GAM: a web-service for integrated transcriptional and metabolic network analysis.

Novel techniques for high-throughput steady-state metabolomic profiling yield information about changes of nearly thousands of metabolites. Such metabolomic profiles, when analyzed together with transcriptional profiles, can reveal novel insights about underlying biological processes. While a number of conceptual approaches have been developed for data integration, easily accessible tools for integrated analysis of mammalian steady-state metabolomic and transcriptional data are lacking. Here we present GAM ('genes and metabolites'): a web-service for integrated network analysis of transcriptional and steady-state metabolomic data focused on identification of the most changing metabolic subnetworks between two conditions of interest. In the web-service, we have pre-assembled metabolic networks for humans, mice, Arabidopsis and yeast and adapted exact solvers for an optimal subgraph search to work in the context of these metabolic networks. The output is the most regulated metabolic subnetwork of size controlled by false discovery rate parameters. The subnetworks are then visualized online and also can be downloaded in Cytoscape format for subsequent processing. The web-service is available at: https://artyomovlab.wustl.edu/shiny/gam/.

Environment Impacts the Metabolic Dependencies of Ras-Driven Non-Small Cell Lung Cancer.

Cultured cells convert glucose to lactate, and glutamine is the major source of tricarboxylic acid (TCA)-cycle carbon, but whether the same metabolic phenotype is found in tumors is less studied. We infused mice with lung cancers with isotope-labeled glucose or glutamine and compared the fate of these nutrients in tumor and normal tissue. As expected, lung tumors exhibit increased lactate production from glucose. However, glutamine utilization by both lung tumors and normal lung was minimal, with lung tumors showing increased glucose contribution to the TCA cycle relative to normal lung tissue. Deletion of enzymes involved in glucose oxidation demonstrates that glucose carbon contribution to the TCA cycle is required for tumor formation. These data suggest that understanding nutrient utilization by tumors can predict metabolic dependencies of cancers in vivo. Furthermore, these data argue that the in vivo environment is an important determinant of the metabolic phenotype of cancer cells.

Network integration of parallel metabolic and transcriptional data reveals metabolic modules that regulate macrophage polarization.

Macrophage polarization involves a coordinated metabolic and transcriptional rewiring that is only partially understood. By using an integrated high-throughput transcriptional-metabolic profiling and analysis pipeline, we characterized systemic changes during murine macrophage M1 and M2 polarization. M2 polarization was found to activate glutamine catabolism and UDP-GlcNAc-associated modules. Correspondingly, glutamine deprivation or inhibition of N-glycosylation decreased M2 polarization and production of chemokine CCL22. In M1 macrophages, we identified a metabolic break at Idh, the enzyme that converts isocitrate to alpha-ketoglutarate, providing mechanistic explanation for TCA cycle fragmentation. (13)C-tracer studies suggested the presence of an active variant of the aspartate-arginosuccinate shunt that compensated for this break. Consistently, inhibition of aspartate-aminotransferase, a key enzyme of the shunt, inhibited nitric oxide and interleukin-6 production in M1 macrophages, while promoting mitochondrial respiration. This systems approach provides a highly integrated picture of the physiological modules supporting macrophage polarization, identifying potential pharmacologic control points for both macrophage phenotypes.

Hexokinase 2 is required for tumor initiation and maintenance and its systemic deletion is therapeutic in mouse models of cancer.

Accelerated glucose metabolism is a common feature of cancer cells. Hexokinases catalyze the first committed step of glucose metabolism. Hexokinase 2 (HK2) is expressed at high level in cancer cells, but only in a limited number of normal adult tissues. Using Hk2 conditional knockout mice, we showed that HK2 is required for tumor initiation and maintenance in mouse models of KRas-driven lung cancer, and ErbB2-driven breast cancer, despite continued HK1 expression. Similarly, HK2 ablation inhibits the neoplastic phenotype of human lung and breast cancer cells in vitro and in vivo. Systemic Hk2 deletion is therapeutic in mice bearing lung tumors without adverse physiological consequences. Hk2 deletion in lung cancer cells suppressed glucose-derived ribonucleotides and impaired glutamine-derived carbon utilization in anaplerosis.

Context and force field dependence of the loss of protein backbone entropy upon folding using realistic denatured and native state ensembles.

The loss of conformational entropy is the largest unfavorable quantity affecting a protein's stability. We calculate the reduction in the number of backbone conformations upon folding using the distribution of backbone dihedral angles (ϕ,ψ) obtained from an experimentally validated denatured state model, along with all-atom simulations for both the denatured and native states. The average loss of entropy per residue is TΔS(BB)(U-N) = 0.7, 0.9, or 1.1 kcal·mol(-1) at T = 298 K, depending on the force field used, with a 0.6 kcal·mol(-1) dispersion across the sequence. The average equates to a decrease of a factor of 3-7 in the number of conformations available per residue (f = Ω(Denatured)/Ω(Native)) or to a total of f(tot) = 3(n)-7(n) for an n residue protein. Our value is smaller than most previous estimates where f = 7-20, that is, our computed TΔS(BB)(U-N) is smaller by 10-100 kcal mol(-1) for n = 100. The differences emerge from our use of realistic native and denatured state ensembles as well as from the inclusion of accurate local sequence preferences, neighbor effects, and correlated motions (vibrations), in contrast to some previous studies that invoke gross assumptions about the entropy in either or both states. We find that the loss of entropy primarily depends on the local environment and less on properties of the native state, with the exception of α-helical residues in some force fields.

Automated real-space refinement of protein structures using a realistic backbone move set.

Crystals of many important biological macromolecules diffract to limited resolution, rendering accurate model building and refinement difficult and time-consuming. We present a torsional optimization protocol that is applicable to many such situations and combines Protein Data Bank-based torsional optimization with real-space refinement against the electron density derived from crystallography or cryo-electron microscopy. Our method converts moderate- to low-resolution structures at initial (e.g., backbone trace only) or late stages of refinement to structures with increased numbers of hydrogen bonds, improved crystallographic R-factors, and superior backbone geometry. This automated method is applicable to DNA-binding and membrane proteins of any size and will aid studies of structural biology by improving model quality and saving considerable effort. The method can be extended to improve NMR and other structures. Our backbone score and its sequence profile provide an additional standard tool for evaluating structural quality.

Functional genomics reveal that the serine synthesis pathway is essential in breast cancer.

Cancer cells adapt their metabolic processes to drive macromolecular biosynthesis for rapid cell growth and proliferation. RNA interference (RNAi)-based loss-of-function screening has proven powerful for the identification of new and interesting cancer targets, and recent studies have used this technology in vivo to identify novel tumour suppressor genes. Here we developed a method for identifying novel cancer targets via negative-selection RNAi screening using a human breast cancer xenograft model at an orthotopic site in the mouse. Using this method, we screened a set of metabolic genes associated with aggressive breast cancer and stemness to identify those required for in vivo tumorigenesis. Among the genes identified, phosphoglycerate dehydrogenase (PHGDH) is in a genomic region of recurrent copy number gain in breast cancer and PHGDH protein levels are elevated in 70% of oestrogen receptor (ER)-negative breast cancers. PHGDH catalyses the first step in the serine biosynthesis pathway, and breast cancer cells with high PHGDH expression have increased serine synthesis flux. Suppression of PHGDH in cell lines with elevated PHGDH expression, but not in those without, causes a strong decrease in cell proliferation and a reduction in serine synthesis. We find that PHGDH suppression does not affect intracellular serine levels, but causes a drop in the levels of α-ketoglutarate, another output of the pathway and a tricarboxylic acid (TCA) cycle intermediate. In cells with high PHGDH expression, the serine synthesis pathway contributes approximately 50% of the total anaplerotic flux of glutamine into the TCA cycle. These results reveal that certain breast cancers are dependent upon increased serine pathway flux caused by PHGDH overexpression and demonstrate the utility of in vivo negative-selection RNAi screens for finding potential anticancer targets.

Mimicking the folding pathway to improve homology-free protein structure prediction.

Since the demonstration that the sequence of a protein encodes its structure, the prediction of structure from sequence remains an outstanding problem that impacts numerous scientific disciplines, including many genome projects. By iteratively fixing secondary structure assignments of residues during Monte Carlo simulations of folding, our coarse-grained model without information concerning homology or explicit side chains can outperform current homology-based secondary structure prediction methods for many proteins. The computationally rapid algorithm using only single (phi,psi) dihedral angle moves also generates tertiary structures of accuracy comparable with existing all-atom methods for many small proteins, particularly those with low homology. Hence, given appropriate search strategies and scoring functions, reduced representations can be used for accurately predicting secondary structure and providing 3D structures, thereby increasing the size of proteins approachable by homology-free methods and the accuracy of template methods that depend on a high-quality input secondary structure.

How the thymus designs antigen-specific and self-tolerant T cell receptor sequences.

T lymphocytes (T cells) orchestrate adaptive immune responses that clear pathogens from infected hosts. T cells recognize short peptides (p) derived from antigenic proteins bound to protein products of the MHC genes. Recognition occurs when T cell receptor (TCR) proteins expressed on T cells bind sufficiently strongly to antigen-derived pMHC complexes on the surface of antigen-presenting cells. A diverse repertoire of self-pMHC-tolerant TCR sequences is shaped during development of T cells in the thymus by processes called positive and negative selection. Combining computational models and analysis of experimental data, we parsed the contributions of positive and negative selection to the design of TCR sequences that recognize antigenic peptides with specificity, yet also exhibit cross-reactivity. A dominant role for negative selection in mediating antigen specificity of mature T cells and a molecular mechanism for TCR recognition of antigen are described.

Benchmarking implicit solvent folding simulations of the amyloid beta(10-35) fragment.

A pathogenetic feature of Alzhemier disease is the aggregation of monomeric beta-amyloid proteins (Abeta) to form oligomers. Usually these oligomers of long peptides aggregate on time scales of microseconds or longer, making computational studies using atomistic molecular dynamics models prohibitively expensive and making it essential to develop computational models that are cheaper and at the same time faithful to physical features of the process. We benchmark the ability of our implicit solvent model to describe equilibrium and dynamic properties of monomeric Abeta(10-35) using all-atom Langevin dynamics (LD) simulations, since Alphabeta(10-35) is the only fragment whose monomeric properties have been measured. The accuracy of the implicit solvent model is tested by comparing its predictions with experiment and with those from a new explicit water MD simulation, (performed using CHARMM and the TIP3P water model) which is approximately 200 times slower than the implicit water simulations. The dependence on force field is investigated by running multiple trajectories for Alphabeta(10-35) using the CHARMM, OPLS-aal, and GS-AMBER94 force fields, whereas the convergence to equilibrium is tested for each force field by beginning separate trajectories from the native NMR structure, a completely stretched structure, and from unfolded initial structures. The NMR order parameter, S2, is computed for each trajectory and is compared with experimental data to assess the best choice for treating aggregates of Alphabeta. The computed order parameters vary significantly with force field. Explicit and implicit solvent simulations using the CHARMM force fields display excellent agreement with each other and once again support the accuracy of the implicit solvent model. Alphabeta(10-35) exhibits great flexibility, consistent with experiment data for the monomer in solution, while maintaining a general strand-loop-strand motif with a solvent-exposed hydrophobic patch that is believed to be important for aggregation. Finally, equilibration of the peptide structure requires an implicit solvent LD simulation as long as 30 ns.

Solvation effect on conformations of 1,2:dimethoxyethane: charge-dependent nonlinear response in implicit solvent models.

The physical content of and, in particular, the nonlinear contributions from the Langevin-Debye model are illustrated using two applications. First, we provide an improvement in the Langevin-Debye model currently used in some implicit solvent models for computer simulations of solvation free energies of small organic molecules, as well as of biomolecular folding and binding. The analysis is based on the implementation of a charge-dependent Langevin-Debye (qLD) model that is modified by subsequent corrections due to Onsager and Kirkwood. Second, the physical content of the model is elucidated by discussing the general treatment within the LD model of the self-energy of a charge submerged in a dielectric medium for three different limiting conditions and by considering the nonlinear response of the medium. The modified qLD model is used to refine an implicit solvent model (previously applied to protein dynamics). The predictions of the modified implicit solvent model are compared with those from explicit solvent molecular dynamics simulations for the equilibrium conformational populations of 1,2-dimethoxyethane (DME), which is the shortest ether molecule to reproduce the local conformational properties of polyethylene oxide, a polymer with tremendous technological importance and a wide variety of applications. Because the conformational population preferences of DME change dramatically upon solvation, DME is a good test case to validate our modified qLD model. The present analysis of the modified qLD model provides the motivation and tools for studying a wide variety of other interesting systems with heterogeneous dielectric properties and spatial anisotropy.

Reduced C(beta) statistical potentials can outperform all-atom potentials in decoy identification.

We developed a series of statistical potentials to recognize the native protein from decoys, particularly when using only a reduced representation in which each side chain is treated as a single C(beta) atom. Beginning with a highly successful all-atom statistical potential, the Discrete Optimized Protein Energy function (DOPE), we considered the implications of including additional information in the all-atom statistical potential and subsequently reducing to the C(beta) representation. One of the potentials includes interaction energies conditional on backbone geometries. A second potential separates sequence local from sequence nonlocal interactions and introduces a novel reference state for the sequence local interactions. The resultant potentials perform better than the original DOPE statistical potential in decoy identification. Moreover, even upon passing to a reduced C(beta) representation, these statistical potentials outscore the original (all-atom) DOPE potential in identifying native states for sets of decoys. Interestingly, the backbone-dependent statistical potential is shown to retain nearly all of the information content of the all-atom representation in the C(beta) representation. In addition, these new statistical potentials are combined with existing potentials to model hydrogen bonding, torsion energies, and solvation energies to produce even better performing potentials. The ability of the C(beta) statistical potentials to accurately represent protein interactions bodes well for computational efficiency in protein folding calculations using reduced backbone representations, while the extensions to DOPE illustrate general principles for improving knowledge-based potentials.

Polypeptide motions are dominated by peptide group oscillations resulting from dihedral angle correlations between nearest neighbors.

To identify basic local backbone motions in unfolded chains, simulations are performed for a variety of peptide systems using three popular force fields and for implicit and explicit solvent models. A dominant "crankshaft-like" motion is found that involves only a localized oscillation of the plane of the peptide group. This motion results in a strong anticorrelated motion of the phi angle of the ith residue (phi(i)) and the psi angle of the residue i - 1 (psi(i-1)) on the 0.1 ps time scale. Only a slight correlation is found between the motions of the two backbone dihedral angles of the same residue. Aside from the special cases of glycine and proline, no correlations are found between backbone dihedral angles that are separated by more than one torsion angle. These short time, correlated motions are found both in equilibrium fluctuations and during the transit process between Ramachandran basins, e.g., from the beta to the alpha region. A residue's complete transit from one Ramachandran basin to another, however, occurs in a manner independent of its neighbors' conformational transitions. These properties appear to be intrinsic because they are robust across different force fields, solvent models, nonbonded interaction routines, and most amino acids.

Minimalist representations and the importance of nearest neighbor effects in protein folding simulations.

In order to investigate the level of representation required to simulate folding and predict structure, we test the ability of a variety of reduced representations to identify native states in decoy libraries and to recover the native structure given the advanced knowledge of the very broad native Ramachandran basin assignments. Simplifications include the removal of the entire side-chain or the retention of only the Cbeta atoms. Scoring functions are derived from an all-atom statistical potential that distinguishes between atoms and different residue types. Structures are obtained by minimizing the scoring function with a computationally rapid simulated annealing algorithm. Results are compared for simulations in which backbone conformations are sampled from a Protein Data Bank-based backbone rotamer library generated by either ignoring or including a dependence on the identity and conformation of the neighboring residues. Only when the Cbeta atoms and nearest neighbor effects are included do the lowest energy structures generally fall within 4 A of the native backbone root-mean square deviation (RMSD), despite the initial configuration being highly expanded with an average RMSD > or = 10 A. The side-chains are reinserted into the Cbeta models with minimal steric clash. Therefore, the detailed, all-atom information lost in descending to a Cbeta-level representation is recaptured to a large measure using backbone dihedral angle sampling that includes nearest neighbor effects and an appropriate scoring function.

Statistical coil model of the unfolded state: resolving the reconciliation problem.

An unfolded state ensemble is generated by using a self-avoiding statistical coil model that is based on backbone conformational frequencies in a coil library, a subset of the Protein Data Bank. The model reproduces two apparently contradicting behaviors observed in the chemically denatured state for a variety of proteins, random coil scaling of the radius of gyration and the presence of significant amounts of local backbone structure (NMR residual dipolar couplings). The most stretched members of our unfolded ensemble dominate the residual dipolar coupling signal, whereas the uniformity of the sign of the couplings follows from the preponderance of polyproline II and beta conformers in the coil library. Agreement with the NMR data substantially improves when the backbone conformational preferences include correlations arising from the chemical and conformational identity of neighboring residues. Although the unfolded ensembles match the experimental observables, they do not display evidence of native-like topology. By providing an accurate representation of the unfolded state, our statistical coil model can be used to improve thermodynamic and kinetic modeling of protein folding.

Helix, sheet, and polyproline II frequencies and strong nearest neighbor effects in a restricted coil library.

A central issue in protein folding is the degree to which each residue's backbone conformational preferences stabilize the native state. We have studied the conformational preferences of each amino acid when the amino acid is not constrained to be in a regular secondary structure. In this large but highly restricted coil library, the backbone preferentially adopts dihedral angles consistent with the polyproline II conformation rather than alpha or beta conformations. The preference for the polyproline II conformation is independent of the degree of solvation. In conjunction with a new masking procedure, the frequencies in our coil library accurately recapitulate both helix and sheet frequencies for the amino acids in structured regions, as well as polyproline II propensities. Therefore, structural propensities for alpha-helices and beta-sheets and for polyproline II conformations in unfolded peptides can be rationalized solely by local effects. In addition, these propensities are often strongly affected by both the chemical nature and the conformation of neighboring residues, contrary to the Flory isolated residue hypothesis.