Luiz Pedro Sorio de Carvalho

Transition metal complexes provide rigid, shape-defined molecular architectures in which the coordination geometry precisely organizes ligand functionality in three dimensions, making them attractive scaffolds for target-based chemical probe discovery. Yet, their broader and prospective use in structure-based virtual screening has remained largely unexplored, due to the absence of structurally diverse, computationally tractable screening libraries. Here, we address this by curating MOSAIC (Metal complex library for Structure-Aided In silico screening), a first collection of 16,901 neutral, drug-like metal complexes assembled directly from experimentally determined geometries deposited in the Cambridge Structural Database. Using MOSAIC , we perform a high-throughput virtual screening campaign against the allosteric pocket of methionine adenosyltransferase 2A (MAT2A), identifying the copper complex Cu1-Complex , for which nano differential scanning fluorimetry (nanoDSF) confirmed strong binding in vitro . The diamagnetic zinc analogue Zn1-Comple x allowed orthogonal binding characterization with ligand-observed and competition NMR studies. Along with Hydrogen–deuterium exchange mass spectrometry (HDX-MS), site-directed mutagenesis and molecular dynamics (MD), these studies show that Zn1-Complex engages the MAT2A through an allosteric pocket, inhibiting enzymatic activity. Together, this establishes MOSAIC as a scalable and generalizable virtual screening platform for transition metal complexes and demonstrates the potential of these metal-based scaffolds as chemical probes to uncover new modes of protein recognition.

The genus Mycobacterium comprises over 200 species, many of which now have complete genome sequences. Some are major pathogens causing diseases like tuberculosis and leprosy, while others are harmless environmental organisms with useful abilities such as degrading pollutants. Environmental mycobacteria are often seen as metabolic generalists, able to utilise a wider range of carbon sources than host-associated species, which are typically more specialised due to their restricted habitats. This metabolic versatility has been proposed to stem from differences in nutrient uptake capabilities rather than catabolic pathways. In order to test this explanation, we developed and validated genome-scale metabolic models for five Mycobacterium species with varying lifestyles and growth rates, creating a computational approach enabled by CarveMe that allows rapid construction of models from genome information. By combining these models with microbiology experiments the study showed that the capacity of the bacteria to transport nutrients into the cell is indeed key to metabolic versatility. We notably found through load-partition experiments that, if a transporter is present but cannot take up its substrate at a rate sufficient for growth, the supply of multiple substrates can mitigate this rate-limiting step. This suggests that mycobacterial species have evolved high-affinity, low-rate systems for nutrient uptake in their ecological niches. More generally, our results demonstrate that a combination of automated annotation methods and straightforward bacterial physiology experiments allow the reconstruction of metabolic models of good predictive quality for hitherto little studied mycobacterial species.

Despite the exponential growth in genome sequencing, the functional annotation of genes, particularly the discovery of novel enzymatic activities, remains a formidable challenge. This gap is exacerbated by annotation biases that propagate assumptions about enzyme function across homologous sequences. Here, we report the identification of an unannotated enzymatic activity within a presumed well-characterized enzyme family. The gene rv2531 c, under strong purifying selection across Mycobacterium tuberculosis strains, has been annotated as a member of the lysine-ornithine-arginine (KOR) decarboxylase superfamily, which includes over 26,000 sequences. Contrary to this annotation, we show that Rv2531c does not decarboxylate KOR substrates. Instead, using an integrative approach combining bioinformatics, microbiology, metabolomics, and enzymology, we demonstrate that Rv2531c is a L-glutamate decarboxylase that sustains carbon flux through the γ-aminobutyric acid shunt in M. tuberculosis . This newly identified subfamily of enzymes is conserved across Bacteria, Archaea, and Eukarya and exhibits distinct structural and kinetic features, including an additional domain, hysteresis, and strong positive cooperativity. These characteristics differentiate it from canonical, enterobacterial KOR decarboxylases. More broadly, our findings challenge the narrow substrate and functional scope traditionally assigned to the KOR-DC superfamily. We propose that many members of this large and diverse enzyme family catalyze distinct reactions and participate in previously unrecognized metabolic pathways, revealing a broader and more nuanced role for pyridoxal 5′-phosphate-dependent enzymes in microbial physiology and evolution.

Transition metal complexes occupy regions of three-dimensional chemical space that organic scaffolds cannot easily reach, yet their use in target-based drug discovery has been held back by the absence of structurally diverse, readily available screening libraries. We addressed this by curating MOSAIC ( M etal c o mplex library for S tructure- Ai ded I n silico s c reening), a first collection of 16,901 neutral, drug-like metal complexes assembled directly from experimentally determined geometries deposited in the Cambridge Structural Database. Docking this library against the allosteric site of the oncology target MAT2A (methionine adenosyltransferase 2A) with MetalDock returned the copper complex Cu1-Complex , from which we designed the diamagnetic zinc analogue Zn1-Complex for detailed study. A combination of nanoDSF, ligand-observed and competition NMR, HDX-MS, site-directed mutagenesis and molecular dynamics placed Zn1-Complex in the allosteric pocket and showed that it engages the site differently from the established inhibitor Pf-9366 , while functionally inhibiting the enzyme. Together, these results establish MOSAIC as a validated, generalisable platform for target-based metal complex discovery as the first scalable strategy for systematically exploiting metal complexes against any well-defined binding site.

In this study, we investigated how iron limitation alters central metabolism in Mycobacterium tuberculosis using metabolomics and stable isotope tracing. Our findings reveal a well-orchestrated metabolic programme to enable Krebs cycle activity despite the inefficient action of its iron-dependent enzymes. Under such conditions, carbon flux through the oxidative branch of the Krebs cycle is stalled, resulting in the accumulation of metabolites that are partially secreted. As a result, carbon flux from glycolysis is partially diverted to the reductive branch of the Krebs cycle to support the production of oxaloacetate and malate through the activity of phosphoenolpyruvate carboxykinase and pyruvate carboxylase. Both branches terminate with the synthesis of malate, which is secreted. This unprecedented split of the Krebs cycle and malate secretion in a bacterial pathogen facilitates the continuous flow of carbon through the core of carbon metabolism, overcoming the metabolic stalling triggered by iron starvation.

Mycobacteria comprise an uneven number of genes coding for biotin carboxylase (BC) and acyl-CoA carboxyl transferase (CT) activities and an ε-subunit of unknown function. To unravel the composition of acyl-CoA carboxylase (ACCase) holoenzymes under native conditions, we pulled down endogenously expressed biotinylated proteins from Mycobacterium smegmatis . The extract included an unexpected 16-subunit ACCase holoenzyme, exhibiting distinct long-chain (LC) and short-chain (SC) CT activities. It consists of a central two-layered CT (AccD4)₂/(AccD5)₄ assembly that interacts with a C-terminal AccE5 dimer within its central hole. On each CT hexamer face, an asymmetric BC AccA3 tetramer is connected through a flexible AccE5 segment. This is preceded by a folded AccE5 segment that contributes to an asymmetric nine-stranded β−barrel with contributions from all four AccA3 subunits. This barrel allows the BC assembly to rotate ~90 ^o in the presence of acyl-CoA substrates. These data demonstrate that asymmetric, multi-substrate ACCases differ fundamentally from symmetric, single-substrate ACCases.

With the continuous rise in antibiotic resistance, novel methods that can reveal currently unknown antibiotic resistance mechanisms are essential to prepare and inform health responses and novel antibiotic discovery campaigns. Here, we built a library of species representative of the genus Mycobacterium and determined their antibiotic resistance profiles, allowing for the first time systematic multispecies comparisons. Analyzing antibiotic resistance in the context of other closely related yet diverse organisms revealed species with truly exceptional traits as well as general principles underpinning antibiotic resistance. Among these, we reveal that intrabacterial accumulation of antibiotics does not correlate with their potency at the species level. Our data also reveals that rifamycin resistance in mycobacteria is dominantly caused by antibiotic modification, contrary to what has been observed in Mycobacterium tuberculosis . Our data provides a solid starting point for the exploration of novel determinants of antibiotic resistance. We illustrate the utility of this species-level approach to discovery of novel traits by characterizing a previously unrecognized rifamycin-inactivating enzyme group that is present in a wide range of bacterial genera.

In this study, we investigated how iron limitation alters central metabolism in Mycobacterium tuberculosis using metabolomics and stable isotope tracing. Our findings reveal a well-orchestrated metabolic program to enable Krebs cycle activity despite the inefficient action of its iron-dependent enzymes. Under such conditions, carbon flux through the oxidative branch of the Krebs cycle is stalled, resulting in the accumulation of metabolites that are partially secreted. As a result, carbon flux from glycolysis is partially diverted to the reductive branch of the Krebs cycle to support the production of oxaloacetate and malate through the activity of phosphoenolpyruvate carboxykinase and pyruvate carboxylase. Both branches terminate with the synthesis of malate, which is secreted. This unprecedented split of the Krebs cycle and malate secretion in a bacterial pathogen facilitates the continuous flow of carbon through the core of carbon metabolism, overcoming the metabolic stalling triggered by iron starvation.

Isocitrate lyase 2 (ICL2) from Mycobacterium tuberculosis undergoes dramatic conformational rearrangements upon binding to the allosteric effector acetyl-CoA. Time-resolved cryo-EM captured conformational states along the ICL2 activation trajectory, revealing how acetyl-CoA binding at the allosteric sites leads to asymmetric, half-of-site activity at the catalytic centres. These findings support a conformational selection model of allostery, whereby acetyl-CoA binding shifts the pre-existing equilibrium towards an active state of the enzyme.

The increasing prevalence of antimicrobial resistance is an important challenge that warrants new approaches to antibiotic development. Currently, all antibiotics inhibit biological processes. To explore whether activation of a biochemical pathway can elicit bactericidal effects we engineered variants of Mycobacterium tuberculosis ATP-phosphoribosyltransferase (ATP-PRT) that are resistant to allosteric inhibition by l -histidine, leading to supraphysiological activation of ATP-PRT and l -histidine overproduction. Upregulation of L-histidine biosynthesis significantly reduces the growth of M. tuberculosis in culture and causes a loss of fitness owing to nutrient and energy depletion. Moreover, the expression of allosteric variants in M. tuberculosis significantly reduced infections in human macrophages and in a mouse model of infection. Thus, metabolic activation represents a new mycobactericidal mechanism that could be applied to antimycobacterial drug discovery.