The study of proteins and their function is key to understanding the intrinsic properties of the cell both in normal and disease states. An important part of this analysis involves the prediction and elucidation of three-dimensional protein structures, and the interactions they undertake, through the use of computational techniques. The majority of work in this thesis focuses on the effects that non-specific charge interactions have on these structures. Firstly, sets of proteins that select a single partner from closely related alternatives were analysed using an empirical binding model in order to identify a determinant for binding specificity. Here, we predicted that charge interactions are more favourable in the majority of cognate pairs, compared to other energetic and geometric properties. In addition to this, charged protein side-chains were found to be important with respect to phosphorylation sites that lie in disordered regions of proteins. The analysis of charge environments around these sites indicated a propensity for a subset of residues to be phosphorylated when surrounded by charged residues. This was especially true for proteins involved in RNA processing.An investigation of protein-mRNA interactions also identified a role for charge interactions that occur within translational control mechanisms. The correlation seen between positively charged disordered regions of specific regulatory proteins and the secondary structure of target mRNAs revealed a potential control mechanism that is partially influenced by polyelectrostatic interactions.

Quase 100 anos após a descoberta do Trypanosoma cruzi, agente etiológico da doença de Chagas, não existem tratamentos efetivos para esta doença tanto em sua fase aguda quanto crônica. Neste trabalho é apresentado um estudo em grande escala das proteínas preditas a partir de seqüências genômicas de T. cruzi, através de técnicas de bioinformática e modelagem comparativa, visando investigar e discutir os seguintes aspectos: (i) quais seqüências podem ter suas estruturas tridimensionais (3D) geradas e qualificadas por técnicas de modelagem comparativa, usando o workflow MHOLline; (ii) associar a estes modelos protéicos uma classificação enzimática (EC), através da ferramenta ECNGet (desenvolvida neste trabalho); (iii) através desta associação enzimática aos modelos, usá-los para inferir possíveis funções biológicas às suas seqüências alvos originalmente anotadas como hipotéticas, e/ou sugerir uma possível reanotação de seqüências; (iv) investigar dentre as enzimas com estruturas 3D geradas quais possuem similaridades (ortologias) com enzimas humanas (genoma de Homo sapiens); (v) determinar quais das enzimas identificadas fazem parte de alguma via metabólica representativa depositada em bancos de dados específicos. Dentre as enzimas de T. cruzi consideradas como alvos moleculares para o tratamento da doença de Chagas, as cisteíno proteases têm sido extensivamente estudadas experimentalmente. Neste presente estudo, as isoformas 1 e 2 da cruzipaína foram investigadas por simulações de dinâmica molecular (DM), nas temperaturas de 25C e 37C, tendo como controle a papaína, enzima representativa da família C1 de cisteíno proteases. Analisando 20.679 seqüências protéicas preditas a partir de CDS (Seqüência Codificante) não redundantes do genoma de T. cruzi (cepa CL Brener), 4.719 seqüências foram associadas a proteínas de membrana e foram gerados 2.786 modelos estruturais protéicos. Destes modelos, 1.888 foram associados a um EC, dos quais aproximadamente 79,5% foram classificados como tendo uma qualidade suficientemente boa para serem usados em estudos de desenho racional de fármacos baseado em estrutura, através de métodos de modelagem molecular. Dos 1.876 modelos enzimáticos com, pelo menos, uma sub-subclasse enzimática, foram identificadas 99 seqüências possivelmente análogas a enzimas humanas, 1.776 como similares e 1 bifuncional como análoga/similar. Também foi possível detectar um subgrupo de 10 modelos considerados como prováveis enzimas específicas de T. cruzi em relação a H. sapiens, não descartando possíveis relações destas seqüências com outros organismos. Nos estudos de DM realizados, o principal resultado obtido indica que a presença de resíduos ácidos na posição 158 (numeração da papaína) no sítio catalítico, pode induzir uma reorganização estrutural, suscetível a variações de temperatura, dos resíduos catalíticos em cisteíno proteases da família da papaína. Esta reorganização estrutural gera uma conformação semelhante à apresentada pela tríade catalítica (ASP/HIS/SER) em serino proteases.

At present, the most accurate approach to predicting protein structure, comparative modelling, builds a model of a target sequence using known protein structures as templates. Comparative modelling becomes markedly less accurate in the ‘twilight zone’, where the target protein shares little sequence identity with all known templates. There are two main causes of this inaccuracy: first, it becomes difficult to identify good structural templates; second, it becomes difficult to determine which amino acids in the template are structurally equivalent to those in the target. These are problems of fold recognition and target-template alignment respectively. In this thesis, new approaches are developed to address both these problems. The alignment problem is investigated in the special case of membrane proteins. These are key modelling targets as they resist structure determination and are pharmaceutically important. The approach taken here is to use ‘environment specific substitution tables’ (ESSTs)– that is, to alter the alignment scoring system for each local environment of the template structure. We show how ESSTs can be made for membrane proteins, tested for robustness of construction, and used to infer the most important evolutionary pressures acting on protein structure. The incorporation of ESSTs into a multiple sequence alignment method leads to more accurate alignments of membrane proteins, and so to more accurate models. Recently, algorithms have been developed that predict contacts in protein structures from a multiple sequence alignment of homologous sequences. We explore the potential of these predictions for fold recognition by developing an algorithm that makes no use of amino acid identity, and so should be agnostic to the existence of a ‘twilight zone’. We show that whilst this is not the case, our method is complementary to state-of-the-art approaches.

Made available in DSpace on 2015-03-04T18:50:54Z (GMT). No. of bitstreams: 1 dissertacao Priscila Capriles.pdf: 11456267 bytes, checksum: 5add21ce6ff9ebf894b553efdf4c5125 (MD5) Previous issue date: 2007-03-15 / Fundação Carlos Chagas Filho de Amparo a Pesquisa do Estado do Rio de Janeiro / Quase 100 anos após a descoberta do Trypanosoma cruzi, agente etiológico da doença de Chagas, não existem tratamentos efetivos para esta doença tanto em sua fase aguda quanto crônica. Neste trabalho é apresentado um estudo em grande escala das proteínas preditas a partir de seqüências genômicas de T. cruzi, através de técnicas de bioinformática e modelagem comparativa, visando investigar e discutir os seguintes aspectos: (i) quais seqüências podem ter suas estruturas tridimensionais (3D) geradas e qualificadas por técnicas de modelagem comparativa, usando o workflow MHOLline; (ii) associar a estes modelos protéicos uma classificação enzimática (EC), através da ferramenta ECNGet (desenvolvida neste trabalho); (iii) através desta associação enzimática aos modelos, usá-los para inferir possíveis funções biológicas às suas seqüências alvos originalmente anotadas como hipotéticas, e/ou sugerir uma possível reanotação de seqüências; (iv) investigar dentre as enzimas com estruturas 3D geradas quais possuem similaridades (ortologias) com enzimas humanas (genoma de Homo sapiens); (v) determinar quais das enzimas identificadas fazem parte de alguma via metabólica representativa depositada em bancos de dados específicos. Dentre as enzimas de T. cruzi consideradas como alvos moleculares para o tratamento da doença de Chagas, as cisteíno proteases têm sido extensivamente estudadas experimentalmente. Neste presente estudo, as isoformas 1 e 2 da cruzipaína foram investigadas por simulações de dinâmica molecular (DM), nas temperaturas de 25°C e 37°C, tendo como controle a papaína, enzima representativa da família C1 de cisteíno proteases. Analisando 20.679 seqüências protéicas preditas a partir de CDS (Seqüência Codificante) não redundantes do genoma de T. cruzi (cepa CL Brener), 4.719 seqüências foram associadas a proteínas de membrana e foram gerados 2.786 modelos estruturais protéicos. Destes modelos, 1.888 foram associados a um EC, dos quais aproximadamente 79,5% foram classificados como tendo uma qualidade suficientemente boa para serem usados em estudos de desenho racional de fármacos baseado em estrutura, através de métodos de modelagem molecular. Dos 1.876 modelos enzimáticos com, pelo menos, uma sub-subclasse enzimática, foram identificadas 99 seqüências possivelmente análogas a enzimas humanas, 1.776 como similares e 1 bifuncional como análoga/similar. Também foi possível detectar um subgrupo de 10 modelos considerados como prováveis enzimas específicas de T. cruzi em relação a H. sapiens, não descartando possíveis relações destas seqüências com outros organismos. Nos estudos de DM realizados, o principal resultado obtido indica que a presença de resíduos ácidos na posição 158 (numeração da papaína) no sítio catalítico, pode induzir uma reorganização estrutural, suscetível a variações de temperatura, dos resíduos catalíticos em cisteíno proteases da família da papaína. Esta reorganização estrutural gera uma conformação semelhante à apresentada pela tríade catalítica (ASP/HIS/SER) em serino proteases.

Threading is a protein structure prediction method that uses a library of template protein structures in the following steps: first the target sequence is matched to the template library and the best template structure is selected, secondly the predicted target structure of the target sequence is modeled by this selected template structure. The deceleration of new folds which are added to the protein data bank promises completion of the template structure library. This thesis uses a new set of template-specific weights to improve the energy function for sequence-to-structure alignment in the template selection step of the threading process. The weights are estimated using least squares methods with the quality of the modelling step in the threading process as the label. These new weights show an average 12.74% improvement in estimating the label. Further family analysis show a correlation between the performance of the new weights to the number of seeds in pFam.

Thiostrepton (TS: TS; C72H85N19O18S5) is a thiazoline antibiotic that is effective against Gram-positive bacteria and the malarial parasite, Plasmodium falciparum. Tight binding of TS to the bacterial L11-23S ribosomal RNA (rRNA) complex of the large 50S ribosomal unit inhibits protein biosynthesis. The TS producing organism, Streptomyces azureus, biosynthesizes thiostrepton-resistance methyltransferase (TSR), an enzyme that uses S-adenosyl-L-methionine (AdoMet) as a methyl donor, to modify the TS target site. Methylation of A1067 (Escherichia coli ribosome numbering) by TSR circumvents TS binding. The S. azureus tsr gene was overexpressed in E. coli and the protein purified for biochemical characterization. Although the recombinant protein was produced in a soluble form, its tendency to aggregate made handling a challenge during the initial stages of establishing a purification protocol. Different purification conditions were screened to generate an isolation protocol that yields milligram quantities of protein with little aggregation and sufficient purity for crystallographic studies. Enzymological characterization of TSR was carried out using an assay to monitor AdoMet-dependent ([methyl-3H]-AdoMet) methylation of the rRNA substrate by liquid scintillation counting. During the optimization of assay, it was found that, although this method is frequently employed, it is very time and labour intensive. A scintillation proximity assay was investigated to evaluate whether it could be a method for collecting kinetic data, and was found that further optimization is required. Comparative sequence analysis of TSR has shown it to be a member of the novel Class IV SpoUT family of AdoMet-dependent MTases. Members of this class possess a non-canonical AdoMet binding site containing a deep trefoil knot. Selected SpoUT family proteins were used as templates to develop a TSR homology model for monomeric and dimeric forms. Validation of the homology models was performed with structural validation servers and the model was then used as the basis of ongoing mutagenesis experiments. The X-ray crystal structure of TSR bound with AdoMet (2.45 Å) was elucidated by our collaborators, Drs. Mark Dunstan and Graeme Conn (University of Manchester). This structure confirms TSR MTase’s membership in the SpoUT MTase family with a deep trefoil knot in the catalytic domain. The AdoMet bound in the crystal structure is in an extended conformation not previously observed in SpoUT MTases. RNA docking simulations revealed some features that may be relevant to binding and recognition of TSR to the L11 binding domain of the RNA substrate. Two structure-activity studies were conducted to investigate the TS-rRNA interaction and TS solubility. Computational analyses of TS conformations, molecular orbitals and dynamics provided insight into the possible modes of TS binding to rRNA. Single-site modification of TS was attempted, targeting the dehydroalanine and dehydrobutyrine residues of the antibiotic. These moieties were modified using the polar thiol, 2-mercaptoethanesulfonic acid (2-MESNA). Similar modifications had been previously used to improve solubility and bioavailability of antibiotics. The resulting analogue was structurally characterized (NMR and mass spectrometry) and showed antimicrobial activity against Bacillus subtilis and Staphylococcus aureus.

Threading is a protein structure prediction method that uses a library of template protein structures in the following steps: first the target sequence is matched to the template library and the best template structure is selected, secondly the predicted target structure of the target sequence is modeled by this selected template structure. The deceleration of new folds which are added to the protein data bank promises completion of the template structure library. This thesis uses a new set of template-specific weights to improve the energy function for sequence-to-structure alignment in the template selection step of the threading process. The weights are estimated using least squares methods with the quality of the modelling step in the threading process as the label. These new weights show an average 12.74% improvement in estimating the label. Further family analysis show a correlation between the performance of the new weights to the number of seeds in pFam.

Thiostrepton (TS: TS; C72H85N19O18S5) is a thiazoline antibiotic that is effective against Gram-positive bacteria and the malarial parasite, Plasmodium falciparum. Tight binding of TS to the bacterial L11-23S ribosomal RNA (rRNA) complex of the large 50S ribosomal unit inhibits protein biosynthesis. The TS producing organism, Streptomyces azureus, biosynthesizes thiostrepton-resistance methyltransferase (TSR), an enzyme that uses S-adenosyl-L-methionine (AdoMet) as a methyl donor, to modify the TS target site. Methylation of A1067 (Escherichia coli ribosome numbering) by TSR circumvents TS binding. The S. azureus tsr gene was overexpressed in E. coli and the protein purified for biochemical characterization. Although the recombinant protein was produced in a soluble form, its tendency to aggregate made handling a challenge during the initial stages of establishing a purification protocol. Different purification conditions were screened to generate an isolation protocol that yields milligram quantities of protein with little aggregation and sufficient purity for crystallographic studies. Enzymological characterization of TSR was carried out using an assay to monitor AdoMet-dependent ([methyl-3H]-AdoMet) methylation of the rRNA substrate by liquid scintillation counting. During the optimization of assay, it was found that, although this method is frequently employed, it is very time and labour intensive. A scintillation proximity assay was investigated to evaluate whether it could be a method for collecting kinetic data, and was found that further optimization is required. Comparative sequence analysis of TSR has shown it to be a member of the novel Class IV SpoUT family of AdoMet-dependent MTases. Members of this class possess a non-canonical AdoMet binding site containing a deep trefoil knot. Selected SpoUT family proteins were used as templates to develop a TSR homology model for monomeric and dimeric forms. Validation of the homology models was performed with structural validation servers and the model was then used as the basis of ongoing mutagenesis experiments. The X-ray crystal structure of TSR bound with AdoMet (2.45 Å) was elucidated by our collaborators, Drs. Mark Dunstan and Graeme Conn (University of Manchester). This structure confirms TSR MTase’s membership in the SpoUT MTase family with a deep trefoil knot in the catalytic domain. The AdoMet bound in the crystal structure is in an extended conformation not previously observed in SpoUT MTases. RNA docking simulations revealed some features that may be relevant to binding and recognition of TSR to the L11 binding domain of the RNA substrate. Two structure-activity studies were conducted to investigate the TS-rRNA interaction and TS solubility. Computational analyses of TS conformations, molecular orbitals and dynamics provided insight into the possible modes of TS binding to rRNA. Single-site modification of TS was attempted, targeting the dehydroalanine and dehydrobutyrine residues of the antibiotic. These moieties were modified using the polar thiol, 2-mercaptoethanesulfonic acid (2-MESNA). Similar modifications had been previously used to improve solubility and bioavailability of antibiotics. The resulting analogue was structurally characterized (NMR and mass spectrometry) and showed antimicrobial activity against Bacillus subtilis and Staphylococcus aureus.