to structural biology methods, presumably because they are predicted to have fewer disordered regions and a higher proportion of salt bridges on the surface.[20][21] As such, thermophilic proteins may be more stable, may crystallize more readily, and may also be more amenable to NMR analysis.
In this study, 68 different pairs of homologous bacterial proteins from Escherichia coli and Thermotoga maritima, respectively, were cloned into bacterial expression vectors for studies of expression, purification and crystallization, and NMR. The results of this study support the notion that the use of a series of sequence homologues will increase the probability of obtaining a structural sample for at least one member of a given protein family. However, the data do not demonstrate a clear advantage for the use of proteins from a thermophilic organism. The results also highlight the complementarity of NMR and X-ray crystallography approaches.
| | | | | |
An all-versus-all implementation of BLAST was used to identify pairs of orthologous proteins in E. coli (EC) and T. maritima (TM). Candidate protein sequences were scanned for homologues in the Protein Database using liberal thresholds of E < 0.01 and ID 
30% or higher. ORFs that had significant BLAST hits to any other protein in either organism within these thresholds were excluded from the candidate list. Proteins with transmembrane regions were identified using a hidden Markov model algorithm and excluded as targets. Proteins harboring secretory signal sequences were identified using a neural network approach and also excluded.[22] The remaining sequences from the E. coli and T. maritima genomes were then compared for homology against each other. In this case, BLAST hits (E < 0.0001 and ID 20% and higher) were included in the candidate list, resulting in a list of pairs of proteins, one from each genome. We further selected protein pairs that had similar lengths and functional annotations and, therefore, are likely to have similar structures and functions. This selection produced 68 pairs of E. coli and T. maritima orthologous proteins (Table I) that were used for further studies. All but two of the 136 proteins were smaller than 320 residues.
 Paired orthologs are presented with E. coli results above T. maritima results, sorted by number of amino acids of E. coli orthologs. Annotation and short descriptions of the ortholog pairs are based on annotations available in NCBI database. Expression and solubility levels, as well as similarity between orthologs, were determined as described in Materials and Methods. Parentheses indicate results obtained at 30°C. The first 23 ortholog pairs were tested for structural studies using both crystal and NMR screening techniques. The final 45 pairs, with greater than 180 amino acids, were tested using crystal screens only. The results of the crystal screening: no, protein was not screened for crystallization due to precipitation or degradation during the purification procedure (see Results and Discussion); tr, protein was screened for crystallization, but no crystallization was detected; cl, initial crystallization conditions were determined. The HSQC results are annotated as pr for  poor and gd for good HSQC profiles. |
Target genes were amplified from genomic DNA using primers designed to create NdeI and BamHI sites upstream from the initiation and stop codons, respectively. In cases where genes contained either of these restriction sites, they were replaced by AceI for NdeI and BglII for BamHI. The PCR of the target samples was performed in a 96-well format using Pfx Polymerase (Invitrogen, La Jolla, CA). The amplification products were cloned as previously described.[23]
PCR reactions were optimized for each set of ortholog genes based on the results for an initial set of 10 genes. The open reading frames were cloned into a modified pET15b T7 RNA polymerase-based expression vector (Invitrogen) that provided an N-terminal hexahistidine fusion (Fig. 1). The sequence encoding the thrombin cleavage site (LVPR
GS) in the pET15b cloning vector was replaced by sequence encoding the cleavage site (ENLYFQG) for the TEV protease for two reasons. First, the TEV protease is available in recombinant form (Invitrogen), and second, it is available in a histidine-tagged version (Science Reagents), which facilitates the removal of the enzyme (see Zhang et al.[23]). The 3
end of the coding region was also modified. Two consecutive ochre stop codons (TAATAA) were introduced immediately downstream from the 3 BamHI restriction site, and the stop codon was omitted from the PCR-amplified coding region. This strategy provides an advantage in that the same PCR fragment can be cloned, if necessary, in a different expression vector that appends a C-terminal hexahistidine fusion. The disadvantage of using the single PCR fragment is that two additional amino acids are added onto the C-terminus of the N-terminally-tagged recombinant proteins because of the addition of a BamHI site in the coding region of the 3 primer (Fig. 1).
Figure 1. Cloning site of the modified pET15b vector starting from the translation start point for the N-terminal tag fusion protein. DNA sequence coding for the NdeI and BamHI restriction sites are underlined. Amino acid sequence recognized by TEV protease is in bold and underlined.  , the point of cleavage.[Normal View 9K | Magnified View 23K] |
Protein Expression, Solubility, and Purification
Clones were transformed into E. coli BL21-Gold (DE3) (Stratagene, La Jolla, CA), which harbor an extra plasmid (pMgk) encoding three rare tRNAs (AGG and AGA for Arg, ATA for Ile).[24] Two to three colonies of each clone were grown in a 24-well format at 37°C in 3 ml Luria Broth supplemented with kanamycin and ampicillin (0.1 mg/ml each). The cultures were grown (37°C, 220 rpm) until an OD600 0.6 was reached. Protein expression was induced by the addition of 0.4 mM IPTG followed by overnight growth at 15°C or 30°C. Two 300-
l aliquots of the culture were transferred to separate 96-well plates and centrifuged to obtain cell pellets (20 min at 3,000 rpm; Beckman Coulter Allegra 6R centrifuge). The cell pellets of one plate were resuspended in denaturing buffer and kept as the whole cell fraction. The fractions of the other plate were flash-frozen in liquid N2 and the soluble protein was extracted by the addition of 100 l of BugBuster (Novagen, Madison, WI) followed by centrifugation (20 min at 3,000 rpm). The resulting supernatant, representing the soluble protein fraction, was compared against the whole cell fraction by denaturing gel electrophoresis in order to determine the size and the relative expression level of each protein.
Large-scale expression and purification was performed as described in Zhang et al.[23] for the proteins destined for crystallization samples and as described in Yee et al.[1] for the proteins destined for NMR spectroscopy.
The primary crystallization conditions were determined using a sparse crystallization matrix (Hampton Research kits) at room temperature using the sitting drop vapor diffusion technique in 96-well plates (Hampton Research). The protein samples were set up at a concentration of up to 10 mg/ml by mixing 2 l of sample with 2 l of the reservoir solution and equilibrated with 100 l of the reservoir solution. Crystals selected for native and MAD data collection were flash-frozen in the crystallization buffer plus an empirically determined cryoprotectant. The diffraction data were collected at the Advanced Photon Source at Argonne National Laboratories (Argonne, Il).
Suitability of proteins for NMR analysis was evaluated as described by Yee et al.[1] Briefly, we employed a rapid batch purification of polyhistidine-tagged 15N-labeled protein followed by a rapid 
screening
of labeled proteins by 1H-15N heteronuclear correlation (HSQC) spectroscopy. The HSQC spectra were classified as good or poor. The good spectra showed dispersion of peaks with roughly equal intensity and in the number expected from the sequence of the protein. These spectra indicated that the protein was readily amenable to structure determination by NMR methods.
| | | | | |
All recombinant clones were tested for expression and solubility in E. coli. The levels of expression and solubility of the recombinant products were analyzed by comparing the total cell and soluble protein fractions obtained after small-scale (3 ml) growths using a standard expression protocol. Protein expression was induced overnight at 15°C in cells harboring a plasmid encoding three rare tRNAs. These generic expression conditions yielded the maximum expression of the greatest number of samples for structural studies in a 400-protein expression study of proteins from Methanobacterium thermoautotrophicum.[17] The levels of over-expression were graded from 0 (no detectable expression) to 5 (dominant protein in extract). The levels of solubility were also graded from 0 (completely insoluble) to 5 (completely soluble) (Table I).
Three EC genes and eight TM genes were not expressed under the standard expression conditions. Two more EC genes and fourteen TM genes were expressed, but in insoluble form. For TM, 12 of the 38 (32%) proteins under 200 residues and 10 of the 30 (33%) of the TM proteins over 200 residues were not expressed or expressed in insoluble form. For EC, two of the 32 (6%) proteins under 200 residues and three of 36 (8%) of the EC proteins over 200 residues were insoluble. From these analyses, EC proteins appear to be more likely to be expressed in soluble form compared with the orthologous proteins from TM.
We were concerned that the larger percentage of insoluble TM proteins might have been due to the relatively low temperature of induction (15°C), given that the optimal temperature for this organism is 80°C. The insoluble TM and EC proteins were, therefore, tested for induction at a higher (30°C) temperature (Table I). Four of 22 TM proteins and 1 of 5 EC proteins showed significantly higher expression and solubility levels when expressed at higher temperatures, and increasing the temperature of induction to 37°C did not further improve the expression and solubility of the proteins (data not shown). Although some improvement of solubility was seen after induction at 30°C, we conclude that the inability to express these specific TM proteins in soluble form is not due to the temperature of growth or induction. The recombinant proteins derived from EC genes were more highly expressed and more soluble than the corresponding proteins from TM, perhaps because the EC proteins were expressed in a homologous expression system.
In no instances were both orthologs insoluble (Table I, Fig. 2) showing that the addition of an ortholog does increase the probability of generating a soluble protein.
| Figure 2. Distribution of the ortholog clones according to the results of the test for expression and solubility in small scale. The recombinant proteins are classified as not expressed (I), expressed and soluble (II), and expressed, but insoluble (III). Each protein is represented as a number corresponding to the number of its ortholog pair in Table I. The clones that demonstrated increased solubility after over-expression at higher temperature (30°C) are shown in italic. The protein samples that were soluble during small-scale experiments but nevertheless did not generate samples suitable for structural studies due to precipitation or degradation are underlined. The orthologs from two genomes, for which similar results have been obtained, are marked in red. [Normal View 38K | Magnified View 128K] |
| Figure 3. A: Distribution of the subset of soluble orthologous proteins according to results of screening for NMR samples by HSQC. B: Distribution of soluble orthologous proteins according to results of screening for crystallization conditions. The proteins are clustered by providing good HSQC and obtained initial crystallization conditions (I), or by providing poor HSQC and no crystallization in initial screens (II). Each protein is represented as a number corresponding to the number of its ortholog pair in Table I. The orthologs representing the same pair, for which similar results have been obtained, are marked in red. [Normal View 46K | Magnified View 158K] |
The 64 soluble EC and 50 soluble TM proteins were grown on a larger scale for purification for structural studies. Of the 64 EC proteins, 53 (83%) could be purified in a form suitable for structural analysis, namely in sufficient yield (>2 mg/L of culture) and with no obvious precipitation at higher concentration (>2 mg/ml). Eleven EC proteins were either degraded or precipitated during the purification procedure. Of the 50 TM proteins, 46 (92%) could be purified and concentrated for structural studies. Thus, there did not appear to be a significant difference between the genomes in the proportion of soluble proteins that could be purified and concentrated for structural studies.
A total of 38 of the 68 pairs of orthologs could be purified from both TM and EC sources in a form suitable for structural studies, and six could not be purified from either source (Table I, Fig. 2). A total of 17 proteins could be purified and concentrated only from EC clones and 8 could be purified and concentrated only from TM clones. Thus, we observed a 25-50% increase in the number of soluble, concentratable samples by the addition of one ortholog of the protein of interest.
We did not observe a correlation between whether a protein could be purified from the EC gene and whether it could be purified and concentrated from the TM gene. In EC, 53 of 68 proteins (78%) could be purified and concentrated for structural studies. In TM, the corresponding number is 46 of 68 (68%). If the ability of a protein from one organism to be purified and concentrated is not related to that of its ortholog from another organism, then we would expect that the number of proteins soluble and able to be concentrated from both sources would be the product of the two probabilities (0.78 × 0.68 × 68 proteins), or 34 proteins, which is very close to 36, the number we observed. If this fact can be extended to orthologs from multiple organisms, then the probability P of deriving a purified, concentrated sample for at least one out of N protein samples would be:
| 1 |
Interestingly, we did not observe a significant size-dependence on the ability to purify and concentrate a sample. Both smaller (<200 residues) and larger (>200 residues) proteins had similar ratios of proteins that were unable to be purified and concentrated.
One of our aims was to explore the advantages of using different orthologs for deriving samples for structural biology. For soluble ortholog pairs under 180 amino acids (Table I), we assessed the suitability for NMR structure determination by employing a rapid batch purification of polyhistidine-tagged 15N-labeled protein followed by a rapid screening of labeled proteins by 1H-15N heteronuclear correlation (HSQC) spectroscopy. The HSQC spectrum provides a diagnostic fingerprint of a protein, and the quality of the spectra can be used to assess the suitability for NMR structure determination.[1]. Twenty-three pairs of proteins were labeled with 15N and targeted for NMR analysis. In 10 instances, both the EC and the TM protein yielded a poor NMR sample, and in four instances both the EC and TM orthologs had good HSQC spectra [Fig. 3(A)]. In four cases, a suitable NMR sample could be derived only from the EC ortholog and in another two instances one could only be obtained for the TM ortholog.
In this study, the analysis of proteins from either single genome generated eight (EC) or six (TM) NMR samples and the combination of genomes added only two or four samples to the total. This marginal increase in samples for structural studies for this small dataset may suggest that small homologous proteins may be more likely to share similar biophysical properties. This analysis also showed that, at least for this sample set, there was no clear advantage to using proteins from thermophiles for NMR spectroscopy in terms of obtaining a sample with favorable NMR properties at 25°C. However, the thermophilic orthologs may in certain cases provide an advantage for NMR data collection at higher temperatures, which may result in better sensitivity due to improved NMR relaxation properties.
All purified proteins, including those for which NMR spectra were collected, were screened for crystallization under a standard set of conditions. This corresponded to 53 EC proteins and 44 TM proteins [Fig. 3(B)]. Of the 53 EC proteins, 14 (26%) formed crystals. For TM proteins, 16/50, or 32%, crystallized. Thus, although the success rate of achieving crystals from the starting gene set was similar (14/68 for EC and 16/68 for TM), the proportion of purified and concentrated TM proteins that crystallized was higher. These data suggests that purified TM proteins might be easier to crystallize, although more difficult to obtain in soluble form, than their EC orthologs. In total, the combination of the two effects resulted in an equivalent number of samples for structural studies attained per gene cloned. However, one must be careful not to conclude that the effects we observe are solely a result of the thermal stability properties of the TM proteins. The differences observed between EC and TM proteins may simply arise because they represent the differences that might be observed between any two sources of orthologous proteins.
Overlap of NMR and Crystallization
We compared the effectiveness of NMR and crystallization to generate samples for structural studies of 46 of the smaller proteins (23 pairs of orthologs). We were able to generate either initial crystallization conditions or a good NMR spectrum for 24 of 46 proteins, corresponding to at least one member of 15 of 23 pairs. Of 34 proteins for which both crystallization and NMR data could be collected, only 3 proteins both crystallized and had good NMR properties. Crystals were obtained for seven proteins with poor NMR spectra and good NMR spectra could be obtained for ten proteins that failed to crystallize in our particular crystallization trials. It is evident that NMR methods and protein crystallization are complementary rather than redundant if the aim is to determine the structures of small proteins. In this sample set, the use of NMR alone would have generated 13 samples for structural studies. The use of crystallization alone would have generated ten samples with defined initial crystallization conditions. However, the use of both methods increased the number of unique samples for structural studies to 20.
| | | | | |
Strategy for Structural Proteomics of Small Proteins
The most efficient strategy for the structural proteomics of small proteins may favor the use of NMR. NMR has four clear advantages over crystallization approaches. First, after one has an expression clone that produces a soluble protein, it only takes a few days to characterize its suitability for NMR spectroscopy, whereas it could take days to months to grow a protein crystal. Second, the results of the NMR spectroscopy are usually decisive. Excellent samples can be identified immediately; poor samples can immediately be eliminated from the process. With crystal trials, it is very difficult to make an informed decision based on lack of a crystal. Third, the efficient determination of a crystal structure depends on the presence of methionine in the protein sequence. The smaller the protein, the lower the probability that the protein contains a methionine. Finally, in this study, although we found that an equal number of proteins generated excellent NMR spectra as did crystallize, it is not really accurate to equate these two metrics. An excellent NMR spectrum is highly correlated with the ability to determine its solution structure.[1] However, the growth of a crystal, while an important step towards determining a crystal structure and the most convenient and rapid parameter to measure, does not guarantee that a structure can be solved. A large percentage of protein crystals, in our hands over 50% are difficult to optimize, can be difficult to crystallize as a selenomethionine-labeled protein or are not well-enough ordered to diffract X-rays to high resolution. Therefore, we conclude that a coordinated combination of crystallography with NMR spectroscopy should provide the most efficient path to the structure determination of small proteins.
| References |
| | | | | |
| 1 | Yee A, Chang X, Pineda-Lucena A, Wu B, Semesi A, Le B, Ramelot T, Lee GM, Bhattacharyya S, Gutierrez P, Denisov A, Lee CH, Cort JR, Kozlov G, Liao J, Finak G, Chen L, Wishart D, Lee W, McIntosh LP, Gehring K, Kennedy MA, Edwards AM, Arrowsmith CH. An NMR approach to structural proteomics. Proc Natl Acad Sci USA 2002; 99: 1825-1830. |
| 2 | Christendat, D, Yee A, Dharamsi A, Kluger Y, Gerstein M, Arrowsmith CH, Edwards AM. Structural proteomics: prospects for high throughput sample preparation. Prog Biophys Mol Biol , 2000; 73: 339-345. |
| 3 | Ho Y, Gruhler A, Heilbut A, Bader GD, Moore L, Adams SL, Millar A, Taylor P, Bennett K, Boutilier K, Yang L, Wolting C, Donaldson I, Schandorff S, Shewnarane J, Vo M, Taggart J, Goudreault M, Muskat B, Alfarano C, Dewar D, Lin Z, Michalickova K, Willems AR, Sassi H, Nielsen PA, Rasmussen KJ, Andersen JR, Johansen LE, Hansen LH, Jespersen H, Podtelejnikov A, Nielsen E, Crawford J, Poulsen V, Sorensen BD, Matthiesen J, Hendrickson RC, Gleeson F, Pawson T, Moran MF, Durocher D, Mann M, Hogue CW, Figeys D, Tyers M. Systematic identification of protein complexes in Saccharomyces cerevisiae by mass spectrometry. Nature 2002; 415: 180-183. |
| 4 | Grant SG, Husi H. Proteomics of multiprotein complexes: answering fundamental questions in neuroscience. Trends Biotechnol 2001; 19: S49-S54. |
| 5 | Simpson RJ, Dorow DS. Cancer proteomics: from signaling networks to tumor markers. Trends Biotechnol 2001; 19: S40-S48. |
| 6 | Wojcik J, Schachter V. Protein-protein interaction map inference using interacting domain profile pairs. Bioinformatics 2001; 17: S296-S305. |
| 7 | Gerstein M, Lin J, Hegyi H. Protein folds in the worm genome. Pac Symp Biocomput , 2000; 30-41. |
| 8 | Jhoti H. High-throughput structural proteomics using x-rays. Trends Biotechnol 2001; 19: S67-S71. |
| 9 | Qian J, Stenger B, Wilson CA, Lin J, Jansen R, Teichmann SA, Park J, Krebs WG, Yu H, Alexandrov V, Echols N, Gerstein M. PartsList: a web-based system for dynamically ranking protein folds based on disparate attributes, including whole-genome expression and interaction information. Nucleic Acids Res 2001; 29: 1750-1764. |
| 10 | Gygi SP, Rochon Y, Franza BR, Aebersold R. Correlation between protein and mRNA abundance in yeast. Mol Cell Biol 1999; 19: 1720-1730. |
| 11 | Greenbaum L, Gozlan Y, Schwartz D, Katcoff DJ, Malik Z. Nuclear distribution of porphobilinogen deaminase (PBGD) in glioma cells: a regulatory role in cancer transformation? Br J Cancer 2002; 86: 1006-1011. |
| 12 | Regnier FE, Riggs L, Zhang R, Xiong L, Liu P, Chakraborty A, Seeley E, Sioma C, Thompson RA. Comparative proteomics based on stable isotope labeling and affinity selection. J Mass Spectrom 2002; 37: 133-145. |
| 13 | Cohen AM, Rumpel K, Coombs GH, Wastling JM. Characterisation of global protein expression by two-dimensional electrophoresis and mass spectrometry: proteomics of Toxoplasma gondii. Int J Parasitol 2002; 32: 39-51. |
| 14 | Xing T, Ouellet T, Miki BL. Towards genomic and proteomic studies of protein phosphorylation in plant-pathogen interactions. Trends Plant Sci 2002; 7: 224-230. |
| 15 | Harrison PM, Hegyi H, Balasubramanian S, Luscombe NM, Bertone P, Echols N, Johnson T, Gerstein M. Molecular fossils in the human genome: identification and analysis of the pseudogenes in chromosomes 21 and 22. Genome Res 2002; 12: 272-280. |
| 16 | Sali A. Comparative protein modeling by satisfaction of spatial restraints. Mol Med Today 1995; 1: 270-277. |
| 17 | Christendat D, Yee A, Dharamsi A, Kluger Y, Savchenko A, Cort JR, Booth V, Mackereth CD, Saridakis V, Ekiel I, Kozlov G, Maxwell KL, Wu N, McIntosh LP, Gehring K, Kennedy MA, Davidson AR, Pai EF, Gerstein M, Edwards AM, Arrowsmith CH. Structural proteomics of an archaeon. Nat Struct Biol 2000; 7: 903-909. |
| 18 | Hammarstrom, M, Hellgren N, van Den Berg S, Berglund H, Hard T. Rapid screening for improved solubility of small human proteins produced as fusion proteins in Escherichia coli. Protein Sci 2002; 11: 313-321. |
| 19 | Zhang P, Li MZ, Elledge SJ. Towards genetic genome projects: genomic library screening and gene-targeting vector construction in a single step. Nat Genet 2002; 30: 31-39. |
| 20 | Wesson L Eisenberg D. Atomic solvation parameters applied to molecular dynamics of proteins in solution. Prot Sci 1992; 1: 227-235. |
| 21 | Chakravarty S Varadarajan R. Elucidation of factors responsible for enhanced thermal stability of proteins: a structural genomics based study. Biochemistry 2002; 41: 8152-8161. |
| 22 | Nielsen H, Engelbrecht J, Brunak S, von Heijne G. A neural network method for identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites. Int J Neural Syst 1997; 8: 581-599. |
| 23 | Zhang, RG, Kim Y, Skarina T, Beasley S, Laskowski R, Arrowsmith C, Edwards A, Joachimiak A, Savchenko A. Crystal structure of Thermotoga maritima 0065, a member of the IclR transcriptional factor family. J Biol Chem 2002; 277: 19183-19190. |
| 24 | Christendat D, Saridakis V, Dharamsi A, Bochkarev A, Pai EF, Arrowsmith CH, Edwards AM. Crystal structure of dTDP-4-keto-6-deoxy-D-hexulose 3,5-epimerase from Methanobacterium thermoautotrophicum complexed with dTDP. J Biol Chem 2000; 275: 24608-24612. |
