Mapping Mitochondrial Channel Toxicity: A Case Study for Predicting Mito-Target Interactions for the Per- and Poly-Fluoroalkyl Compounds on the Zebrafish Voltage-Dependent Anion Channel 2

Mapping Mitochondrial Channel Toxicity: A Case Study for Predicting Mito-Target Interactions for the Per- and Poly-Fluoroalkyl Compounds on the Zebrafish Voltage-Dependent Anion Channel 2 | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Mapping Mitochondrial Channel Toxicity: A Case Study for Predicting Mito-Target Interactions for the Per- and Poly-Fluoroalkyl Compounds on the Zebrafish Voltage-Dependent Anion Channel 2 Michael González-Durruthy, Amit Kumar Halder, Ana Silveira Moura, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4362510/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The significance of effective and reliable prediction of ecotoxicity, particularly across various trophic levels, including humans, is gaining increasing prominence as ecosystems face new threats and challenges. Computational ecotoxicological predictive approaches have already been deemed as a swifter and economical feasible answer. This work presents a new proposal in that context, integrating structure-based virtual screening and quantitative structure-activity relationship (QSAR) methodologies to address the ecotoxicity of per- and poly-fluoroalkyl substances (PFAS) in aquatic organisms, such as zebrafish. By focusing on the interaction between PFAS and the zebrafish mitochondrial voltage-dependent anion channel (zfVDAC2), resembling bioaccumulation in low concentrations, we analyzed 123 PFAS compounds. Our findings reveal that the top-ranked docked PFAS exhibits a predominant affinity for van der Waal interactions, followed by fluorine (F)-halogen bonds and hydrogen bonds interactions. The latter suggests that PFAS interaction strength may influence mitochondrial ATP transport via zfVDAC2. Similarly, the derived QSAR models identified packing density index, a descriptor linked to van der Waal interactions, as the most significant PFAS factor. Moreover, the high predictive power and statistical robustness of these models positioning them as valuable tools for environmental risk assessment in PFAS applications, while offering mechanistic insights into ecotoxicity. Environmental risk assessment Poly-fluoroalkyl substances Zebrafish mitochondrial voltage-dependent anion channel Virtual screening Quantitative Structure-Activity Relationships modeling Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1. Introduction Per- and poly-fluoroalkyl substances (PFAS), firstly defined by Buck et al . in 2011, are considered a highly fluorinated aliphatic family of compounds which are formed by one or more carbon atoms where all the hydrogen substituents have been replaced by atoms of fluorine. Structurally speaking and in general terms, the nomenclature can be represented as C n F 2n+1 being n ≥ 1, with the molecular structure containing at least one CF 3 – group (Alderete et al., 2019 ; Scheringer et al., 2014 ; OECD/UNEP, 2018 ; http://www.pops.int/ . 2004). Despite the current industrial value of PFAS, scientific evidence has raised concerns regarding their potential toxicity to wildlife and ecosystems, as well as to their potential role as inductive agents for physiological disorders on human metabolism, including diabetes, hypertension, and even various types of cancer (Bischel et al., 2018; Poothong et al., 2017 ; Brendel et al., 2018 ). A major characteristic of PFAS toxicity is their persistence in biological systems and the environment for extended periods in their unchanged forms. A recent investigation revealed that some PFAS not only accumulate but also are bio-transformed extensively into a wide range of structurally diverse chemicals capable of persisting for long time in the environment. (Tamara et al., 2021; Andersen et al., 2006 ) Furthermore, the continuous release of PFAS contaminants in the aqueous ecosystems can cause a secondary contamination of natural water leading to the spread of contamination in water levels of parts per trillion (Bischel et al., 2018; Wang et al., 2015 ; Krafft et al., 2015; Bowman et al., 2015; Butenhoff et al., 2012 ). In fact, according to the Stockholm Convention, PFAS have been considered as one of the most controversial among the category ‘persistent organic pollutant’ (POP) for the anthropologically manufactured worldwide environmental contaminants ( http://www.pops.int/ , 2004). As such, the need for sophisticated bio-remediation strategies in order to mitigate their environmental impact is urgent (Scheringer et al., 2014 ). Indeed, proper knowledge regarding PFAS-bioaccumulation parameters alongside with appropriate environmental risk assessment models that addressed PFAS ecotoxicity (in particular when main food-chains are in question) are still amiss (Vestergren et al., 2009). In particular, this knowledge is absent for a wide variety of these main-food chains organisms, such as bacteria, algae, crustaceans, ciliates, fish, yeasts and nematodes. Nevertheless, research has evidenced a direct association between the bioaccumulation potential of PFAS and the length of their perfluoroalkyl chain, as PFAS analogs of longer chains, with at least six carbons, presented a higher bioconcentration potential (Bischel et al., 2018; Poothong et al., 2017 ; Brendel et al., 2018 ; Tamara et al., 2021; Andersen et al., 2006 ). These chemical attributes are related with increasing concentration of environmental pollution the PFAS cause through their exposure routes, which could involve different levels of trophic chains, including humans, and diverse environmental compartments, such as water, soil or air (Wang et al., 2015 ; Krafft et al., 2015; Butenhoff et al., 2012 ). An interesting bio-target for a PFAS ecotoxicological predictive/assessment model in aqueous ecosystems is the zebrafish ( Danio rerio ), since the zebrafish model can be efficiently used to predict/study mechanisms of toxicity for both the macroscopic and cellular level of this organism regarding PFAS induced damages (Wang et al., 2015 ; Krafft et al., 2015; Bowman et al., 2015; Butenhoff et al., 2012 ; Vestergren et al., 2009; Hagenaars et al., 2013 ; O'Brien et al., 2004; Mashayekhi et al., 2015 ). Interestingly, among the most PFAS – induced macroscopic damages in the zebrafish models, reports include the endocrine disruption (thyroids), altered swimming, developmental and reproductive toxicity in sub-lethal doses. Furthermore, subcellular and biochemical level damages include cytotoxicity, perturbations in lipid metabolism, oxidative stress leading to genotoxicity, and mitochondrial dysfunction, such as ATP bioenergetics perturbations, and channel mitotoxicity. A hypothesis that we propose in this paper, departing from the analysis of the published works (Wang et al., 2015 ; Krafft et al., 2015; Bowman et al., 2015; Butenhoff et al., 2012 ; Vestergren, et al., 2009; Hagenaars et al., 2013 ; O'Brien et al., 2004; Mashayekhi et al., 2015 ), is that PFAS could be bioaccumulated at mitochondrial level ( i.e. , at the mitochondrial matrix) after interacting with the zebrafish voltage-dependent anion channel 2 (zfVDAC2), or selectively induce significant perturbations in the native structure and physiological function of zfVDAC2, or selectively induce significant perturbations in the native structure and physiological function of zfVDAC2 by triggering mitotoxic events directly associated to the opening of the mitochondria permeability transition pore-based zfVDAC2 channel dysfunction (zfVDAC2-MPTP). The latter is directly involved in ROS induction, cardiolipin peroxidation, ADP/ATP transport perturbations, cytochrome c release, decreased ATP levels. From a phylogenetic point of view, and regardless of the animal species under study, it is considered that VDAC channel (zfVDAC2) is highly conserved and is involved in the mitochondrial bioenergetics and homeostasis regulation, as well as in the control of intracellular metabolism (Vestergren et al., 2009; Choi et al., 2017 ). With this accepted evidence as cornerstone, we explore a computational model for the mitochondrial channel dysfunction as an eco-toxicodynamic mechanism in zebrafish. In addition, this departing concept is sustained by the zfVDAC2 being the most abundant and densely localized channel protein in the outer mitochondrial membrane of all cells in the zebrafish (Dreier et al., 2019 ), thus representing the main route of communication between the mitochondrial intermembrane space and the cytosol. It should be noted that the zfVDAC2 channel has an N-terminal α-helix segment located horizontally inside the pore, which is aligned nearly parallel to the membrane plane, causing a partial narrowing of the zfVDAC2 pore (Shimizu et al., 2015 , Choi et al., 2017 ). In particular, the N-terminal α-helix segment can adopt different conformations within the voltage-gating channel depending on the internal/or external stimuli (Shimizu et al., 2015 , Choi et al., 2017 ). Further, the importance of this channel covers many mitochondrial essential aspects, such as ideal equilibrium dynamics of the intracellular [Ca 2+ ], or the ADP plus Pi/ATP influx/efflux transport which favor key biochemical functions as glycolysis and ATP-bioenergetics regulation (Shimizu et al., 2015 , Choi et al., 2017 , Zafeer et al., 2018 , Liao et al., 2020 , Girdhar et al., 2020 ). Notwithstanding, the fact is that there is a data gap regarding the biological hazardousness factor per PFAS group and versus specific bio-targets (Cousins et al., 2020 , Cousins, et al., 2022 ), something that is even more noticeable when considering mitochondria channel toxicity versus zfVDAC2 toxicity, as the evidence scenario, either experimental or in silico, is scarce if not inexistent. As such, within the in silico model proposed in this work regarding PFAs versus zfVDAC2 toxicity, validation versus experimental evidence can only be achieved when there is availability of such results. Therefore, the focus will be in the intrinsic in silico validation methodologies, that have been proven effective in other ecotoxicological approaches that were also validated versus experimental evidence. The development of an in silico ecotoxicological predictive technique in PFAS environmental risk assessment, which is currently rather limited to our best knowledge, not only offers great advantages in time and money savings but is also aligned with the need of implementing effective predictive animal-free testing methodologies (Schredelseker et al., 2014 ; Cheng et al., 2019; Sima et al., 2021). Furthermore, in silico methods are strongly recommended in the very early stages of the predictive toxicological (ecotoxicological) investigation, only to be followed by in vitro and in vivo experimental validation after the computational and theoretical assessment. As such, a synergy between classic computational approaches as structure-based virtual screening (SB-VS) and predictive Quantitative Structure-Activity Relationships (QSAR) modeling could be efficiently implemented to precisely assess the potential ecotoxicological effects of PFAS in aquatic organisms (Vestergren et al., 2009; Choi et al., 2017 ; Ankley et al., 2021 ; Schredelseker et al., 2014 ; Cheng et al., 2019). These in silico methodologies have demonstrated the ability to simultaneously evaluate the molecular mechanisms of interaction of a large number of compounds, including PFAS, while maintaining sensitivity, specificity, and accuracy to realistically reproduce the key interactions of traditional or emerging pollutants (Vestergren et al., 2009; Choi et al., 2017 ; Ankley et al., 2021 ; Schredelseker et al., 2014 ). In fact, while molecular docking software allows for the rapid calculation of docking scores, the integration of the 3D-QSAR approach offers valuable additional insights and complements the docking analysis in our study, namely in two major vectors: the identification of relevant specific 2D molecular descriptors, and a wider range of prediction. Regarding the first vector, while the identification of specific 2D molecular descriptors, i.e. , structural features of molecules in a two-dimensional representation, allows to gain insight into the molecular properties that may contribute to the observed activity or properties of the compounds, the 3D-docked poses, i.e. , the predicted spatial arrangements of the PFAS ligand molecules within the binding site of the target protein obtained through molecular docking simulations, represent potential binding configurations and help assess the binding affinity between the ligand and protein. Though the docking score obtained from the 3D-docked poses serves as an indicator of the binding affinity between the PFAS compounds and the zfVDAC2 target channel, it does not provide detailed information about the specific structural physico-chemical based molecular descriptors responsible for the PFAS binding activity on the zfVDAC2 channel. Therefore, the recourse to the QSAR methodology aims to elucidate the structure-mitotarget interactions activity relationship of the PFAS compounds on zfVDAC by identifying relevant specific 2D molecular descriptors derived from the 3D-docked poses. Regarding the second vector, the proposed QSAR model will allow us to quantitatively predict the ecotoxicological properties of a larger set of PFAS compounds beyond those specifically docked in our study. This broader perspective enhances our understanding of the ecotoxic potential of various PFAS analogs, guiding future experimental efforts and providing insights for environmental risk assessment in the context of an ecotoxicological target as zfVDAC2. In addition, employing the QSAR model enables the analysis of the contributions of individual PFAS molecular descriptors and their respective weights, providing a more detailed understanding of the underlying molecular mechanisms governing PFAS ecotoxicity. The present study proposes an effective methodology for assessing the molecular interactions between per- and poly-fluoroalkyl compounds and the zfVDAC2 channel protein. This is achieved through a synergistic approach combining structure-based virtual screening and 2D/3D-QSAR models, introducing a novel hybrid in silico risk assessment technique. This approach represents a qualitative advancement in ecotoxicological predictive techniques within the field of Computational Toxicology. Notwithstanding, the framework of the proposed approach is overall supported by well-recognized structural and thermodynamic techniques to delve into the potential mitochondrial toxicity response induced by PFAS compounds in the zebrafish. In addition, considering the importance of understanding mitochondrial dysfunctions in generating hypotheses across various biological levels, from molecular mechanisms to population health, in the context of Environmental Pollutants (Dreier et al., 2019 ), this work not only presents novelty but also holds significant relevance. 2. Methodology section The methodology had the stages that follow: (i) Selection of PFAS compounds and prediction of zfVDAC2 binding-sites; (ii) Setting up the flexibility profile of zfVDAC2; (iii) Performing structure-based virtual screening on PFAS compounds with zfVDAC2; and (iv) Performing QSAR modeling. 2.1. Selection of PFAS compounds and prediction of zfVDAC2 binding-sites. As stated previously, the field of PFAS interaction with mitochondria toxicity is an emergent field, and there is still absence to clear, precise, or mandatory criteria regarding the selection of PFAS for a dataset aiming to develop an in silico ecotoxicological tool for the mitochondrial channel toxicity, namely the voltage-dependent anion channel 2 (VDAC2) mitochondrial channel toxicity. Therefore, aiming to be a first approximation to the ecotoxicological profile of PFAS interaction with mitochondria channels, the criteria for the selection of the PFAS focused the ecotoxicological evidence presented by EPA and OECD lists of PFAS chemicals. As such, a representative dataset, with well-known 123 PFAS compounds, was selected, corresponding to a combination of two subsets (Chelcea et al., 2020 ): i) a 75 PFAS list of a EPA prioritized subset (Patlewicz et al., 2019 ) of a larger chemical inventory ( https://comptox.epa.gov/dashboard/chemical_lists/EPAPFAS75S1 ) and the remaining 48 belonging to ii) a list of PFAS chemicals in the OECD New Comprehensive Global Database ( https://comptox.epa.gov/dashboard/chemical_lists/PFASOECD ), as per the indications of OECD recommendations (OECD, 2018). It should be noted that both selections followed rigorous criteria in terms of nomenclature for commonly recognized per- and poly-fluoroalkyl substances and, as such, other highly fluorinated substances that match the definition of PFASs, but still have not been recognized as such by the OECD were excluded from the dataset in the present study. Nevertheless, these could be included in future predictive purpose studies once they are recognized by the OECD, if considering the PFASs nomenclature and also considering the potential mitochondrial toxicity (OECD, 2022 ; OECD 2015). The prediction for the best-ranked zfVDAC2 binding-site resourced to DeepSite (Jimenez et al., 2017 ). The DeepSite tool considers all the molecular descriptors related to the proteins zfVDAC2 by using a machine-learning algorithm based on 3D-deep convolutional neural networks (DCNN) (Jimenez et al., 2017 ). Moreover, DeepSite has an extensive validation test set with > 6500 proteins belonging to the scPDB database, thus allowing for unequivocal prediction of the main catalytic binding sites in zfVDAC2. One of the most crucial steps to ensure quality of modeling results is the accurate prediction/identification of the suitable zfVDAC2 channel binding-sites coupled with an appropriate crystallographic structural validation, and the flexibility properties of the targeted protein. By itself, the search space that was set up for the zfVDAC2 channel using the fpocket tool within the DeepSite validation procedure (Le Guilloux et al., 2009 ; Jimenez et al., 2017 ) is a cubic grid box with a size of 20 × 20 × 20 Å 3 , centered at X = −20.35 Å, Y = + 21.99 Å, and Z = + 6.13 Å, and a discretization of 0.25 Å. To begin with, predictions of possible zfVDAC2 pocket binding-sites the Voronoi tessellation algorithm was applied to detect small cavities in zfVDAC2, such as protuberances, and collecting information-based crystallographic descriptors on the pocket geometry and topology associated with high ligand binding probability (Jimenez et al., 2017 ). Such algorithm allows to gather the cartesian coordinates like volumetric maps of possible binding-site pockets on the zfVDAC2 channel. Both the zfVDAC2 mesh volumetric maps gathered, along with the best-ranked binding site prediction obtained from DeepSite, were used to define the search space ( i.e. , the cubic grid box) for then carrying out the structure-based virtual screening (Jimenez et al., 2017 ). After this, a structural validation of the zfVDAC2 channel was carried out by a Ramachandran analysis (Chen et al., 2010 ). In so doing, false positives for the docking zfVDAC2-PFAS generated complexes are prevented by verifying the absence of restricted flexibility of the residues of each zfVDAC2 binding site found. 2.2. Setting up the flexibility profile of zfVDAC2. Elastic Network models (ENM) were used to determine and characterize the flexibility properties of zfVDAC2. This computational algorithm allows to evaluate the directions of movement associated with elastic normal modes of fluctuations by representing the C-(α)-atoms belonging to the whole zfVDAC2 as a residues network communication according to a Hookean potential (or “springs”). A more in-depth description of the ENM algorithm can be found elsewhere (Yang et al., 2018; Tama et al., 2008; Chennubhotla et al., 2007; Guedes et al., 2021 ). In the present study, the zfVDAC2 flexibility profile was set up by considering all the flexibility fluctuation patterns in which the molecular structure of the zfVDAC2 protein will oscillate around its equilibrium conformation and collective motions of its residues network in the unbound state. For more clarity, only the results of the following simulation conditions are presented: (i) unbound zfVDAC2 as the control simulation experiment; (ii) zfVDAC2-ATP as reference control of simulation, since the ATP molecule represents the native substrate of the zfVDAC2 channel and is used here for comparison purposes with (iii) and (iv); (iii) the top-ranked zfVDAC2-PFAS docking complex − i.e. , as maximum level of zfVDAC2 ecotoxicity; (iv) the worst-ranked zfVDAC2-PFAS docking complex − i.e. , as minimum level of zfVDAC2 ecotoxicity. Additionally, the zfVDAC2 flexibility properties, collective motions like anisotropic fluctuations generated from the zfVDAC2 inter-residue network and communication efficiency within, were also evaluated from the whole zfVDAC2 structure (Tama et al., 2008; Chennubhotla et al., 2007; Guedes et al., 2021 ). 2.3. Performing structure-based virtual screening on PFAS compounds with zfVDAC2 The next step is to get the free energies of binding (FEB) for the complexes formed between the zfVDAC2 protein and the PFAS family of compounds by virtual screening. As human health has been the main concern when probing the PFAS molecular mechanisms, in order to reproduce with a certain level of precision the PFAS molecular interactions with the zfVDAC2 channel, our structure-based virtual screening results mimic low-concentration conditions for each compound. This is achieved since these interactions are between one PFAS molecule and one zfVDAC2 protein and not involving potential ligand-ligand interactions (Laskowski et al., 2011), i.e. , PFAS plus PFAS interactions, as it would happen in mixtures released into the complex aquatic environment at high-concentrations. The DockThor virtual screening platform (Guedes et al., 2021 ) was employed for docking the targeted PFAS within the zfVDAC2 channel. The DockThor-VS code makes use of an accurate phenotypic crowding-based multiple solution steady-state genetic algorithm as searching tool. The latter allowed us to identify multiple minima solutions in a single docking run, keeping the population diversity of the PFAS ligand structures. Default parameters were set on each of these runs, as the number of evaluations (500000 evaluations per docking run), the population size (750 individuals), the number of runs (24), and maximum of cluster leaders (20). The scoring function applied in DockThor (Guedes et al., 2021 ) to score the various docked poses is based on computing their total affinity ( E total , kcal/mol) as a sum of intermolecular and intramolecular interaction energy terms from the MMFF94S49 force field. All the needed interaction parameters for both zfVDAC2 and the PFAS ligands were explicitly set up by resorting to the DockThor tools MMFF-Ligand and PdbThorBox (Guedes et al., 2021 ). The total affinity values obtained for the docking poses were then categorized like energetically unfavorable when ∆ G bind ≥ 0 kcal/mol, indicating a complete absence of affinity of zfVDAC2 to that particular PFAS structure; otherwise, those were categorized as favorable interactions. In addition, affinity values less than the one of ATP with zfVDAC2 ( E total = −7.4 kcal/mol) were classified as strong docking interactions considering that ATP is its native substrate. 2.4. Performing QSAR modeling The 3D-docked poses ( i.e. , docked poses with the highest docking score) of the PFAS were used as input structures for calculating 2D descriptors using the Dragon software (Mauri et al., 2006 ). The PFAS dataset was randomly divided into a training set (80%) and a test set (20%) with help of our in-house tool SFS-QSAR (available to download at https://github.com/ncordeirfcup/SFS-QSAR-tool ) (Halder et al., 2022 ). Details about these PFAS sets are given in Table S1 of the Supporting Information (SI). A multiple linear regression (MLR) approach was adopted for setting up the 2D-QSAR models, by applying as variable selection procedure either the sequential forward selection algorithm (Raschka S., 2018 ) (SFS) or the genetic algorithm (GA) − the latter using the open-source code NanoBRIDGES (Ambure et al., 2015 ). In both cases, pre-treatment of descriptors was carried out with a correlation (Pearson R ) cut-off of 0.95 and variance cut-off of 0.001. Selection of SFS-MLR models was done by examining usual statistical parameters − i.e. : the determination coefficient ( R 2 ) and the negative mean absolute error (NMAE), and the variables selected without cross-validation (0-fold CV) or by 5-fold CV. As to the GA-MLR models, these were generated by applying default parameters as the mutation probability (= 0.3) and the total number of iterations per run (= 100), being the best final GA-MLR model selected from 20 of these runs. The quality of both type of 2D-QSAR models was evaluated by well-known statistical metrics such as the R 2 coefficient, adjusted R 2 ( R 2 Adj ), and the Fisher’s ratio ( F ) (Gramatica P., 2007 ). Similarly, the commonly used leave-one-out (LOO) cross-validation regression coefficient ( Q 2 LOO ), along with two r m 2 metrics ( i.e .: r m 2 (LOO) and its associated standard deviation D r m 2 (LOO) ; see (Eq. 1 ) and (Eq. 2 ) below) were employed to estimate their internal predictivity and analogous metrics ( i.e .: R 2 Pred , r m 2 (test) , and Δ r m 2 (test) ) to evaluate their external predictivity (Gramatica P., 2007 ; Gramatica P., 2013 ; Tetko et al., 2001 ; Roy et al., 2013 ; Golbraikh et al., 2002). $${{{r}^{2}}_{m}}_{\left(\text{L}\text{O}\text{O}\right)}= {R}^{2}(1-\sqrt{({R}^{2 }-{R}_{0}^{2})})$$ 1 $${{{\varDelta r}^{2}}_{m}}_{\left(\text{L}\text{O}\text{O}\right)}=|{R}_{m}^{2}-{R}_{m}^{2}{\prime } |$$ 2 In Eq. 1, R 2 and \({R}_{0}^{2}\) are the determination coefficients between the observed and the LOO-predicted scores of the compounds on the training dataset without and with the intercept set to zero, respectively. In Eq. 2 , for \({ R}_{m}^{2}\) , the observed values are considered in the Y -axis and the predicted values in the X -axis whereas for \({R}_{m}^{2}{\prime }\) , the values of these axes are exchanged. The inter-collinearity of the 2D-QSAR model’s descriptors were checked by examining the cross-correlation matrix. To further assess the multi-collinearity of the model, the variation inflation factor (VIF) of each descriptor was also computed using the following equation (Yoo et al., 2014 ): $$VIF=1/(1-{R}_{i}^{2})$$ 3 where, \({R}_{i}^{2}\) is the correlation coefficient ( R 2 ) determined by regressing the i th descriptor on the other descriptors. Additionally, the Y -based randomization technique was applied (1000 response randomization runs), and the metric c R P 2 computed to infer the uniqueness of the models (Ojha et al., 2011). Lastly, the applicability domain of the 2D-QSAR models − i.e ., the response and data structure space in which the models afford predictions with a given reliability, was assessed by plotting the leverage vs . standardized residuals for each dataset case – the so-called William’s plot, which allows to detect possible structural and response outliers (Gramatica P., 2007 ; Serra et al., 2020 ). The Open3DQSAR tool (Tosco et al., 2011) was used for setting up predictive 3D-QSAR models using the same PFAS training and test sets combination. This software uses a carbon and volume-less positively (+ 1) charged probe for calculating steric and electrostatic fields of the query chemicals, respectively. The pre-treatment of the fields was carried out after setting a smart region definition (SRD)-cut-off level of 2.0 and also by removing N -level variables − i.e. : removal of variables with less variations in order to avoid biasing of the model (Tosco et al., 2011). The Open3DQSAR uses SRD for variable grouping, based on the closeness of variables in 3D space, as well as two different variable selection algorithms, namely: Fractional Factorial Design-based variable selection (FFD-SEL), and the Uninformative Variable Elimination-based partial least square (UVE-PLS) (Tosco et al., 2011; Halder et al., 2017 ). The predictive quality of the 3D-QSAR-based PLS models generated from the PFAS was examined using the same statistical metrics as the ones referred to above for the 2D-QSAR models. 3. Results and Discussion 3.1. Structure-based virtual screening on zfVDAC2-PFAS interactions The identification of the best ranked zfVDAC2 binding sites followed according to the details in the Methods and model section. The results pertaining to the zfVDAC2 flexibility properties, collective motions like anisotropic fluctuations generated from the zfVDAC2 inter-residue network and communication efficiency are depicted in Fig. 1 . The crystallographic validation of the zfVDAC2 channel 3D X-ray structure ensued by applying the Ramachandran analysis with quality assessment, as displayed in Supplementary Figure S1 . Ramachandran analysis corresponds to a 2D-projection onto the plane from the 3D-crystallographic structure of zfVDAC2 (pdb model) where all its possible residue conformations are defined in the plot based on the dihedral angles ( φ and ψ ) around their peptide-bonds (Chen et al., 2010 ). As such, the allowed φ vs. ψ torsion values of one particular residue, found within the Ramachandran contour (colored purple; see Figure S1 ) were taken as conformational-favored residues of zfVDAC2. Note that the zfVDAC2 sterically disallowed residue LEU 227, identified as a conformationally non-favored residue, is not part of the predicted best-ranked binding site that thus ensures the absence of false positive VS results. Figure 2 . In the top, A) van der Waals representation of the predicted top-ranked binding site for zfVDAC2 in its unbound state (depicted in blue). B) Representation of the zfVDAC2-ATP complex, simulating the theoretical physiological condition. C) Docking complex of zfVDAC2 with PFAS#25, the best-ranked among the docking complexes. D) Docking complex of zfVDAC2 with PFAS#29, ranked lowest among the docking complexes. In the bottom, E) , F) , and G) depict corresponding 2D-lig-plot diagrams illustrating interactions within these docking complexes. For B) - D) , regions depicted in 3D van der Waals representations labeled in blue represent high solvent accessibility surfaces, those in green denote low solvent accessibility surfaces, and those in gray indicate moderate solvent accessibility surfaces. Analyses on these outputs point to a general tendency of the tested PFASs, whether with short or long chain length, to being able to interact in the predicted binding site placed in the middle of the zfVDAC2 channel. Further, the interaction occurs after a spontaneous thermodynamics process, according to the negative FEB values of binding affinity obtained in all the cases (refer to Table S1 ). Focusing on the best ranked PFAS#25 (Fig. 2 F) and the worst ranked PFAS#29 (Fig. 2 G) docking complexes, a prevalence of van der Waals interactions was identified between the evaluated PFAS ligands and the zfVDAC2 with approximately the same number of van der Waals contacts (#vdW as light-green spheres). The identified PFAS ligands present approximately the same number of van der Waals contacts (#vdW as light-green spheres) with relevant amino acid residues of zfVDAC2 binding site as PFAS#25 (#vdW = 6) ≈ PFAS#29 (#vdW = 5). Though the difference between these numbers is small, it carries a significant impact from a quantitative perspective since the most relevant contribution is provided by the binding properties of the interacting residues involved, such as the druggability − i.e ., the ability to bind ligands. Albeit the best-ranked and worst-ranked PFASs both interact at the same biophysical environment in the zfVDAC2 binding site than the ATP molecule − i.e ., the natural zfVDAC2 substrate (Hagenaars et al., 2013 ; O'Brien et al., 2004; Mashayekhi et al., 2015 ; Ankley et al., 2021 ; Schredelseker et al., 2014 ), they share key ATP-binding residues that are responsible for the ADP/ATP influx/efflux through the zfVDAC2 channel. Indeed, PFAS#25 interacts with residues GLY191, GLY192, ALA209, LEU242, and VAL273, while PFAS#29 interacts just with the HIS181 residue of protein zfVDAC2, thus being the first more directly perturbed by these ATP-binding residues. The latter explains the difference between the obtained van der Waals energy values for complexes with these PFAS ligands, namely: E vdW (zfVDAC2-PFAS#25) = −17.303 kcal/mol; E vdW (zfVDAC2-PFAS#29) = + 0.304 kcal/mol. It should be noted here that the FEB was decomposed into different energy/entropy items by applying the equations used in the work of González-Durruthy et al., namely equations 3 and 4 (González-Durruthy et al., 2017 ). The higher number of fluorine bond interactions found for the best-ranked PFAS, PFAS#25 (see Fig. 2 E), contrasts also to the just two (F)-halogen-bond interactions found for the worst-ranked ligand, PFAS#29 (with residues LEU10 and TYR7) (Schredelseker et al., 2014 ). It is noteworthy that TYR7 and LEU10 residues do not seem to have a direct influence on the inhibition of the ATP-translocation through the zfVDAC2 channel, which could be a consequence of those not being involved in the stabilization of the ATP molecule during the passage across such channel. These differences in the number and type of halogen-bond interactions have great relevance, since those embody the eco-toxicophoric moiety of the PFAS family and give rise to the ecotoxicity based-affinity for complex zfVDAC2-PFAS#25 (= −8.207 kcal/mol) being slightly higher than that of the zfVDAC2-ATP complex (= −7.400 kcal/mol), either for the evaluated zfVDAC2 target and other protein targets potentially exposed to PFAS. When comparing the VS score-based affinity values of the generated docking complexes for long-chain PFAS compounds, typically containing ≥ 6 carbons (as in the case of the best-ranked complex zfVDAC2-PFAS#25), with corresponding short-chain PFAS compounds, there is evidence that the length of the chain of per-and-polyfluoroalkyl is a key eco-toxicophoric attribute inducing more strong inhibition on the zfVDAC2 channel (Bischel et al., 2018; Poothong et al., 2017 ; Brendel et al., 2018 ). As seen in Fig. 2 B, the extended conformation of the long-chain PFAS compounds within the zfVDAC2 binding site could significantly affect the correct fit of ATP, if both molecules would interact simultaneously in the same biophysical environment (see Supplementary Figure S2 ). Regarding the hydrogen-bond interactions, we identify the presence of a high degree of stabilization for the best-ranked PFAS vs. the worst-ranked PFAS docking complexes originated from three classical H-bond interactions involving the ATP-binding residues. The latter are residues ASN183, ASN207, and LEU208 in the case of PFAS#25, whereas only two conventional H-bond interactions are detected with non-target ATP-binding residues in the worst-ranked PFAS#29 ( i.e. : LEU10 and the regulatory-phosphorylation residue TYR7). Additionally, the presence of one non-classical C-H bond interaction from PFAS#25 with the ATP-binding residue (LYS236) might affect the correct ATP conformational-fit of the adenosine-moiety into the zfVDAC2 binding site, which could interfere with the voltage-sensing N-terminal α-helix segment that regulates the ATP-efflux translocation from the mitochondrial matrix to the cytosol through zfVDAC2 channel. Indeed, this perturbation near to the N-terminal α-helix segment could theoretically induce changes in the zebrafish, such as mitochondrial swelling or the electronic transport chain in zebrafish mitochondrial complex (I, III, IV), compromising its physiological process of ATP/ADP cycles through the zfVDAC2 channel (Ankley et al., 2021 ; Schredelseker et al., 2014 ; Choi et al., 2017 ). 3.2. zfVDAC2 inter-residue network communication efficiency To probe how the communication efficiency in the inter-residue network of zfVDAC2 in the unbound and bound state, PFAS#25 and PFAS#29, an anisotropic ENM was implemented by considering local perturbations induced by the ligands in the zfVDAC2 residue network ( Cf. Figure 1 , E and F ). The results are displayed in Fig. 3 . The local perturbation response scanning (LPRS) map approach tries to offer a conceptual vision on the local perturbations induced by the PFAS ligands interacting with zfVDAC in the bound state, which is a novelty result regarding this context. The local perturbations are defined by the collective anisotropic fluctuations in the intra-segment C α -C α atomic residue distances by the presence of a given ligand, such as ATP or PFAS, that could be biochemically relevant or unfavorable from a structural point of view whenever affecting the binding site native conformation and function of the zfVDAC2 (Tama et al., 2008; Chennubhotla et al., 2007; Guedes et al., 2021 ; Lee et al., 2017 ). Thus, the anisotropic changes induced by the ligands ATP, long-chain PFAS#25, and short-chain PFAS#29 in the inter-residue network of the zfVDAC2 were determined using the LPRS maps focusing on the zfVDAC2 binding site − i.e ., on the voltage-sensing N-terminal α-helix segment (marked as white arrows in Fig. 3 ). The physiological patterns generated by the ATP molecule when interacts with the zfVDAC2 target sensor residues − yellow to dark-red color in the corresponding LPRS map, promote significant conformational chain flexibilization from large blocks of such consecutive residues, contrasting with the labeled-blue regions of the LPRS map that denote the presence of blocks of consecutive zfVDAC2 effector residues with conformational rigidification. Comparing the ATP bound vs. unbound zfVDAC2 LPRS maps, the proper fit of the ATP molecule is assured. In Fig. 4 , panel B, the result of the local perturbation induced by the ATP molecule on the zfVDAC channel corresponds to the modeling of the physiological binding condition. As such, the binding interaction of the ATP intra the zfVDAC channel can be considered as a reversible biochemical process when acknowledging that the ATP necessarily transits the interior of the channel interacting with key residues in a reversible way (González-Durruthy et al., 2020 ). Therefore, the theoretical physiological patterns generated by the ATP molecule when interacts with the zfVDAC2 target sensor residues - yellow to dark-red color in the corresponding LPRS map, promote significant conformational chain flexibilization from large blocks of such consecutive residues, contrasting with the labeled-blue regions of the LPRS map that denote the presence of blocks of consecutive zfVDAC2 effector residues with conformational rigidification. Comparing the ATP bound vs . unbound zfVDAC2 LPRS maps, the proper fit of the ATP molecule is assured. Regarding the bound state with the evaluated PFAS, both ligands evidence promotion of local perturbations in the zfVDAC2 binding site, i.e .: potential ecotoxicity, which are mainly associated to non-covalent interactions (see Fig. 2 , E, F, and G ) (Laskowski et al., 2011), though a different pattern can be observed considering the length of their chains. In the case of the long-chain PFAS compound − PFAS#25, the local perturbations ( i.e. , interactions with the target zfVDAC2 sensor residues) can potentially activate similar blocks of consecutives zfVDAC2 residues. This strongly suggests that the extended docked conformation of the long-chain PFAS compounds within the zfVDAC2 binding site could significantly affect the correct fit of the ATP molecule if it is considered that both molecules could interact at the same biophysical environment partially sharing the same binding residues (see Fig. 3 , B and C ), something also fitting with the results obtained in the 2D-lig-plot interaction diagrams (refer to Fig. 2 E, F and Supplementary Figure S2 ). For the case of the worst-ranked PFAS, the short-chain PFAS#29, the local perturbations induce just light modifications in the inter-residue network of the zfVDAC2 binding site ( Cf. Figure 2 D), showing a pattern remarkably close to that obtained for the unbound zfVDAC ( Cf. Figure 2 A), a result which also fits well with the corresponding 2D-lig-plot interaction diagram ( Cf. Figure 2 G). Final validation of the results from the LPRS maps was the differences found between the topology of the anisotropic network of signal communication efficiency of zfVDAC2 (Fig. 4 ) . Further, the obtained LPRS maps for the unbound state and bound states ( Cf. Figure 3 A and 3 B-D, respectively) are in both cases strongly dependent on the zfVDAC2 flexibility properties, which are directly associated with the allosteric anisotropic network topology ( vide Fig. 4 A-C). This allosteric anisotropic network topology takes place in the inter-residue communication network formed by the backbone–backbone atom contacts between two sequence-adjacent residues in the evaluated region (zfVDAC2 binding site) under the unbound and bound states, where a different pattern of allosteric signal-based local flexibility perturbations was identified involved sensor-effector residues ( i , j ) of zfVDAC2 (Schredelseker et al., 2014 ; Tama et al., 2008; Chennubhotla et al., 2007; Guedes et al., 2021 ; Lee et al., 2017 ). As such, slight changes in the residue environment of the binding site could theoretically affect the molecular mechanisms of the ATP complementarity (or functional coupling) in the zfVDAC2 pocket and the mitochondrial ATP-transport in zfVDAC2 in the presence of PFAS compounds. Finally, it should be noted the option for resourcing to local perturbation response scanning (LPRS) maps versus molecular dynamics (MD) simulations for evaluating PFAS-induced conformational changes in the mitochondrial zfVDAC2 channel resulted from LPRS maps can be a preferable choice over MD simulations, though these are commonly used to study conformational changes in biomolecules, in a situation as the one of the present work, where localized insights, computational efficiency, and quick exploratory analyses are priorities. This explains why LPRS maps have been successfully applied in several studies to investigate conformational changes in proteins. 3.3. 2D-QSAR model The generated 2D-QSAR models attempt to explore quantitative relationships between the targeted ecotoxicological endpoint, i.e. , the binding properties of the evaluated ligands under interaction with the zfVDAC2 channel (affinity score ≈ FEB), and the PFAS molecular structures plus physicochemical properties. The latter were described by a limited number of 2D descriptors obtained from the PFAS ligands, namely: constitutional, ring, 2D functional group counts, atom-centered fragments, CATS2D, 2D-atom pairs, molecular properties, and drug-like indices, in order to assure 2D-QSAR models with good quality (Muratov et al., 2020 ). Note also that these 0D-2D descriptors provide a more comprehensible interpretation regarding the structural requirements of the compounds. The most predictive model was found using the SFS algorithm for variable selection and it is given below along with its MLR statistical parameters. N training = 98; R 2 = 0.787; R 2 Adj = 0.763; F (10,87) = 32.219; Q 2 LOO = 0.734; r m 2 (LOO) = 0.637; ∆ r m 2 (LOO) = 0.145; c R P 2 (1000 runs) = 0.737; N test = 25; R 2 Pred = 0.697; r m 2 (test) = 0.673; ∆ r m 2 (test) = 0.173 (For plots related to this model, please see Supplementary Figure S2 ). This model demonstrates strong internal predictive capability, as evidenced by the statistical parameters: Q 2 LOO (0.734), r m 2 (LOO) (0.637), and ∆ r m 2 (LOO) (0.145). Similarly, the good external predictivity of the model is also ensured from the attained values for R 2 Pred , r m 2 (test) , and ∆ r m 2 (test) . The high value found for c R P 2 further confirms the uniqueness of the model, ruling out the possibility of chance occurrence. Moreover, besides evaluating statistical predictivity, the reliability of the 2D-QSAR model was assessed using additional parameters. For instance, the maximum intercollinearity among the variables was found to be 0.701, and importantly, all variables had VIF values below 5 (Cf. Supplementary Table S2 ). This reaffirms that the 2D-QSAR model is free from multicollinearity problems. In addition, the affinity scores of PFAS in the dataset are between −5.990 to −8.207 kcal/mol (Table S1 ). As the ecotoxicity based-affinity of this work is referred to the formation/obtention of a thermodynamically stable docking complex between the PFAS compounds and the target zfVDAC protein, since the affinity binding scores of PFAS in the dataset are between the later values, there is a clear in silico evidence strongly suggesting the PFASs can spontaneously interact with the zfVDAC at the same biophysical environment that the ATP, i.e. , the native zfVDAC substrate. A close look at the William’s plot of the 2D-QSAR model (see Supplementary Figure S2 ), being the applicability domain of the latter established inside a squared area within ± 3 standard deviations and a leverage threshold h * of 0.337 ( h * = 3 p / N , with p the number of descriptors plus 1 and N the number of data-points in the training set), shows that none of these compounds is found to be a structural influential chemical and only two of them are to be considered response outliers (Gramatica P., 2007 ; Serra et al., 2020 ). Based on the standardized coefficients, the molecular descriptors of this QSAR model have the following descending order of predictive significance: PDI > CATS2D_07_AL > B08[C−O] > B05[N−S] > B02[O−O] > S−110 > B10[C−C] > CATS2D_03_AA > F06[S−F] > F09[O−F]. As such, the most significant descriptor of the model is the packing density index (PDI) that represents a key molecular property of PFAS. In particular, this descriptor for a given PFAS compound is closely related to its molecular van der Waals volume determining relevant interactions with the protein receptor zfVDAC2 (Motoc et al., 1985 ; Todeschini et al., 2000; Todeschini et al., 2009 ). In this case, a negative relation between this parameter and the docking scores indicates that an increased packing density associated to the van der Waals volume improves the interactions of PFASs with the zfVDAC2 channel. The second most significant descriptor of the model is CATS2D_07_AL. Chemically advanced template search (CATS) descriptors are based upon finding topological distances between two pharmacophore features in the molecules and CATS2D_07_AL accounts specifically for one hydrogen bond acceptor and lipophilic groups that are separated by a topological distance of 7 (Reutlinger et al., 2013 ). Further, the positive contribution of CATS2D_07_AL suggests that higher values of this descriptor should be found in compounds with low docking scores. Generally, CATS molecular descriptors are often used for scaffold hopping and, in the current dataset, CATS2D_07_AL mainly points toward compounds containing the sulphone moiety that acts as a strong hydrogen bond acceptor. The positive contribution of CATS2D_07_AL comparatively suggests that, in long-chain PFAS compounds ( i.e. , with a topological distance of 7), the presence of sulfone moiety significantly reduces the binding affinity of PFAS to zfVDAC2. Similar information about the role of sulfone moieties is also derived from the S−110 descriptor, which stands for the atom-centered fragment R-SO2-R, also positively correlated with the docking scores-based binding affinity. This implies that the presence of a sulfone moiety clearly reduces the interactions of PFAS with zfVDAC2 but, since the importance of CATS2D_07_AL is higher than that of S−110, the simultaneous presence of lipophilic features at a specific distance with an acceptor feature like sulfone may further reduce such interactions. Remarkable is also the fact that this 2D-QSAR model comprises a substantial number of atom pairs descriptors, which account for the presence of two type of atoms in the PFAS compounds at a given topological distance. Among these, the presence of carbon-oxygen at topological distance 8, of carbon-carbon (distance 10), and the frequency of sulfur-fluorine (distance 6) were found to increase the binding affinity of the evaluated PFAS. Moreover, these descriptors involve large topological distances typical of long chain length PFAS, implying that the presence of these 2D atom pairs in such type of PFAS help in boosting the affinity interaction with zfVDAC2 associated to their potential ecotoxicity. Instead, the presence of nitrogen-sulfur (at distance 5) and oxygen-oxygen (at distance 2), as well as the frequency of oxygen-fluorine (at distance 9), negatively influence the binding affinity values. Finally, the CATS2D_03_AA descriptor stands for PFAS structures where two acceptor groups are at a topological distance 3, and in contrast to CATS2D_07_AL, this descriptor positively contributes towards higher docking scores-based affinity of the PFAS. Notwithstanding, PFAS with polar atoms (oxygen) belonging to the poly-fluoroalkyl chains had higher values for this descriptor showing their roles in the binding interactions between PFAs and the zfVDAC2 channel, in accordance with the virtual screening results. 3.4. 3D-QSAR-PLS model To further explore the full potential of 3D-descriptors in characterizing the interactions of PFASs with zfVDAC2, as well as to visualize the 3D contour maps that determine the binding behavior of the PFAS, two different feature selection algorithms were used, namely Fractional Factorial Design based selection (FFD-SEL), and the Uninformative Variable Elimination based partial least squares (UVE-PLS). The optimized 3D-QSAR-PLS models were then generated with five components, but, as that would diminish the internal predictivity of models − judging from the reduction in Q 2 LOO values, the number of components was limited to 4. The statistical results of these models are depicted in Table 1 . Table 1 Statistical parameters for the 3D-QSAR-PLS models generated with 4 components. Parameter FFD-SEL UVE-PLS N training 98 98 R 2 0.923 0.914 F 277.064 246.55 Q 2 LOO 0.835 0.803 r m 2 (LOO) 0.768 0.727 ∆ r m 2 (LOO) 0.100 0.108 N test 25 25 R 2 Pred 0.795 0.773 r m 2 (test) 0.695 0.657 ∆ r m 2 (test) 0.163 0.180 Since the FFD-SEL technique yielded the most predictive 3D-QSAR model, with Q 2 LOO of 0.835 and R 2 Pred of 0.795, its observed vs . predicted values are shown in Supplementary Figure S3 . Therefore, both internal and external predictivity of this model are better than the previously obtained 2D-QSAR models. Furthermore, the leverage values of the training dataset compounds were calculated with Open3DQSAR (Tosco et al., 2011) and lead to a range of 0.812−0.983, which is a considerably lower variation than that obtained in the 2D-QSAR analysis for the training set ( i.e. : 0.047−0.332). Therefore, it should be assumed that similar to 2D-QSAR model, no structural outliers are present for the 3D-QSAR model as well. Upon confirmation of this, analysis of the the electrostatic and steric contour maps generated by the later model ensued (Fig. 5 ). From the results of the generated contour maps comprising electrostatic and steric fields using as examples the representative PFAS docked poses, namely: best-ranked PFAS#25, PFAS#53, PFAS#96, PFAS#108, and worst-ranked PFAS#29, it is evidenced that the score of a PFAS gradually decreases as it distances away from the favorable steric maps. Actually, PFAS# 25 depicted the highest docking score against zfVDAC2 (FEB = −8.207 kcal/mol) with the highest potential ecotoxicity, whereas the least docking score was obtained for the worst ranked PFAS#29 (FEB = −5.999 kcal/mol), i.e. , the PFAS compound with the lowest potential ecotoxicity on zfVDAC2. Regarding the contour maps of PFAS#25, the poly-fluoroalkyl chain is closer to the favorable steric fields, and due to lack of any polar moiety it is benefitted by the uninterrupted interaction with this field. The long chain of this PFAS also distances itself from the large positive electrostatic field to avoid unfavorable interactions. The terminal polar moiety of PFAS#25, on the other hand, perfectly locates itself near electrostatic fields increasing its chance for favorable polar docking interactions. When comparing the contour maps of PFAS#25 with the less active PFAS#53 compound (FEB = −7.850 kcal/mol), the latter fails to utilize the steric favorable field, in spite of having a relatively long aliphatic chain, probably due to the presence of its sulfonamide moiety that tends to drag the polar part of this chain towards the positive electrostatic field. Specifically, the partial interaction of the sulfonamide moiety with this positive electrostatic field may not compensate for the lack of interaction with the favorable steric field. This information complies with the fact that, according to the 2D-QSAR model, the sulphone moiety is found to be a detrimental factor for higher affinity interaction with the zfVAC2 protein. Still, analogous to PFAS#25, the polar terminal residue of PFAS#96 locates near electrostatic fields. However, as compared to compounds PFAS#25 and PFAS#53, the chain of compound PFAS#96 is smaller and at the same time, it also contains multiple polar oxygen atoms. Therefore, it lies in close proximity to the large positive electrostatic field that is away from the favorable steric moiety. Thus, the favorable interactions between the electronegative atoms with positive electrostatic field may partially compensate for the lack of favorable hydrophobic interactions. Note that similar information is conveyed by two molecular descriptors integrating the 2D-QSAR model, namely by descriptors B02[O-O] and CATS2D_03_AA. However, the reduced score of PFAS#96 may also be attributed to the penalty it faces due to close proximity to the unfavorable steric field. Finally, both the PFAS#108 and the worst ranked PFAS#29 compounds preferably locate themselves in a small pocket containing the electrostatic fields for favorable interactions with their terminal polar moieties (PFAS eco-toxicophoric groups). Due to their small structures, both these PFAS however lack favorable steric interactions. 3.5. Transversal Summary of Biochemical Mechanistic Implications Mitochondrial toxicity and its mechanistic explanation have gained momentum within in silico research, due to its implications for environmental assessment and biomedical applications (Nelms et al., 2015 ; Ebert et al., 2022). While compelling evidence exists for the mitochondrial toxicity of PFOA (Hagenaars et al., 2013 ; O'Brien et al., 2004; Mashayekhi et al., 2015 ; Choi et al., 2017 ), the present study focuses on selected subsets of PFAS compounds relevant to mitochondrial toxicity, drawing from available in silico tools and literature (Nelms et al., 2015 ; Ebert et al., 2022), fostering the hypothesis-driven methodology implemented, i.e. an in-depth computational study of the zfVDAC2/PFASs through a hybrid structure-based virtual screening ( HSB-VS ) plus Quantitative Structure-Activity Relationship ( QSAR ) approach. Though cautioning that the outlined methodology does not intend to study the relation between the affinity docking scores and certain macroscopic ecotoxicity (rather the study was carried at the molecular level to explore the toxicodynamic behavior of PFAS compounds in a relevant molecular target belonging to the well-recognized ecotoxicological animal model, i.e . the zebrafish), the approach allows the emergence of the mitochondria (eco)toxicity-based affinity as a new in silico concept associated to molecular docking interactions with zfVDAC. Moreover, it goes beyond traditional docking approaches by incorporating local perturbations under the unbound and bound states of zfVDAC2. This additional feature expands the interpretability of interaction mechanisms. Beyond not only exclusively quantifying the energy of interaction (FEB values) and the binding participating amino-acid residues, typically found in traditional docking experiments, it also offers relevant information about the signals propagation in the zfVDAC2, defined by the collective anisotropic fluctuations in the intra-segment Cα-Cα atomic residue distances by the presence of a given ligand, such as ATP or PFAS, which could be biochemically relevant or unfavorable from a structural point of view whenever affecting the binding site native conformation and function of the zfVDAC2. Additionally, our molecular docking results were further supported by previous structural and functional analyses (Fig. 1 ), including: (i) prediction for the best-ranked zfVDAC2 binding using DeepSite, (ii) performing the zfVDAC2 flexibility profile, and (iii) 2D-matrix of communication efficiency in the zfVDAC2 residue network illustrating the correlated motions. As mechanistic information based in mitochondria dysfunctions has been considered relevant to generate new hypotheses through the different levels of organization as molecular, cellular, organismal and population health in the context of Environmental Pollutants, see for example the critical review and analysis by Dreier et al. ( 2019 ), this provides ground for the importance and relevance of our results and its implications. Table 2 displays a summary of the most relevant results and mechanistic implications, which were discussed in minutiae in the previous sections, per in silico tool, while referencing the experimental and/or literature evidence of the finding. Table 2 Transversal summary of main biochemical mechanistic implications for the hybrid HSB-VS plus QSAR model results versus experimental/literature evidence. In silico strategy Objectives Method Result Implications Hybrid structure-based virtual screening (HSB-VS) Predict best-ranked zfVDAC2 binding-site DeepSite tool (machine-learning algorithm based on 3D-deep convolutional neural networks) Identification of the zfVDAC2 cavities/PFAs binding sites Maximum cavity volume : 686.40Å 3 zfVDAC2 cavity/PFAs likely binding site involves voltage-gating N-terminal α-helix segment in the middle of the channel Validate zfVDAC2 channel structure Ramachandran analysis (elimination of false positives for the docking zfVDAC2-PFAS generated complexes) Predicted best-ranked binding sites verified as conformational-favored residues of zfVDAC2 Conformationally non-favored residue LEU 227 is not part of the predicted best-ranked binding site Absence of false positives in the predicted best-ranked binding sites Determine/characterize zfVDAC2 flexibility properties Elastic Network models (ENM) (zfVDAC2 flexibility profile) Local perturbation response scanning (LPRS) map approach (anisotropic changes induced by ATP, long-chain PFAS#25, and short-chain PFAS#29 versus zfVDAC2 voltage-sensing N-terminal α-helix segment) Proper fit of ATP molecule is assured by comparing the ATP bound vs. unbound zfVDAC2 LPRS maps Promotion of ATP/zfVDAC2 target sensor large blocks residues conformational chain flexibilization (versus conformational rigidification of consecutive zfVDAC2 effector residues blocks) Determine complexes zfVDAC2 protein/ PFAS free energies of binding (FEB) DockThor virtual screening (phenotypic crowding-based multiple solution steady-state genetic algorithm searching tool) General tendency for the tested PFAS to interact with predicted zfVDAC2 binding site zfVDAC2 protein/ PFAS interaction occurs after a spontaneous thermodynamics process Similar #vdW contacts between identified PFAS ligands and relevant zfVDAC2 binding site amino acid residues [PFAS#25 (#vdW = 6) ≈ PFAS#29 (#vdW = 5)] PFAS#25 interacts with residues GLY191, GLY192, ALA209, LEU242, and VAL273 PFAS#29 interacts just with the HIS181 residue of protein zfVDAC2 # fluorine bond interactions : best-ranked PFAS#25 = 8 worst-ranked PFAS#29 = 2 b) #hydrogen bond interactions : best-ranked PFAS#25 = 3 worst-ranked PFAS#29 = 2 non-classical C-H bond interaction from PFAS#25 with the ATP-binding residue (LYS236) Best (long-chain complex) a) zfVDAC2-PFAS#25 versus worst ranked PFAS docking complexes, i.e. respectively maximum and minimum level of zfVDAC2 ecotoxicity, indicates a prevalence of van der Wals interactions (#vdW) Long-chain b) PFAS/zfVDAC2 binding site could significantly affect the correct fit of ATP , if both molecules would interact simultaneously in the same biophysical environment PFAS#29 interaction with the HIS181 residue of protein zfVDAC2 explains the difference between the complexes’ obtained van der Waals energy values TYR7 and LEU10 residues do not seem to have a direct influence on the inhibition of the ATP-translocation through the zfVDAC2 channel Differences in the number and type of halogen-bond interactions : (1) embody the eco-toxicophoric moiety of the PFAS family (2) give rise to the ecotoxicity based-affinity for complex zfVDAC2-PFAS#25 being slightly higher than that of the zfVDAC2-ATP complex C-H bond interaction PFAS#25/ LYS236 might affect the correct ATP conformational-fit of the adenosine-moiety into the zfVDAC2 binding site , and interfere ATP-efflux translocation regulation from the mitochondrial matrix to the cytosol through zfVDAC2 channel c) QSAR approach Calculate 2D QSAR descriptors Explore quantitative relationships between the binding properties of the evaluated ligands under interaction with the zfVDAC2 channel and the PFAS molecular structures (2D QSAR model) Dragon software Multiple linear regression (MLR) Two most significant descriptors: (1) Packing Density Index (PDI) : Molecular PFAS property closely related with molecular van der Wals volume (2) Chemically advanced template search for one hydrogen bond acceptor and lipophilic groups separated by a topological distance of 7 (CATS2D_07_AL): mainly compounds containing sulphone moiety acting as a strong hydrogen bond acceptor Significant descriptors of carbon-oxygen at topological distance 8, B08[C−O], of carbon-carbon (distance 10), B10[C−C], and the frequency of sulfur-fluorine (distance 6), F06[S−F], increase the binding affinity of the evaluated PFAS Presence of nitrogen-sulfur (at distance 5), B05[N−S], and oxygen-oxygen (at distance 2), B02[O−O], as well as the frequency of oxygen-fluorine (at distance 9), F09[O−F], negatively influence the binding affinity values CATS2D_03_AA descriptor stands for PFAS structures where two acceptor groups are at a topological distance 3, and positively contributes towards higher docking scores-based affinity of the PFAS PDI determines relevant interactions with the protein receptor zfVDAC2 d) Negative relation of PDI and docking scores implicates increased PDI improves PFASs/ zfVDAC2 channel interactions High values of CATS2D_07_AL should be found in compounds with low docking scores In long-chain PFAS compounds ( i.e. , with a topological distance of 7), the presence of sulfone moiety significantly reduces the binding affinity of PFAS to zfVDAC2 Simultaneous presence of lipophilic features at a specific distance with an acceptor feature like sulfone may further reduce such interactions Presence of B08[C−O], B10[C−C], and F06[S−F] in long chain length PFAS increase affinity interaction with zfVDAC2 associated to their potential ecotoxicity PFAS with polar atoms (oxygen) belonging to the poly-fluoroalkyl chains had higher values for CATS2D_03_AA, implying their roles in the PFAs/zfVDAC2 channel binding interactions Develop predictive 3D QSARs Probe 3D-Descriptors potential in characterizing the interactions of PFASs with zfVDAC2 Determine PFAS binding behavior Open3DQSAR 3D Countour maps Fractional Factorial Design based selection (FFD-SEL) Uninformative Variable Elimination based partial least squares (UVE-PLS) PFAS score gradually decreases as it distances away from the favorable steric maps PFAS# 25: poly-fluoroalkyl chain is closer to the favorable steric fields , and due to lack of any polar moiety it is benefitted by the uninterrupted interaction with this field (countour map) PFAS#29: worst ranked compounds preferably locate themselves in a small pocket containing the electrostatic fields for favorable interactions with their terminal polar moieties (countour map) PFAS# 25: highest docking score against zfVDAC2 with the highest potential ecotoxicity PFAS# 29: lowest docking score against zfVDAC2 with the lowest potential ecotoxicity PFAS# 25: poly-fluoroalkyl long chain distances itself from the large positive electrostatic field to avoid unfavorable interactions, terminal polar moiety perfectly locates itself near electrostatic fields increasing its chance for favorable polar docking interactions Experimental/literature supporting evidence: a) Bischel et al., 2018; Poothong et al., 2017 ; Brendel et al., 2018 ; b) Hagenaars et al., 2013 ; O'Brien et al., 2004; Mashayekhi et al., 2015 ; Ankley et al., 2021 ; Schredelseker et al., 2014 ; c) Ankley et al., 2021 ; Schredelseker et al., 2014 ; Choi et al., 2017 ; d) Motoc et al., 1985 ; Todeschini et al., 2000; Todeschini et al., 2009 . 4. Conclusions The ecotoxicological profile of per- and poly-fluoroalkyl substances versus mitochondria channel, specifically the VADC2, is still challenging and emergent, with scarce experimental evidence. This work proposes a first tentative approach to this context, presenting a novel in silico tool, to be contrasted versus future experimental data on the subject. This study investigates the interaction of PFAS with the ecotoxicologically relevant zebrafish mitotarget (zfVDA2), by resourcing to in silico approaches, due to their broad regulatory acceptance for alternative ecotoxicity assessments. Hybrid structure-based virtual screening plus predictive 2D/3D-QSAR models are proposed as a novel approach in order to theoretically explore the mechanistic interactions of PFAS with potential environmental impact mimicking PFAS bioaccumulation in low-concentration of exposure at the sub-cellular level (PFAS ecotoxicity). The structural validation-based Ramachandran plots revealed that the zfVDAC2 channel protein can be efficiently modeled with a high crystallographic quality (> 85%) and absence of flexibility restrictions for the zfVDAC2 binding residues. In addition, this hybrid approach does not rely exclusively on the docking results, rather it also incorporates computational work based on local perturbations under the unbound and bound state of the zfVDAC2. Such a feature extends the possibilities in terms of interpretability of interaction mechanisms beyond not only exclusively quantifying the energy of interaction (FEB values) and the binding participating amino-acid residues which is usually found in the traditional docking experiments, but also offering relevant information about the signals propagation in the zfVDAC2, defined by the collective anisotropic fluctuations in the intra-segment Cα-Cα atomic residue distances by the presence of a given ligand, such as ATP or PFAS. These latter could be biochemically relevant or unfavorable from a structural point of view whenever affecting the binding site native conformation and function of the zfVDAC2. The obtained results for the best and worst-ranked docked PFAS from a data set of 123 PFAS compounds showed a spontaneous thermodynamic binding process, with prevalence for non-covalent interactions mainly associated to non-covalent hydrophobic van der Waal interactions, followed by fluorine (F)-halogen-bond interactions and then hydrogen-bond interactions. Furthermore, different interaction patterns observed by the 2D-lig-plot interaction diagrams associated to the PFAS chain length, strongly suggests that long-chain PFAS (≥ 6 carbons) present a higher zebrafish ecotoxicity when compared with their PFAS analogs of short-chains. Results also indicate that PFAS could affect the communication efficiency in the network of binding residues in the voltage-sensing N-terminal α-helix segment involved in the conformational fit and regulation of the mitochondrial transport of the ATP molecule through zfVDAC2. What is more, the most significant PFAS’ molecular descriptor is the packing density index that is directly associated to their most dominant van der Waal interactions. This work demonstrated that some PFAS may have substantial potencies to bind on the zebrafish mitochondrial voltage-dependent anion channel (zfVDAC2). Therefore, these in silico evidences open new horizons to improve the rational-design of PFAS with the required low-ecotoxicity, as well as forward better bioremediation strategies to prevent the negative impact of PFAS in aquatic organisms under environmental exposure, and to ensure a safe and sustainable use of PFAS applications. Declarations Supplementary Information (SI) Supplementary material to this article can be found online at DOI: …. Funding This work received financial support from FCT/MCTES (UIDB/50006/2020 DOI 10.54499/UIDB/50006/2020) through national funds. Acknowledgments This work received support and help from FCT/MCTES (LA/P/0008/2020 DOI 10.54499/LA/P/0008/2020 and UIDP/50006/2020 DOI 10.54499/UIDP/50006/2020), through national funds. Ana S. Moura further acknowledges FCT/MECS for the contract IF CEECIND/03631/2017. Competing Interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Credit author statement Michael González-Durruthy : Conceptualization; Methodology; Investigation; Formal analysis; Validation; Visualization; Writing-original draft. Amit Kumar Halder : Methodology; Investigation; Formal analysis; Validation; Visualization; Writing review & editing. Ana Silveira Moura : Validation; Writing review & editing. Maria Natália Dias Soeiro Cordeiro : Conceptualization; Methodology; Validation; Supervision; Funding acquisition; Writing review & editing. Compliance with Ethical Standards Ethical approval: Not applicable. Consent to Participate: Not applicable. Consent to Publish: Not applicable. References Alderete, T. L., Jin, R., Walker, D. I., Valvi, D., Chen, Z. H., et al., 2019. Perfluoroalkyl substances, metabolomic profiling, and alterations in glucose homeostasis among overweight and obese Hispanic children: A proof-of-concept analysis. Environ. Int. 126, 445−453. https://doi.org/10.1016/j.envint.2019.02.047 Ambure, P., Aher, R. B., Gajewicz, A., Puzyn, T., Roy, K. 2015. "NanoBRIDGES" software: Open access tools to perform QSAR and nano-QSAR modeling. Chemometr. Intell. Lab. 147, 1−13. https://doi.org/10.1016/j.chemolab.2015.07.007 Andersen, M. E., Clewell, H. J., Tan, Y. M., Butenhoff, J. L., Olsen, G. W., 2006. Pharmacokinetic modeling of saturable, renal resorption of perfluoroalkylacids in monkeys - Probing the determinants of long plasma half-lives. Toxicology 227, 156−164. https://doi.org/10.1016/j.tox.2006.08.004 Ankley, G. T., Cureton, P., Hoke, R. A., Houde, M., Kumar, A., et al., 2021. Assessing the Ecological Risks of Per- and Polyfluoroalkyl Substances: Current State-of-the Science and a Proposed Path Forward. Environ. Toxicol. Chem. 40, 564−605. https://doi.org/10.1002/etc.4869 Bischel, H. N., MacManus-Spencer, L. A., Luthy, R. G., 2010. Noncovalent Interactions of Long-Chain Perfluoroalkyl Acids with Serum Albumin. Environ. Sci. Technol. 44, 5263−5269. https://doi.org/10.1021/es101334s Bowman, J. S., 2015. Fluorotechnology Is Critical to Modern Life: The FluoroCouncil Counterpoint to the Madrid Statement. Environ. Health Persp. 123, A112−A113. https://doi.org/10.1289/ehp.1509910 Brendel, S., Fetter, E., Staude, C., Vierke, L., Biegel-Engler, A., 2018. Short-chain perfluoroalkyl acids: Environmental concerns and a regulatory strategy under REACH. Environ. Sci. Eur. 30, 9. https://doi.org/10.1186/s12302-018-0134-4 Buck, R. C., Franklin, J., Berger, U., Conder, J. M., Cousins, I. T., et al., 2011. Perfluoroalkyl and polyfluoroalkyl substances in the environment: terminology, classification, and origins. Integr. Environ. Assess Manag. 7, 513−541. https://doi.org/10.1002/ieam.258 Butenhoff, J. L., Chang, S. C., Olsen, G. W., Thomford, P. J., 2012. Chronic dietary toxicity and carcinogenicity study with potassium perfluorooctanesulfonate in Sprague Dawley rats. Toxicology 293, 1−15. https://doi.org/10.1016/j.tox.2012.01.003 Chelcea, I. C., Ahrens L., Örn S., Mucs D., Andersson P. L., 2020. Investigating the OECD database of per- and polyfluoroalkyl substances – chemical variation and applicability of current fate models. Environmental Chemistry 17, 498−508. https://doi.org/10.1071/EN19296 Chen, V. B., Arendall, W. B., Headd, J. J., Keedy, D. A., Immormino, R. M., Kapral, G. J., Murray, L. W., Richardson, J. S., Richardson, D. C., 2010. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr D Biol Crystallogr. 66, 12−21. https://doi.org/10.1107/S0907444909042073 Cheng, W. X., Ng, C. A., 2019. Using Machine Learning to Classify Bioactivity for 3486 Per- and Polyfluoroalkyl Substances (PFASs) from the OECD List. Environ. Sci. Technol. 53, 13970−13980. https://doi.org/10.1021/acs.est.9b04833 Chennubhotla, C., Bahar, I., 2007. Signal propagation in proteins and relation to equilibrium fluctuations. PLoS Comput. Biol. 3, 1716−1726. https://doi.org/10.1371/journal.pcbi.0030172 Choi, E. M., Suh, K. S., Rhee, S. Y., Oh, S., Woo, J. T., et al., 2017. Perfluorooctanoic acid induces mitochondrial dysfunction in MC3T3-E1 osteoblast cells. J. Environ. Sci. Heal. A 52, 281−289. https://doi.org/10.1080/10934529.2016.1253402 Cousins, I. T., DeWitt, J. C., Glüge, J., Goldenman, G., Herzke, D., Lohmann, R., Miller, M., Ng, C. A., Scheringer, M., Vierke, L., Wang, Z., 2020. Strategies for Grouping Per- and Polyfluoroalkyl Substances (PFAS) to Protect Human and Environmental Health. Environ. Sci.: Processes Impacts, 22 (7), 1444−1460. https://doi.org/10.1039/D0EM00147C Cousins, I. T., Johansson, J. H., Salter, M. E., Sha, B., Scheringer, M., 2022. Outside the Safe Operating Space of a New Planetary Boundary for Per- and Polyfluoroalkyl Substances (PFAS). Environ. Sci. Technol., 56, 11172−11179. https://doi.org/10.1021/acs.est.2c02765 Dreier, D. A., Mello, D. F., Meyer, J. N., Martyniuk, C. J., 2019. Linking Mitochondrial Dysfunction to Organismal and Population Health in the Context of Environmental Pollutants: Progress and Considerations for Mitochondrial Adverse Outcome Pathways. Environ Toxicol Chem. 38, 1625−1634. https://doi.org/10.1002/etc.4453 Ebert, A., Goss, K.U., 2022. Screening of 6000 Compounds for Uncoupling Activity: A Comparison Between a Mechanistic Biophysical Model and the Structural Alert Profiler Mitotox. Toxicol. Sci. 185, 208–219. https://doi.org/10.1093/toxsci/kfab139 Girdhar, A., Bharathi, V., Tiwari, V. R., Abhishek, S., Deeksha, W., Mahawar, U. S., Raju, G., Singh, S. K., Prabusankar, G., Rajakumara, E., Patel, B. K. 2020. Computational insights into mechanism of AIM4-mediated inhibition of aggregation of TDP-43 protein implicated in ALS and evidence for in vitro inhibition of liquid-liquid phase separation (LLPS) of TDP-432C-A315T by AIM4. Int. J. Biol. Macromol. 147, 117-130. https://doi.org/10.1016/j.ijbiomac.2020.01.032. Golbraikh, A., Tropsha, A., 2002. Beware of q2! J. Mol. Graph. Model. 20, 269−276. https://doi.org/10.1016/S1093-3263(01)00123-1 González-Durruthy, M., Giri, A. K., Moreira, I., Concu, R., Melo, A., Ruso, J. M., Cordeiro, M. N. D. S., 2020. Computational Modeling on Mitochondrial Channel Nanotoxicity. Nano Today. 34, 100913. https://doi.org/10.1016/j.nantod.2020.100913 González-Durruthy, M., Werhli, A.V., Seus, V. et al., 2017. Decrypting Strong and Weak Single-Walled Carbon Nanotubes Interactions with Mitochondrial Voltage-Dependent Anion Channels Using Molecular Docking and Perturbation Theory. Sci Rep 7, 13271. https://doi.org/10.1038/s41598-017-13691-8 Gramatica, P. 2007. Principles of QSAR models validation: internal and external. QSAR Comb. Sci. 2007, 26, 694-701. https://doi.org/10.1002/qsar.200610151 Gramatica, P. 2013. On the development and validation of QSAR models. Methods. Mol. Biol. 930, 499–526. https://doi.org/10.1007/978-1-62703-059-5_21 Guedes, I. A., Costa, L. S. C., dos Santos, K. B., Karl, A. L. M., Rocha, G. K., et al., 2021. Drug design and repurposing with DockThor-VS web server focusing on SARS-CoV-2 therapeutic targets and their non-synonym variants. Sci. Rep. 11, 5543. https://doi.org/10.1038/s41598-021-84700-0 Hagenaars, A., Vergauwen, L., Benoot, D., Laukens, K., Knapen, D., 2013. Mechanistic toxicity study of perfluorooctanoic acid in zebrafish suggests mitochondrial dysfunction to play a key role in PFOA toxicity. Chemosphere 91, 844–856. https://doi.org/10.1016/j.chemosphere.2013.01.056 Halder, A. K., Amin, S. A., Jha, T., Gayen, S., 2017. Insight into the structural requirements of pyrimidine-based phosphodiesterase 10A (PDE10A) inhibitors by multiple validated 3D QSAR approaches. SAR QSAR Environ. Res. 28, 253–273. https://doi.org/10.1080/1062936X.2017.1302991 Halder, A. K., Delgado, A. H. S., Cordeiro, M. N. D. S., 2022. First multi-target QSAR model for predicting the cytotoxicity of acrylic acid-based dental monomers. Dental Mater. 38, 333–346. https://doi.org/10.1016/j.dental.2021.12.014 https://comptox.epa.gov/dashboard/chemical_lists/pfasmaster. Jimenez, J., Doerr, S., Martinez-Rosell, G., Rose, A. S., De Fabritiis, G., 2017. DeepSite: protein-binding site predictor using 3D-convolutional neural networks. Bioinformatics 33, 3036–3042. https://doi.org/10.1093/bioinformatics/btx350 Krafft, M. P., Riess, J. G., 2015. Per- and polyfluorinated substances (PFASs): Environmental challenges. Curr. Opin. Colloid. Interface Sci. 20, 192–212. https://doi.org/10.1016/j.cocis.2015.07.004 Laskowski, R. A., Swindells, M. B., 2011. LigPlot+: Multiple Ligand-Protein Interaction Diagrams for Drug Discovery. J. Chem. Inf. Model. 51, 2778–2786. https://doi.org/10.1021/ci200227u Lee, B. H., Seo, S., Kim, M. H., Kim, Y., Jo, S., et al., 2017. Normal mode-guided transition pathway generation in proteins. PloS One 12, e0185658. https://doi.org/10.1371/journal.pone.0185658 Le Guilloux, V., Schmidtke, P., Tuffery, P., 2009. Fpocket: An open source platform for ligand pocket detection. BMC Bioinformatics 10, 168. https://doi.org/10.1186/1471-2105-10-168 Liao, B., Zhou, C., Liu, T., Dai, Y., Huang, H. 2020. A novel Hericium erinaceus polysaccharide: Structural characterization and prevention of H2O2-induced oxidative damage in GES-1 cells. Int. J. Biol. Macromol. 154, 1460-1470. https://doi.org/10.1016/j.ijbiomac.2019.11.027. Mashayekhi, V., Tehrani, K. H. M. E., Hashemzaei, M., Tabrizian, K., Shahraki, J., et al., 2015. Mechanistic approach for the toxic effects of perfluorooctanoic acid on isolated rat liver and brain mitochondria. Hum. Exp. Toxicol. 34, 985–996. https://doi.org/10.1177/0960327114565492 Mauri, A., Consonni, V., Pavan, M., Todeschini, R., 2006. Dragon software: An easy approach to molecular descriptor calculations. Match-Commun Math. Co. 56, 237–248. Motoc, I., Marshall, G. R., Dammkoehler, R. A., Labanowski, J., 1985. Molecular Shape Descriptors .1. Three-Dimensional Molecular Shape Descriptor. Z Naturforsch A, 40, 1108–1113. Muratov, E. N., Bajorath, J., Sheridan, R. P., Tetko, I. V., Filimonov, D., et al., 2020. QSAR without borders. Chem. Soc. Rev. 49, 3716–3716. https://doi.org/10.1039/D0CS00098A Nelms, M. D., Mellor, C.L., Cronin, M. T. D., Madden, J. C., Enoch, S. J., 2015. Development of an in Silico Profiler for Mitochondrial Toxicity. Chem. Res. Toxicol. 28, 10, 1891–1902. https://doi.org/10.1021/acs.chemrestox.5b00275 OECD. Fact Cards of Major Groups of Per- and Polyfluoroalkyl Substances (PFASs). Series on Risk Management No. 68. Organisation for Economic Cooperation and Development, 2022. OECD/UNEP. Toward a new comprehensive global database of per- and poly-fluoroalkyl substances (PFASs): Summary report on updating the OECD 2007 list of per- and poly-fluoroalkyl substances (PFASs), Organisation for Economic Cooperation and Development, 2018. OECD/UNEP, Working towards a global emission inventory of PFASs: focus on PFCAS - status quo and the way forward. OECD Environment, Health and Safety Publications Series on Risk Management No. 30. Organisation for Economic Cooperation and Development, 2015. O'Brien, T. M., Wallace, K. B., 2004. Mitochondrial permeability transition as the critical target of N-acetyl perfluorooctane sulfonamide toxicity in vitro. Toxicol. Sci. 82, 333–340. https://doi.org/10.1093/toxsci/kfh244 Ojha, P. K., Roy, K., 2011. Comparative QSARs for antimalarial endochins: Importance of descriptor-thinning and noise reduction prior to feature selection. Chemometr. Intell. Lab. 109, 146–161. https://doi.org/10.1016/j.chemolab.2011.08.007 Patlewicz, G., Richard, A. M., Williams, A. J., Grulke, C. M., Sams, R., Lambert, J., Noyes, P. D., DeVito,M. J., Hines, R. N., Strynar, M., Guiseppi-Elie, A., Thomas, R. S., 2019. A Chemical Category-Based Prioritization Approach for Selecting 75 Per- and Polyfluoroalkyl Substances (PFAS) for Tiered Toxicity and Toxicokinetic Testing. Environ. Health Perspect. 127, 014501. https://doi.org/10.1289/EHP4555 Poothong, S., Thomsen, C., Padilla-Sanchez, J. A., Papadopoulou, E., Haug, L. S., 2017. Distribution of Novel and Well-Known Poly- and Perfluoroalkyl Substances (PFASs) in Human Serum, Plasma, and Whole Blood. Environ. Sci. Technol. 51, 13388–13396. https://doi.org/10.1021/acs.est.7b03299 Raschka, S., 2018. MLxtend: Providing machine learning and data science utilities and extensions to Python's scientific computing stack. J. Open. Source Software 3, 638. https://doi.org/10.21105/joss.00638 Reutlinger, M., Koch, C. P., Reker, D., Todoroff, N., Schneider, P., et al., 2013. Chemically Advanced Template Search (CATS) for Scaffold-Hopping and Prospective Target Prediction for 'Orphan' Molecules. Mol. Inform. 32, 133-138. Roy, K., Chakraborty, P., Mitra, I., Ojha, P. K., et al., 2013. Some case studies on application of "r(m)2" metrics for judging quality of quantitative structure-activity relationship predictions: emphasis on scaling of response data. J. Comput. Chem. 34, 1071-1082. https://doi.org/10.1002/jcc.23231 Scheringer, M., Trier, X., Cousins, I. T., de Voogt, P., Fletcher, T., et al., 2014. Helsingør statement on poly- and perfluorinated alkyl substances (PFASs). Chemosphere 114, 337-339. https://doi.org/10.1016/j.chemosphere.2014.05.044 Schredelseker, J., Paz, A., Lopez, C. J., Altenbach, C., Leung, C. S., et al., 2014. High Resolution Structure and Double Electron-Electron Resonance of the Zebrafish Voltage-dependent Anion Channel 2 Reveal an Oligomeric Population. J. Biol. Chem. 289, 12566-12577. https://doi.org/10.1074/jbc.M113.497438 Serra, A., Onlu, S., Festa, P., Fortino, V., Greco, D., 2020. MaNGA: A novel multi-niche multi-objective genetic algorithm for QSAR modelling. Bioinformatics 36, 145-153. https://doi.org/10.1093/bioinformatics/btz521 Sima, M. W., Jaffe, P. R., 2021. A critical review of modeling Poly- and Perfluoroalkyl Substances (PFAS) in the soil-water environment. Sci. Total Environ. 757, 143793. https://doi.org/10.1016/j.scitotenv.2020.143793 Shimizu, H. et al. (2015) Mitochondrial Ca 2+ uptake by the voltage-dependent anion channel 2 regulates cardiac rhythmicity. eLife 4, e04801. https://doi.org/10.7554/eLife.04801 Stockholm convention on persistent organic pollutants (POPs). http://www.pops.int/. 2004. Tama, F., Gadea, F. X., Marques, O., Sanejouand, Y. H., 2000. Building-block approach for determining low-frequency normal modes of macromolecules. Proteins 41, 1-7. https://doi.org/10.1002/1097-0134(20001001)41:13.0.CO;2-P Tamara, T., Vogs, C., 2021. Invited Perspective: PFAS Bioconcentration and Biotransformation in Early Life Stage Zebrafish and Its Implications for Human Health Protection. Environ. Health Persp. 129, 071304. https://doi.org/10.1289/EHP9625 Tetko, I. V., Tanchuk, V. Y., Villa, A. E. P., 2001. Prediction of n-octanol/water partition coefficients from PHYSPROP database using artificial neural networks and E-state indices. J. Chem. Inf. Comp. Sci. 41, 1407-1421. https://doi.org/10.1021/ci010368v Todeschini, R., Consonni, V., Todeschini, R., Molecular descriptors for chemoinformatics. 2nd ed., Wiley-VCH ,John Wiley distributor: Weinheim Chichester, 2009. Todeschini, R., Consonni, V., Handbook of molecular descriptors. Wiley-VCH: Weinheim, New York, 2000. Tosco, P., Balle, T., 2011. Open3DQSAR: a new open-source software aimed at high-throughput chemometric analysis of molecular interaction fields. J. Mol. Model. 17, 201-208. https://doi.org/10.1007/s00894-010-0684-x Wang, Z. Y., Cousins, I. T., Scheringer, M., Hungerbuehler, K., 2015. Hazard assessment of fluorinated alternatives to long-chain perfluoroalkyl acids (PFAAs) and their precursors: Status quo, ongoing challenges and possible solutions. Environ. Int. 75, 172-179. https://doi.org/10.1016/j.envint.2014.11.013 Vestergren, R., Cousins, I. T., 2009. Tracking the Pathways of Human Exposure to Perfluorocarboxylates. Environ. Sci. Technol. 43, 5565-5575. https://doi.org/10.1021/es900228k Yang, L. W., Chng, C. P., 2008. Coarse-grained models reveal functional dynamics-I. Elastic network models--theories, comparisons and perspectives. Bioinform. Biol. Insights 2, 25-45. https://doi.org/10.4137/bbi.s460 Yoo, W., Mayberry, R., Bae, S., Singh, K., He, Q.P., Lillard, J. W. Jr. 2014. A study of efects of multicollinearity in the multivariable analysis. Int. J. Appl. Sci. Technol. 4, 9-19. Zafeer, M. F., Firdaus, F., Ahmad, F., Ullah, R., Anis, E., Waseem, M., Ali, A., Hossain M. M. 2018. Perillyl alcohol alleviates amyloid-b peptides-induced mitochondrial dysfunction and cytotoxicity in SH-SY5Y cells. Int. J. Biol. Macromol. 109, 1029-1038. https://doi.org/10.1016/j.ijbiomac.2017.11.082. Supplementary Files SupplementaryInformationSI.docx TableS1.xlsx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4362510","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":313135327,"identity":"85fc6ac1-0809-45a6-b97c-f81e79c5aa51","order_by":0,"name":"Michael González-Durruthy","email":"","orcid":"","institution":"REQUIMTE LAQV Porto","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"","lastName":"González-Durruthy","suffix":""},{"id":313135328,"identity":"00123086-6c73-406c-ae5d-8b0a1ea516d7","order_by":1,"name":"Amit Kumar Halder","email":"","orcid":"","institution":"REQUIMTE LAQV Porto","correspondingAuthor":false,"prefix":"","firstName":"Amit","middleName":"Kumar","lastName":"Halder","suffix":""},{"id":313135329,"identity":"c3b70d54-a44b-4118-92c8-5eef85bbbb1c","order_by":2,"name":"Ana Silveira Moura","email":"","orcid":"","institution":"REQUIMTE LAQV Porto","correspondingAuthor":false,"prefix":"","firstName":"Ana","middleName":"Silveira","lastName":"Moura","suffix":""},{"id":313135330,"identity":"93ca2311-260d-4a40-9863-03f0c73695d0","order_by":3,"name":"Maria Natália Dias Soeiro Cordeiro","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAuklEQVRIiWNgGAWjYLACHgYJBn7mA6RqkWxLgHGI08LAYHCMWC3m7IePfXhTYZFvfIz5ATPvDoY8e0JaLHvSkmfOOSNhue0YmwEz7xmGYoK2GNzgMWbmbZMwMLvfwwBkMCT2EKfln4SBcRsPSVoaJAwM2IjWciYtmXHOMQkDCaBfDs49I1HMc4CQluOHDzO8qakz4G9jfvjg7Q6bPPYGQtYggwOMDRIJpGgAAsYGBlK1jIJRMApGwQgAAFEyNQeQ+LTLAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-3375-8670","institution":"REQUIMTE LAQV Porto","correspondingAuthor":true,"prefix":"","firstName":"Maria","middleName":"Natália Dias Soeiro","lastName":"Cordeiro","suffix":""}],"badges":[],"createdAt":"2024-05-03 07:23:53","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4362510/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4362510/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":59524588,"identity":"57b5bd98-c355-4cb7-b288-b7ce08ac8847","added_by":"auto","created_at":"2024-07-02 20:47:57","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":332258,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA, B)\u003c/strong\u003e Lateral and top views of the \u003cem\u003efpocket\u003c/em\u003e tool prediction for the best-ranked zfVDAC2 binding site, validated by DeepSite (Jimenez et al. 2017), within the zfVDAC2 channel. The depicted maximum cavity volume measures 686.40Å\u003csup\u003e3\u003c/sup\u003e (light blue), encompassing the voltage-gating N-terminal α-helix segment situated at the channel's midpoint. \u003cstrong\u003eC\u003c/strong\u003e,\u003cstrong\u003e D) \u003c/strong\u003eLateral and top view representations of the zfVDAC2 flexibility profile depicted as a 3D-colored structure, reflecting variations in fluctuation magnitude across the structure from low-flexibility regions (blue) to high-flexibility regions (red). \u003cstrong\u003eE)\u003c/strong\u003e Schematic portrayal of the zfVDAC2 residue network and corresponding vectors indicating intrinsic fluctuation movements (in the unbound state). \u003cstrong\u003eF)\u003c/strong\u003e Two-dimensional matrix showcasing communication efficiency within the zfVDAC2 residue network, illustrating correlated motions (red color bar, ranging from 0 to 1) and anti-correlated motions (blue color bar, ranging from -1 to 0).\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-4362510/v1/842f40d6a01049b66ecc5647.png"},{"id":59524586,"identity":"f04ef581-57dd-4e6b-93df-b8a50065c63c","added_by":"auto","created_at":"2024-07-02 20:47:57","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":293784,"visible":true,"origin":"","legend":"\u003cp\u003eIn the top, \u003cstrong\u003eA)\u003c/strong\u003evan der Waals representation of the predicted top-ranked binding site for zfVDAC2 in its unbound state (depicted in blue). \u003cstrong\u003eB)\u003c/strong\u003e Representation of the zfVDAC2-ATP complex, simulating the theoretical physiological condition. \u003cstrong\u003eC)\u003c/strong\u003eDocking complex of zfVDAC2 with PFAS#25, the best-ranked among the docking complexes. \u003cstrong\u003eD)\u003c/strong\u003e Docking complex of zfVDAC2 with PFAS#29, ranked lowest among the docking complexes. In the bottom, \u003cstrong\u003eE)\u003c/strong\u003e, \u003cstrong\u003eF)\u003c/strong\u003e, and \u003cstrong\u003eG)\u003c/strong\u003e depict corresponding 2D-lig-plot diagrams illustrating interactions within these docking complexes. For \u003cstrong\u003eB)\u003c/strong\u003e-\u003cstrong\u003eD)\u003c/strong\u003e, regions depicted in 3D van der Waals representations labeled in blue represent high solvent accessibility surfaces, those in green denote low solvent accessibility surfaces, and those in gray indicate moderate solvent accessibility surfaces.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-4362510/v1/65ce5946d94b17099d77cf80.png"},{"id":59524589,"identity":"0f82f210-164b-4366-9360-c4b6c687ad92","added_by":"auto","created_at":"2024-07-02 20:47:57","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":588567,"visible":true,"origin":"","legend":"\u003cp\u003eTwo-dimensional matrix representation illustrating local perturbation response scanning (LPRS) maps for the zfVDAC2 channel under various simulated conditions. \u003cstrong\u003eA)\u003c/strong\u003e Unbound zfVDAC2. \u003cstrong\u003eB)\u003c/strong\u003e zfVDAC2-ATP complex, representing the theoretical physiological condition. \u003cstrong\u003eC)\u003c/strong\u003e Docking complex of zfVDAC2 with PFAS#25, ranked as the best among the docking complexes. \u003cstrong\u003eD)\u003c/strong\u003eDocking complex of zfVDAC2 with PFAS#29, ranked as the worst among the docking complexes. In the LPRS maps, local perturbations in residues belonging to the ATP-binding site situated within the voltage-sensing N-terminal a-helix segment of zfVDAC2 are indicated by white arrows.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-4362510/v1/d901bfc56f6edaa054d8bf65.png"},{"id":59525805,"identity":"f2574d24-bc56-457f-a1e6-a8a4c97f5177","added_by":"auto","created_at":"2024-07-02 20:55:57","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":317253,"visible":true,"origin":"","legend":"\u003cp\u003eRepresentation of the diverse anisotropic network topology observed in the evaluated docking systems. The top section illustrates the embedded-network based receptor-ligand interaction model, encompassing the voltage-sensing N-terminal a-helix segment of zfVDAC2. The bottom section depicts isolated receptor-ligand interaction networks for: \u003cstrong\u003eA)\u003c/strong\u003e zfVDAC2-ATP docking complex, representing the theoretical physiological condition; \u003cstrong\u003eB)\u003c/strong\u003eBest-ranked zfVDAC2-PFAS#25 docking complex; and \u003cstrong\u003eC)\u003c/strong\u003e Worst-ranked zfVDAC2-PFAS#29 docking complex.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-4362510/v1/fdcd92127afdad45a671c03c.png"},{"id":59524591,"identity":"2a86da94-ae98-474a-bb23-0b37155cda00","added_by":"auto","created_at":"2024-07-02 20:47:57","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":388342,"visible":true,"origin":"","legend":"\u003cp\u003eDepiction of contour maps illustrating electrostatic and steric fields for representative PFAS compounds: \u003cstrong\u003eA)\u003c/strong\u003e Unbound zfVDAC2 binding site, \u003cstrong\u003eB)\u003c/strong\u003e Best-ranked PFAS#25, \u003cstrong\u003eC)\u003c/strong\u003e PFAS#53, \u0026nbsp;\u003cstrong\u003eD)\u003c/strong\u003e PFAS#96, \u003cstrong\u003eE)\u003c/strong\u003e PFAS#108, and \u003cstrong\u003eF)\u003c/strong\u003e Worst-ranked PFAS#29. In all cases, volumetric regions labeled with colors indicate: green for steric positive, yellow for steric negative, blue for electrostatic positive, and red for electrostatic negative.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-4362510/v1/f27aaced7d7433e5c48246f2.png"},{"id":67172256,"identity":"b3f018ae-01b3-4054-8264-db9b197487f6","added_by":"auto","created_at":"2024-10-22 03:58:40","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3538645,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4362510/v1/197b6ca2-d285-45ec-8f63-df14235f184a.pdf"},{"id":59525806,"identity":"d7840d0f-52fc-4bca-b05b-5167b18cf31f","added_by":"auto","created_at":"2024-07-02 20:55:57","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":290404,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryInformationSI.docx","url":"https://assets-eu.researchsquare.com/files/rs-4362510/v1/90144da064985de8fd86a2c0.docx"},{"id":59524587,"identity":"43391a70-4896-4c0e-bcdf-ceaae0b49702","added_by":"auto","created_at":"2024-07-02 20:47:57","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":21003,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-4362510/v1/25253772c55191a8c655f666.xlsx"}],"financialInterests":"","formattedTitle":"Mapping Mitochondrial Channel Toxicity: A Case Study for Predicting Mito-Target Interactions for the Per- and Poly-Fluoroalkyl Compounds on the Zebrafish Voltage-Dependent Anion Channel 2","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003ePer- and poly-fluoroalkyl substances (PFAS), firstly defined by Buck \u003cem\u003eet al\u003c/em\u003e. in 2011, are considered a highly fluorinated aliphatic family of compounds which are formed by one or more carbon atoms where all the hydrogen substituents have been replaced by atoms of fluorine. Structurally speaking and in general terms, the nomenclature can be represented as C\u003csub\u003en\u003c/sub\u003eF\u003csub\u003e2n+1\u003c/sub\u003e being n\u0026thinsp;\u0026ge;\u0026thinsp;1, with the molecular structure containing at least one CF\u003csub\u003e3\u003c/sub\u003e\u0026ndash; group (Alderete et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Scheringer et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; OECD/UNEP, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.pops.int/\u003c/span\u003e\u003cspan address=\"http://www.pops.int/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. 2004). Despite the current industrial value of PFAS, scientific evidence has raised concerns regarding their potential toxicity to wildlife and ecosystems, as well as to their potential role as inductive agents for physiological disorders on human metabolism, including diabetes, hypertension, and even various types of cancer (Bischel et al., 2018; Poothong et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Brendel et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). A major characteristic of PFAS toxicity is their persistence in biological systems and the environment for extended periods in their unchanged forms. A recent investigation revealed that some PFAS not only accumulate but also are bio-transformed extensively into a wide range of structurally diverse chemicals capable of persisting for long time in the environment. (Tamara et al., 2021; Andersen et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2006\u003c/span\u003e) Furthermore, the continuous release of PFAS contaminants in the aqueous ecosystems can cause a secondary contamination of natural water leading to the spread of contamination in water levels of parts per trillion (Bischel et al., 2018; Wang et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Krafft et al., 2015; Bowman et al., 2015; Butenhoff et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). In fact, according to the Stockholm Convention, PFAS have been considered as one of the most controversial among the category \u0026lsquo;persistent organic pollutant\u0026rsquo; (POP) for the anthropologically manufactured worldwide environmental contaminants (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.pops.int/\u003c/span\u003e\u003cspan address=\"http://www.pops.int/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, 2004). As such, the need for sophisticated bio-remediation strategies in order to mitigate their environmental impact is urgent (Scheringer et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIndeed, proper knowledge regarding PFAS-bioaccumulation parameters alongside with appropriate environmental risk assessment models that addressed PFAS ecotoxicity (in particular when main food-chains are in question) are still amiss (Vestergren et al., 2009). In particular, this knowledge is absent for a wide variety of these main-food chains organisms, such as bacteria, algae, crustaceans, ciliates, fish, yeasts and nematodes. Nevertheless, research has evidenced a direct association between the bioaccumulation potential of PFAS and the length of their perfluoroalkyl chain, as PFAS analogs of longer chains, with at least six carbons, presented a higher bioconcentration potential (Bischel et al., 2018; Poothong et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Brendel et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Tamara et al., 2021; Andersen et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). These chemical attributes are related with increasing concentration of environmental pollution the PFAS cause through their exposure routes, which could involve different levels of trophic chains, including humans, and diverse environmental compartments, such as water, soil or air (Wang et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Krafft et al., 2015; Butenhoff et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2012\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAn interesting bio-target for a PFAS ecotoxicological predictive/assessment model in aqueous ecosystems is the zebrafish (\u003cem\u003eDanio rerio\u003c/em\u003e), since the zebrafish model can be efficiently used to predict/study mechanisms of toxicity for both the macroscopic and cellular level of this organism regarding PFAS induced damages (Wang et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Krafft et al., 2015; Bowman et al., 2015; Butenhoff et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Vestergren et al., 2009; Hagenaars et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; O'Brien et al., 2004; Mashayekhi et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Interestingly, among the most PFAS \u0026ndash; induced macroscopic damages in the zebrafish models, reports include the endocrine disruption (thyroids), altered swimming, developmental and reproductive toxicity in sub-lethal doses. Furthermore, subcellular and biochemical level damages include cytotoxicity, perturbations in lipid metabolism, oxidative stress leading to genotoxicity, and mitochondrial dysfunction, such as ATP bioenergetics perturbations, and channel mitotoxicity.\u003c/p\u003e \u003cp\u003eA hypothesis that we propose in this paper, departing from the analysis of the published works (Wang et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Krafft et al., 2015; Bowman et al., 2015; Butenhoff et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Vestergren, et al., 2009; Hagenaars et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; O'Brien et al., 2004; Mashayekhi et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), is that PFAS could be bioaccumulated at mitochondrial level (\u003cem\u003ei.e.\u003c/em\u003e, at the mitochondrial matrix) after interacting with the zebrafish voltage-dependent anion channel 2 (zfVDAC2), or selectively induce significant perturbations in the native structure and physiological function of zfVDAC2, or selectively induce significant perturbations in the native structure and physiological function of zfVDAC2 by triggering mitotoxic events directly associated to the opening of the mitochondria permeability transition pore-based zfVDAC2 channel dysfunction (zfVDAC2-MPTP). The latter is directly involved in ROS induction, cardiolipin peroxidation, ADP/ATP transport perturbations, cytochrome c release, decreased ATP levels. From a phylogenetic point of view, and regardless of the animal species under study, it is considered that VDAC channel (zfVDAC2) is highly conserved and is involved in the mitochondrial bioenergetics and homeostasis regulation, as well as in the control of intracellular metabolism (Vestergren et al., 2009; Choi et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). With this accepted evidence as cornerstone, we explore a computational model for the mitochondrial channel dysfunction as an eco-toxicodynamic mechanism in zebrafish.\u003c/p\u003e \u003cp\u003eIn addition, this departing concept is sustained by the zfVDAC2 being the most abundant and densely localized channel protein in the outer mitochondrial membrane of all cells in the zebrafish (Dreier et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), thus representing the main route of communication between the mitochondrial intermembrane space and the cytosol. It should be noted that the zfVDAC2 channel has an N-terminal α-helix segment located horizontally inside the pore, which is aligned nearly parallel to the membrane plane, causing a partial narrowing of the zfVDAC2 pore (Shimizu et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2015\u003c/span\u003e, Choi et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). In particular, the N-terminal α-helix segment can adopt different conformations within the voltage-gating channel depending on the internal/or external stimuli (Shimizu et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2015\u003c/span\u003e, Choi et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Further, the importance of this channel covers many mitochondrial essential aspects, such as ideal equilibrium dynamics of the intracellular [Ca\u003csup\u003e2+\u003c/sup\u003e], or the ADP \u003cem\u003eplus\u003c/em\u003e Pi/ATP influx/efflux transport which favor key biochemical functions as glycolysis and ATP-bioenergetics regulation (Shimizu et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2015\u003c/span\u003e, Choi et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2017\u003c/span\u003e, Zafeer et al., \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2018\u003c/span\u003e, Liao et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, Girdhar et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eNotwithstanding, the fact is that there is a data gap regarding the biological hazardousness factor per PFAS group and versus specific bio-targets (Cousins et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, Cousins, et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), something that is even more noticeable when considering mitochondria channel toxicity versus zfVDAC2 toxicity, as the evidence scenario, either experimental or in silico, is scarce if not inexistent. As such, within the \u003cem\u003ein silico\u003c/em\u003e model proposed in this work regarding PFAs versus zfVDAC2 toxicity, validation versus experimental evidence can only be achieved when there is availability of such results. Therefore, the focus will be in the intrinsic \u003cem\u003ein silico\u003c/em\u003e validation methodologies, that have been proven effective in other ecotoxicological approaches that were also validated versus experimental evidence.\u003c/p\u003e \u003cp\u003eThe development of an \u003cem\u003ein silico\u003c/em\u003e ecotoxicological predictive technique in PFAS environmental risk assessment, which is currently rather limited to our best knowledge, not only offers great advantages in time and money savings but is also aligned with the need of implementing effective predictive animal-free testing methodologies (Schredelseker et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Cheng et al., 2019; Sima et al., 2021). Furthermore, \u003cem\u003ein silico\u003c/em\u003e methods are strongly recommended in the very early stages of the predictive toxicological (ecotoxicological) investigation, only to be followed by \u003cem\u003ein vitro\u003c/em\u003e and \u003cem\u003ein vivo\u003c/em\u003e experimental validation after the computational and theoretical assessment.\u003c/p\u003e \u003cp\u003eAs such, a synergy between classic computational approaches as structure-based virtual screening (SB-VS) and predictive Quantitative Structure-Activity Relationships (QSAR) modeling could be efficiently implemented to precisely assess the potential ecotoxicological effects of PFAS in aquatic organisms (Vestergren et al., 2009; Choi et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Ankley et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Schredelseker et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Cheng et al., 2019). These in silico methodologies have demonstrated the ability to simultaneously evaluate the molecular mechanisms of interaction of a large number of compounds, including PFAS, while maintaining sensitivity, specificity, and accuracy to realistically reproduce the key interactions of traditional or emerging pollutants (Vestergren et al., 2009; Choi et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Ankley et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Schredelseker et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn fact, while molecular docking software allows for the rapid calculation of docking scores, the integration of the 3D-QSAR approach offers valuable additional insights and complements the docking analysis in our study, namely in two major vectors: the identification of relevant specific 2D molecular descriptors, and a wider range of prediction.\u003c/p\u003e \u003cp\u003eRegarding the first vector, while the identification of specific 2D molecular descriptors, \u003cem\u003ei.e.\u003c/em\u003e, structural features of molecules in a two-dimensional representation, allows to gain insight into the molecular properties that may contribute to the observed activity or properties of the compounds, the 3D-docked poses, \u003cem\u003ei.e.\u003c/em\u003e, the predicted spatial arrangements of the PFAS ligand molecules within the binding site of the target protein obtained through molecular docking simulations, represent potential binding configurations and help assess the binding affinity between the ligand and protein. Though the docking score obtained from the 3D-docked poses serves as an indicator of the binding affinity between the PFAS compounds and the zfVDAC2 target channel, it does not provide detailed information about the specific structural physico-chemical based molecular descriptors responsible for the PFAS binding activity on the zfVDAC2 channel. Therefore, the recourse to the QSAR methodology aims to elucidate the structure-mitotarget interactions activity relationship of the PFAS compounds on zfVDAC by identifying relevant specific 2D molecular descriptors derived from the 3D-docked poses.\u003c/p\u003e \u003cp\u003eRegarding the second vector, the proposed QSAR model will allow us to quantitatively predict the ecotoxicological properties of a larger set of PFAS compounds beyond those specifically docked in our study. This broader perspective enhances our understanding of the ecotoxic potential of various PFAS analogs, guiding future experimental efforts and providing insights for environmental risk assessment in the context of an ecotoxicological target as zfVDAC2. In addition, employing the QSAR model enables the analysis of the contributions of individual PFAS molecular descriptors and their respective weights, providing a more detailed understanding of the underlying molecular mechanisms governing PFAS ecotoxicity.\u003c/p\u003e \u003cp\u003eThe present study proposes an effective methodology for assessing the molecular interactions between per- and poly-fluoroalkyl compounds and the zfVDAC2 channel protein. This is achieved through a synergistic approach combining structure-based virtual screening and 2D/3D-QSAR models, introducing a novel hybrid \u003cem\u003ein silico\u003c/em\u003e risk assessment technique. This approach represents a qualitative advancement in ecotoxicological predictive techniques within the field of Computational Toxicology. Notwithstanding, the framework of the proposed approach is overall supported by well-recognized structural and thermodynamic techniques to delve into the potential mitochondrial toxicity response induced by PFAS compounds in the zebrafish. In addition, considering the importance of understanding mitochondrial dysfunctions in generating hypotheses across various biological levels, from molecular mechanisms to population health, in the context of Environmental Pollutants (Dreier et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), this work not only presents novelty but also holds significant relevance.\u003c/p\u003e"},{"header":"2. Methodology section","content":"\u003cp\u003eThe methodology had the stages that follow: (i) Selection of PFAS compounds and prediction of zfVDAC2 binding-sites; (ii) Setting up the flexibility profile of zfVDAC2; (iii) Performing structure-based virtual screening on PFAS compounds with zfVDAC2; and (iv) Performing QSAR modeling.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Selection of PFAS compounds and prediction of zfVDAC2 binding-sites.\u003c/h2\u003e \u003cp\u003eAs stated previously, the field of PFAS interaction with mitochondria toxicity is an emergent field, and there is still absence to clear, precise, or mandatory criteria regarding the selection of PFAS for a dataset aiming to develop an \u003cem\u003ein silico\u003c/em\u003e ecotoxicological tool for the mitochondrial channel toxicity, namely the voltage-dependent anion channel 2 (VDAC2) mitochondrial channel toxicity. Therefore, aiming to be a first approximation to the ecotoxicological profile of PFAS interaction with mitochondria channels, the criteria for the selection of the PFAS focused the ecotoxicological evidence presented by EPA and OECD lists of PFAS chemicals.\u003c/p\u003e \u003cp\u003eAs such, a representative dataset, with well-known 123 PFAS compounds, was selected, corresponding to a combination of two subsets (Chelcea et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2020\u003c/span\u003e): i) a 75 PFAS list of a EPA prioritized subset (Patlewicz et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) of a larger chemical inventory (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://comptox.epa.gov/dashboard/chemical_lists/EPAPFAS75S1\u003c/span\u003e\u003cspan address=\"https://comptox.epa.gov/dashboard/chemical_lists/EPAPFAS75S1\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) and the remaining 48 belonging to ii) a list of PFAS chemicals in the OECD New Comprehensive Global Database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://comptox.epa.gov/dashboard/chemical_lists/PFASOECD\u003c/span\u003e\u003cspan address=\"https://comptox.epa.gov/dashboard/chemical_lists/PFASOECD\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), as per the indications of OECD recommendations (OECD, 2018). It should be noted that both selections followed rigorous criteria in terms of nomenclature for commonly recognized per- and poly-fluoroalkyl substances and, as such, other highly fluorinated substances that match the definition of PFASs, but still have not been recognized as such by the OECD were excluded from the dataset in the present study. Nevertheless, these could be included in future predictive purpose studies once they are recognized by the OECD, if considering the PFASs nomenclature and also considering the potential mitochondrial toxicity (OECD, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; OECD 2015).\u003c/p\u003e \u003cp\u003eThe prediction for the best-ranked zfVDAC2 binding-site resourced to DeepSite (Jimenez et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). The DeepSite tool considers all the molecular descriptors related to the proteins zfVDAC2 by using a machine-learning algorithm based on 3D-deep convolutional neural networks (DCNN) (Jimenez et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Moreover, DeepSite has an extensive validation test set with \u0026gt;\u0026thinsp;6500 proteins belonging to the scPDB database, thus allowing for unequivocal prediction of the main catalytic binding sites in zfVDAC2. One of the most crucial steps to ensure quality of modeling results is the accurate prediction/identification of the suitable zfVDAC2 channel binding-sites coupled with an appropriate crystallographic structural validation, and the flexibility properties of the targeted protein. By itself, the search space that was set up for the zfVDAC2 channel using the fpocket tool within the DeepSite validation procedure (Le Guilloux et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Jimenez et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) is a cubic grid box with a size of 20 \u0026times; 20 \u0026times; 20 \u0026Aring;\u003csup\u003e3\u003c/sup\u003e, centered at \u003cem\u003eX\u003c/em\u003e\u0026thinsp;=\u0026thinsp;\u0026minus;20.35 \u0026Aring;, \u003cem\u003eY\u003c/em\u003e\u0026thinsp;=\u0026thinsp;+\u0026thinsp;21.99 \u0026Aring;, and \u003cem\u003eZ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;+\u0026thinsp;6.13 \u0026Aring;, and a discretization of 0.25 \u0026Aring;.\u003c/p\u003e \u003cp\u003eTo begin with, predictions of possible zfVDAC2 pocket binding-sites the Voronoi tessellation algorithm was applied to detect small cavities in zfVDAC2, such as protuberances, and collecting information-based crystallographic descriptors on the pocket geometry and topology associated with high ligand binding probability (Jimenez et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Such algorithm allows to gather the cartesian coordinates like volumetric maps of possible binding-site pockets on the zfVDAC2 channel. Both the zfVDAC2 mesh volumetric maps gathered, along with the best-ranked binding site prediction obtained from DeepSite, were used to define the search space (\u003cem\u003ei.e.\u003c/em\u003e, the cubic grid box) for then carrying out the structure-based virtual screening (Jimenez et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAfter this, a structural validation of the zfVDAC2 channel was carried out by a Ramachandran analysis (Chen et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). In so doing, false positives for the docking zfVDAC2-PFAS generated complexes are prevented by verifying the absence of restricted flexibility of the residues of each zfVDAC2 binding site found.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Setting up the flexibility profile of zfVDAC2.\u003c/h2\u003e \u003cp\u003eElastic Network models (ENM) were used to determine and characterize the flexibility properties of zfVDAC2. This computational algorithm allows to evaluate the directions of movement associated with elastic normal modes of fluctuations by representing the C-(α)-atoms belonging to the whole zfVDAC2 as a residues network communication according to a Hookean potential (or \u0026ldquo;springs\u0026rdquo;). A more in-depth description of the ENM algorithm can be found elsewhere (Yang et al., 2018; Tama et al., 2008; Chennubhotla et al., 2007; Guedes et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn the present study, the zfVDAC2 flexibility profile was set up by considering all the flexibility fluctuation patterns in which the molecular structure of the zfVDAC2 protein will oscillate around its equilibrium conformation and collective motions of its residues network in the unbound state. For more clarity, only the results of the following simulation conditions are presented: (i) unbound zfVDAC2 as the control simulation experiment; (ii) zfVDAC2-ATP as reference control of simulation, since the ATP molecule represents the native substrate of the zfVDAC2 channel and is used here for comparison purposes with (iii) and (iv); (iii) the top-ranked zfVDAC2-PFAS docking complex \u0026minus; \u003cem\u003ei.e.\u003c/em\u003e, as maximum level of zfVDAC2 ecotoxicity; (iv) the worst-ranked zfVDAC2-PFAS docking complex \u0026minus; \u003cem\u003ei.e.\u003c/em\u003e, as minimum level of zfVDAC2 ecotoxicity.\u003c/p\u003e \u003cp\u003eAdditionally, the zfVDAC2 flexibility properties, collective motions like anisotropic fluctuations generated from the zfVDAC2 inter-residue network and communication efficiency within, were also evaluated from the whole zfVDAC2 structure (Tama et al., 2008; Chennubhotla et al., 2007; Guedes et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Performing structure-based virtual screening on PFAS compounds with zfVDAC2\u003c/h2\u003e \u003cp\u003eThe next step is to get the free energies of binding (FEB) for the complexes formed between the zfVDAC2 protein and the PFAS family of compounds by virtual screening. As human health has been the main concern when probing the PFAS molecular mechanisms, in order to reproduce with a certain level of precision the PFAS molecular interactions with the zfVDAC2 channel, our structure-based virtual screening results mimic low-concentration conditions for each compound. This is achieved since these interactions are between one PFAS molecule and one zfVDAC2 protein and not involving potential ligand-ligand interactions (Laskowski et al., 2011), \u003cem\u003ei.e.\u003c/em\u003e, PFAS plus PFAS interactions, as it would happen in mixtures released into the complex aquatic environment at high-concentrations.\u003c/p\u003e \u003cp\u003eThe DockThor virtual screening platform (Guedes et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) was employed for docking the targeted PFAS within the zfVDAC2 channel. The DockThor-VS code makes use of an accurate phenotypic crowding-based multiple solution steady-state genetic algorithm as searching tool. The latter allowed us to identify multiple minima solutions in a single docking run, keeping the population diversity of the PFAS ligand structures. Default parameters were set on each of these runs, as the number of evaluations (500000 evaluations per docking run), the population size (750 individuals), the number of runs (24), and maximum of cluster leaders (20).\u003c/p\u003e \u003cp\u003eThe scoring function applied in DockThor (Guedes et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) to score the various docked poses is based on computing their total affinity (\u003cem\u003eE\u003c/em\u003e\u003csub\u003etotal\u003c/sub\u003e, kcal/mol) as a sum of intermolecular and intramolecular interaction energy terms from the MMFF94S49 force field. All the needed interaction parameters for both zfVDAC2 and the PFAS ligands were explicitly set up by resorting to the DockThor tools MMFF-Ligand and PdbThorBox (Guedes et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The total affinity values obtained for the docking poses were then categorized like energetically unfavorable when ∆\u003cem\u003eG\u003c/em\u003e\u003csub\u003ebind\u003c/sub\u003e \u0026ge; 0 kcal/mol, indicating a complete absence of affinity of zfVDAC2 to that particular PFAS structure; otherwise, those were categorized as favorable interactions. In addition, affinity values less than the one of ATP with zfVDAC2 (\u003cem\u003eE\u003c/em\u003e\u003csub\u003etotal\u003c/sub\u003e = \u0026minus;7.4 kcal/mol) were classified as strong docking interactions considering that ATP is its native substrate.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4. Performing QSAR modeling\u003c/h2\u003e \u003cp\u003eThe 3D-docked poses (\u003cem\u003ei.e.\u003c/em\u003e, docked poses with the highest docking score) of the PFAS were used as input structures for calculating 2D descriptors using the Dragon software (Mauri et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). The PFAS dataset was randomly divided into a training set (80%) and a test set (20%) with help of our \u003cem\u003ein-house\u003c/em\u003e tool SFS-QSAR (available to download at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/ncordeirfcup/SFS-QSAR-tool\u003c/span\u003e\u003cspan address=\"https://github.com/ncordeirfcup/SFS-QSAR-tool\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) (Halder et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Details about these PFAS sets are given in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e of the Supporting Information (SI).\u003c/p\u003e \u003cp\u003eA multiple linear regression (MLR) approach was adopted for setting up the 2D-QSAR models, by applying as variable selection procedure either the sequential forward selection algorithm (Raschka S., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) (SFS) or the genetic algorithm (GA) \u0026minus; the latter using the open-source code NanoBRIDGES (Ambure et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). In both cases, pre-treatment of descriptors was carried out with a correlation (Pearson \u003cem\u003eR\u003c/em\u003e) cut-off of 0.95 and variance cut-off of 0.001. Selection of SFS-MLR models was done by examining usual statistical parameters \u0026minus; \u003cem\u003ei.e.\u003c/em\u003e: the determination coefficient (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e) and the negative mean absolute error (NMAE), and the variables selected without cross-validation (0-fold CV) or by 5-fold CV. As to the GA-MLR models, these were generated by applying default parameters as the mutation probability (=\u0026thinsp;0.3) and the total number of iterations \u003cem\u003eper\u003c/em\u003e run (=\u0026thinsp;100), being the best final GA-MLR model selected from 20 of these runs. The quality of both type of 2D-QSAR models was evaluated by well-known statistical metrics such as the \u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e coefficient, adjusted \u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eAdj\u003c/sub\u003e), and the Fisher\u0026rsquo;s ratio (\u003cem\u003eF\u003c/em\u003e) (Gramatica P., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). Similarly, the commonly used leave-one-out (LOO) cross-validation regression coefficient (\u003cem\u003eQ\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLOO\u003c/sub\u003e), along with two \u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e metrics (\u003cem\u003ei.e\u003c/em\u003e.: \u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(LOO)\u003c/sub\u003e and its associated standard deviation D\u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(LOO)\u003c/sub\u003e; see (Eq.\u0026nbsp;\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) and (Eq.\u0026nbsp;\u003cspan refid=\"Equ2\" class=\"InternalRef\"\u003e2\u003c/span\u003e) below) were employed to estimate their internal predictivity and analogous metrics (\u003cem\u003ei.e\u003c/em\u003e.: \u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003ePred\u003c/sub\u003e, \u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(test)\u003c/sub\u003e, and Δ\u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(test)\u003c/sub\u003e) to evaluate their external predictivity (Gramatica P., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Gramatica P., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Tetko et al., \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Roy et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Golbraikh et al., 2002).\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$${{{r}^{2}}_{m}}_{\\left(\\text{L}\\text{O}\\text{O}\\right)}= {R}^{2}(1-\\sqrt{({R}^{2 }-{R}_{0}^{2})})$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$${{{\\varDelta r}^{2}}_{m}}_{\\left(\\text{L}\\text{O}\\text{O}\\right)}=|{R}_{m}^{2}-{R}_{m}^{2}{\\prime } |$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eIn Eq.\u0026nbsp;1, \u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({R}_{0}^{2}\\)\u003c/span\u003e\u003c/span\u003e are the determination coefficients between the observed and the LOO-predicted scores of the compounds on the training dataset without and with the intercept set to zero, respectively. In Eq.\u0026nbsp;\u003cspan refid=\"Equ2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, for \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({ R}_{m}^{2}\\)\u003c/span\u003e\u003c/span\u003e, the observed values are considered in the \u003cem\u003eY\u003c/em\u003e-axis and the predicted values in the \u003cem\u003eX\u003c/em\u003e-axis whereas for \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({R}_{m}^{2}{\\prime }\\)\u003c/span\u003e\u003c/span\u003e, the values of these axes are exchanged.\u003c/p\u003e \u003cp\u003eThe inter-collinearity of the 2D-QSAR model\u0026rsquo;s descriptors were checked by examining the cross-correlation matrix. To further assess the multi-collinearity of the model, the variation inflation factor (VIF) of each descriptor was also computed using the following equation (Yoo et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2014\u003c/span\u003e):\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$$VIF=1/(1-{R}_{i}^{2})$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({R}_{i}^{2}\\)\u003c/span\u003e\u003c/span\u003e is the correlation coefficient (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e) determined by regressing the \u003cem\u003ei\u003c/em\u003e\u003csup\u003eth\u003c/sup\u003e descriptor on the other descriptors.\u003c/p\u003e \u003cp\u003eAdditionally, the \u003cem\u003eY\u003c/em\u003e-based randomization technique was applied (1000 response randomization runs), and the metric \u003csup\u003e\u003cem\u003ec\u003c/em\u003e\u003c/sup\u003e\u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003eP\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e computed to infer the uniqueness of the models (Ojha et al., 2011). Lastly, the applicability domain of the 2D-QSAR models \u0026minus; \u003cem\u003ei.e\u003c/em\u003e., the response and data structure space in which the models afford predictions with a given reliability, was assessed by plotting the leverage \u003cem\u003evs\u003c/em\u003e. standardized residuals for each dataset case \u0026ndash; the so-called William\u0026rsquo;s plot, which allows to detect possible structural and response outliers (Gramatica P., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Serra et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe Open3DQSAR tool (Tosco et al., 2011) was used for setting up predictive 3D-QSAR models using the same PFAS training and test sets combination. This software uses a carbon and volume-less positively (+\u0026thinsp;1) charged probe for calculating steric and electrostatic fields of the query chemicals, respectively. The pre-treatment of the fields was carried out after setting a smart region definition (SRD)-cut-off level of 2.0 and also by removing \u003cem\u003eN\u003c/em\u003e-level variables \u0026minus; \u003cem\u003ei.e.\u003c/em\u003e: removal of variables with less variations in order to avoid biasing of the model (Tosco et al., 2011). The Open3DQSAR uses SRD for variable grouping, based on the closeness of variables in 3D space, as well as two different variable selection algorithms, namely: Fractional Factorial Design-based variable selection (FFD-SEL), and the Uninformative Variable Elimination-based partial least square (UVE-PLS) (Tosco et al., 2011; Halder et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). The predictive quality of the 3D-QSAR-based PLS models generated from the PFAS was examined using the same statistical metrics as the ones referred to above for the 2D-QSAR models.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results and Discussion","content":"\u003cdiv id=\"Sec8\"\u003e\n \u003ch2\u003e3.1. Structure-based virtual screening on zfVDAC2-PFAS interactions\u003c/h2\u003e\n \u003cp\u003eThe identification of the best ranked zfVDAC2 binding sites followed according to the details in the Methods and model section. The results pertaining to the zfVDAC2 flexibility properties, collective motions like anisotropic fluctuations generated from the zfVDAC2 inter-residue network and communication efficiency are depicted in Fig.\u0026nbsp;\u003cspan\u003e1\u003c/span\u003e.\u003c/p\u003e\n \u003cp\u003eThe crystallographic validation of the zfVDAC2 channel 3D X-ray structure ensued by applying the Ramachandran analysis with quality assessment, as displayed in Supplementary \u003cstrong\u003eFigure \u003cspan\u003eS1\u003c/span\u003e\u003c/strong\u003e. Ramachandran analysis corresponds to a 2D-projection onto the plane from the 3D-crystallographic structure of zfVDAC2 (pdb model) where all its possible residue conformations are defined in the plot based on the dihedral angles (\u003cem\u003e\u0026phi;\u003c/em\u003e and \u003cem\u003e\u0026psi;\u003c/em\u003e) around their peptide-bonds (Chen et al., \u003cspan\u003e2010\u003c/span\u003e). As such, the allowed \u003cem\u003e\u0026phi; vs. \u0026psi;\u003c/em\u003e torsion values of one particular residue, found within the Ramachandran contour (colored purple; see \u003cstrong\u003eFigure \u003cspan\u003eS1\u003c/span\u003e\u003c/strong\u003e) were taken as conformational-favored residues of zfVDAC2. Note that the zfVDAC2 sterically disallowed residue LEU 227, identified as a conformationally non-favored residue, is not part of the predicted best-ranked binding site that thus ensures the absence of false positive VS results.\u003c/p\u003e\n \u003cp\u003eFigure \u003cspan\u003e2\u003c/span\u003e. In the top, \u003cstrong\u003eA)\u003c/strong\u003e van der Waals representation of the predicted top-ranked binding site for zfVDAC2 in its unbound state (depicted in blue). \u003cstrong\u003eB)\u003c/strong\u003e Representation of the zfVDAC2-ATP complex, simulating the theoretical physiological condition. \u003cstrong\u003eC)\u003c/strong\u003e Docking complex of zfVDAC2 with PFAS#25, the best-ranked among the docking complexes. \u003cstrong\u003eD)\u003c/strong\u003e Docking complex of zfVDAC2 with PFAS#29, ranked lowest among the docking complexes. In the bottom, \u003cstrong\u003eE)\u003c/strong\u003e, \u003cstrong\u003eF)\u003c/strong\u003e, and \u003cstrong\u003eG)\u003c/strong\u003e depict corresponding 2D-lig-plot diagrams illustrating interactions within these docking complexes. For \u003cstrong\u003eB)\u003c/strong\u003e-\u003cstrong\u003eD)\u003c/strong\u003e, regions depicted in 3D van der Waals representations labeled in blue represent high solvent accessibility surfaces, those in green denote low solvent accessibility surfaces, and those in gray indicate moderate solvent accessibility surfaces.\u003c/p\u003e\n \u003cp\u003eAnalyses on these outputs point to a general tendency of the tested PFASs, whether with short or long chain length, to being able to interact in the predicted binding site placed in the middle of the zfVDAC2 channel. Further, the interaction occurs after a spontaneous thermodynamics process, according to the negative FEB values of binding affinity obtained in all the cases (refer to \u003cstrong\u003eTable \u003cspan\u003eS1\u003c/span\u003e\u003c/strong\u003e). Focusing on the best ranked PFAS#25 (Fig.\u0026nbsp;\u003cspan\u003e2\u003c/span\u003eF) and the worst ranked PFAS#29 (Fig.\u0026nbsp;\u003cspan\u003e2\u003c/span\u003eG) docking complexes, a prevalence of van der Waals interactions was identified between the evaluated PFAS ligands and the zfVDAC2 with approximately the same number of van der Waals contacts (#vdW as light-green spheres). The identified PFAS ligands present approximately the same number of van der Waals contacts (#vdW as light-green spheres) with relevant amino acid residues of zfVDAC2 binding site as PFAS#25 (#vdW\u0026thinsp;=\u0026thinsp;6)\u0026thinsp;\u003cem\u003e\u0026asymp;\u003c/em\u003e\u0026thinsp;PFAS#29 (#vdW\u0026thinsp;=\u0026thinsp;5). Though the difference between these numbers is small, it carries a significant impact from a quantitative perspective since the most relevant contribution is provided by the binding properties of the interacting residues involved, such as the druggability \u0026minus; \u003cem\u003ei.e\u003c/em\u003e., the ability to bind ligands. Albeit the best-ranked and worst-ranked PFASs both interact at the same biophysical environment in the zfVDAC2 binding site than the ATP molecule \u0026minus; \u003cem\u003ei.e\u003c/em\u003e., the natural zfVDAC2 substrate (Hagenaars et al., \u003cspan\u003e2013\u003c/span\u003e; O\u0026apos;Brien et al., 2004; Mashayekhi et al., \u003cspan\u003e2015\u003c/span\u003e; Ankley et al., \u003cspan\u003e2021\u003c/span\u003e; Schredelseker et al., \u003cspan\u003e2014\u003c/span\u003e), they share key ATP-binding residues that are responsible for the ADP/ATP influx/efflux through the zfVDAC2 channel. Indeed, PFAS#25 interacts with residues GLY191, GLY192, ALA209, LEU242, and VAL273, while PFAS#29 interacts just with the HIS181 residue of protein zfVDAC2, thus being the first more directly perturbed by these ATP-binding residues. The latter explains the difference between the obtained van der Waals energy values for complexes with these PFAS ligands, namely: \u003cem\u003eE\u003c/em\u003e\u003csub\u003evdW\u003c/sub\u003e (zfVDAC2-PFAS#25)\u0026thinsp;=\u0026thinsp;\u0026minus;17.303 kcal/mol; \u003cem\u003eE\u003c/em\u003e\u003csub\u003evdW\u003c/sub\u003e (zfVDAC2-PFAS#29)\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.304 kcal/mol. It should be noted here that the FEB was decomposed into different energy/entropy items by applying the equations used in the work of Gonz\u0026aacute;lez-Durruthy et al., namely equations \u003cspan\u003e3\u003c/span\u003e and \u003cspan\u003e4\u003c/span\u003e (Gonz\u0026aacute;lez-Durruthy et al., \u003cspan\u003e2017\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eThe higher number of fluorine bond interactions found for the best-ranked PFAS, PFAS#25 (see Fig.\u0026nbsp;\u003cspan\u003e2\u003c/span\u003eE), contrasts also to the just two (F)-halogen-bond interactions found for the worst-ranked ligand, PFAS#29 (with residues LEU10 and TYR7) (Schredelseker et al., \u003cspan\u003e2014\u003c/span\u003e). It is noteworthy that TYR7 and LEU10 residues do not seem to have a direct influence on the inhibition of the ATP-translocation through the zfVDAC2 channel, which could be a consequence of those not being involved in the stabilization of the ATP molecule during the passage across such channel. These differences in the number and type of halogen-bond interactions have great relevance, since those embody the eco-toxicophoric moiety of the PFAS family and give rise to the ecotoxicity based-affinity for complex zfVDAC2-PFAS#25 (=\u0026thinsp;\u0026minus;8.207 kcal/mol) being slightly higher than that of the zfVDAC2-ATP complex (=\u0026thinsp;\u0026minus;7.400 kcal/mol), either for the evaluated zfVDAC2 target and other protein targets potentially exposed to PFAS.\u003c/p\u003e\n \u003cp\u003eWhen comparing the VS score-based affinity values of the generated docking complexes for long-chain PFAS compounds, typically containing\u0026thinsp;\u0026ge;\u0026thinsp;6 carbons (as in the case of the best-ranked complex zfVDAC2-PFAS#25), with corresponding short-chain PFAS compounds, there is evidence that the length of the chain of per-and-polyfluoroalkyl is a key eco-toxicophoric attribute inducing more strong inhibition on the zfVDAC2 channel (Bischel et al., 2018; Poothong et al., \u003cspan\u003e2017\u003c/span\u003e; Brendel et al., \u003cspan\u003e2018\u003c/span\u003e). As seen in Fig.\u0026nbsp;\u003cspan\u003e2\u003c/span\u003eB, the extended conformation of the long-chain PFAS compounds within the zfVDAC2 binding site could significantly affect the correct fit of ATP, if both molecules would interact simultaneously in the same biophysical environment (see Supplementary \u003cstrong\u003eFigure S2\u003c/strong\u003e).\u003c/p\u003e\n \u003cp\u003eRegarding the hydrogen-bond interactions, we identify the presence of a high degree of stabilization for the best-ranked PFAS \u003cem\u003evs.\u003c/em\u003e the worst-ranked PFAS docking complexes originated from three classical H-bond interactions involving the ATP-binding residues. The latter are residues ASN183, ASN207, and LEU208 in the case of PFAS#25, whereas only two conventional H-bond interactions are detected with non-target ATP-binding residues in the worst-ranked PFAS#29 (\u003cem\u003ei.e.\u003c/em\u003e: LEU10 and the regulatory-phosphorylation residue TYR7). Additionally, the presence of one non-classical C-H bond interaction from PFAS#25 with the ATP-binding residue (LYS236) might affect the correct ATP conformational-fit of the adenosine-moiety into the zfVDAC2 binding site, which could interfere with the voltage-sensing N-terminal \u0026alpha;-helix segment that regulates the ATP-efflux translocation from the mitochondrial matrix to the cytosol through zfVDAC2 channel. Indeed, this perturbation near to the N-terminal \u0026alpha;-helix segment could theoretically induce changes in the zebrafish, such as mitochondrial swelling or the electronic transport chain in zebrafish mitochondrial complex (I, III, IV), compromising its physiological process of ATP/ADP cycles through the zfVDAC2 channel (Ankley et al., \u003cspan\u003e2021\u003c/span\u003e; Schredelseker et al., \u003cspan\u003e2014\u003c/span\u003e; Choi et al., \u003cspan\u003e2017\u003c/span\u003e).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\"\u003e\n \u003ch2\u003e3.2. zfVDAC2 inter-residue network communication efficiency\u003c/h2\u003e\n \u003cp\u003eTo probe how the communication efficiency in the inter-residue network of zfVDAC2 in the unbound and bound state, PFAS#25 and PFAS#29, an anisotropic ENM was implemented by considering local perturbations induced by the ligands in the zfVDAC2 residue network (\u003cem\u003eCf.\u003c/em\u003e Figure\u0026nbsp;\u003cspan\u003e1\u003c/span\u003e, E and \u003cstrong\u003eF\u003c/strong\u003e). The results are displayed in Fig.\u0026nbsp;\u003cspan\u003e3\u003c/span\u003e.\u003c/p\u003e\n \u003cp\u003eThe local perturbation response scanning (LPRS) map approach tries to offer a conceptual vision on the local perturbations induced by the PFAS ligands interacting with zfVDAC in the bound state, which is a novelty result regarding this context. The local perturbations are defined by the collective anisotropic fluctuations in the intra-segment C\u003cem\u003e\u0026alpha;\u003c/em\u003e-C\u003cem\u003e\u0026alpha;\u003c/em\u003e atomic residue distances by the presence of a given ligand, such as ATP or PFAS, that could be biochemically relevant or unfavorable from a structural point of view whenever affecting the binding site native conformation and function of the zfVDAC2 (Tama et al., 2008; Chennubhotla et al., 2007; Guedes et al., \u003cspan\u003e2021\u003c/span\u003e; Lee et al., \u003cspan\u003e2017\u003c/span\u003e). Thus, the anisotropic changes induced by the ligands ATP, long-chain PFAS#25, and short-chain PFAS#29 in the inter-residue network of the zfVDAC2 were determined using the LPRS maps focusing on the zfVDAC2 binding site \u0026minus; \u003cem\u003ei.e\u003c/em\u003e., on the voltage-sensing N-terminal \u0026alpha;-helix segment (marked as white arrows in Fig.\u0026nbsp;\u003cspan\u003e3\u003c/span\u003e).\u003c/p\u003e\n \u003cdiv\u003e\n \u003cp\u003eThe physiological patterns generated by the ATP molecule when interacts with the zfVDAC2 target sensor residues \u0026minus; yellow to dark-red color in the corresponding LPRS map, promote significant conformational chain flexibilization from large blocks of such consecutive residues, contrasting with the labeled-blue regions of the LPRS map that denote the presence of blocks of consecutive zfVDAC2 effector residues with conformational rigidification. Comparing the ATP bound vs. unbound zfVDAC2 LPRS maps, the proper fit of the ATP molecule is assured.\u003c/p\u003e\n \u003cp\u003eIn Fig.\u0026nbsp;\u003cspan\u003e4\u003c/span\u003e, panel B, the result of the local perturbation induced by the ATP molecule on the zfVDAC channel corresponds to the modeling of the physiological binding condition. As such, the binding interaction of the ATP intra the zfVDAC channel can be considered as a reversible biochemical process when acknowledging that the ATP necessarily transits the interior of the channel interacting with key residues in a reversible way (Gonz\u0026aacute;lez-Durruthy et al., \u003cspan\u003e2020\u003c/span\u003e). Therefore, the theoretical physiological patterns generated by the ATP molecule when interacts with the zfVDAC2 target sensor residues - yellow to dark-red color in the corresponding LPRS map, promote significant conformational chain flexibilization from large blocks of such consecutive residues, contrasting with the labeled-blue regions of the LPRS map that denote the presence of blocks of consecutive zfVDAC2 effector residues with conformational rigidification. Comparing the ATP bound \u003cem\u003evs\u003c/em\u003e. unbound zfVDAC2 LPRS maps, the proper fit of the ATP molecule is assured.\u003c/p\u003e\n \u003cp\u003eRegarding the bound state with the evaluated PFAS, both ligands evidence promotion of local perturbations in the zfVDAC2 binding site, \u003cem\u003ei.e\u003c/em\u003e.: potential ecotoxicity, which are mainly associated to non-covalent interactions (see Fig.\u0026nbsp;\u003cspan\u003e2\u003c/span\u003e, E, F, and \u003cstrong\u003eG\u003c/strong\u003e) (Laskowski et al., 2011), though a different pattern can be observed considering the length of their chains. In the case of the long-chain PFAS compound \u0026minus; PFAS#25, the local perturbations (\u003cem\u003ei.e.\u003c/em\u003e, interactions with the target zfVDAC2 sensor residues) can potentially activate similar blocks of consecutives zfVDAC2 residues. This strongly suggests that the extended docked conformation of the long-chain PFAS compounds within the zfVDAC2 binding site could significantly affect the correct fit of the ATP molecule if it is considered that both molecules could interact at the same biophysical environment partially sharing the same binding residues (see Fig.\u0026nbsp;\u003cspan\u003e3\u003c/span\u003e, B and \u003cstrong\u003eC\u003c/strong\u003e), something also fitting with the results obtained in the 2D-lig-plot interaction diagrams (refer to Fig.\u0026nbsp;\u003cspan\u003e2\u003c/span\u003eE, F and Supplementary \u003cstrong\u003eFigure S2\u003c/strong\u003e). For the case of the worst-ranked PFAS, the short-chain PFAS#29, the local perturbations induce just light modifications in the inter-residue network of the zfVDAC2 binding site (\u003cem\u003eCf.\u003c/em\u003e Figure\u0026nbsp;\u003cspan\u003e2\u003c/span\u003eD), showing a pattern remarkably close to that obtained for the unbound zfVDAC (\u003cem\u003eCf.\u003c/em\u003e Figure\u0026nbsp;\u003cspan\u003e2\u003c/span\u003eA), a result which also fits well with the corresponding 2D-lig-plot interaction diagram (\u003cem\u003eCf.\u003c/em\u003e Figure\u0026nbsp;\u003cspan\u003e2\u003c/span\u003eG). Final validation of the results from the LPRS maps was the differences found between the topology of the anisotropic network of signal communication efficiency of zfVDAC2 (Fig.\u0026nbsp;\u003cspan\u003e4\u003c/span\u003e\u003cstrong\u003e)\u003c/strong\u003e. Further, the obtained LPRS maps for the unbound state and bound states (\u003cem\u003eCf.\u003c/em\u003e Figure\u0026nbsp;\u003cspan\u003e3\u003c/span\u003eA and \u003cspan\u003e3\u003c/span\u003eB-D, respectively) are in both cases strongly dependent on the zfVDAC2 flexibility properties, which are directly associated with the allosteric anisotropic network topology (\u003cem\u003evide\u003c/em\u003e Fig.\u0026nbsp;\u003cspan\u003e4\u003c/span\u003eA-C). This allosteric anisotropic network topology takes place in the inter-residue communication network formed by the backbone\u0026ndash;backbone atom contacts between two sequence-adjacent residues in the evaluated region (zfVDAC2 binding site) under the unbound and bound states, where a different pattern of allosteric signal-based local flexibility perturbations was identified involved sensor-effector residues (\u003cem\u003ei\u003c/em\u003e, \u003cem\u003ej\u003c/em\u003e) of zfVDAC2 (Schredelseker et al., \u003cspan\u003e2014\u003c/span\u003e; Tama et al., 2008; Chennubhotla et al., 2007; Guedes et al., \u003cspan\u003e2021\u003c/span\u003e; Lee et al., \u003cspan\u003e2017\u003c/span\u003e). As such, slight changes in the residue environment of the binding site could theoretically affect the molecular mechanisms of the ATP complementarity (or functional coupling) in the zfVDAC2 pocket and the mitochondrial ATP-transport in zfVDAC2 in the presence of PFAS compounds.\u003c/p\u003e\n \u003cp\u003eFinally, it should be noted the option for resourcing to local perturbation response scanning (LPRS) maps versus molecular dynamics (MD) simulations for evaluating PFAS-induced conformational changes in the mitochondrial zfVDAC2 channel resulted from LPRS maps can be a preferable choice over MD simulations, though these are commonly used to study conformational changes in biomolecules, in a situation as the one of the present work, where localized insights, computational efficiency, and quick exploratory analyses are priorities. This explains why LPRS maps have been successfully applied in several studies to investigate conformational changes in proteins.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\"\u003e\n \u003ch2\u003e3.3. 2D-QSAR model\u003c/h2\u003e\n \u003cp\u003eThe generated 2D-QSAR models attempt to explore quantitative relationships between the targeted ecotoxicological endpoint, \u003cem\u003ei.e.\u003c/em\u003e, the binding properties of the evaluated ligands under interaction with the zfVDAC2 channel (affinity score\u0026thinsp;\u0026asymp;\u0026thinsp;FEB), and the PFAS molecular structures plus physicochemical properties. The latter were described by a limited number of 2D descriptors obtained from the PFAS ligands, namely: constitutional, ring, 2D functional group counts, atom-centered fragments, CATS2D, 2D-atom pairs, molecular properties, and drug-like indices, in order to assure 2D-QSAR models with good quality (Muratov et al., \u003cspan\u003e2020\u003c/span\u003e). Note also that these 0D-2D descriptors provide a more comprehensible interpretation regarding the structural requirements of the compounds.\u003c/p\u003e\n \u003cp\u003eThe most predictive model was found using the SFS algorithm for variable selection and it is given below along with its MLR statistical parameters.\u003c/p\u003e\n \u003cdiv id=\"Equ4\"\u003e\n \u003cdiv\u003e\u003cimg src=\"https://myfiles.space/user_files/122228_c8a1650c59388082/122228_custom_files/img1719477597.png\"\u003e\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cem\u003eN\u003c/em\u003e \u003csub\u003etraining\u003c/sub\u003e = 98; \u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.787; \u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eAdj\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.763; \u003cem\u003eF\u003c/em\u003e(10,87)\u0026thinsp;=\u0026thinsp;32.219; \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLOO\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.734; \u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(LOO)\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.637; ∆\u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(LOO)\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.145; \u003csup\u003e\u003cem\u003ec\u003c/em\u003e\u003c/sup\u003e\u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003eP\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e (1000 runs)\u0026thinsp;=\u0026thinsp;0.737; \u003cem\u003eN\u003c/em\u003e\u003csub\u003etest\u003c/sub\u003e = 25; \u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003ePred\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.697; \u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(test)\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.673; ∆\u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(test)\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.173 (For plots related to this model, please see Supplementary \u003cstrong\u003eFigure S2\u003c/strong\u003e).\u003c/p\u003e\n \u003cp\u003eThis model demonstrates strong internal predictive capability, as evidenced by the statistical parameters: \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLOO\u003c/sub\u003e (0.734), \u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(LOO)\u003c/sub\u003e (0.637), and ∆\u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(LOO)\u003c/sub\u003e (0.145). Similarly, the good external predictivity of the model is also ensured from the attained values for \u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003ePred\u003c/sub\u003e, \u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(test)\u003c/sub\u003e, and ∆\u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(test)\u003c/sub\u003e. The high value found for \u003csup\u003e\u003cem\u003ec\u003c/em\u003e\u003c/sup\u003e\u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003eP\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e further confirms the uniqueness of the model, ruling out the possibility of chance occurrence. Moreover, besides evaluating statistical predictivity, the reliability of the 2D-QSAR model was assessed using additional parameters. For instance, the maximum intercollinearity among the variables was found to be 0.701, and importantly, all variables had VIF values below 5 (Cf. Supplementary Table \u003cspan\u003eS2\u003c/span\u003e). This reaffirms that the 2D-QSAR model is free from multicollinearity problems.\u003c/p\u003e\n \u003cp\u003eIn addition, the affinity scores of PFAS in the dataset are between \u0026minus;5.990 to \u0026minus;8.207 kcal/mol (Table \u003cspan\u003eS1\u003c/span\u003e). As the ecotoxicity based-affinity of this work is referred to the formation/obtention of a thermodynamically stable docking complex between the PFAS compounds and the target zfVDAC protein, since the affinity binding scores of PFAS in the dataset are between the later values, there is a clear \u003cem\u003ein silico\u003c/em\u003e evidence strongly suggesting the PFASs can spontaneously interact with the zfVDAC at the same biophysical environment that the ATP, \u003cem\u003ei.e.\u003c/em\u003e, the native zfVDAC substrate. A close look at the William\u0026rsquo;s plot of the 2D-QSAR model (see Supplementary Figure \u003cspan\u003eS2\u003c/span\u003e), being the applicability domain of the latter established inside a squared area within \u0026plusmn;\u0026thinsp;3 standard deviations and a leverage threshold \u003cem\u003eh\u003c/em\u003e* of 0.337 (\u003cem\u003eh\u003c/em\u003e* = 3\u003cem\u003ep\u003c/em\u003e/\u003cem\u003eN\u003c/em\u003e, with \u003cem\u003ep\u003c/em\u003e the number of descriptors plus 1 and \u003cem\u003eN\u003c/em\u003e the number of data-points in the training set), shows that none of these compounds is found to be a structural influential chemical and only two of them are to be considered response outliers (Gramatica P., \u003cspan\u003e2007\u003c/span\u003e; Serra et al., \u003cspan\u003e2020\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eBased on the standardized coefficients, the molecular descriptors of this QSAR model have the following descending order of predictive significance: PDI\u0026thinsp;\u0026gt;\u0026thinsp;CATS2D_07_AL\u0026thinsp;\u0026gt;\u0026thinsp;B08[C\u0026minus;O]\u0026thinsp;\u0026gt;\u0026thinsp;B05[N\u0026minus;S]\u0026thinsp;\u0026gt;\u0026thinsp;B02[O\u0026minus;O]\u0026thinsp;\u0026gt;\u0026thinsp;S\u0026minus;110\u0026thinsp;\u0026gt;\u0026thinsp;B10[C\u0026minus;C]\u0026thinsp;\u0026gt;\u0026thinsp;CATS2D_03_AA\u0026thinsp;\u0026gt;\u0026thinsp;F06[S\u0026minus;F]\u0026thinsp;\u0026gt;\u0026thinsp;F09[O\u0026minus;F]. As such, the most significant descriptor of the model is the packing density index (PDI) that represents a key molecular property of PFAS. In particular, this descriptor for a given PFAS compound is closely related to its molecular van der Waals volume determining relevant interactions with the protein receptor zfVDAC2 (Motoc et al., \u003cspan\u003e1985\u003c/span\u003e; Todeschini et al., 2000; Todeschini et al., \u003cspan\u003e2009\u003c/span\u003e). In this case, a negative relation between this parameter and the docking scores indicates that an increased packing density associated to the van der Waals volume improves the interactions of PFASs with the zfVDAC2 channel. The second most significant descriptor of the model is CATS2D_07_AL. Chemically advanced template search (CATS) descriptors are based upon finding topological distances between two pharmacophore features in the molecules and CATS2D_07_AL accounts specifically for one hydrogen bond acceptor and lipophilic groups that are separated by a topological distance of 7 (Reutlinger et al., \u003cspan\u003e2013\u003c/span\u003e). Further, the positive contribution of CATS2D_07_AL suggests that higher values of this descriptor should be found in compounds with low docking scores. Generally, CATS molecular descriptors are often used for scaffold hopping and, in the current dataset, CATS2D_07_AL mainly points toward compounds containing the sulphone moiety that acts as a strong hydrogen bond acceptor. The positive contribution of CATS2D_07_AL comparatively suggests that, in long-chain PFAS compounds (\u003cem\u003ei.e.\u003c/em\u003e, with a topological distance of 7), the presence of sulfone moiety significantly reduces the binding affinity of PFAS to zfVDAC2. Similar information about the role of sulfone moieties is also derived from the S\u0026minus;110 descriptor, which stands for the atom-centered fragment R-SO2-R, also positively correlated with the docking scores-based binding affinity. This implies that the presence of a sulfone moiety clearly reduces the interactions of PFAS with zfVDAC2 but, since the importance of CATS2D_07_AL is higher than that of S\u0026minus;110, the simultaneous presence of lipophilic features at a specific distance with an acceptor feature like sulfone may further reduce such interactions.\u003c/p\u003e\n \u003cp\u003eRemarkable is also the fact that this 2D-QSAR model comprises a substantial number of atom pairs descriptors, which account for the presence of two type of atoms in the PFAS compounds at a given topological distance. Among these, the presence of carbon-oxygen at topological distance 8, of carbon-carbon (distance 10), and the frequency of sulfur-fluorine (distance 6) were found to increase the binding affinity of the evaluated PFAS. Moreover, these descriptors involve large topological distances typical of long chain length PFAS, implying that the presence of these 2D atom pairs in such type of PFAS help in boosting the affinity interaction with zfVDAC2 associated to their potential ecotoxicity. Instead, the presence of nitrogen-sulfur (at distance 5) and oxygen-oxygen (at distance 2), as well as the frequency of oxygen-fluorine (at distance 9), negatively influence the binding affinity values. Finally, the CATS2D_03_AA descriptor stands for PFAS structures where two acceptor groups are at a topological distance 3, and in contrast to CATS2D_07_AL, this descriptor positively contributes towards higher docking scores-based affinity of the PFAS. Notwithstanding, PFAS with polar atoms (oxygen) belonging to the poly-fluoroalkyl chains had higher values for this descriptor showing their roles in the binding interactions between PFAs and the zfVDAC2 channel, in accordance with the virtual screening results.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\"\u003e\n \u003ch2\u003e3.4. 3D-QSAR-PLS model\u003c/h2\u003e\n \u003cp\u003eTo further explore the full potential of 3D-descriptors in characterizing the interactions of PFASs with zfVDAC2, as well as to visualize the 3D contour maps that determine the binding behavior of the PFAS, two different feature selection algorithms were used, namely Fractional Factorial Design based selection (FFD-SEL), and the Uninformative Variable Elimination based partial least squares (UVE-PLS). The optimized 3D-QSAR-PLS models were then generated with five components, but, as that would diminish the internal predictivity of models \u0026minus; judging from the reduction in \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLOO\u003c/sub\u003e values, the number of components was limited to 4. The statistical results of these models are depicted in Table\u0026nbsp;\u003cspan\u003e1\u003c/span\u003e.\u003c/p\u003e\n \u003cdiv\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 1\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eStatistical parameters for the 3D-QSAR-PLS models generated with 4 components.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"3\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eParameter\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eFFD-SEL\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eUVE-PLS\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eN\u003c/em\u003e\u003csub\u003etraining\u003c/sub\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e98\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.923\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.914\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eF\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e277.064\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e246.55\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eQ\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLOO\u003c/sub\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.835\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.803\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(LOO)\u003c/sub\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.768\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.727\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e∆\u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(LOO)\u003c/sub\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.108\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eN\u003c/em\u003e\u003csub\u003etest\u003c/sub\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003ePred\u003c/sub\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.795\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.773\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(test)\u003c/sub\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.695\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.657\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e∆\u003cem\u003er\u003c/em\u003e\u003csub\u003e\u003cem\u003em\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e(test)\u003c/sub\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.163\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.180\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eSince the FFD-SEL technique yielded the most predictive 3D-QSAR model, with \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLOO\u003c/sub\u003e of 0.835 and \u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u003csub\u003ePred\u003c/sub\u003e of 0.795, its observed \u003cem\u003evs\u003c/em\u003e. predicted values are shown in Supplementary \u003cstrong\u003eFigure S3\u003c/strong\u003e. Therefore, both internal and external predictivity of this model are better than the previously obtained 2D-QSAR models. Furthermore, the leverage values of the training dataset compounds were calculated with Open3DQSAR (Tosco et al., 2011) and lead to a range of 0.812\u0026minus;0.983, which is a considerably lower variation than that obtained in the 2D-QSAR analysis for the training set (\u003cem\u003ei.e.\u003c/em\u003e: 0.047\u0026minus;0.332). Therefore, it should be assumed that similar to 2D-QSAR model, no structural outliers are present for the 3D-QSAR model as well. Upon confirmation of this, analysis of the the electrostatic and steric contour maps generated by the later model ensued (Fig.\u0026nbsp;\u003cspan\u003e5\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eFrom the results of the generated contour maps comprising electrostatic and steric fields using as examples the representative PFAS docked poses, namely: best-ranked PFAS#25, PFAS#53, PFAS#96, PFAS#108, and worst-ranked PFAS#29, it is evidenced that the score of a PFAS gradually decreases as it distances away from the favorable steric maps. Actually, PFAS# 25 depicted the highest docking score against zfVDAC2 (FEB\u0026thinsp;=\u0026thinsp;\u0026minus;8.207 kcal/mol) with the highest potential ecotoxicity, whereas the least docking score was obtained for the worst ranked PFAS#29 (FEB\u0026thinsp;=\u0026thinsp;\u0026minus;5.999 kcal/mol), \u003cem\u003ei.e.\u003c/em\u003e, the PFAS compound with the lowest potential ecotoxicity on zfVDAC2.\u003c/p\u003e\n \u003cp\u003eRegarding the contour maps of PFAS#25, the poly-fluoroalkyl chain is closer to the favorable steric fields, and due to lack of any polar moiety it is benefitted by the uninterrupted interaction with this field. The long chain of this PFAS also distances itself from the large positive electrostatic field to avoid unfavorable interactions. The terminal polar moiety of PFAS#25, on the other hand, perfectly locates itself near electrostatic fields increasing its chance for favorable polar docking interactions. When comparing the contour maps of PFAS#25 with the less active PFAS#53 compound (FEB\u0026thinsp;=\u0026thinsp;\u0026minus;7.850 kcal/mol), the latter fails to utilize the steric favorable field, in spite of having a relatively long aliphatic chain, probably due to the presence of its sulfonamide moiety that tends to drag the polar part of this chain towards the positive electrostatic field. Specifically, the partial interaction of the sulfonamide moiety with this positive electrostatic field may not compensate for the lack of interaction with the favorable steric field. This information complies with the fact that, according to the 2D-QSAR model, the sulphone moiety is found to be a detrimental factor for higher affinity interaction with the zfVAC2 protein. Still, analogous to PFAS#25, the polar terminal residue of PFAS#96 locates near electrostatic fields. However, as compared to compounds PFAS#25 and PFAS#53, the chain of compound PFAS#96 is smaller and at the same time, it also contains multiple polar oxygen atoms. Therefore, it lies in close proximity to the large positive electrostatic field that is away from the favorable steric moiety. Thus, the favorable interactions between the electronegative atoms with positive electrostatic field may partially compensate for the lack of favorable hydrophobic interactions. Note that similar information is conveyed by two molecular descriptors integrating the 2D-QSAR model, namely by descriptors B02[O-O] and CATS2D_03_AA. However, the reduced score of PFAS#96 may also be attributed to the penalty it faces due to close proximity to the unfavorable steric field. Finally, both the PFAS#108 and the worst ranked PFAS#29 compounds preferably locate themselves in a small pocket containing the electrostatic fields for favorable interactions with their terminal polar moieties (PFAS eco-toxicophoric groups). Due to their small structures, both these PFAS however lack favorable steric interactions.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\"\u003e\n \u003ch2\u003e3.5. Transversal Summary of Biochemical Mechanistic Implications\u003c/h2\u003e\n \u003cp\u003eMitochondrial toxicity and its mechanistic explanation have gained momentum within \u003cem\u003ein silico\u003c/em\u003e research, due to its implications for environmental assessment and biomedical applications (Nelms et al., \u003cspan\u003e2015\u003c/span\u003e; Ebert et al., 2022). While compelling evidence exists for the mitochondrial toxicity of PFOA (Hagenaars et al., \u003cspan\u003e2013\u003c/span\u003e; O\u0026apos;Brien et al., 2004; Mashayekhi et al., \u003cspan\u003e2015\u003c/span\u003e; Choi et al., \u003cspan\u003e2017\u003c/span\u003e), the present study focuses on selected subsets of PFAS compounds relevant to mitochondrial toxicity, drawing from available \u003cem\u003ein silico\u003c/em\u003e tools and literature (Nelms et al., \u003cspan\u003e2015\u003c/span\u003e; Ebert et al., 2022), fostering the hypothesis-driven methodology implemented, \u003cem\u003ei.e.\u003c/em\u003e an in-depth computational study of the zfVDAC2/PFASs through a hybrid \u003cem\u003estructure-based virtual screening\u003c/em\u003e (\u003cem\u003eHSB-VS\u003c/em\u003e) \u003cem\u003eplus Quantitative Structure-Activity Relationship\u003c/em\u003e (\u003cem\u003eQSAR\u003c/em\u003e) approach.\u003c/p\u003e\n \u003cp\u003eThough cautioning that the outlined methodology does not intend to study the relation between the affinity docking scores and certain macroscopic ecotoxicity (rather the study was carried at the molecular level to explore the toxicodynamic behavior of PFAS compounds in a relevant molecular target belonging to the well-recognized ecotoxicological animal model, \u003cem\u003ei.e\u003c/em\u003e. the zebrafish), the approach allows the emergence of the mitochondria (eco)toxicity-based affinity as a new \u003cem\u003ein silico\u003c/em\u003e concept associated to molecular docking interactions with zfVDAC. Moreover, it goes beyond traditional docking approaches by incorporating local perturbations under the unbound and bound states of zfVDAC2. This additional feature expands the interpretability of interaction mechanisms. Beyond not only exclusively quantifying the energy of interaction (FEB values) and the binding participating amino-acid residues, typically found in traditional docking experiments, it also offers relevant information about the signals propagation in the zfVDAC2, defined by the collective anisotropic fluctuations in the intra-segment C\u0026alpha;-C\u0026alpha; atomic residue distances by the presence of a given ligand, such as ATP or PFAS, which could be biochemically relevant or unfavorable from a structural point of view whenever affecting the binding site native conformation and function of the zfVDAC2. Additionally, our molecular docking results were further supported by previous structural and functional analyses (Fig.\u0026nbsp;\u003cspan\u003e1\u003c/span\u003e), including: (i) prediction for the best-ranked zfVDAC2 binding using DeepSite, (ii) performing the zfVDAC2 flexibility profile, and (iii) 2D-matrix of communication efficiency in the zfVDAC2 residue network illustrating the correlated motions.\u003c/p\u003e\n \u003cp\u003eAs mechanistic information based in mitochondria dysfunctions has been considered relevant to generate new hypotheses through the different levels of organization as molecular, cellular, organismal and population health in the context of Environmental Pollutants, see for example the critical review and analysis by Dreier et al. (\u003cspan\u003e2019\u003c/span\u003e), this provides ground for the importance and relevance of our results and its implications. Table\u0026nbsp;\u003cspan\u003e2\u003c/span\u003e displays a summary of the most relevant results and mechanistic implications, which were discussed in minutiae in the previous sections, \u003cem\u003eper in silico\u003c/em\u003e tool, while referencing the experimental and/or literature evidence of the finding.\u003c/p\u003e\n \u003cdiv\u003e\n \u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 2\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eTransversal summary of main biochemical mechanistic implications for the hybrid HSB-VS plus QSAR model results versus experimental/literature evidence.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eIn silico strategy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eObjectives\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMethod\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eResult\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eImplications\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"4\"\u003e\n \u003cp\u003e\u003cstrong\u003eHybrid structure-based virtual screening (HSB-VS)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ePredict\u003c/em\u003e \u003cstrong\u003ebest-ranked zfVDAC2 binding-site\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eDeepSite tool\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e(machine-learning algorithm based on 3D-deep convolutional neural networks)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eIdentification\u003c/strong\u003e of the\u003c/p\u003e\n \u003cp\u003ezfVDAC2 cavities/PFAs binding sites\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eMaximum cavity volume\u003c/strong\u003e: 686.40\u0026Aring;\u003csup\u003e3\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ezfVDAC2 cavity/PFAs likely binding site \u003cem\u003einvolves voltage-gating N-terminal \u0026alpha;-helix segment in the middle of the channel\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eValidate\u003c/em\u003e zfVDAC2 channel structure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eRamachandran analysis\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e(elimination of false positives for the docking zfVDAC2-PFAS generated complexes)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePredicted best-ranked binding sites \u003cstrong\u003everified\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eas conformational-favored residues of zfVDAC2\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eConformationally non-favored residue LEU 227 \u003cstrong\u003eis not part\u003c/strong\u003e of the predicted best-ranked binding site\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eAbsence of false positives\u003c/em\u003e in the predicted best-ranked binding sites\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eDetermine/characterize\u003c/em\u003e zfVDAC2 flexibility properties\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eElastic Network models (ENM)\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e(zfVDAC2 flexibility profile)\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eLocal perturbation response scanning (LPRS) map approach\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e(anisotropic changes induced by ATP, long-chain PFAS#25, and short-chain PFAS#29 versus zfVDAC2 voltage-sensing N-terminal \u0026alpha;-helix segment)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eProper fit of ATP molecule is assured\u003c/strong\u003e by comparing the ATP bound vs. unbound zfVDAC2 LPRS maps\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ePromotion of ATP/zfVDAC2 target sensor large blocks residues conformational chain flexibilization\u003c/em\u003e (versus conformational rigidification of consecutive zfVDAC2 effector residues blocks)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eDetermine\u003c/em\u003e complexes zfVDAC2 protein/ PFAS free energies of binding (FEB)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eDockThor virtual screening\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e(phenotypic crowding-based multiple solution steady-state genetic algorithm searching tool)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGeneral tendency for the \u003cstrong\u003etested PFAS to interact with predicted zfVDAC2 binding site\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003ezfVDAC2 protein/ PFAS \u003cstrong\u003einteraction occurs after a spontaneous thermodynamics process\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eSimilar #vdW contacts\u003c/strong\u003e between identified PFAS ligands and relevant zfVDAC2 binding site amino acid residues\u003c/p\u003e\n \u003cp\u003e[PFAS#25 (#vdW\u0026thinsp;=\u0026thinsp;6)\u0026thinsp;\u0026asymp;\u0026thinsp;PFAS#29 (#vdW\u0026thinsp;=\u0026thinsp;5)]\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003ePFAS#25 interacts with residues GLY191, GLY192, ALA209, LEU242, and VAL273\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003ePFAS#29 interacts just with the HIS181 residue\u003c/strong\u003e of protein zfVDAC2\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e# fluorine bond interactions\u003c/strong\u003e:\u003c/p\u003e\n \u003cp\u003ebest-ranked PFAS#25\u0026thinsp;=\u0026thinsp;8\u003c/p\u003e\n \u003cp\u003eworst-ranked PFAS#29\u0026thinsp;=\u0026thinsp;2\u003csup\u003eb)\u003c/sup\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e#hydrogen bond interactions\u003c/strong\u003e:\u003c/p\u003e\n \u003cp\u003ebest-ranked PFAS#25\u0026thinsp;=\u0026thinsp;3\u003c/p\u003e\n \u003cp\u003eworst-ranked PFAS#29\u0026thinsp;=\u0026thinsp;2\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003enon-classical C-H bond interaction from PFAS#25 with the ATP-binding residue (LYS236)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBest (long-chain complex) \u003csup\u003ea)\u003c/sup\u003e zfVDAC2-PFAS#25 versus worst ranked PFAS docking complexes, \u003cem\u003ei.e.\u003c/em\u003e respectively \u003cem\u003emaximum and minimum level of zfVDAC2 ecotoxicity, indicates a prevalence of van der Wals interactions (#vdW)\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eLong-chain\u003c/em\u003e\u003csup\u003e\u003cem\u003eb)\u003c/em\u003e\u003c/sup\u003e \u003cem\u003ePFAS/zfVDAC2 binding site could significantly affect the correct fit of ATP\u003c/em\u003e, if both molecules would interact simultaneously in the same biophysical environment\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ePFAS#29 interaction with the HIS181 residue of protein zfVDAC2 explains the difference between the complexes\u0026rsquo; obtained van der Waals energy values\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eTYR7 and LEU10 residues do not seem to have a direct influence on the inhibition of the ATP-translocation through the zfVDAC2 channel\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eDifferences in the number and type of halogen-bond interactions\u003c/em\u003e:\u003c/p\u003e\n \u003cp\u003e(1) embody the \u003cem\u003eeco-toxicophoric moiety\u003c/em\u003e of the PFAS family\u003c/p\u003e\n \u003cp\u003e(2) give rise to the \u003cem\u003eecotoxicity based-affinity for complex zfVDAC2-PFAS#25 being slightly higher\u003c/em\u003e than that of the zfVDAC2-ATP complex\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eC-H bond interaction PFAS#25/ LYS236 might affect the correct ATP conformational-fit of the adenosine-moiety into the zfVDAC2 binding site\u003c/em\u003e, and interfere ATP-efflux translocation regulation from the mitochondrial matrix to the cytosol through zfVDAC2 channel\u003csup\u003ec)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003eQSAR approach\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eCalculate\u003c/em\u003e 2D QSAR descriptors\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eExplore\u003c/em\u003e quantitative relationships between the binding properties of the evaluated ligands under interaction with the zfVDAC2 channel and the PFAS molecular structures (2D QSAR model)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eDragon software\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eMultiple linear regression (MLR)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTwo most significant descriptors:\u003c/p\u003e\n \u003cp\u003e(1) \u003cstrong\u003ePacking Density Index (PDI)\u003c/strong\u003e:\u003c/p\u003e\n \u003cp\u003eMolecular PFAS property closely related with molecular van der Wals volume\u003c/p\u003e\n \u003cp\u003e(2) \u003cstrong\u003eChemically advanced template search for one hydrogen bond acceptor and lipophilic groups separated by a topological distance of 7\u003c/strong\u003e (CATS2D_07_AL): mainly compounds containing sulphone moiety acting as a strong hydrogen bond acceptor\u003c/p\u003e\n \u003cp\u003eSignificant descriptors of carbon-oxygen at topological distance 8, B08[C\u0026minus;O], of carbon-carbon (distance 10), B10[C\u0026minus;C], and the frequency of sulfur-fluorine (distance 6), F06[S\u0026minus;F], \u003cstrong\u003eincrease the binding affinity of the evaluated PFAS\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003ePresence of nitrogen-sulfur (at distance 5), B05[N\u0026minus;S], and oxygen-oxygen (at distance 2), B02[O\u0026minus;O], as well as the frequency of oxygen-fluorine (at distance 9), F09[O\u0026minus;F], \u003cstrong\u003enegatively influence the binding affinity values\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eCATS2D_03_AA descriptor stands for PFAS structures where two acceptor groups are at a topological distance 3, and \u003cstrong\u003epositively contributes towards higher docking scores-based affinity of the PFAS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ePDI determines relevant interactions with the protein receptor zfVDAC2\u003c/em\u003e \u003csup\u003ed)\u003c/sup\u003e\u003c/p\u003e\n \u003cp\u003eNegative relation of PDI and docking scores implicates \u003cem\u003eincreased PDI improves PFASs/ zfVDAC2 channel interactions\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eHigh values of CATS2D_07_AL should be found in compounds with low docking scores\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003eIn long-chain PFAS compounds (\u003cem\u003ei.e.\u003c/em\u003e, with a topological distance of 7), \u003cem\u003ethe presence of sulfone moiety significantly reduces the binding affinity of PFAS to zfVDAC2\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eSimultaneous presence of lipophilic features at a specific distance with an acceptor feature like sulfone may further reduce such interactions\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003ePresence of B08[C\u0026minus;O], B10[C\u0026minus;C], and F06[S\u0026minus;F] in long chain length PFAS \u003cem\u003eincrease affinity interaction with zfVDAC2 associated to their potential ecotoxicity\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003ePFAS with polar atoms (oxygen) belonging to the poly-fluoroalkyl chains had higher values for CATS2D_03_AA, \u003cem\u003eimplying their roles in the PFAs/zfVDAC2 channel binding interactions\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eDevelop\u003c/em\u003e predictive 3D QSARs\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eProbe\u003c/em\u003e 3D-Descriptors potential in characterizing the interactions of PFASs with zfVDAC2\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eDetermine\u003c/em\u003e PFAS binding behavior\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eOpen3DQSAR\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e3D Countour maps\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eFractional Factorial Design based selection (FFD-SEL)\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eUninformative Variable Elimination based partial least squares (UVE-PLS)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePFAS score gradually decreases as it distances away from the favorable steric maps\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003ePFAS# 25: poly-fluoroalkyl chain is closer to the favorable steric fields\u003c/strong\u003e, and due to lack of any polar moiety it is benefitted by the uninterrupted interaction with this field (countour map)\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003ePFAS#29: worst ranked compounds preferably locate themselves in a small pocket containing the electrostatic fields for favorable interactions with their terminal polar moieties (countour map)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePFAS# 25: \u003cem\u003ehighest docking score against zfVDAC2 with the highest potential ecotoxicity\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003ePFAS# 29: \u003cem\u003elowest docking score against zfVDAC2 with the lowest potential ecotoxicity\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003ePFAS# 25: poly-fluoroalkyl long chain distances itself from the large positive electrostatic field to avoid unfavorable interactions, \u003cem\u003eterminal polar moiety perfectly locates itself near electrostatic fields increasing its chance for favorable polar docking interactions\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"5\"\u003e\n \u003cp\u003eExperimental/literature supporting evidence:\u003c/p\u003e\n \u003cp\u003e\u003csup\u003ea)\u003c/sup\u003e Bischel et al., 2018; Poothong et al., \u003cspan\u003e2017\u003c/span\u003e; Brendel et al., \u003cspan\u003e2018\u003c/span\u003e;\u003c/p\u003e\n \u003cp\u003e\u003csup\u003eb)\u003c/sup\u003e Hagenaars et al., \u003cspan\u003e2013\u003c/span\u003e; O\u0026apos;Brien et al., 2004; Mashayekhi et al., \u003cspan\u003e2015\u003c/span\u003e; Ankley et al., \u003cspan\u003e2021\u003c/span\u003e; Schredelseker et al., \u003cspan\u003e2014\u003c/span\u003e;\u003c/p\u003e\n \u003cp\u003e\u003csup\u003ec)\u003c/sup\u003e Ankley et al., \u003cspan\u003e2021\u003c/span\u003e; Schredelseker et al., \u003cspan\u003e2014\u003c/span\u003e; Choi et al., \u003cspan\u003e2017\u003c/span\u003e;\u003c/p\u003e\n \u003cp\u003e\u003csup\u003ed)\u003c/sup\u003e Motoc et al., \u003cspan\u003e1985\u003c/span\u003e; Todeschini et al., 2000; Todeschini et al., \u003cspan\u003e2009\u003c/span\u003e.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e"},{"header":"4. Conclusions","content":"\u003cp\u003eThe ecotoxicological profile of per- and poly-fluoroalkyl substances versus mitochondria channel, specifically the VADC2, is still challenging and emergent, with scarce experimental evidence. This work proposes a first tentative approach to this context, presenting a novel \u003cem\u003ein silico\u003c/em\u003e tool, to be contrasted versus future experimental data on the subject.\u003c/p\u003e \u003cp\u003eThis study investigates the interaction of PFAS with the ecotoxicologically relevant zebrafish mitotarget (zfVDA2), by resourcing to \u003cem\u003ein silico\u003c/em\u003e approaches, due to their broad regulatory acceptance for alternative ecotoxicity assessments. Hybrid structure-based virtual screening \u003cem\u003eplus\u003c/em\u003e predictive 2D/3D-QSAR models are proposed as a novel approach in order to theoretically explore the mechanistic interactions of PFAS with potential environmental impact mimicking PFAS bioaccumulation in low-concentration of exposure at the sub-cellular level (PFAS ecotoxicity). The structural validation-based Ramachandran plots revealed that the zfVDAC2 channel protein can be efficiently modeled with a high crystallographic quality (\u0026gt; 85%) and absence of flexibility restrictions for the zfVDAC2 binding residues. In addition, this hybrid approach does not rely exclusively on the docking results, rather it also incorporates computational work based on local perturbations under the unbound and bound state of the zfVDAC2. Such a feature extends the possibilities in terms of interpretability of interaction mechanisms beyond not only exclusively quantifying the energy of interaction (FEB values) and the binding participating amino-acid residues which is usually found in the traditional docking experiments, but also offering relevant information about the signals propagation in the zfVDAC2, defined by the collective anisotropic fluctuations in the intra-segment Cα-Cα atomic residue distances by the presence of a given ligand, such as ATP or PFAS. These latter could be biochemically relevant or unfavorable from a structural point of view whenever affecting the binding site native conformation and function of the zfVDAC2.\u003c/p\u003e \u003cp\u003eThe obtained results for the best and worst-ranked docked PFAS from a data set of 123 PFAS compounds showed a spontaneous thermodynamic binding process, with prevalence for non-covalent interactions mainly associated to non-covalent hydrophobic van der Waal interactions, followed by fluorine (F)-halogen-bond interactions and then hydrogen-bond interactions. Furthermore, different interaction patterns observed by the 2D-lig-plot interaction diagrams associated to the PFAS chain length, strongly suggests that long-chain PFAS (≥ 6 carbons) present a higher zebrafish ecotoxicity when compared with their PFAS analogs of short-chains. Results also indicate that PFAS could affect the communication efficiency in the network of binding residues in the voltage-sensing N-terminal α-helix segment involved in the conformational fit and regulation of the mitochondrial transport of the ATP molecule through zfVDAC2. What is more, the most significant PFAS’ molecular descriptor is the packing density index that is directly associated to their most dominant van der Waal interactions.\u003c/p\u003e \u003cp\u003eThis work demonstrated that some PFAS may have substantial potencies to bind on the zebrafish mitochondrial voltage-dependent anion channel (zfVDAC2). Therefore, these \u003cem\u003ein silico\u003c/em\u003e evidences open new horizons to improve the rational-design of PFAS with the required low-ecotoxicity, as well as forward better bioremediation strategies to prevent the negative impact of PFAS in aquatic organisms under environmental exposure, and to ensure a safe and sustainable use of PFAS applications.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eSupplementary Information (SI)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSupplementary material to this article can be found online at DOI: \u0026hellip;.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work received financial support from FCT/MCTES (UIDB/50006/2020 DOI 10.54499/UIDB/50006/2020) through national funds.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work received support and help from FCT/MCTES (LA/P/0008/2020 DOI 10.54499/LA/P/0008/2020 and UIDP/50006/2020 DOI 10.54499/UIDP/50006/2020), through national funds. Ana S. Moura further acknowledges FCT/MECS for the contract IF CEECIND/03631/2017.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCredit author statement\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMichael Gonz\u0026aacute;lez-Durruthy\u003c/strong\u003e: Conceptualization; Methodology; Investigation; Formal analysis; Validation; Visualization; Writing-original draft. \u003cstrong\u003eAmit Kumar Halder\u003c/strong\u003e: Methodology; Investigation; Formal analysis; Validation; Visualization; Writing review \u0026amp; editing. \u003cstrong\u003eAna Silveira Moura\u003c/strong\u003e: Validation; Writing review \u0026amp; editing. \u003cstrong\u003eMaria Nat\u0026aacute;lia Dias Soeiro Cordeiro\u003c/strong\u003e: Conceptualization; Methodology; Validation; Supervision; Funding acquisition; Writing review \u0026amp; editing.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eCompliance with Ethical Standards\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eEthical approval:\u0026nbsp;\u003c/strong\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eConsent to Participate:\u0026nbsp;\u003c/strong\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eConsent to Publish:\u0026nbsp;\u003c/strong\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAlderete, T. L., Jin, R., Walker, D. I., Valvi, D., Chen, Z. H., et al., 2019. Perfluoroalkyl substances, metabolomic profiling, and alterations in glucose homeostasis among overweight and obese Hispanic children: A proof-of-concept analysis. Environ. Int. 126, 445\u0026minus;453. https://doi.org/10.1016/j.envint.2019.02.047\u003c/li\u003e\n\u003cli\u003eAmbure, P., Aher, R. B., Gajewicz, A., Puzyn, T., Roy, K. 2015. \u0026quot;NanoBRIDGES\u0026quot; software: Open access tools to perform QSAR and nano-QSAR modeling. Chemometr. Intell. Lab. 147, 1\u0026minus;13. https://doi.org/10.1016/j.chemolab.2015.07.007 \u003c/li\u003e\n\u003cli\u003eAndersen, M. E., Clewell, H. J., Tan, Y. M., Butenhoff, J. L., Olsen, G. W., 2006. Pharmacokinetic modeling of saturable, renal resorption of perfluoroalkylacids in monkeys - Probing the determinants of long plasma half-lives. Toxicology 227, 156\u0026minus;164. https://doi.org/10.1016/j.tox.2006.08.004 \u003c/li\u003e\n\u003cli\u003eAnkley, G. T., Cureton, P., Hoke, R. A., Houde, M., Kumar, A., et al., 2021. Assessing the Ecological Risks of Per- and Polyfluoroalkyl Substances: Current State-of-the Science and a Proposed Path Forward. Environ. Toxicol. Chem. 40, 564\u0026minus;605. https://doi.org/10.1002/etc.4869 \u003c/li\u003e\n\u003cli\u003eBischel, H. N., MacManus-Spencer, L. A., Luthy, R. G., 2010. Noncovalent Interactions of Long-Chain Perfluoroalkyl Acids with Serum Albumin. Environ. Sci. Technol. 44, 5263\u0026minus;5269. https://doi.org/10.1021/es101334s \u003c/li\u003e\n\u003cli\u003eBowman, J. S., 2015. Fluorotechnology Is Critical to Modern Life: The FluoroCouncil Counterpoint to the Madrid Statement. Environ. Health Persp. 123, A112\u0026minus;A113. https://doi.org/10.1289/ehp.1509910 \u003c/li\u003e\n\u003cli\u003eBrendel, S., Fetter, E., Staude, C., Vierke, L., Biegel-Engler, A., 2018. Short-chain perfluoroalkyl acids: Environmental concerns and a regulatory strategy under REACH. Environ. Sci. Eur. 30, 9. https://doi.org/10.1186/s12302-018-0134-4 \u003c/li\u003e\n\u003cli\u003eBuck, R. C., Franklin, J., Berger, U., Conder, J. M., Cousins, I. T., et al., 2011. Perfluoroalkyl and polyfluoroalkyl substances in the environment: terminology, classification, and origins. Integr. Environ. Assess Manag. 7, 513\u0026minus;541. https://doi.org/10.1002/ieam.258 \u003c/li\u003e\n\u003cli\u003eButenhoff, J. L., Chang, S. C., Olsen, G. W., Thomford, P. J., 2012. Chronic dietary toxicity and carcinogenicity study with potassium perfluorooctanesulfonate in Sprague Dawley rats. Toxicology 293, 1\u0026minus;15. https://doi.org/10.1016/j.tox.2012.01.003 \u003c/li\u003e\n\u003cli\u003eChelcea, I. C., Ahrens L., \u0026Ouml;rn S., Mucs D., Andersson P. L., 2020. Investigating the OECD database of per- and polyfluoroalkyl substances \u0026ndash; chemical variation and applicability of current fate models. Environmental Chemistry 17, 498\u0026minus;508. https://doi.org/10.1071/EN19296 \u003c/li\u003e\n\u003cli\u003eChen, V. B., Arendall, W. B., Headd, J. J., Keedy, D. A., Immormino, R. M., Kapral, G. J., Murray, L. W., Richardson, J. S., Richardson, D. C., 2010. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr D Biol Crystallogr. 66, 12\u0026minus;21. https://doi.org/10.1107/S0907444909042073 \u003c/li\u003e\n\u003cli\u003eCheng, W. X., Ng, C. A., 2019. Using Machine Learning to Classify Bioactivity for 3486 Per- and Polyfluoroalkyl Substances (PFASs) from the OECD List. Environ. Sci. Technol. 53, 13970\u0026minus;13980. https://doi.org/10.1021/acs.est.9b04833 \u003c/li\u003e\n\u003cli\u003eChennubhotla, C., Bahar, I., 2007. Signal propagation in proteins and relation to equilibrium fluctuations. PLoS Comput. Biol. 3, 1716\u0026minus;1726. https://doi.org/10.1371/journal.pcbi.0030172 \u003c/li\u003e\n\u003cli\u003eChoi, E. M., Suh, K. S., Rhee, S. Y., Oh, S., Woo, J. T., et al., 2017. Perfluorooctanoic acid induces mitochondrial dysfunction in MC3T3-E1 osteoblast cells. J. Environ. Sci. Heal. A 52, 281\u0026minus;289. https://doi.org/10.1080/10934529.2016.1253402 \u003c/li\u003e\n\u003cli\u003eCousins, I. T., DeWitt, J. C., Glüge, J., Goldenman, G., Herzke, D., Lohmann, R., Miller, M., Ng, C. A., Scheringer, M., Vierke, L., Wang, Z., 2020. Strategies for Grouping Per- and Polyfluoroalkyl Substances (PFAS) to Protect Human and Environmental Health. Environ. Sci.: Processes Impacts, 22 (7), 1444\u0026minus;1460. https://doi.org/10.1039/D0EM00147C \u003c/li\u003e\n\u003cli\u003eCousins, I. T., Johansson, J. H., Salter, M. E., Sha, B., Scheringer, M., 2022. Outside the Safe Operating Space of a New Planetary Boundary for Per- and Polyfluoroalkyl Substances (PFAS). Environ. Sci. Technol., 56, 11172\u0026minus;11179. https://doi.org/10.1021/acs.est.2c02765 \u003c/li\u003e\n\u003cli\u003eDreier, D. A., Mello, D. F., Meyer, J. N., Martyniuk, C. J., 2019. Linking Mitochondrial Dysfunction to Organismal and Population Health in the Context of Environmental Pollutants: Progress and Considerations for Mitochondrial Adverse Outcome Pathways. Environ Toxicol Chem. 38, 1625\u0026minus;1634. https://doi.org/10.1002/etc.4453 \u003c/li\u003e\n\u003cli\u003eEbert, A., Goss, K.U., 2022. Screening of 6000 Compounds for Uncoupling Activity: A Comparison Between a Mechanistic Biophysical Model and the Structural Alert Profiler Mitotox. Toxicol. Sci. 185, 208\u0026ndash;219. https://doi.org/10.1093/toxsci/kfab139 \u003c/li\u003e\n\u003cli\u003eGirdhar, A., Bharathi, V., Tiwari, V. R., Abhishek, S., Deeksha, W., Mahawar, U. S., Raju, G., Singh, S. K., Prabusankar, G., Rajakumara, E., Patel, B. K. 2020. Computational insights into mechanism of AIM4-mediated inhibition of aggregation of TDP-43 protein implicated in ALS and evidence for in vitro inhibition of liquid-liquid phase separation (LLPS) of TDP-432C-A315T by AIM4. Int. J. Biol. Macromol. 147, 117-130. https://doi.org/10.1016/j.ijbiomac.2020.01.032.\u003c/li\u003e\n\u003cli\u003eGolbraikh, A., Tropsha, A., 2002. Beware of q2! J. Mol. Graph. Model. 20, 269\u0026minus;276. https://doi.org/10.1016/S1093-3263(01)00123-1 \u003c/li\u003e\n\u003cli\u003eGonz\u0026aacute;lez-Durruthy, M., Giri, A. K., Moreira, I., Concu, R., Melo, A., Ruso, J. M., Cordeiro, M. N. D. S., 2020. Computational Modeling on Mitochondrial Channel Nanotoxicity. Nano Today. 34, 100913. https://doi.org/10.1016/j.nantod.2020.100913 \u003c/li\u003e\n\u003cli\u003eGonz\u0026aacute;lez-Durruthy, M., Werhli, A.V., Seus, V. et al., 2017. Decrypting Strong and Weak Single-Walled Carbon Nanotubes Interactions with Mitochondrial Voltage-Dependent Anion Channels Using Molecular Docking and Perturbation Theory. Sci Rep 7, 13271. https://doi.org/10.1038/s41598-017-13691-8 \u003c/li\u003e\n\u003cli\u003eGramatica, P. 2007. Principles of QSAR models validation: internal and external. QSAR Comb. Sci. 2007, 26, 694-701. https://doi.org/10.1002/qsar.200610151 \u003c/li\u003e\n\u003cli\u003eGramatica, P. 2013. On the development and validation of QSAR models. Methods. Mol. Biol. 930, 499\u0026ndash;526. https://doi.org/10.1007/978-1-62703-059-5_21 \u003c/li\u003e\n\u003cli\u003eGuedes, I. A., Costa, L. S. C., dos Santos, K. B., Karl, A. L. M., Rocha, G. K., et al., 2021. Drug design and repurposing with DockThor-VS web server focusing on SARS-CoV-2 therapeutic targets and their non-synonym variants. Sci. Rep. 11, 5543. https://doi.org/10.1038/s41598-021-84700-0 \u003c/li\u003e\n\u003cli\u003eHagenaars, A., Vergauwen, L., Benoot, D., Laukens, K., Knapen, D., 2013. Mechanistic toxicity study of perfluorooctanoic acid in zebrafish suggests mitochondrial dysfunction to play a key role in PFOA toxicity. Chemosphere 91, 844\u0026ndash;856. https://doi.org/10.1016/j.chemosphere.2013.01.056 \u003c/li\u003e\n\u003cli\u003eHalder, A. K., Amin, S. A., Jha, T., Gayen, S., 2017. Insight into the structural requirements of pyrimidine-based phosphodiesterase 10A (PDE10A) inhibitors by multiple validated 3D QSAR approaches. SAR QSAR Environ. Res. 28, 253\u0026ndash;273. https://doi.org/10.1080/1062936X.2017.1302991 \u003c/li\u003e\n\u003cli\u003eHalder, A. K., Delgado, A. H. S., Cordeiro, M. N. D. S., 2022. First multi-target QSAR model for predicting the cytotoxicity of acrylic acid-based dental monomers. Dental Mater. 38, 333\u0026ndash;346. https://doi.org/10.1016/j.dental.2021.12.014 \u003c/li\u003e\n\u003cli\u003ehttps://comptox.epa.gov/dashboard/chemical_lists/pfasmaster.\u003c/li\u003e\n\u003cli\u003eJimenez, J., Doerr, S., Martinez-Rosell, G., Rose, A. S., De Fabritiis, G., 2017. DeepSite: protein-binding site predictor using 3D-convolutional neural networks. Bioinformatics 33, 3036\u0026ndash;3042. https://doi.org/10.1093/bioinformatics/btx350 \u003c/li\u003e\n\u003cli\u003eKrafft, M. P., Riess, J. G., 2015. Per- and polyfluorinated substances (PFASs): Environmental challenges. Curr. Opin. Colloid. Interface Sci. 20, 192\u0026ndash;212. https://doi.org/10.1016/j.cocis.2015.07.004 \u003c/li\u003e\n\u003cli\u003eLaskowski, R. A., Swindells, M. B., 2011. LigPlot+: Multiple Ligand-Protein Interaction Diagrams for Drug Discovery. J. Chem. Inf. Model. 51, 2778\u0026ndash;2786. https://doi.org/10.1021/ci200227u \u003c/li\u003e\n\u003cli\u003eLee, B. H., Seo, S., Kim, M. H., Kim, Y., Jo, S., et al., 2017. Normal mode-guided transition pathway generation in proteins. PloS One 12, e0185658. https://doi.org/10.1371/journal.pone.0185658 \u003c/li\u003e\n\u003cli\u003eLe Guilloux, V., Schmidtke, P., Tuffery, P., 2009. Fpocket: An open source platform for ligand pocket detection. BMC Bioinformatics 10, 168. https://doi.org/10.1186/1471-2105-10-168 \u003c/li\u003e\n\u003cli\u003eLiao, B., Zhou, C., Liu, T., Dai, Y., Huang, H. 2020. A novel Hericium erinaceus polysaccharide: Structural characterization and prevention of H2O2-induced oxidative damage in GES-1 cells. Int. J. Biol. Macromol. 154, 1460-1470. https://doi.org/10.1016/j.ijbiomac.2019.11.027. \u003c/li\u003e\n\u003cli\u003eMashayekhi, V., Tehrani, K. H. M. E., Hashemzaei, M., Tabrizian, K., Shahraki, J., et al., 2015. Mechanistic approach for the toxic effects of perfluorooctanoic acid on isolated rat liver and brain mitochondria. Hum. Exp. Toxicol. 34, 985\u0026ndash;996. https://doi.org/10.1177/0960327114565492 \u003c/li\u003e\n\u003cli\u003eMauri, A., Consonni, V., Pavan, M., Todeschini, R., 2006. Dragon software: An easy approach to molecular descriptor calculations. Match-Commun Math. Co. 56, 237\u0026ndash;248.\u003c/li\u003e\n\u003cli\u003eMotoc, I., Marshall, G. R., Dammkoehler, R. A., Labanowski, J., 1985. Molecular Shape Descriptors .1. Three-Dimensional Molecular Shape Descriptor. Z Naturforsch A, 40, 1108\u0026ndash;1113.\u003c/li\u003e\n\u003cli\u003eMuratov, E. N., Bajorath, J., Sheridan, R. P., Tetko, I. V., Filimonov, D., et al., 2020. QSAR without borders. Chem. Soc. Rev. 49, 3716\u0026ndash;3716. https://doi.org/10.1039/D0CS00098A \u003c/li\u003e\n\u003cli\u003eNelms, M. D., Mellor, C.L., Cronin, M. T. D., Madden, J. C., Enoch, S. J., 2015. Development of an in Silico Profiler for Mitochondrial Toxicity. Chem. Res. Toxicol. 28, 10, 1891\u0026ndash;1902. https://doi.org/10.1021/acs.chemrestox.5b00275 \u003c/li\u003e\n\u003cli\u003eOECD. Fact Cards of Major Groups of Per- and Polyfluoroalkyl Substances (PFASs). Series on Risk Management No. 68. Organisation for Economic Cooperation and Development, 2022.\u003c/li\u003e\n\u003cli\u003eOECD/UNEP. Toward a new comprehensive global database of per- and poly-fluoroalkyl substances (PFASs): Summary report on updating the OECD 2007 list of per- and poly-fluoroalkyl substances (PFASs), Organisation for Economic Cooperation and Development, 2018.\u003c/li\u003e\n\u003cli\u003eOECD/UNEP, Working towards a global emission inventory of PFASs: focus on PFCAS - status quo and the way forward. OECD Environment, Health and Safety Publications Series on Risk Management No. 30. Organisation for Economic Cooperation and Development, 2015.\u003c/li\u003e\n\u003cli\u003eO\u0026apos;Brien, T. M., Wallace, K. B., 2004. Mitochondrial permeability transition as the critical target of N-acetyl perfluorooctane sulfonamide toxicity in vitro. Toxicol. Sci. 82, 333\u0026ndash;340. https://doi.org/10.1093/toxsci/kfh244 \u003c/li\u003e\n\u003cli\u003eOjha, P. K., Roy, K., 2011. Comparative QSARs for antimalarial endochins: Importance of descriptor-thinning and noise reduction prior to feature selection. Chemometr. Intell. Lab. 109, 146\u0026ndash;161. https://doi.org/10.1016/j.chemolab.2011.08.007 \u003c/li\u003e\n\u003cli\u003ePatlewicz, G., Richard, A. M., Williams, A. J., Grulke, C. M., Sams, R., Lambert, J., Noyes, P. D., DeVito,M. J., Hines, R. N., Strynar, M., Guiseppi-Elie, A., Thomas, R. S., 2019. A Chemical Category-Based Prioritization Approach for Selecting 75 Per- and Polyfluoroalkyl Substances (PFAS) for Tiered Toxicity and Toxicokinetic Testing. Environ. Health Perspect. 127, 014501. https://doi.org/10.1289/EHP4555 \u003c/li\u003e\n\u003cli\u003ePoothong, S., Thomsen, C., Padilla-Sanchez, J. A., Papadopoulou, E., Haug, L. S., 2017. Distribution of Novel and Well-Known Poly- and Perfluoroalkyl Substances (PFASs) in Human Serum, Plasma, and Whole Blood. Environ. Sci. Technol. 51, 13388\u0026ndash;13396. https://doi.org/10.1021/acs.est.7b03299 \u003c/li\u003e\n\u003cli\u003eRaschka, S., 2018. MLxtend: Providing machine learning and data science utilities and extensions to Python\u0026apos;s scientific computing stack. J. Open. Source Software 3, 638. https://doi.org/10.21105/joss.00638 \u003c/li\u003e\n\u003cli\u003eReutlinger, M., Koch, C. P., Reker, D., Todoroff, N., Schneider, P., et al., 2013. Chemically Advanced Template Search (CATS) for Scaffold-Hopping and Prospective Target Prediction for \u0026apos;Orphan\u0026apos; Molecules. Mol. Inform. 32, 133-138. \u003c/li\u003e\n\u003cli\u003eRoy, K., Chakraborty, P., Mitra, I., Ojha, P. K., et al., 2013. Some case studies on application of \u0026quot;r(m)2\u0026quot; metrics for judging quality of quantitative structure-activity relationship predictions: emphasis on scaling of response data. J. Comput. Chem. 34, 1071-1082. https://doi.org/10.1002/jcc.23231 \u003c/li\u003e\n\u003cli\u003eScheringer, M., Trier, X., Cousins, I. T., de Voogt, P., Fletcher, T., et al., 2014. Helsing\u0026oslash;r statement on poly- and perfluorinated alkyl substances (PFASs). Chemosphere 114, 337-339. https://doi.org/10.1016/j.chemosphere.2014.05.044 \u003c/li\u003e\n\u003cli\u003eSchredelseker, J., Paz, A., Lopez, C. J., Altenbach, C., Leung, C. S., et al., 2014. High Resolution Structure and Double Electron-Electron Resonance of the Zebrafish Voltage-dependent Anion Channel 2 Reveal an Oligomeric Population. J. Biol. Chem. 289, 12566-12577. https://doi.org/10.1074/jbc.M113.497438 \u003c/li\u003e\n\u003cli\u003eSerra, A., Onlu, S., Festa, P., Fortino, V., Greco, D., 2020. MaNGA: A novel multi-niche multi-objective genetic algorithm for QSAR modelling. Bioinformatics 36, 145-153. https://doi.org/10.1093/bioinformatics/btz521 \u003c/li\u003e\n\u003cli\u003eSima, M. W., Jaffe, P. R., 2021. A critical review of modeling Poly- and Perfluoroalkyl Substances (PFAS) in the soil-water environment. Sci. Total Environ. 757, 143793. https://doi.org/10.1016/j.scitotenv.2020.143793 \u003c/li\u003e\n\u003cli\u003eShimizu, H. et al. (2015) Mitochondrial Ca\u003csup\u003e2+\u003c/sup\u003e uptake by the voltage-dependent anion channel 2 regulates cardiac rhythmicity. eLife 4, e04801. https://doi.org/10.7554/eLife.04801 \u003c/li\u003e\n\u003cli\u003eStockholm convention on persistent organic pollutants (POPs). http://www.pops.int/. 2004.\u003c/li\u003e\n\u003cli\u003eTama, F., Gadea, F. X., Marques, O., Sanejouand, Y. H., 2000. Building-block approach for determining low-frequency normal modes of macromolecules. Proteins 41, 1-7. https://doi.org/10.1002/1097-0134(20001001)41:1\u0026lt;1::AID-PROT10\u0026gt;3.0.CO;2-P \u003c/li\u003e\n\u003cli\u003eTamara, T., Vogs, C., 2021. Invited Perspective: PFAS Bioconcentration and Biotransformation in Early Life Stage Zebrafish and Its Implications for Human Health Protection. Environ. Health Persp. 129, 071304. https://doi.org/10.1289/EHP9625 \u003c/li\u003e\n\u003cli\u003eTetko, I. V., Tanchuk, V. Y., Villa, A. E. P., 2001. Prediction of n-octanol/water partition coefficients from PHYSPROP database using artificial neural networks and E-state indices. J. Chem. Inf. Comp. Sci. 41, 1407-1421. https://doi.org/10.1021/ci010368v \u003c/li\u003e\n\u003cli\u003eTodeschini, R., Consonni, V., Todeschini, R., Molecular descriptors for chemoinformatics. 2nd ed., Wiley-VCH ,John Wiley distributor: Weinheim Chichester, 2009.\u003c/li\u003e\n\u003cli\u003eTodeschini, R., Consonni, V., Handbook of molecular descriptors. Wiley-VCH: Weinheim, New York, 2000.\u003c/li\u003e\n\u003cli\u003eTosco, P., Balle, T., 2011. Open3DQSAR: a new open-source software aimed at high-throughput chemometric analysis of molecular interaction fields. J. Mol. Model. 17, 201-208. https://doi.org/10.1007/s00894-010-0684-x \u003c/li\u003e\n\u003cli\u003eWang, Z. Y., Cousins, I. T., Scheringer, M., Hungerbuehler, K., 2015. Hazard assessment of fluorinated alternatives to long-chain perfluoroalkyl acids (PFAAs) and their precursors: Status quo, ongoing challenges and possible solutions. Environ. Int. 75, 172-179. https://doi.org/10.1016/j.envint.2014.11.013 \u003c/li\u003e\n\u003cli\u003eVestergren, R., Cousins, I. T., 2009. Tracking the Pathways of Human Exposure to Perfluorocarboxylates. Environ. Sci. Technol. 43, 5565-5575. https://doi.org/10.1021/es900228k \u003c/li\u003e\n\u003cli\u003eYang, L. W., Chng, C. P., 2008. Coarse-grained models reveal functional dynamics-I. Elastic network models--theories, comparisons and perspectives. Bioinform. Biol. Insights 2, 25-45. https://doi.org/10.4137/bbi.s460 \u003c/li\u003e\n\u003cli\u003eYoo, W., Mayberry, R., Bae, S., Singh, K., He, Q.P., Lillard, J. W. Jr. 2014. A study of efects of multicollinearity in the multivariable analysis. Int. J. Appl. Sci. Technol. 4, 9-19.\u003c/li\u003e\n\u003cli\u003eZafeer, M. F., Firdaus, F., Ahmad, F., Ullah, R., Anis, E., Waseem, M., Ali, A., Hossain M. M. 2018. Perillyl alcohol alleviates amyloid-b peptides-induced mitochondrial dysfunction and cytotoxicity in SH-SY5Y cells. Int. J. Biol. Macromol. 109, 1029-1038. https://doi.org/10.1016/j.ijbiomac.2017.11.082.\u003cbr\u003e \u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Environmental risk assessment, Poly-fluoroalkyl substances, Zebrafish mitochondrial voltage-dependent anion channel, Virtual screening, Quantitative Structure-Activity Relationships modeling","lastPublishedDoi":"10.21203/rs.3.rs-4362510/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4362510/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe significance of effective and reliable prediction of ecotoxicity, particularly across various trophic levels, including humans, is gaining increasing prominence as ecosystems face new threats and challenges. Computational ecotoxicological predictive approaches have already been deemed as a swifter and economical feasible answer. This work presents a new proposal in that context, integrating structure-based virtual screening and quantitative structure-activity relationship (QSAR) methodologies to address the ecotoxicity of per- and poly-fluoroalkyl substances (PFAS) in aquatic organisms, such as zebrafish. By focusing on the interaction between PFAS and the zebrafish mitochondrial voltage-dependent anion channel (zfVDAC2), resembling bioaccumulation in low concentrations, we analyzed 123 PFAS compounds. Our findings reveal that the top-ranked docked PFAS exhibits a predominant affinity for van der Waal interactions, followed by fluorine (F)-halogen bonds and hydrogen bonds interactions. The latter suggests that PFAS interaction strength may influence mitochondrial ATP transport via zfVDAC2. Similarly, the derived QSAR models identified packing density index, a descriptor linked to van der Waal interactions, as the most significant PFAS factor. Moreover, the high predictive power and statistical robustness of these models positioning them as valuable tools for environmental risk assessment in PFAS applications, while offering mechanistic insights into ecotoxicity.\u003c/p\u003e","manuscriptTitle":"Mapping Mitochondrial Channel Toxicity: A Case Study for Predicting Mito-Target Interactions for the Per- and Poly-Fluoroalkyl Compounds on the Zebrafish Voltage-Dependent Anion Channel 2","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-02 20:47:51","doi":"10.21203/rs.3.rs-4362510/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"b11cc268-293f-4da4-8e3a-cf52a220f8a7","owner":[],"postedDate":"July 2nd, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-10-22T03:50:31+00:00","versionOfRecord":[],"versionCreatedAt":"2024-07-02 20:47:51","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4362510","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4362510","identity":"rs-4362510","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}