One of the main problems faced by the current paradigm of drug discovery in the therapeutic area of infection is in identification of a valid drug target, followed by an understanding of the concentration dynamics of the metabolites in response to its inhibition. A valid target at the genetic level is the one that is essential for the survival of the pathogen. At the functional level, the target protein structure and/or its biological activity is substantially different from that of the host. However, not all targets with these criteria yield inhibitors that kill the pathogen specifically through the mechanism intended. This necessitates generating a large number of targets suitable for high throughput screening (HTS).1
A recent review of anti-infective HTS by Payne et al indicated that the success rate is abysmally low even with a large number of HTS campaigns.2
Hence there is an urgent need to strategize and modify guidelines that define an ideal anti-infective drug target.
Many potent inhibitors of essential enzymes have been reported to lack antibacterial activity.3
In most cases, efflux or lack of permeability across the bacterial cell wall has been suspected to be a principal factor responsible for this phenomenon. This explanation has remained mainly unsubstantiated since it is difficult to determine the exact cause. However a quantitative relationship between enzyme inhibition and its effect on the growth of the pathogen has been delineated. Eisenthal and Cornish-Bowden have postulated two basic mechanisms by which the enzyme inhibition leads to the cessation of growth of an organism: “either the flux through an essential metabolic pathway can be decreased to a point where life is no longer possible or metabolite concentration can be increased to toxic levels”.4
Recent evidence also indicates that independent of the target inhibited, bacterial cell death ultimately occurs through a generalized mechanism involving modulation of multiple pathways that leads to generation of free radicals.5
Thus there is a need to interconnect inhibition of target enzyme in vitro and inside the cell. It is well known that inhibition or over expression of an enzyme in a metabolic pathway does not guarantee either a change in flux or a concomitant increase or decrease in the metabolite concentration. The activity of any enzyme inside the cell depends on its own substrate concentration and kinetic properties (eg, Km,Vmax) as well as those of other enzymes in that particular pathway. On the other hand, the degree of enzyme inhibition inside the cell is related to the mode of enzyme inhibition (MOI) as well as to the enzyme-inhibitor dissociation constant (Ki). Therefore, it is essential to consider a network of enzymes or pathways around a particular target enzyme so as to predict the effect of its inhibition on the overall cell metabolism. Since target gene knockouts mimic complete inhibition inside the cell, it may not be possible to achieve the same inhibition through compound-mediated inhibition. It is thus necessary to assess the effect of partial inhibition through regulated expression of a target gene.6
Since this involves extensive experimentation, a better alternative is to predict cell vulnerability in response to a specific target enzyme inhibition through in silico simulations. Vulnerability is defined as the extent of inhibition of a target required to have a negative impact on growth leading to cell death.7
Towards this objective, we have built an in silico dynamic network of 8 major pathways in Escherichia coli, including glycolysis, pentose phosphate pathway, and tricarboxylic acid (TCA) cycle along with glyoxylate shunt, fatty acid metabolism, and biosyntheses of branched chain amino acids, pantothenic acid, and coenzyme A (CoA) (). These pathways were chosen based on their connection to central carbon metabolism and their potential for possessing drug targets, for example, isocitrate lyase or pantothenate kinase, in drug discovery against pathogens like Mycobacterium tuberculosis. Published data on enzyme kinetics, pathway flux distribution, operon structure, and gene regulation were used to build this platform. The pathway dynamics were simulated by interconnecting ordinary differential equations describing kinetic behavior of each enzyme in the pathway. Such a kinetic model is thus a computational and mathematical framework, built by using intracellular enzyme as well as metabolite concentration and other kinetic parameters. This platform is expected to elicit responses to perturbations in a fashion similar to the way the natural system in question would. This type of modeling has been referred to as an “impossible” problem, primarily because of the dearth of parametric data required to give meaning to flux equations, and, secondly, the absence of a software that can simulate and give stable solutions to systems comprising thousands of ordinary differential equations (ODEs). The proprietary Cellworks (Cellworks Group, Inc., Saratoga, CA, USA) technology, iC-PHYS™, can be used to create and simulate such a platform. Cell growth could be simulated by providing glucose or acetate as a carbon source (C-source) along with essential metabolites (metabolites not synthesized on this platform) such as beta-alanine, threonine, and cysteine. Altogether, 189 genes and 449 biochemical reactions covering 434 metabolites were modeled on this platform. The model was aligned (see Materials and Methods – Alignment of the platform) with data reported in the literature and “frozen” so that the subsequent validation studies did not require any alteration in the model parameters or equations. Validation studies were carried out using published data for E. coli. Various enzymes were evaluated as potential drug targets and those that were either vulnerable or relatively immune to inhibition of a specific type were identified and some experimentally verified. Since there is a significant overlap in the metabolic pathways among various bacteria, we tested the predictive capability of the in silico platform by correlating model predictions with experiments carried out on Mycobacterium bovis BCG as a surrogate for M. tuberculosis.
Schematic representation of pathways modeled on this platform. Each pathway is represented by central reactions or metabolites, truncated such for simplistic representation.