Toxicology in the broadest sense is the study of the adverse effects of drugs or chemicals on living systems. The questions posed by this discipline include what compounds are toxic, how and why toxicity is manifested, and how might toxicity be predicted, treated, or prevented. Some of the toxicity effects are receptor-mediated, i.e. the toxic chemical binds to a receptor that initiates a cellular response. Ligand-induced modifications of the physicochemical properties or conformational changes of the receptor can trigger transcription processes or signal-transduction cascades. Computational toxicology aims to use rules, models and algorithms based on prior data for specific endpoints, to enable the prediction of whether a molecule is or is not toxic. Computational molecular modeling methods are at the core of mechanistic toxicology, allowing to understand the mechanisms through which a given chemical induces an adverse outcome pathway. We can say that the first step of the detoxification mechanism is the xenobiotic detection by the receptors, which include: Aryl hydrocarbon Receptor (AhR) and Pregnane X Receptor (PXR). Both receptors act as transcription factor activated by ligands. The functional domain responsible for ligand recognition is the Ligand Binding Domain (LBD). Xenobiotic receptors are highly promiscuous, because they have to bind a large variety of chemicals in order to eliminate toxic compounds, which can be very diverse. Promiscuity is often achieved increasing protein flexibility and plasticity. In case of AhR, we modeled the LBD by homology modeling and managed to include flexibility during both the docking step, and the post-docking step, by a short MD simulation. This new protocol allowed the identification of three groups of ligands, each binding in a specific site inside the cavity. The differences observed in the ligand-protein interactions could result in differential effects downstream in the AhR signaling pathway, thus these findings could help to explain the toxicity of some agonists. In case of PXR, given that many experimental structures are available, we firstly included flexibility using the ensemble docking approach with different X-ray structures. Results confirmed that often these structures are biased toward the native co-crystallized ligand geometry. In contrast, the explicit inclusion of flexibility using MD-based methods exhibited no initial bias and lead us to identify the entrance path of ligands to the PXR binding cavity and to rationalize the multiple SR12813 binding modes. Promiscuous receptors often display species-specificity; the modeling of these protein structures shed light on the molecular determinants of the observed different responses among different species. We built the first homology models of the invertebrate C. elegans and the amphibian G. multiplicata AhR LBDs. The first one showed a peculiar internal cavity, that is probably unable to bind any of the classical AhR ligands. The second helped in elucidating that the low-sensitivity of amphibians to TCDD arose before their divergence from the common lineage. Finally, the direct comparison between mouse/rat and human AhR LBDs highlighted the differences and similarities in binding to the different receptors of both agonists and selective modulators. The goal of computationally predict the potential activity of a ligand given a protein sequence or structure is clearly a big challenge, but we can conclude that molecular modeling, coupled with experimental techniques, is extremely useful to give mechanistic insights about biological and toxicological events. The elucidation of these events will greatly improve the possible application of in silico methods in ecological and human risk assessment. Moreover the understanding of the mechanism of action could contribute to a better estimation of interspecies scaling factor between animal models and human, increasing the reliability of prediction based on these data.

La tossicologia in senso ampio è lo studio degli effetti collaterali di farmaci o altri composti sui sistemi viventi. Le domande che questa disciplina si pone includono quali composti sono tossici, come e perché manifestano tossicità, e come questa tossicità possa essere predetta, trattata e prevenuta. Alcuni effetti di tossicità sono mediati da recettori, ossia il tossico lega un recettore che scatena una risposta cellulare. Le modificazioni, indotte dal ligando, delle proprietà chimico-fisiche o conformazionali del recettore possono innescare processi trascrizionali o cascate di trasduzione del segnale. La tossicologia computazionale mira a usare regole, modelli e algoritmi basati su dati con scopi specifici, per consentire il discernimento tra molecole tossiche e non tossiche. I metodi di modellistica molecolare sono al cuore della tossicologia meccanicistica, ossia che permette di comprendere il meccanismo attraverso il quale un composto espleta la sua tossicità. Possiamo dire che il primo passo del meccanismo di detossificazione è l’individuazione del tossico da parte dei recettori, questi recettori sono: Aryl hydrocarbon Receptor (AhR) e Pregnane X Receptor (PXR). Entrambi agiscono come fattori trascrizionali attivati da ligandi. Il dominio funzionale responsabile per il riconoscimento dei ligandi è il Ligand Binding Domain (LBD). I recettori degli xenobiotici sono molto promiscui, perché devono legare una vasta gamma di composti per poter eliminare quelli tossici, essi possono essere molto diversi. La promiscuità è ottenuta spesso aumentando la flessibilità e la plasticità della proteina. Nel caso di AhR, abbiamo modellato il LBD tramite homology modeling e incluso la flessibilità sia durante la fase di docking, sia durante il post-docking, grazie ad una breve simulazione di dinamica molecolare. Abbiamo identificato tre gruppi di ligandi, ciascuno che occupa un sito diverso all’interno della cavità di legame. Le differenze osservate nelle interazioni ligando-proteina potrebbero portare ad avere effetti diversi a valle del pathway di AhR, aiutando quindi a spiegare come mai alcuni composti sono tossici e altri no. Nel caso di PXR abbiamo incluso esplicitamente la flessibilità usando metodi basati su dinamica molecolare non ha mostrato alcun bias iniziale e ci ha permesso di identificare il path di ingresso alla cavità di legame e di razionalizzare le diverse geometria di binding di SR12813. I recettori promiscui spesso presentano anche una notevole specie-specificità; la modellazione di queste strutture ha chiarito i determinanti molecolari della diversità nelle risposte osservate in ciascuna specie. Abbiamo proposto il primo modello per AhR LBD dell’invertebrato C. elegans e dell’anfibio G. multiplicata. Il primo ha una cavità interna peculiare, che probabilmente non può legare i ligandi classici degli AhR vertebrati. Il secondo ha aiutato a capire che la bassa sensibilità degli anfibi alla diossina si è evoluta prima della divergenza dalla linea evolutiva comune. Infine, la comparazione diretta tra AhR di ratto/topo e umano ha sottolineato le differenze e le similitudini nel binding sia di agonisti, sia di modulatori selettivi. L’obiettivo di predire computazionalmente l’attiva di un ligando data la sequenza e/o la struttura di una proteina è chiaramente una grande sfida, ma possiamo concludere che la modellistica molecolare, unitamente a tecniche sperimentali, è estremamente utile per dare indizi meccanicistici riguardo eventi biologici e tossicologici. La delucidazione di questi eventi aumenterà di molto la possibilità di applicare metodi in silico nella valutazione del rischio ecologica e umana. Inoltre, la comprensione del meccanismo d’azione può contribuire a migliorare la stima dei fattori di scala interspecie tra animali modello (sia in vitro, sia in vivo) e uomo, aumentando l’affidabilità delle predizioni basate su questi dati.

(2019). Computational approaches to study binding of xenobiotic molecules to receptors. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2019).

Computational approaches to study binding of xenobiotic molecules to receptors

GIANI TAGLIABUE, SARA
2019

Abstract

Toxicology in the broadest sense is the study of the adverse effects of drugs or chemicals on living systems. The questions posed by this discipline include what compounds are toxic, how and why toxicity is manifested, and how might toxicity be predicted, treated, or prevented. Some of the toxicity effects are receptor-mediated, i.e. the toxic chemical binds to a receptor that initiates a cellular response. Ligand-induced modifications of the physicochemical properties or conformational changes of the receptor can trigger transcription processes or signal-transduction cascades. Computational toxicology aims to use rules, models and algorithms based on prior data for specific endpoints, to enable the prediction of whether a molecule is or is not toxic. Computational molecular modeling methods are at the core of mechanistic toxicology, allowing to understand the mechanisms through which a given chemical induces an adverse outcome pathway. We can say that the first step of the detoxification mechanism is the xenobiotic detection by the receptors, which include: Aryl hydrocarbon Receptor (AhR) and Pregnane X Receptor (PXR). Both receptors act as transcription factor activated by ligands. The functional domain responsible for ligand recognition is the Ligand Binding Domain (LBD). Xenobiotic receptors are highly promiscuous, because they have to bind a large variety of chemicals in order to eliminate toxic compounds, which can be very diverse. Promiscuity is often achieved increasing protein flexibility and plasticity. In case of AhR, we modeled the LBD by homology modeling and managed to include flexibility during both the docking step, and the post-docking step, by a short MD simulation. This new protocol allowed the identification of three groups of ligands, each binding in a specific site inside the cavity. The differences observed in the ligand-protein interactions could result in differential effects downstream in the AhR signaling pathway, thus these findings could help to explain the toxicity of some agonists. In case of PXR, given that many experimental structures are available, we firstly included flexibility using the ensemble docking approach with different X-ray structures. Results confirmed that often these structures are biased toward the native co-crystallized ligand geometry. In contrast, the explicit inclusion of flexibility using MD-based methods exhibited no initial bias and lead us to identify the entrance path of ligands to the PXR binding cavity and to rationalize the multiple SR12813 binding modes. Promiscuous receptors often display species-specificity; the modeling of these protein structures shed light on the molecular determinants of the observed different responses among different species. We built the first homology models of the invertebrate C. elegans and the amphibian G. multiplicata AhR LBDs. The first one showed a peculiar internal cavity, that is probably unable to bind any of the classical AhR ligands. The second helped in elucidating that the low-sensitivity of amphibians to TCDD arose before their divergence from the common lineage. Finally, the direct comparison between mouse/rat and human AhR LBDs highlighted the differences and similarities in binding to the different receptors of both agonists and selective modulators. The goal of computationally predict the potential activity of a ligand given a protein sequence or structure is clearly a big challenge, but we can conclude that molecular modeling, coupled with experimental techniques, is extremely useful to give mechanistic insights about biological and toxicological events. The elucidation of these events will greatly improve the possible application of in silico methods in ecological and human risk assessment. Moreover the understanding of the mechanism of action could contribute to a better estimation of interspecies scaling factor between animal models and human, increasing the reliability of prediction based on these data.
BONATI, LAURA
modellistica molecol; diossina; contaminanti; recettori; tossicologia
molecular modeling; TCDD; dioxin; receptors; tossicologia
CHIM/02 - CHIMICA FISICA
English
20-feb-2019
SCIENZE CHIMICHE, GEOLOGICHE E AMBIENTALI - 94R
31
2017/2018
open
(2019). Computational approaches to study binding of xenobiotic molecules to receptors. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2019).
File in questo prodotto:
File Dimensione Formato  
phd_unimib_720079.pdf

Accesso Aperto

Descrizione: tesi di dottorato
Dimensione 10.94 MB
Formato Adobe PDF
10.94 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10281/241321
Citazioni
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
Social impact