Research in the Unit is focussed on two main themes: the mechanisms by which proteins fold into their functional state, either in a test tube or in a cell, protein engineering applied to a variety of theoretical or biotechnological problems.
Protein folding in vitro (Michel Goldberg)
These studies aim at understanding, at a fundamental level, the mechanisms that enable a protein to acquire, in a few seconds at most, the complex three-dimensional structure, named "native", that endows it with its biological properties. The knowledge thus acquired is used to improve protein folding, in particular in industrial processes related to biotechnologies.
Studies on early folding intermediates (Alain Chaffotte - Michel Goldberg)
We went on, this year, studying the coupling between between the formation of long range interactions (disulfide bonds) and of local structures (the alpha-helices and beta-strands of the secondary structure) during the early folding steps of a model protein, hen lysozyme. Starting from the observation that the lysozyme secondary structure forms very rapidly when the 4 disulfide bonds are preestablished, while it does not form in their absence, we tried to identify those of the disulfide bonds that are essential for the very rapid formation of the secondary structure. For that purpose, we constructed four double mutants in which the two cysteins engaged in an individual disulfide bonds were been replaced by alanines.
The effect of the suppression of one S-S bond in three proteins, two of our double mutants and the partly reduced normal protein have thus far been characterized by a variety of spectroscopic, hydrodynamic and functional crieteria which confirmed the very strong structural similarity between the natural, fully oxidized enzyme (with 4 S-S bonds) and the modified enzymes (with 3 S-S bonds). The various phases of their refolding have been monitored using a rapid mixing device that triggers refolding in less than 4 milliseconds. We thus could observe that suppression of the disulfide bond between residues number 30 and 115, as well as of the disulfide between residues 6 and 127, does not prevent the rapid formation of secondary structure which, like in normal fully oxidized lysozyme, occurs in less than 4 milliseconds. However, suppression of one or the other of these two S-S bonds speeds up the folding of these mutants by preventing the formation of a "trapped", partly folded species that slows down the folding of the natural protein. Suppressing the S-S bond between residues 76 and 94 does not either prevent the burst of "native" secondary structure in less than 4 milliseconds. However, the trapped, partly folded intermediate can still be formed, which slows down the renaturation of this mutant. These observations represent a set of solid experimental evidence supporting the "energy landscape" model recently proposed by theoriticians to explain the rapid folding of polypeptide chains.
b- Production and study of a recombinant protein, candidate vaccine against malaria (Alain Chaffotte)
Shirley Longacre's team (Unit of Biology of Host-Parasite Interactions) has shown that a polypeptide fragment of about 100 aminoacids corresponding to the C-terminal end of a Plasmodium membrane protein was a good candidate vaccine against malaria. These studies were initially performed on P. cinomolgi, weakly pathogenic for humans. The three-dimensional structure of the fragment has been solved in the Unit of Structural Immunology. Transposition of these results to human vaccination requires to produce and charactezrize the equivalent polypeptide fragment isolated from P. falciparum, the main parasite involved in the human disease. Attempts have been made to produce this polypeptide fragment in large amounts and at low cost in recombinant bacteria. However, the protein produced in bacteria can not form the 6 disulfide bonds that are requested for its proper folding and for its immunoligical properties.
In the Unit, preliminary attempts to oxidize the reduced protein have shown that it is difficult to obtain, in vitro, the native protein with the correct disulfide bonds. We have therfefore attempted to express the C-terminal fragment of the P. falciparum protein under conditions where the natural oxidation of disulfide bonds is favoured. Conditions where one of the genetic constructs we prepared is expressed in significant amounts have been found. The oxidation state and the antigenicity of the protein thus produced are currently under investigation.
c- Kinetics of association-dissociation and folding of the oligomeric R67 DHFR (Annick Méjean)
Our studies on the kinetic aspects of the dimer-tetramer equilibrium in R67 DHFR, and more specifically on the mechanisms by which the pH controls this equilibrium, have been completed this year. They confirmed that the triggering event for the dissociation is the protonation of the tetramer. Conversely, the triggering event of the association has been shown to be the deprotonation of the dimeric species. An original model, in which the main pathways used for association and dissociation are different, has been established to quantitatively account for these observations. It includes the properties of mutant DHFRs produced for the modeling studies reported above.
In order to identify the molecular events that initiate the folding of R67 DHFR, we try to find out whether the assembly of two disordered monomers precedes or follows the folding of the monomers. For that purpose, the folding kinetics of R67 DHFR were investigated at pH 5, where it refolds as stable dimers. By observing in a stopped-flow machine the kinetics of regain of the native spectral properties of the protein (circular dichroism in the far UV, fluorescence) three phases were observed. A very fast phase, completed in less than 4 milliseconds. A rapid one, with a half life of about half a second. And a slow one, with a half life of about 30 seconds. The latter was shown to be accelerated by the presence of prolyl-isomerase, and hence corresponds to the cis-trans isomerization of a prolyl residue. We are currently attempting to identify which of these three phases is associated to the dimerization step.
II- Molecular modeling of the energy of association between proteins(Arnaud Blondel)
This group focusses on the experimental and theoretical study of the associations between biological macromolecules. The motivations are the following:
- Develop precise and fast method to model the interaction energies between proteins that would be reliable enough to allow bypassing some intermediate experiments in biological science development. For the design of new drugs in particular, the role of modeling is ever increasing. Therefore, it is an important issue to render the methods for modeling the associations involved in various biological processes more powerful and reliable.
- Reach a fine and correct knowledge of the physics of biomolecules on which modeling methods could be based in order to make them as general as possible. The systematic study of associations between biological macromolecules allows the quantitative testing of this knowledge. This is again an important issue since the post-genomic era is calling for an unprecedented amount of bioinformatics data processing and predictions which will not be satisfied without such a knowledge.
The study of the energy of protein-protein association was chosen because it can at the same time be:
- characterized with direct experimental measures
- modeled with a rigorous theoretical formalism.
This allows a rigorous testing of the theoretical prediction methods. With the same concern about rigour, the group has developed its own experimental tools to ensure that the molecular processes that are to be modeled are:
- mechanistically well defined
- characterized with great precision.
The R67 dihydrolate reductase, which confers antibiotic resistance to certain bacteria and is formed by the association of 4 identical subunits, was therefore chosen for our studies. Mutations were introduced in that protein by genetic construction to probe the importance of various contacts for the association. Association between variants were identified with a combinatorial test developped in the laboratory. The energy of these associations were then characterized with precise physical-chemical methods also developed in the laboratory. Finally, in order to test the structural effect of the mutations, several variants were crystallized in the isolated or associated form and the crystallographic refinement of the atomic structure of these molecules was undertaken.
In parallel, in order to test the theoretical methods, the energies involved in those associations were calculated by modeling. For that purpose, a method to calculate free energy differences with a significant reduction of uncertainties was conceived and developed in the laboratory. This method was extended to take water molecules and long range electrostatics into account (see figure). It was introduced in the academic version of the CHARMM program and implemented in the laboratory to run on unix workstations and massively parallel super-computers.
The results gathered in 2000 complete those of the preceding years:
- At the experimental level, the association mechanism was studied with kinetics and equilibrium measurements. This lead to the development of a mathematical model of all the reactions involved and allowed the determination of the affinity constants of the complexes. The robustness of this mathematical model was tested for all the examples where it was applied. Thus, the effect of 15 types of modifications between 6 types of complexes are known for a well defined association mechanism. Crystals were obtained for a few variants either in the associated or the dissociated form. The preliminary crystallographic refinement showed little structural changes in the associated form as compared with the wild type protein.
- At the modeling level, several series of calculations were performed. They showed that it is essential to take structural relaxation due to mutations into account to make precise calculations. A simulation protocol was developed for that purpose. It was also found with redundant calculations that our method had good convergence properties (~0.4 kcal/mol per calculation). Finally, the agreement with experimental data was good (~0.4 kcal/mol average difference). Testing of the method is pursued on other types of associations.
It is worth pointing out that a series of calculations takes currently about 2 months on a workstation, but that on super-computers of tomorrow ( > 1 Tflops), the same calculations should take 2-3 days, which means that it will be a lot faster than the experimental procedures...
The research of the last few years which were completed this year yielded the following methods and results:
- A combinatorial test of protein-protein interactions.
- A new measurement method for heterodimer associations.
- A set of precise experimental data to test predictions on protein-protein interactions.
- A new approach to calculate free energy with molecular dynamics that significantly reduces calculation uncertainties while taking into account water molecules and long-range electrostatics.
- A simulation protocol to take structural relaxations into account in the free energy calculations.
Two important conclusions can be reached from this year's results:
- It now seems possible to predict, by molecular modeling, the effects of mutations on protein association with a precision comparable to that of experiments.
- Within the limits of the calculations that were performed, the force field and modeling methods used in the CHARMM program give a good representation of the physics of biological macromolecules and thus seem to be applicable to other modeling problems.
III- Control of the folding of bacterial envelope proteins(Jean-Michel Betton - Nathalie Sassoon - Jean-Philippe Arié - Sabine Hunke - Mireille Hervé)
The compartmentalization of Gram-negative bacteria implies the existence of specific mechanisms, monitoring protein folding in the envelope, that detect the presence of misfolded proteins in this compartment and transmit this information across the inner membrane. These mechanisms of control of protein folding are linked to the extracytoplasmic-stress response. This response is regulated by two distinct pathways : the CpxAR two component system constituting of the membrane-localized sensor CpxA and the cognate response regulator CpR, and the heat-shock regulon controlled by the sigma factor s>E and the anti-sigma factor RseA. The precise nature of activating signals which specifically induces the CpX and s>E pathways is unknown, but they activate the transcription of the heat-shock genes like degP or fkpA which encode a serine protease degrading misfolded envelope proteins and a peptidyl-prolyl isomerase, respectively. Moreover, these genes play a determinant role in the virulence of many pathogenic Gram-negative bacteriaprotease degrading misfolded envelope proteins and a peptidyl-prolyl isomerase, respectively. Moreover, these genes play a determinant role in the virulence of many pathogenic Gram-negative bacteria.
To understand how bacterial cells recognize, signal and respond to the presence of misfolded envelope proteins, we use the overexpression of a mutant of maltose-binding protein or MalE31 displaying a defective folding pathway and leading to the formation of inclusion bodies. We examine the effects of degP and fkpA gene deletions on the induction of both Cpx and s>E signaling pathways by using the degP promoter activity (with a transcriptional fusion to lacZ) and a biochemical approach based on the subcellular partition between soluble and insoluble proteins. This technique that we set up allows the quantification of kinetics between productive (towards native state) and unproductive (towards aggregated states) pathways which determine the periplasmic fate of misfolded MalE31. Now, we are studying the effect of temperature on the bacterial growth. When overproduced at 30°C, MalE31 did not interfere with the bacterial physiology, but at 37°C the aggregation of MalE31 becomes toxic and causes lethality. Contrary to the expectation, heat-shock conditions (growth at 42°C) rescue this lethal phenotype by increasing the degradation of MalE31.
We also study the structure-function relationships of two heat-shock proteins; FkpA and DegP that are important players in the quality control of envelope proteins. We just demonstrated that the chaperone activity of FkpA, which is independent of its PPIases activity, suppresses the formation of inclusion bodies from misfolded MalE31. For DegP, we hypothesized that the PDZ domains, which form the C-terminus part of this protease, could play a critical role in substrate recognition or by mediating self-assembly of the protease. To test this hypothesis, we are constructing several DegP mutants lacking one and both PDZ domains, and assessing their oligomeric structure and proteolytic activity.
In parallel with these studies, we are interested by alternative systems to produced recombinant proteins. These experiments started by participating to the evaluation of a new system (Rapid Translation System or RTS500 from Roche Diagnostics) designed for protein expression by a coupled in vitro transcription translation reaction. The key technology is the cell-free continuous exchange of substrates and energy components via a semipermeable membrane. Previously, we showed that MalE31 could be expressed in this system at 0.35 mg /ml and remained fully soluble and active, even at 37°C. This year, in collaboration with the laboratory of "Structural Chemistry of Macromolecule", we assessed the selective incorporation of heteronucleus 15N or 13C into specific amino acid residues. Although bacterial expression remains the most economical method from producing uniformly labeled proteins, selective labeling of proteins with one or more15N/13C enriched amino acid in E. coli is not always possible due to amino acid metabolism. We compared the efficient incorporation of 15N/13C Asp, Gly and Arg into MalE-wt between bacterial and RTS production. This isotope labeling scheme, tested with only three amino acids, was representative of amino acid metabolic pathways of E. coli. In contrast to bacterial expression, no scrambling or dilution of isotope labels was observed with RTS production. High efficient incorporation of selective labels in proteins produced by RTS500 provides the means to resolve and assign the side-chain resonances in NMR spectra of larger proteins.
IV- Protein Engineering (URL www.pasteur.fr/units/bcel/peng) (Hugues Bedouelle)
Our research is focused on the relations between the three-dimensional structure of proteins, their conformational stability and their mechanism of action. We use the multidisciplinary approach of protein engineering, including in vitro molecular evolution. During the year 2000, we have completed our studies on the relations between the conformational dynamics and function of two bacterial enzymes, tyrosyl-tRNA synthetase and tryptophan synthase, continued our work on the transformation of antibodies into reagentless optical biosensors, and studied the mechanisms of recognition between a protective antibody and different serotypes of the dengue virus.
a- Construction of biosensors from antibodies (Martial Renard Laurent Belkadi Patrick England Hugues Bedouelle)
A biosensor comprises two major components: a biological receptor, which specifically recognizes a ligand, and a tranducer, which detects the recognition event and tranforms it into a measurable signal. The monoclonal antibodies seem ideally suited to provide the biological receptor of biosensors, since they can be directed against most haptens and macromolecules. We are developing a set of approaches to transform antibodies into reagentless optical biosensors. We chose monoclonal antibody mAbD1.3, directed against hen lysozyme, as an experimental system because detailed structural data are available. These approaches will be extended to antibodies for which no structural data exist, then for other types of receptors. These biosensors could have numerous applications in diagnostics, pharmacology and industry, in particular as protein chips.
b- Structural bases of immune cross-reactions (Laurent Belkadi Patrick England Hugues Bedouelle)
Many viruses possess several serotypes, which can be distinguished by specific antibodies. Reciprocally, some monoclonal antibodies can recognize several serotypes . We have analyzed this phenomenon for a monoclonal antibody which is directed against the dengue virus and recognizes its 4 serotypes with varying efficiencies. Each residue of the CDR3 hypervariable loop was changed into alanine. The variations of affinity for the envelope proteins of the DEN1 and DEN2 serotypes of the virus, and the corresponding rates of interaction were measured by competition ELISA and BIAcore. We found that the residues of the antibody that were involved in the recognition of the DEN2 serotype, constituted a sub-set of those for DEN1, and that their energetic contributions were weaker. Nevertheless, the residues that were preponderant for the recognition of the antigen, were the same for both serotypes. Some deletions of side chains improved the recognition of the DEN2 serotype without affecting that of DEN1. These results will help us to modify the antibody so that it can recognize the different serotypes with the same affinity, for therapeutic purposes.
c- Modular structure of tyrosyl-tRNA synthetase (Valérie Guez Carole Gaillard Hugues Bedouelle ; in collaboration with the Unit of NMR of Biomolecules)
Many natural proteins were formed by an assembling preexisting modules or domains, and this observation has inspired many approaches in protein engineeering. Tyrosyl-tRNA synthetase (TyrRS) is a homodimer. Each subunit comprises three structural domains: an N-terminal domain (residues 1-220 in B. stearothermophilus) which possesses the characteristic fold of the class I synthetases, an alpha-helical domain with unknown function, and a C-terminal domain (residues 320-419) which binds the anticodon of the tRNA and is disordered in the crystal structure.
We determined the three-dimensional structure of the C-terminal domain of TyrRS from B. stearothermophilus, in collaboration with the NMR Unit. This structure belongs to the superfamily of the S4 ribosomal protein but is new among the aminoacyl-tRNA syntetases. Thus, TyrRS and other proteins have recruted the same domain to bind their RNA ligand.
We performed a scanning of the peptide that links the alpha-helical and C-terminal domains of TyrRS by mutagenesis. We found that residue Phe323 and the upstream sequence were important for the charging of tRNATyr but not the downstream sequence, which could be made flexible. The aromatic character of Phe323 was important for the stability of the initial complex between TyrRS and tRNATyr, and still more important for the stability of their complex in the transition state. These results have indicated that Phe323 belongs to the alpha-helical domain of TyrRS, and that the recognition of the anticodon arm of tRNATyr involves residues of its two idiosynchratic domains (alpha-helical and C-terminal).
d- Allostery and substrate channelling in tryptophan synthase (Philippe Rondard Hugues Bedouelle)
Tryptophan synthase from Escherichia coli is a tetrameric enzyme, with a TrA.TrpB.TrpB.TrpA structure. Structural studies have identified residues 273-283 of TrpB as a potentially important region for the allosteric communication between the TrpA and TrpB subunits, and for the transport of indole between their active sites, across a hydrophobic tunnel. We constructed 19 mutations of residues 273-283 of TrpB to explore the functional role of this region. The mutations could be divided into 4 classes according to their effects on the tryptophan synthase and serine deaminase activities of the TrpB2 dimer, either in the presence or in the absence of the TrpA subunit. We thus confirmed the allosteric role of residues 278-282 of TrpB, and described the reaction steps and the intra-molecular contacts which could be affected by each class of mutations.
V- Recombinant toxins of therapeutic and biotechnological interest(Daniel Ladant)
In our group, we have been studying for several years the adenylate cyclase (AC) toxin produced by Bordetella pertussis, the causative agent of whooping cough. This toxin is one of the major virulence factors of this organism: it is secreted by the virulent bacteria and it is able to enter into eukaryotic cells where, upon activation by endogenous calmodulin, it catalyzes high-level synthesis of cAMP that in turn alters cellular physiology. The AC toxin is a 1706 residues-long protein that exhibits several striking features. In particular, the AC toxin has a unique mode of entry into eukaryotic cells, by which the N-terminal, catalytic, domain is translocated into the cytosol directly across the plasma membrane of the target cells.
Our work during the last year has been focused on three different aspects:
a- Structure-function relationships of the AC toxin: Mechanisms of entry into eukaryotic cells and delivery of T-cell epitopes into antigen presenting cells (Cécile Bauche Daniel Ladant)
This work is carried out in collaboration with the team of Claude Leclerc (Biologie des Régulations Immunitaires, Institut Pasteur). From a fundamental perspective, our objective is to decipher the molecular mechanisms of the invasion of target cells by the AC toxin, by combining genetic, biochemical and physico-chemical approaches. This has been pursued over the last years with a particular emphasis on the engineering of recombinant toxins able to deliver into eukaryotic cells, proteins or polypeptides genetically fused to the catalytic domain of the toxin. In particular, we have used recombinant AC toxins to deliver major histocompatibility complex (MHC) class I-associated T-cell epitopes within antigen-presenting cells, in order to trigger specific cytotoxic T- cell (CTL) responses. One major result obtained was the identification of the aMb2 integrin (CD11b/CD18) as the cellular receptor of CyaA. This integrin is expressed by a restricted subset of leukocytes including neutrophils, macrophages and dendritic cells. The presence of a specific receptor for CyaA on dendritic cells could explain the remarkable efficacy of our recombinant toxins in priming CTLs. Besides, we produced and characterized various recombinant CyaA toxins harboring T cell epitopes derived from human melanoma to be tested in a transgenic mice model for induction of CTL responses.
b- Development of a bacterial two-hybrid system to study protein-protein interaction in Escherichia coli (Gouzel Karimova Agnès Ullmann Daniel Ladant)
We have recently set up a novel bacterial two-hybrid system that allows an easy in vivo screening and selection of functional interactions between two proteins in E. coli. This system is based on the functional complementation between the two AC subdomains, fused to polypeptides of interest. In an E. coli cya strain, heterodimerization of these chimeric polypeptides restore AC enzymatic activity and leads to cAMP synthesis, which then, activates transcription of catabolic operons. This can be scored either on indicator plates or on selective media. Recent work was focused on the improvement of the system that might be useful for large scale screening of protein-protein interactions as well as for structure-function studies of macromolecules.
c- A genetic screen for site-specific protease activity in Escherichia coli: Application to the detection of antiprotease-resistant HIV proteases (Nathalie Dautin - Gouzel Karimova - Agnès Ullmann - Daniel Ladant)
We have set up a genetic system that allows in vivo screening or selection of site-specific proteases and of their cognate specific inhibitors in E. coli. This genetic test is based on the specific proteolysis of the AC catalytic domain. As a model system, we tested the HIV protease that is responsible for the proteolytic processing of the HIV polyprotein precursor gag/pol into mature viral proteins. We showed that AC is an exquisitely sensitive reporter system for monitoring the proteolytic activity of the HIV protease and its inhibition by known inhibitors. In particular, we showed that this genetic test is able to distinguish between wild type HIV protease and variants resistant to inhibitors that were isolated from patients under highly active antiretroviral therapy (HAART). This genetic test could represent a powerful approach to detect, in patients undergoing HAART, the emergence of HIV variants harboring antiprotease-resistant proteases.
VI- Contribution of physical-chemical methods from the Unit to various collaboration (Alain Chaffotte - Michel Goldberg - Roland Nageotte)
The Unit offers to the scientific community, in particular the pasteurian one, its equipments and expertise in the physical-chemical studies of proteins in solution and of their interactions. It contributes to the conception, of experiments using analytical ultracentrifugation, fluorescence spectroscopy, circular dichroism, and stopped-flow rapid mixing, it performs, and it interprets the corresponding experiments for the benefit of laboratories on and off campus.
As an example of these many collaborations, this year's study by circular dichroism, performed for V. Redeker (Laboratory of Neurobiology and Cellular Diversity of ESPCI), of 5 bioactive synthetic peptides with sequences related to those of toxins secreted by venimous ants have shown that, unlike the sequence based predictions, these peptides are disorganized in aqueous solution, but show a marked propensity to adopt an alpha-helical conformation in the presence of low trifluoroethanol concentrations.
VII- Teaching and Education
The Unit is in charge of organizing the "Protein Biochemistry" laboratory course (Director: Alain Chaffotte) of the Institut Pasteur, which is associated to the Master's program of the Paris 6, Paris 7 and Orsay Universities, the Ecole Normale Supérieure, the Ecole Polytechnique, and the CEA. A large part of this course has been performed by members this Unit. In particular, two groups of the Unit (Folding in vitro and Protein Folding in the bacterial envelope) have organized 2 full weeks each of laboratory sessions for the 2000 course.
2 Master's students and 4 PhD students were under training in the Unit this year.
Calculation of the effect of mutations on the association constant of R67 DHFR.
a: General view. The secondary structures are represented and colored respectively in red, blue, green, yellow for the four subunits of the protein. The atoms of the modified residues are shown by orange and black spheres for the atoms that are removed and added, respectively. Finally, the O-H bonds of water molecules are shown in blue.
b: Closer view of the modified residues corresponding to the upper half of the general view. The blue and red subunits are represented by secondary structure type and covalent bonds. The residues that are removed or added are shown with CPK colors but with light blue carbons and thin rods, or dark green carbon and thick rods, respectively. Unmodified symmetrical residues are represented by rods and spheres. The surface created by the atoms surrounding those that are modified is represented in translucent white. Finally, O-H bonds of water molecules are shown in blue.