id
stringlengths 24
24
| idx
int64 0
402
| paragraph
stringlengths 106
17.2k
|
|---|---|---|
6760cd4381d2151a02fb5356 | 3 | DigCat's extensive data analysis also provides strong support for standardizing experiments and ensuring data reliability. By comparing the performance of common benchmark samples across multiple studies, users can easily identify outliers in the data and investigate potential causes. This helps streamline experimental procedures and data processing, ensuring more reliable and reproducible results. terials Database. By utilizing large language models to match the fundamental structures and properties of these materials with the demands found in literature, the AI Catalysis Scientist significantly broadens the potential materials range for catalyst prediction. |
6197f536a831ec3301d6c254 | 0 | A vast array of synthetic methods involving nickel catalysis has been developed in recent years. Methods that involve exogeneous reductants often are best accomplished with air-stable Ni(II) catalysts, which are desirable compared with more air-sensitive Ni(0) counterparts. While processes involving phosphine and pyridine ligand frameworks often perform well with Ni(II) precursors, 2 reactions that involve N-heterocyclic carbene ligands are more commonly performed with Ni(COD)2 as the precatalyst. |
6197f536a831ec3301d6c254 | 1 | This choice is due to inefficiencies in the catalyst reduction and formation of catalytically active Ni(0) species. At the same time, in situ formation of Ni(0)-NHC complexes has disadvantages of instability of Ni(0) precatalysts and NHC ligands, inhibitory effects of cyclooctadiene in some classes of catalytic processes, especially C-H activation processes involving LLHT activation mechanisms, and the precise control of metal-ligand stoichiometry, especially on the small scales required for high throughput experimentation. Ni(0) catalysts that are stabilized by simple alkenes have proven effective across many reaction classes, with more electron-deficient alkenes typically providing more stable but less active catalysts compared with the most commonly employed precursors such as Ni(COD)2. Catalyst 1, initially reported by Cavell, includes IMes as the NHC ligand and dimethyl fumarate as the stabilizing p-acid and serves as a prototypical example of the increased stability and diminished reactivity imparted by the electrondeficient alkene additives. Recent work from our lab built upon this template and examined acrylate, fumarate, and methacrylate p-acids to refine the stability-reactivity balance. While catalyst 1 possesses exceptional stability in air and performs effectively in the oxidation of secondary alcohols, we found that C-C bond-forming processes including aldehyde-alkyne reductive couplings and Buchwald Hartwig aminations were not effective due to deactivation of the catalyst by the fumarate. Systematic variation of the NHC and stabilizing p-acid led to the identification of a number of catalysts that participate effectively with these reaction classes and rapidly initiate without a discernable induction period. Our initial observations found that the optimal p-acid depends on the NHC ligand, and the stability-reactivity continuum can be optimized according to the precise catalytic properties and stability desired. Complexes 2-4 were found to display excellent catalytic properties in aldehyde-alkyne reductive couplings (catalysts 2 and 3) and Buchwald-Hartwig aminations (catalyst 4) and are now sold by commercial vendors (Figure ). |
6197f536a831ec3301d6c254 | 2 | Other classes of promising air-stable Ni(0) catalysts have subsequently been disclosed by Cornella and Engle, with Ni(0) centers stabilized by either stilbene or quinone p-acids. These catalysts possess the advantage of enabling modular in situ coordination to different ligands, whereas the NHC/p-acid combinations have the advantage of being a single-component system with pre-defined structure and metalligand stoichiometry, as illustrated with electronic-deficient akenes and other olefin classes. We envision that the latter characteristics will offer unique advantages with NHC catalysts in high-throughput arrayed methods where inefficient mixing and imprecise control of metal-ligand stoichiometry are avoided with a single-component, well-defined catalyst source. Prior studies from numerous laboratories have illustrated that judicious choice of alkene ligands can play a key role in tuning the stability and reactivity of numerous families of Ni(0) catalysts, a current gap in the field is the understanding of how Ni(0) complexes stabilized by simple ligands undergo activation to more active forms of the catalyst. A question that remains unanswered for most Ni(0) precatalysts is whether simple ligand dissociation affords active catalyst forms, or if more complex activation steps involving chemical modification of the alkene are required. We have now studied this question in detail with new fumarate catalysts that build on the design features of an NHC ligand paired with a p-acid selected to balance stability and reactivity. In this study, the fate of the stabilizing alkene ligand and mechanism of catalyst activation are elucidated through experimental and computational studies that evaluated different mechanisms for catalyst activation, including displacement as well as covalent sequestration of the stabilizing p-acid. |
6197f536a831ec3301d6c254 | 3 | In order to test this hypothesis and use this information to improve this family of catalysts, we set out to better understand the chemical principles that govern the relationship of stability and reactivity. The design strategy, based on this thermodynamic rationale, was to tune the fumarate binding affinity to offset the innate electronic stabilization of the electron-deficient alkene by steric interactions with the NHC. |
6197f536a831ec3301d6c254 | 4 | To test whether the thermodynamics of ligand binding were controlling the activation of the Ni(0) NHC complexes, seven different fumarate complexes of IMes (1-2, 5-10, Figure ) with varying electronics and sterics were considered. In the model reaction of 4-fluorobenzaldehyde (12) and phenyl propyne (13) using triethylsilane as the reductant, a potential first step for activation of the Ni complex is displacement of the two fumarate ligands with aldehyde and alkyne. We computed the free energy of fumarate ligand exchange with the aldehyde and alkyne reaction components using seven representative fumarate complexes and compared their binding affinities to Cavell's original complex (1) (Table ). Complex 1 was chosen as a reference point for this series, as it is known to be air-stable and was observed to be unreactive in the reductive couplings of aldehydes and alkynes and in Buchwald Hartwig aminations. We anticipated that if the mechanism of catalyst activation simply involves exchange of the fumarate 16 for the aldehyde and alkyne components (12 and 13), then the catalysts with lowest free energy of exchange will most easily reach the active catalyst state. As seen in Table , the complexes examined were found to have similar or higher fumarate binding affinities, relative to The hypothesis of catalyst activation through purely thermodynamic control considered with the relative binding energies from Table suggests that most fumarate complexes (except for 2) would be as inactive as complex 1 in reductive coupling reactions. Regardless, the increased steric bulk of the fumarates compared to 1 and variations in electronics of the aryl groups of 7-10 provided a significant range of binding energies. Therefore, this set could be used to better understand the relationship between fumarate binding and reactivity, and we experimentally tested a representative set of the compounds evaluated by computation. The model coupling reaction (Figure , top) was performed for representative precatalysts and monitored by 19 F NMR. Across the catalyst series 1-2 and 5-10, all of the tested fumarate complexes except for catalyst 10 were found to be more active than the parent dimethyl fumarate complex 1 (see SI). |
6197f536a831ec3301d6c254 | 5 | Figure shows that the rates and conversions were highest for those catalysts that possessed the bulkiest fumarate substituents (2, 6, 7). The reactivity ordering in Figure showed no clear relationship to the binding affinities detailed in Table . While the most reactive catalyst 2, which possesses a di-(t-butyl) fumarate ligand, also has the most favorable exchange energy for the 14 to 15 conversion, other complexes such o-tol precatalyst 7 exhibited excellent catalytic reactivity at room temperature, despite having a fumarate binding affinity that is too endergonic to undergo the 14 to 15 exchange under the reaction conditions (>14 kcal/mol above that of 1). This means that if ligand exchange were a necessary step for catalyst activation, 7 should be completely inactive. |
6197f536a831ec3301d6c254 | 6 | Unlike catalyst 7, catalyst 11 was found to have a weaker fumarate binding energy that was 4.2 kcal/mol uphill of catalyst 1. The inactivity of 11, despite having the same fumarate as activate catalyst 7, indicated that catalyst activity is not solely dependent on fumarate identity, but may also be affected by the interplay between the NHC ligand and the fumarate ligand. |
6197f536a831ec3301d6c254 | 7 | The above results indicate that ligand exchange is probably not the mechanism of catalyst activation for catalysts such as 7. Based on this information, we propose an alternative hypothesis: a chemical activation event is responsible for converting the precatalyst into the active catalyst. In other words, the fumarate in 7 and other precatalysts must be consumed through a chemical transformation of the p-acidic ligand prior to catalysis. |
6197f536a831ec3301d6c254 | 8 | We set out to test the hypothesis of ligand consumption by identifying the fate of the fumarate in the activation process. Specifically, we examined reactions with an elevated catalyst loading to allow the fate of the fumarate ligand to be tracked. Precatalyst 7 is readily prepared from Ni(COD)2, IMes, and di-(otolyl) fumarate, it possesses excellent stability and reactivity, and its structure (Figure ) is analogous to previously reported catalyst 2. In using 50 mol % loading of catalyst 7 in the three-component coupling of 4-fluorobenzaldehyde, 1-phenyl propyne, and triethylsilane, product 18 was isolated. 18 might result from a four-component reductive cycloaddition including the di-(o-tolyl) fumarate from the nickel catalyst, and byproduct 19 was observed by GCMS analysis (Scheme 1). The process resembles Et3B-mediated reductive cycloaddition involving enoates, alkynes, and aldehydes, 12 but has not been observed or proposed as a mechanism for catalyst activation. Given the complexity and uncertain mechanism of the formation of byproduct 18, we turned to computational reaction pathway evaluation tools to provide a clear explanation for these phenomena. To determine whether a given catalyst goes through activation path A or B, the highest energy transition states of both pathways need to be compared. In the case of catalyst 7, with an IMes ligand, the transition state for ketene elimination (IMes-TS-II-B, 22.0 kcal/mol) in path B is significantly higher in energy than the highest energy transition state in path A (IMes-TS-II-A, 15.1 kcal/mol), which suggests that catalyst 7 undergoes activation via path A. |
6197f536a831ec3301d6c254 | 9 | Additionally, the barrier for isomerization from h 3 bound BAC-III-A (BAC-TS-III-A, 19.5 kcal/mol) is moderately higher than the corresponding barrier for 7 (IMes-TS-III-A, 13.9 kcal/mol). Taken together, The above analysis suggests that the fumarate ligands of IMes precatalyst 7 and BAC precatalyst 11 react via different mechanisms. With this knowledge in hand, we then hypothesized that this difference can explain why 7 is a competent catalyst, but 11 is not. To evaluate this hypothesis, we followed the progression of path A in 7 and path B in 11 along the free energy surface. In the case of catalyst 7, path A provides a means to release a potential active catalyst. Figure details the pathway for catalyst release. Seven-membered metallacycle IMes-IV-A can ligate to an aldehyde (IMes-V-A, Figure ), and can then undergo an aldol reaction (IMes-TS-V-A) to yield complex The quantum chemical results shown in Table and and can be used to explain the competency of catalyst 7 in coupling of 12 and 13. Despite simple ligand exchange being thermodynamically unfeasible, catalyst 7 is competent in the production of allylic alcohol 17. This implies the catalyst activation route involves consumption of the fumarate, and the catalytic activity is predicated on the formation of a byproduct such as 18 (Scheme 1). This observation motivated us to experimentally isolate compound 18 to provide a test of the fumarate consumption hypothesis. A feasible reaction pathway leading to 18 is outlined in Figures and. |
6197f536a831ec3301d6c254 | 10 | In short, catalyst 7 is competent because it can undergo a reaction that removes its (strongly bound) fumarates from solution. This observation also provides a putative reason as to why BAC catalyst 11 is incompetent in similar reductive couplings to form 12 or 13. Computational investigation of the activation pathways for 14 suggest that a ketene-first path is preferred, in contrast to the aldol-first path preferred by 11. As a result of this change, catalyst 11 is can form highly stabilized complex BAC-IV-B (Figure ). |
6197f536a831ec3301d6c254 | 11 | In summary, we introduce a new Ni(0) catalyst ( ) complexed with IMes and two stabilizing di-(otolyl) fumarate ligands, and we demonstrate it to be a competent catalyst in the reductive coupling of aldehydes and alkynes using silanes as the terminal reductant. The catalyst is easily prepared and handled, while undergoing rapid catalyst activation under mild reaction conditions. Computational study of a panel of catalysts that range in stability and catalyst activity illustrated that simple dissociation of fumarate ligand was unlikely to serve as the catalyst activation step, as the thermodynamics of ligand exchange are uncorrelated with catalyst activity. Instead, consumption of the fumarate through a cascade cycloaddition process involving the reaction components was identified as a likely pathway for catalyst activation. |
6197f536a831ec3301d6c254 | 12 | Computational studies elucidated the operative mechanism for the catalyst activation step and provided a predictive model for explaining the divergent reactivity of catalysts that possess similar structures but that undergo different activation mechanisms. This work continues to advance the development of highly active and well-defined Ni(0) catalysts that provide improvements in stability and ease of handling over the corresponding structures obtained through in situ catalyst preparations. |
65cf64d99138d23161435c69 | 0 | Societal adoption of the Internet of Things (IoT) technologies has been both rapid and ubiquitous, with the number of devices predicted to be in use expected to reach more than 29 billion by 2030. These technologies have wide-ranging applications, in both industrial and consumer market places. This rapid growth of the IoT market is driving the need for additional, more functional and reliable sensors that can operate and communicate on IoT networks. Sensor devices typically employ a variety of transducer methods such as optical, electrical, and/or microelectromechanical (MEMS)-based detection mechanisms. However, optical and MEMS sensor devices require their transduced signals to be Electrochemical sensors and systems have been utilised and simulated theoretically throughout the years, however, the majority of this work has been undertaken using quiescent conditions. |
65cf64d99138d23161435c69 | 1 | Incorporation of flow has long been understood to boost electrochemical sensor responses by replenishing oxidised/reduced analyte at a sensor surface. The addition of convection, as an additional mode of mass transport, in combination with Fickian diffusion, facilitates more rapid delivery of an analyte to a sensor surface; leading to more analyte molecules/ions in contact with a sensor surface over a set timeframe. Furthermore, convection also effectively removes oxidised / reduced analyte from a sensor surface. Both these traits lead to the diffusion limitations, commonly observed in quiescent systems, being surmounted. Consequently, the use of fluid flow combined with electrochemical sensors is an interesting area of research. Most commonly, the incorporation of fluid flow, hydrodynamics, with electrochemical sensors has been achieved using a rotating ring-disk electrode set-up. Incorporating microfluidic components on a chip surface is a growing area of research. Concerning the latter, the dominant method employs soft-lithography to define microfluidic elements, typically polydimethylsiloxane (PDMS), as first reported by Xia and Whitesides. In this approach both the substrate and microfluidic element are oxygen plasma treated, to create Si-OH bonds on both surfaces, aligned and brought into conformal contact with each other such that covalent Si-O-Si bonds form between the two components through a condensation reaction. For example, Ko et al, amongst others, employed this method for carbon nanotube-based electrode sensors, patterned onto a polyethylene terephthalate surface. In that work, a PDMS microfluidic device was bonded to their substrate by treating both with O2 plasma, bringing them into conformal contact overnight allowing covalent bonds to form. This approach works well with oxidised surfaces but is problematic when trying to attach to other dielectric surfaces, such as silicon nitride, due to the lack of oxygen atoms in the Si3N4 layer. Nitride is the preferred passivation layer for use with silicon chip based electrochemical sensors, as it is a dense, non-porous layer and, as such, does not allow electrolyte molecules ingress into the layer thereby preventing unwanted parasitic capacitances. To this end, new microfluidic production methods, such as xurography and 3-D printing, are being explored by different research groups. With advances in additive production capabilities, the use of micro-scale channels with electrochemical sensors has now become more facile with less fabrication expertise required compared to the photolithography driven microfabrication techniques previously employed. Often, designs require complex multi-step processes that can be time consuming, Swensen, et al. requiring a multi-step process and the use of hazardous chemicals thereby restricting the approach to the laboratory settings. Progress has also been made in reducing the time required for microfluidic production, such as through the use of xurography as shown by Speller, et al. however, these methods still require multi-step processes. Each step presents a potential source of error with alignment issues, and with human input required for several steps. 3-D printing in particular, addresses the issues of microfluidic fabrication and integration with sensors, and also has the ability to produce multiple reproducible components in parallel. |
65cf64d99138d23161435c69 | 2 | In this work, we employ resin 3-D printing to rapidly create reliable and reusable microfluidic components and combine these with our silicon chip -based electrochemical sensor platform. We minimise potential errors by automating the entire microfluidic production process and developed a single step process for alignment with the on-chip sensors. To understand the effect of flow on solid state electrodes operating in generator-collector mode (analogous to ring-disc systems) we undertake a finite element study on the application of electrochemical pH control under hydrodynamic systems. We then confirm this theoretical study experimentally and demonstrate its feasibility through pH dependent chlorine detection under these hydrodynamic conditions. An iterative feedback approach allowed the microfluidic elements to be tweaked and optimised. An additional advantage of the 3D-printed modular approach is reusability, i.e., it allows the system to be disassembled, cleaned and a new chip inserted as required. This is possible as the elements are not irreversibly covalently bonded to a chip surface (such as in soft-lithography) thereby increasing the sustainability of this approach. A commercial colorimetric free-chlorine detector was used to confirm results. |
65cf64d99138d23161435c69 | 3 | Silicon chip based devices were fabricated as described by Dawson et al, with the subsequent modifications proposed by Seymour et al 71 also included. Each sensor consisted of two combs of gold interdigitated ultramicroelectrodes, as working electrodes, as well as platinum pseudo reference and gold counter electrodes. Chips were designed with a peripheral microSD pinout to permit interfacing with external electronics via a microSD port. All devices were fabricated on 4-inch silicon wafers bearing a thermally grown 300 nm silicon dioxide layer. Blanket metal evaporation of Titanium (10 nm) and Gold (100 nm) (Temescal FC-2000 E-beam evaporator) and lift off technique to yield interdigitated microband comb structures. The first comb (WE1) had 33 tines, while the second comb two (WE2) had 34 tines. Each comb was 1 µm wide, 178 µm long, 100 nm high and had an inter-electrode gap of 2 µm. A second metal evaporation and subsequent lift-off step was undertaken to pattern the interconnection tracks, contact pads and the gold counter electrode (90 µm x 7mm). A third and final metal evaporation was performed to pattern the platinum pseudo reference electrode (90 µm x 7mm). |
65cf64d99138d23161435c69 | 4 | Silicon nitride was then blanket deposited using plasma enhanced chemical vapour deposition across the whole wafer to act as an insulating layer. This insulating layer was used to prevent unwanted electrochemical interactions along the connection tracks. To allow electrolyte access to the ultramicroband electrodes, vias (windows) were opened in the insulating Si3N4 layer (above the sensor electrodes using photolithography and dry etching. Additional vias were opened over the counter and pseudo-reference electrodes as well as the periphery microSD contact pads. Each device contained three separate interdigitated electrode sensors which are separated by 1.88 mm. This separation was maintained to prevent cross talk, i.e., electrochemical reactions occurring at one sensor from interfering with a neighbouring sensor (when operating without the microfluidic component) and also to provide sufficient surface area between the electrodes to facilitate correct integration of the microfluidic channels to prevent leaking and cross talk. Following fabrication, the wafer was diced to produce 28 separate chip dice. |
65cf64d99138d23161435c69 | 5 | A 3D-printed modular microfluidics platform was designed to enable solutions to be flowed over electrode surfaces. This system, shown in greater detail in Figure , consisted of three components: a clear tough resin (Formlabs Inc clear resin) for both the top and bottom layers and a soft flexible resin (Formlabs Inc flexible 80a resin) for the middle 'gasket' layer. The microfluidics channels were defined in this flexible middle layer. All components were designed using Solidworks™ software. The designs were 3D-printed using a Formlabs Inc 3+ printer. Polytetrafluoroethylene (PTFE) tubing (Merck Life Science Ltd) with an outer diameter of 1.5 mm was connected to a stainless steel dispensing tip 20G (Merck Life Science Ltd) to and connected to the microfluidic channel. A 10 mL syringe (Sigma Aldrich) was filled with the solution being analysed and a digital programmable pump (NE-1000) allowed for liquid to be dispensed at controlled volumes and flowrates. The accuracy of this digital pump was confirmed by flowing a known volume of water into a pre-weighed beaker for a set amount of time and then measuring the mass increase and comparing against expected mass using a calibrated four-point balance (OHaus Pioneer PA213) |
65cf64d99138d23161435c69 | 6 | Following fabrication, each chip was inspected using optical microscopy to identify any visible defects or faults and any defective chips discarded. Prior to electrochemical characterisation, chips were first cleaned by rinsing with acetone followed by de-ionized water. The chips were subsequently dried using nitrogen and inserted into a microfluidic holder. Electrochemical analysis was performed using an Autolab Bipotentiostat (MAC80150 with BA Module, Metrohm), within a Faraday cage. Cyclic voltammograms (CV) were performed in the voltage range of -0.15 V to 0.45 V at 50 mV/s at the first interdigitated comb (WE1) in 1 mM/L ferrocene carboxylic acid (FCA, Sigma Aldrich, 97%) while the second interdigitated comb of electrodes, (WE2) was biased at -0.15 V; for the duration of the scans. |
65cf64d99138d23161435c69 | 7 | To accurately calibrate the pH control, an external Ag/AgCl reference was used. A beaker was placed downstream containing the external reference and filled with the appropriate buffer. The outlet tube for the microfluidic device was submerged in these buffer solutions to ensure a closed circuit. Chemically buffered solutions were prepared by mixing various ratios of citric acid monohydrate and monosodium phosphate (Sigma Aldrich). The pH of each solution was determined using a calibrated pH probe (Thermo Scientific Orion star A211 pH meter). A phosphate buffer (PB) solution was prepared using 268.8 mg sodium phosphate monobasic and 110.2 mg sodium phosphate dibasic, then making this solution up to 1 L using deionised water. This generated a pH 7.4 10 mM solution of PB, which was used for all electrochemical pH control buffer experiments where a potential window of -0.2V to 1.6V was used, with a scan rate of 50 mV, varying a potential bias at WE2 for the electrochemical pH control. |
65cf64d99138d23161435c69 | 8 | Artificial drinking water (ADW) samples were prepared to simulate real drinking water conditions but without any residual chlorine present. ADW was prepared by dissolving 1 g of sodium bicarbonate, 0.0654 g of magnesium sulphate (Sigma Aldrich, 99.5% anhydrous), 0.3414 g calcium sulphate dehydrate (Honeywell, 99%), 0.007 g potassium phosphate dibasic (Fluka, 98%), potassium phosphate monobasic (Sigma Aldrich, 99%) and 0.01 g sodium nitrate (Sigma Aldrich, 99%) in 10 L of deionised water. For localised pH control, voltammograms were performed in ADW samples by scanning WE1 from -0.2 V to 1.2 V at 50 mV/s with WE2 biased at ~1.6 V to protonate (acidify) the local environment of the sensing electrodes via hydrolysis of water. The local pH can then be readily controlled by adjusting the applied potential to drive the appropriate current, as shown by O'Sullivan et al. In acidic conditions, the pH can be changed through the water splitting reactions of Eqs. ( ) and (2): |
65cf64d99138d23161435c69 | 9 | Diffusion simulations of proton concentrations in the vicinity of the protonator electrodes (WE2) were performed using Fick's second law. A model was designed to simulate the generation of protons at the protonator electrodes, and their subsequent diffusion from these electrodes, using finite element analysis software, Comsol Multiphysics 6.0 ™, in line with the galvanostatic model shown previously. The model boundary conditions were set at as a box 930 µm high by 5 mm wide, representing the experimental domain. Two sets of interdigitated 1 µm wide microband electrodes (34 protonator electrodes and 33 sensing electrodes), separated by 2 µm were defined at the bottom boundary layer. |
65cf64d99138d23161435c69 | 10 | By fixing the anodic current to the protonator (WE1), a flux of protons was applied at the surface of the protonators, and the flux was assumed to be proportional to the current applied to the electrodes. The initial pH within the experimental domain was set to pH 7. The proton diffusion coefficient used for the simulation was 9.31x10 -5 cm 2 s -1 . Once the initial static diffusion simulations were undertaken, the addition of a single-phase |
65cf64d99138d23161435c69 | 11 | Based on the work of Seymour, et al. the electrochemical pH control method based on oxygen evolution was employed to locally control the pH for chlorine analysis within a channel without any fluid flow. This allowed electrochemical pH control of samples via the hydroylysis of water; leaving a net excess of protons. Similarly, the voltage at which the gold oxide reductive peak minimum occurred was used to approximate the local pH at a sensor. This approach was benchmarked against a set of know standards using a chemically controlled sample altered using different buffers. The targeted pH for chlorine detection was selected to be pH 3 as this was shown to be an area of high specificity for hypochlorous acid detection by Seymour, et al. A fixed potential bias of 1.6V was applied to WE2 to drive the local pH down to 3. Initial scans were performed in a known concentrations of free-chlorine added to ADW, which was electrochemically acidified to pH 3, in order to establish the appropriate cyclic voltammetry parameters. Working chlorine samples were prepared by diluting Milton Sterilising Fluid (2% Sodium Hypochlorite) to the desired concentration. The concentration of free-chlorine in these samples were then measured and confirmed using a standard commercial free-chlorine colorimeter (Pocket Colorimeter II 58700-00 with CL2 Test Kit), prior to electrochemical analysis. CV's were performed in the voltage range 0.85 V to -0.2 V (versus Pt pseudo-reference) at a 50 mV/s scan rate. All scans were repeated in quadruplet, with the first scan disregarded as a conditioning step and the subsequent three scans averaged. Cleaning scans were performed in-between each concentration to prevent any possible electrode fouling and improve reproducibility. These scans consisted of disconnecting WE1 and applying a typical CV 0.85 V to -0.2 V (versus Pt pseudo-reference) was applied to WE2 for 4 scans at a 50 mV/s scan rate. By applying these scan parameters to WE2, any potential fouling was minimised, ensuring a relatively constant electrode surface across all concentrations. WE1 was left disconnected to minimise unnecessary usage and therefore prolong the sensor lifespan. |
65cf64d99138d23161435c69 | 12 | Having optimised the free-chlorine pH control conditions within a channel, under quiescent conditions, the effect of flowrate was investigated. The microfluidic system was connected to a digitally programmable syringe pump to permit quantified flowrates through the channels. All CV parameters remained the same as the static conditions described above, except for the protonator bias, which was increased by 100 mV to 1.7V. This voltage increase was to drive a higher protonator current to offset the hydrodynamic removal of some of the generated protons by flow which was observed to decrease the efficacy of the pH control. By generating extra protons, the pH was shifted back to the desired pH within the vicinity of the sensors; despite the effect of flow. Cleaning scans were again performed between each set of measurements. |
65cf64d99138d23161435c69 | 13 | Figure (A) shows an exploded view of the Solidworks™ schematic for the microfluidic platform. The system consists of three separate parts; a hard base layer with a slot to hold the microchip as well as two pillars to ensure accurate alignment of the sensor chips. a middle gasket component, made of a soft flexible resin, which comes into conformal contact with the chip surface. this gasket had three microfluidic channels designed to align with the three sensor locations on a chip; as well as three inlet and outlet apertures to allow fluid flow. The third, hard, top layer was designed to apply a constant pressure to the middle gasket to prevent leakage. This constant pressure was ensured by locking the top component into place using two pegs to retain downwards pressure on the system, effectively sealing it. |
65cf64d99138d23161435c69 | 14 | Figure (B) shows an optical micrograph plan view of the assembled system containing a chip, and three microfluidic channels in the correct orientation with the inlets and outlets for sample flow over the sensors. Tubes are omitted for clarity. A silicon chip device consisting of three sensors and on-chip counter and reference electrodes is shown in Figure (A) (Supp info) while S1 (B) (Supp info) shows an assembled microfluidic system, with the inlet and outlet tubes removed for clarity. Typically, inlet and outlet tubes would be connected to allow flow through the selected microchannel over a selected sensor. The distal ends contact pads of a microchip were connected to an external SD connector port, as shown. This allowed facile connection to an Autolab potentiostat. |
65cf64d99138d23161435c69 | 15 | The generator was cycled from -0.15V to 0.45V while the collector was held at -0.15V. The generator comb oxidised the FCA to FCA + . The FCA + species then diffused across the gap to the collector electrodes, where it was subsequently reduced back to FCA. This phenomenon is known as redox cycling and can be used to boost signals when using reversible redox molecules; as described by Seymour, et al. The CV steady-state behavioural shape observed for the static FCA scan in Figure (A): inset, arises from REDOX cycling as it surmounts the diffusion limited behaviour typically associated with transport of fresh analyte to the electrode. As a flowrate across an electrode surface is introduced and increased, the measured current response signals were observed to increase with increasing flow rate. However, the steady-state plateau of signal, typically associated with quiescent solutions, is replaced with an elongated current response at higher voltages with increasing flowrate. This is due to the addition of convection as a mode of mass transport, as opposed to the diffusion only conditions experienced in quiescent solutions. The convection allows for constant replenishment of fresh FCA at an electrode surface, thereby increasing the current response. Each voltammogram presented in Figure (A) is an average of three consecutive cycles for each flowrate: inset is a typical CV obtained under quiescent conditions. The generation and subsequent reduction of gold oxide was used as an indicator of pH, as shown previously by Seymour et al. Chemical buffers in the pH range of 3 -7.6 were characterised electrochemically to monitor the position of the reduction peak for each buffer, in both static and flow conditions (100 µL/min); Figure (A) and (B) show these, respectively. Under both static and flowing conditions, little to no difference observed in the minima positions of the gold oxide reduction peaks. |
65cf64d99138d23161435c69 | 16 | This suggests that flowing conditions has little effect on pH. This is further illustrated in Figure where two calibrations of gold oxide reduction potential minima versus pH for both hydrodynamic conditions are plotted and overlaid and exhibit very similar behaviour within the investigated pH range. Once the effects of flowrate on desired pH had been compensated for, electrochemical analysis was then undertaken. Hypochlorous acid was selected as an analyte of interest due to its importance is water sterilisation. A stock solution of 5.3 ppm was prepared and used to assess the impacts of flowrate. |
65cf64d99138d23161435c69 | 17 | For each flowrate, the pH was electrochemically controlled by applying a potential bias of 1.65V to the protonator. This allowed the local pH to be set to ~5. This pH is lower than pH 6 the value where all free-chlorine is converted to hypochlorous acid; see bias, as shown previously in Figure (B) and (C). This suggests that there is great potential for future work on the coupling of these phenomena. |
60e078d25cb3f6238591e424 | 0 | The present outbreak of Covid-19 infection due to SARS-CoV-2, a virus from the coronavirus family, has become a major threat to human healthcare. This occurrence of SARS-CoV-2 is the third highly pathogenic event and large-scale epidemic affecting the human population. It follows the severe acute respiratory syndrome coronavirus (SARS-CoV) in 2003 and the Middle East respiratory syndrome coronavirus (MERS-CoV) in 2012. SARS-CoV 2 was first reported in December 2019 and after that it has spread all over the world infecting almost 44 million people till date. Moreover, it has been fatal to more than 1.1 million people. The higher rate of transmission of SARS-CoV-2 compared to the SARS-CoV is most probably associated with S-glycoprotein in the receptor-binding domain (RBD) region. The spike S glycoprotein of coronavirus facilitates the binding to the target cell and hence the efficiency of receptor binding domain(RBD)/ angiotensin-converting enzyme 2 (ACE2) is determinant to SARS-COV 2 transmissibility. It is a high time for all researchers in drug development to develop or repurpose FDA approved drug or an epitope to circumvent the current situation of pandemic. SARS-CoV2 is an envelope virus and it contains a positive-sense single-stranded RNA (ssRNA) genome (26-32 kb). The genomics data of pathogens are very important to obtain information regarding potential targets relevant for therapy or diagnostics. Hence in order to design and develop drug against such viral pathogens, it is necessary to start with data mining of viral genomes. The SARS-CoV-2 genome is made of less than 30000 nucleotides and contains genes for 29 different proteins and 10 open reading frame (ORF). The ORF1ab constitutes two third of the viral RNA and encodes as many as 16 non-structural proteins. Some of the key proteins encoded by this gene are PL pro (NSP2), 3CL pro (NSP5), RdRp (NSP12), and O-methyl transferase (NSP16) which play a vital role in the replication and transcription. The structural proteins such as membrane protein (M), envelope protein (E), spike protein (S), nucleocapsid protein (N) and other auxiliary proteins are encoded in ORF2-10. The RNA gene is packed within the N protein and M, E and S protein make the viral coat. Further the entry protein i.e Spike protein in involved in host cell recognition and binds specifically to the Angiotensin-converting enzyme 2 (ACE2) mammalian receptor. Therefore, for therapeutic approach all the structured and non-structured viral protein can be considered as potential target. However, the non-structured proteins PL pro (NSP2), 3CL pro (NSP5), RNA-dependent RNA polymerase (NSP12), and O-methyl transferase (NSP16) that are responsible for viral replication as well as transcription are more essential to target. Again, the infectivity and transmission capacity of SARS-CoV-2 to host cell depends largely on the S glycoprotein. Preclinical studies have predicted that glucose plays a vital role in replication of SARS-CoV-2. SARS-CoV-2 increases the demand of glucose in cells by upregulating the metabolic processes in host cell. Two such important metabolic processes in human cells are glycolysis and glycosylation. These features may have led to the idea that drugs inhibiting glycolysis and glycosylation might prove efficient in the context of SARS-CoV-2. One of the most widely studied glycolytic inhibitors is 2-deoxyglucose (2-DG), a synthetic glucose analogue in which the hydroxyl group at the second carbon atom is replaced by hydrogen ( Hence, to understand the mechanism in two different competitive pathways of therapeutic action of 2-deoxy D-glucose on SARS-CoV-2 infected hosts, we need to take into consideration its interaction with non-structured viral proteins involved in translation and replication of SARS-CoV-2 on one hand and its inhibition mechanism of the glycolysis pathway. In this work, we have used our fully automated integrated platform Prescience in silico Multi-Target Multi-Ligand Enhance Sampling Screening (PRinMTML-ESS) to determine the mode of therapeutic action of 2-DG in SARS-CoV-2 treatment. PRinMTML-ESS is a combined computational approach where docking, molecular dynamic simulation and the free energy calculations using enhanced sampling methods are used to explore the energetics of target-ligand complexation associated with the protein's or an enzyme's binding site (including the ligand) in an explicit solvent. To explore the antiviral property of 2-DG its binding interaction has been studied using our platform with the non-structured viral protein PL pro (NSP2), 3CL pro (NSP5), RNA-dependent RNA polymerase (NSP12), and O-methyl transferase (NSP16) which plays a major role in replication/translation of virus. The effect of 2-DG on the glycolysis process that converts glucose into ATP has been probed using the same combined computational approach and targeting the enzyme hexokinase that phosphorylates glucose to glucose-6-phosphate. Due the structural similarities between 2-DG and glucose (Scheme 1), it is expected that 2-DG would act as a competitive inhibitor of glucose metabolism and thus might strongly affect the SARS-CoV-2 virus replication and activity, which largely depends on the ATP generated from glycolysis in human cells. |
60e078d25cb3f6238591e424 | 1 | In our study, we have considered the possibility of 2-deoxy D-glucose (2-DG) working as an anti-viral drug and its role of a competitive inhibitor of glycolysis pathway. A key stage of the SARS-CoV-2 life cycle is the replication of the viral genome within the infected cells. It is a complex process involving the action of several viral and host proteins in order to perform RNA polymerization, proofreading and final capping. To investigate the antiviral property of 2-DG we have selected four viral proteins which are crucial actors of the replicatory machinery of SARS-CoV-2 as targets. The crystal structural data for these 4 viral proteins PL pro (PDB ID: 6wuu) , 3CL pro (PDB ID: 6lu7) , RNA-dependent RNA polymerase (PDB ID: 7bv1) , and O-methyl transferase (PDB ID: 6wkq) of SARS-CoV-2 are available now. The interaction of 2-DG with all the four antiviral proteins was computationally investigated using PRinMTML-ESS platform () which uses the combination of molecular docking at the binding site of protein followed by all atom molecular dynamic simulation of the best-docked pose in water for determining the stability of the protein-ligand bound structure and finally enhanced free energy sampling for better understand the binding interaction of the ligand with the protein. The same methodology has been used for screening large number of new chemical candidates for 3CL pro in our earlier work and the multi-target multi-ligand approach has been used in our high throughput analysis of literature derived repurposing drug candidates for SAR-CoV-2 that can be used to target the genetic regulators known to interact with viral proteins based on experimental and interactome studies In our study, we have initially analysed the interaction of 2-DG with four viral proteins 3C-like protease (3CL pro ), non-structural protein 16 (Nsp16), papain-like cysteine protease (PL pro ) and RNA-dependent RNA polymerase (RdRp) which take part in replication and translation mechanism. Establishment of the viral replication and transcription complex (RTC) that includes, amongst others, RNA-processing and RNA-modifying enzymes and an RNA proofreading function essential for maintaining the integrity of the >30kb coronavirus genome is crucial for virus replication and thus a promising target for antivirals against SARS-CoV-2. One of such target is 3CL pro which resides in non-structural protein-5 (nsp5). 3CL pro releases majority of the non-structural proteins from polyprotein and is crucial for viral life cycle. It's another role is inhibition of interferon signalling. PL pro releases non-structural protein (nsp1, nsp2, nsp3) and the amino terminus of nsp4 from the polyproteins pp1a and pp1ab. PL pro helps SARS-CoV-2 in evading the host innate immune responses by stripping ubiquitin and ISG15 from host-cell proteins. Therefore, targeting PLpro with antiviral drugs may have an advantage for inhibiting viral replication and also inhibiting the dysregulation of signalling cascades in infected cells that may lead to cell death in surrounding, uninfected cells. 2'-O-methyltransferase plays an essential role in immune evasion. Nsp16 achieves this functionality by mimic its human homolog, CMTr1 which methylates mRNA to enhance translation efficiency. However, Nsp16 requires a binding partner, Nsp10, to activate its enzymatic activity unlike CMTr1. The most important part of the coronavirus RTC is RdRP residing in nsp12 and this viral protein is suggested as a promising drug target as it is a crucial enzyme in the virus life cycle both for replication of the viral genome but also for transcription of sgRNAs. Molecular docking was started with automated generation of gridbox at binding site of the viral proteins PL pro (binding region residues: Asp-164, Gly-163, Cys-111, His-272, Gly-271, Tyr-268) obtained from a study that revealed the inhibitory mechanisms and determined the crystal structure of inhibitor VIR250 and VIR251 in complex with SARS-CoV-2 PL pro . A study by Yang et al. determined the crystal structure of 3CL pro of SARS-CoV-2 in complex with the inhibitor (N3). They have predicted the bonding site of inhibitors (binding region residues: Cys-145 to His-41 cavity) which we have used for the grid generation. Yin et al. determined the crystal structure of polymerase bound to RNA and to the drug Remdesivir. The binding site information of RdRP (Gly-683, Ser-682, Asp-684, Asp-760, Ser-759, Ala-685, Tyr-689, Ser-814, Gln-815) has been obtained from this study and used for grid generation. Rosas-Lemus et al. unravelled the crystal structures 2'-O-methyltransferase, the nsp16-nsp10 heterodimer, in complex with an inhibitor Sinefungin. Sinefungin binds to Nsp-16 at the binding site (ASN-6899, ASP-6897, ASP-6928, ASN-6841, CYS-6913, TYR-6930) which has been used for grid generation. In all the protein ligand combination the best docked pose was found be residing at that particular binding site, whose information we have obtained from crystal structure (Figure ). The binding energy of the best docked pose for 3CL pro -2-DG(6lu7-2-DG), NSp-16-2-DG(6wkq-2-DG), PL pro -2-DG(6wuu-2-DG) and RdRp-2-DG(7bv1-2-DG) are listed in Table . Docking scores were found to be -2.35 kcal/mol for 3CL pro -2-DG(6lu7-2-DG) ). Average RMSD value lower than 0.2 nm clearly shows that all the protein ligand combinations are stable. The hydrogen bonding section plots the number of hydrogen bonds between the ligand and the residues at the binding site of the target. NSp16-2-DG(6wkq-2-DG) is found to have the highest number of hydrogen bonding interaction, six followed by two hydrogen bonding in RdRp-2-DG(7bv1-2-DG) and no hydrogen bonding in PL pro -2-DG(6wuu-2-DG), and 3CL pro -2-DG(6lu7-2-DG) combination (Figure ). In the Interaction Energy section, the non-bonded interaction energies, coulomb and van der Waals interaction energies, between the ligand and protein are plotted for each combination in the platform. |
60e078d25cb3f6238591e424 | 2 | The total non-bonded interaction energy was found to -197.7 kJ/mol for NSp16-2-DG(6wkq-2-DG), hence the most favourable interaction of ligand 2-DG occurs with O-methyl transferase (NSp-16) of SARS-CoV-2. Total non-bonded interaction energy for 3CL pro -2-DG(6lu7-2-DG), RdRp-2-DG(7bv1-2-DG) and PL pro -2-DG(6wuu-2-DG) are found to be -97.1 kJ/mol, -66.2 kJ/mol and -10.3 kJ/mol respectively (Table ). These energy values predict that only in NSp16(6wkq) the ligand 2-DG is stable and tends to be in the binding site. However, for the other three viral proteins 3CL pro , RdRp and PL pro the interaction energy is much lower to stabilise the ligand in the binding cavity. The next section of MD analysis is the centre of mass distance which calculates the distance between the centre of mass of the residues in the binding site with the centre of mass of the ligand. Average centre of mass distance if larger than 1 nm would predict the ligand has escaped the binding pocket after MD simulation. |
60e078d25cb3f6238591e424 | 3 | The average centre of mass (COM) distance plot shows that the combinations NSp16-2-DG(6wkq-2-DG) and RdRp-2-DG(7bv1-2-DG) have value 0.47 nm and 0.16 nm respectively, hence lies within the binding site cavity at the equilibrated structure after MD simulation. (Table ) However, in both the combination 3CL pro -2-DG(6lu7-2-DG) and PL pro -2-DG(6wuu-2-DG) the average COM distance are 1.27 nm and 1.08 nm respectively, hence here the ligand 2-DG have escaped the binding sites because of less interaction strength with the binding site residues of protein. The final equilibrated structure of the four combinations and visualization of the trajectory is consistent with the COM distance results. The equilibrated structure of NSp16-2-DG(6wkq-2-DG) have three hydrogen bonding interaction between 2-DG and the binding residues Asp-6897, Asp-6928 and Gly-6871. (Figure ). The equilibrated structure of RdRp-2-DG(7bv1-2-DG) has two hydrogen bonds with Ser-682. Our platform generates a combination score based on the average RMSD, internal energy and hydrogen bonding data obtained from analysis of MD simulation. Based on this score the selection of combinations for running the enhanced sampling free energy simulation is done. Among the four combination we have selected NSp16-2-DG(6wkq-2-DG) and RdRp-2-DG(7bv1-2-DG) combination based on their combination score 0.51 and 0.55 respectively, since only in these two combinations the ligand is still in the binding cavity after MD simulation. The sampling around the binding sites in instance of an MD simulation is not sufficient to predict the stability of the ligand in the binding cavity of protein, as conformations might get stuck in local minima. Therefore, meta-dynamics an enhanced sampling method of simulation for quantitatively predicting ligand binding energy and analysis of changes in the conformation of ligands is important to ascertain the most stable (bound) protein-ligand complex and predicting the inhibition property of the ligand. The equilibrium structures obtained from MD simulations were used as the starting configurations in the enhanced sampling free energy simulations (FES). The average free energy of dissociation obtained from 5 independent dissociation simulations for the combination NSp16-2-DG(6wkq-2-DG) and RdRp-2-DG(7bv1-2-DG) are 29.1 kJ/mol and 26.4 kJ/mol (Table ). For clarity and understanding of the free energy surfaces the individual free energy surfaces of all the independent runs and the average surface is shown in Figure . The free energy surfaces predict that the surfaces have multiple local minima and one global minimum for the ligand bound at the binding cavity. This structural characteristic represents multiple interactions between the ligands and residues of the binding site. The free energy of dissociation of the ligand 2-DG is of same order for both the viral protein NSp16(6wkq) and RdRp(7bv1), however the combination NSp16-2-DG(6wkq-2-DG) has the maximum ligand dissociation free energy. However previous studies show that antiviral drugs like Carfilzomib, Eravacycline, Valrubicin, Lopinavir, Elbasvir and Ritonavir interacts with SARS-COV-2 viral proteins 3CL pro , PL pro , RdRp and spike protein where total binding free energy ranges from -79.4 kJ/mol to -168.1 kJ/mol with the van der Waal On the other hand, several studies have shown glucose plays a major role in proliferation of SARS-CoV-2. This is one of the reasons why obese and diabetic patients with uncontrolled blood glucose level are more prone to develop a severe form of COVID-19. Due to the structural similarity between 2-DG and glucose (Scheme 1), it is expected that 2-DG would act as the competitive inhibitor of glucose metabolism and would strongly effect the metabolic processes like glycolysis and glycosylation which are dependent on glucose. When SARS-CoV-2 attacks our cells, they co-opt both of these metabolic processes to increase its replication and transcription. SARS-CoV-2 induces an anabolic state in their host cell which causes these infected cells to upregulate their production of energy using glycolysis as compared with their healthy neighbours. 2-DG differs from the glucose by removal of an oxygen atom from the 2-position (Scheme 1). In the glycolysis 2-DG is absorbed by the cell and undergoes phosphorylation at the 6 position to generate 2-DG-6P, in the next step unlike glucose-6-phosphate, 2-DG-6P cannot undergo isomerisation by glucose-6-phosphate isomerase hence the glycolytic flux is reduced (Scheme 2). Therefore, in presence of 2-DG, the rate limiting reaction of glycolysis is ATPdependent phosphorylation of glucose to form glucose-6-phosphate (G-6-P) and is catalysed by tissue-specific isoenzymes known as hexokinases. A recent study showed that SARS-COV-2 infection induces higher amount of glucose influx and glycolysis in the infected cells, resulting in selective high accumulation of the fluorescent glucose/2-DG analogue in the viral infected cells. They further observed that mannose inhibit the entry of 2-DG analogue at a very low concentration hence predicted that 2-DG entry in virus-infected cells might be manipulating specific mannose transporter or high-affinity glucose transporter, GLUT3, which was found to be increased on SARS-CoV-2 infection. In our study we have tried to explore the competitive inhibition property of 2-DG on the first step of phosphorylation with the aid of hexokinase enzyme after entering the viral infected cell. The 2-DG and glucose cyclic conformers were docked at the binding site of hexokinase (Thr-210, Thr-232, Ser-155, Lys-173).The binding energy was found to be -7.72 kcal/mol and -7.33 kcal/mol for glucose and 2-DG respectively. The entropic contribution associated with the solvent effect and conformational changes in the docked ligand is not accounted in the docking. The best docked pose and MD simulation is carried out in the next step. The non-bonded total interaction energy in the equilibrated structure were found to be slightly lower in glucose, -330.2 kJ/mol than 2-DG, -257.2 kJ/mol. The Coulomb energy and Van der Walls energy was also found to follow the same trend (Table ). RMSD was computed for the target binding pocket residues is the initial MD configuration and the MD equilibrated structure for both the 2-DG-hexokinase and glucose-hexokinase structure (Table ). The Centre of mass distance 0.50nm and 0.38 nm for glucose and 2-DG indicated that the ligands are stable in the binding cavity. Number of hydrogen bonding interaction in the equilibrated structure in glucose and 2-DG are ten and seven respectively, the interacting residues are indicated in Table and Figure . To increase the sampling around the binding site of the protein and calculate the more accurate binding energy of the ligand considering the changes in conformation of the ligand the equilibrium structure obtained from MD simulations was used as the starting configuration in the enhanced wt-metaD simulation. The average free energy barrier of dissociation obtained from enhanced wt-metaD simulations was found to be 66.3 kJ/mol and 52.4 kJ/mol for glucose and 2-DG (Table ). Hence, it can be predicted that since the free energy barrier is comparable and of same order in both glucose and 2-DG, they have the similar binding affinity to hexokinase enzyme. Therefore, the phosphorylation step of conversion of glucose to glucose-6-phosphate is inhibited in presence of 2-DG at a certain concentration. Thus, competitive inhibition is caused by 2-DG, which structurally similar to glucose and can combine at the same binding site of hexokinase. For clearly understanding the free energy surface the average and the individual surfaces are plotted in PRinMTML-ESS platform as shown in Figure . The free energy surfaces show jagged and uneven surfaces arising due to the multiple interaction of the ligand at the binding site. The MD simulation was extended to 50 ns keeping all other parameters same as the 2 ns simulation. However negligible change in the ligand configuration and interaction of the ligand with the binding residues at the final equilibrated structure. |
60e078d25cb3f6238591e424 | 4 | In this study for carrying out all the calculations and generating the figure the Prescience in silico Multi Target Multi Ligand Enhanced Sampling Screening (PRinMTML-ESS) has been used. In PRinMTML-ESS the process flow is in-build and fully automated, so no manual intervention is required from the user-side. The process flow here consists of molecular docking for docking the ligands in by automated grid generation at the target binding site. To explore the stability of the target-ligand complex all-atom molecular dynamic simulation is carried out in the next step and enhanced free energy sampling at the last stage for final analysis of the binding property of the ligand with the target (protein/enzyme/DNA-RNA). Efficient scoring algorithms are used after Docking and MD simulation for screening at each stages. All molecular docking calculation and MD simulations are performed using Autodock 4 and GROMACS-5.1.4 simulation package. The enhanced sampling using metadynamics (metaD) and its variant well-tempered metadynamics (wt-metaD) using , is widely used in simple and complex molecules due to its advantage over free energy perturbation and thermodynamic integration methods which are computationally more expensive for larger complex system like protein-ligand, protein-protein binding. A time-dependent bias is included to the system in a metadynamic simulation along with some deposited bias like a suitably chosen reaction coordinate(s) which will eventually push the complex away from the minimum energy state, so that the system doesn't get trapped at a local minima for sufficiently long time. Since this method is independent on the choice of reaction coordinate and not very sensitive to the precise choice of biasing parameters (except in the case where the parameters are chosen to be too high or low), any rudimentary reaction coordinate can bias the system helping it to escape the local minima within a small time and generate a qualitative free energy surface (FES). Metadynamics simulation uses a historydependent bias which inhibits the system from repeated revisiting previously visited regions of the phase space. Another, widely used scoring method over molecular dynamic simulation is the molecular mechanics Poisson-Boltzmann surface area (MMPBSA) which uses approximations to calculate enthalpic and entropic contributions using implicit continuum solvent hence less accurate in comparison to enhanced free energy sampling method used in our study. Additionally, as an alternative to MMPBSA we have performed all-atom simulation in explicit solvent medium including dynamics of solvent, solute and ions. Considering the effectiveness and computational viability of enhanced free energy sampling method, it has been used as the final scoring method for calculating free energy barrier for dissociation in PRinMTML-ESS platform. Inhouse codes have been used for generating the analysis in each section of computational methods used. A scheme of the methodologies and analysis used in PRinMTML-ESS platform are shown in Scheme 2. The structure of ligands used in this study has been downloaded from PubChem (). The crystal structure of the viral proteins of SAR-COV2 PLpro (PDB ID: 6wuu), 3CLpro (PDB ID: 6lu7), RNA-dependent RNA polymerase (PDB ID: 7bv1) and the Hexokinase enzyme (PDB ID: 1QHA) are obtained from RCSB Protein Data Bank(). All the protein and enzyme structures were cleaned (removing ligands, ions, water molecules) and missing loop was modelled using our PRinBio platform. A ligand conformational search is carried out generating an automated grid box of the dimension 60*60*60 Å by taking the binding site centroid as a grid center with a spacing of 0.375 Å. The docking was conducted the Lamarckian genetic algorithm (LGA), and a total of 100 GA-LA hybrid runs for performing the conformational search. In the MD simulation the proteins and enzyme were modelled using the CHARMM27 force field 28 parameters. For generating the CHARMM27 force field for all the ligands, SwissParam was used and the force field generation is automated in PRinMTML-ESS platform. The target-ligand systems were solvated in water and equilibrated using MD simulations at room temperature. The systems were first equilibrated using an NVT ensemble at 300 K for 1 ns and extended to the NPT ensemble at 300 K and 1 atm for another 2 ns. The temperature and pressure during the simulations were maintained using a velocity rescaling thermostat and Parrinello-Rahman barostat, respectively. A time step of 1 fs was used to integrate the equation of motion, and a nonbonded cutoff of 10Å was used to perform the MD simulations. We have used a bias V(s,t) in the form of Gaussians with every 500 steps (1 ps) deposition pace with a Gaussian hill-height of 2.0 kJ/mol, width of σ (0.1 nm), bias factor of 15, and temperature (T) of 300 K. For smooth convergence of the system in a wt-metaD the amplitude of the bias is tuned accordingly. Here a tempering factor ΔT (Equation ) is used to adjust the height of the hills, and henceforth a smooth convergence of the free energy landscape is attained. |
60e078d25cb3f6238591e424 | 5 | Once the system converges, free energy F(s) (Eq. ( )) can be extracted by adding the deposited hills along the biased reaction coordinates. The centre of mass distance between the heavy atoms of the ligands and the protein backbone in the surrounding area of the binding pocket is considered as the reaction coordinates in the free energy sampling, since the aim of the calculation is taking into account the ligand dissociation from the binding site. In this study 5 independent simulations are done and an averaging is done for each combination to obtain better sampling and statistically reliable results. |
60e078d25cb3f6238591e424 | 6 | This computational study aims to understand the role of 2-deoxy-D-glucose as an antiviral and glycolysis pathway inhibitor in SAR-CoV-2 affected human body using PRinMTML-ESS platform. Using this platform molecular docking, all-atom molecular dynamics simulations with enhanced free energy simulation was performed and screening was done in each stage using effective scoring algorithm. Here we have explored the inhibition property of 2-DG with the four SAR-CoV-2 viral proteins 3CL pro , PL pro , NSp-16 and RdRp which are important for transmission and replication of SARS-CoV-2. 2-DG has very low dissociation barrier with O-methyl-transferase (Nsp16) and RdRp in free energy simulation. Hence it can be predicted that 2-DG can barely inhibit any viral protein taking part in replication and translation of SARS-CoV-2. On the other hand, PRinMTML-ESS platform predicts 2-DG binds strongly to hexokinase, the enzyme responsible for phosphorylation of glucose to glucose-6-phosphate in glycolysis. Though glucose has higher binding affinity to hexokinase, comparable binding affinity of 2-DG may lead to competitive inhibition of the glycolysis, hence the glycolytic flux is reduced which in turn reduces the replication and transcription of SARS-CoV-2 virus in human cells. Therefore, our study predicts that 2-DG though doesn't show any comparable inhibition for non-structured protein associated with translation and replication of SAR-CoV-2 can be used as a clinical therapy in COVID-19 infected patient for competitive inhibition of glycolysis process in virus infected cells, thereby reducing the rate the replication of virus. |
6720011f98c8527d9ea8fe4a | 0 | α-synuclein is a protein normally involved in synaptic vesicle transport and exocytosis for neurotransmission . Aggregation of α-synuclein is a primary feature of the synucleinopathies, such as Parkinson's Disease (PD), multiple system atrophy (MSA) and Lewy Body dementia (LBD) . Neuropathological examination of PD and LBD brains reveals the formation of intra-cellular inclusions called Lewy Bodies (LBs) . LBs are spherical structures consisting of aggregated α-synuclein in an amyloid conformation , lipids, organelles and other proteins, such as cytoskeletal proteins . While the pathological importance of LBs has been well known for decades, the mechanism behind their formation remains incompletely understood . α-synuclein molecules within LBs are mostly post-translationally modified, with ubiquitination, phosphorylation and truncations being particularly prominent . C-terminal truncations of α-synuclein species are expected to occur post fibril formation, demonstrated in neuronal models capable of recreating LB-like inclusions . The strongly negatively charged C-terminal region of α-synuclein is located outside the fibril core of the amyloid conformation and constitutes part of the so-called "fuzzy coat" . Within recent years an increasing number of amyloid core structures have been determined at an atomic level by cryo-EM techniques . However, the fuzzy coat is effectively invisible or only available in low resolution to modern imaging techniques due to its highly dynamic and disordered nature . This means that, in many cases, a large part of the total amino acid sequence, and more than half of the total sequence in the case of α-synuclein fibrils remains structurally uncharacterized. The fuzzy coat remains important for the biophysical properties of amyloid, having significant effects on kinetic parameters and can even govern the mechanical and adhesive properties of amyloid fibrils . The oligomeric assemblies of α-synuclein, which are suspected to be the drivers of amyloid neurotoxicity, are indeed also formed by fibril-surface catalyzed pathways . Furthermore the disordered flank regions may control the access of chaperones to the fibril core, such as Hsp70 which is able to depolymerize α-synuclein amyloid fibrils under ATP consumption . Increased knowledge of the roles of the fuzzy coat and its implications for amyloid kinetics and cellular interactions would therefore provide an important step towards understanding the role of amyloid in disease as well as identifying new targets for drug development. It is therefore important to establish experimental approaches which allow for systematic investigation of the biophysical properties of the fuzzy coat. Indeed Ulamec and coworkers recently called for the development of assays of protease treated fibrils, with the fuzzy coat "shaven off", to study which parts of the fibril are driving binding and intracellular interactions . |
6720011f98c8527d9ea8fe4a | 1 | Here, we demonstrate an approach which allows for the systematic investigation of biophysical properties of amyloid fibrils subjected to protease treatment. We show that classical bulk measurement approaches of protease treated amyloid fibrils are exceedingly challenging due to flocculation and loss of ThT sensitivity of α-synuclein fibrils, but surface-based biosensing approaches, such as Quartz Cystal Microbalance with Dissipation (QCM-D) provides a suitable experimental platform for such investigations. We compare the growth rate of α-synuclein amyloid fibrils with and without the fuzzy coat "shaven off", and we propose a consistent molecular model that explains the sustained increase in aggregation rate observed after the proteolytic treatment through enhanced secondary processes. |
6720011f98c8527d9ea8fe4a | 2 | To obtain α-synuclein for the experiments, BL21 (DE3) E. Coli cultures, carrying the pT-7 plasmid encoding the WT human αsynuclein gene, were grown overnight. The culture was used to inoculate 1 L of LB-Amp media in a 3 L flask and was grown at 37 • C at 180 rpm shaking. When OD 600 = 0.8 was reached protein expression was induced by adding IPTG to a final concentration of 1 mM. The induced cells were incubated for 4 h at 37 • C at 180 rpm shaking. The cells were harvested by centrifugation at 4 • C, 7000 x g for 20 min and stored at -20 • C until further use. Bacterial pellets corresponding to 1 L of culture were resuspended in 20 mL of 10 mM Tris-HCl, 1 mM EDTA, pH 8.0 with 1 mM PMSF. The suspension is sonicated with a probe sonicator for 2 min at 10 s intervals with 30 s pause at 40% amplitude. 1 µL Benzonase was added to the cell lysate and centrifuged at 4 • C, 20000 x g for 30 min. The supernatant is collected and the solution is boiled for 20 min. The solution was subsequently centrifuged at 4 • C, 20000 x g for 20 min to precipitate the heat-sensitive proteins, α-synuclein remaining in the supernatant. Next, 4 mL saturated (NH 4 ) 2 SO 4 was added for 1 ml supernatant to salt out α-synuclein. The solution was stirred at 4 • C for 15 min and centrifuged at 4 • C, 20000 x g for 20 min to pellet down the protein. The pellet is dissolved in 7 ml of 25 mM Tris-HCl pH 7.7 and 7 µL DTT is added to the final concentration of 1 mM. Next, the protein solution was dialyzed against the same buffer for 16-18 h at 4 • C with a change of the buffer after 12 h of dialysis. The dialyzed protein solution was then subjected to an anion exchange coloumn (AEC) (Hi-Trap Q Hp 5 ml, GE Healthcare) followed by size exclusion chromotagraphy (SEC) (HiLoad 16/600 Superdex 200 pg. column) and eluted in 10 mM of sodium phosphate buffer (pH 7.4). Protein concentrations were assessed by measurement of absorption spectroscopy at 280 nm (using a ProbeDrum instrument, Proba-tionLabs, Lund, Sweden) and the concentrations calculated using theoretical molar extinction coefficients predicted by ProtParam (Expasy, Switzerland). |
6720011f98c8527d9ea8fe4a | 3 | De novo α-synuclein fibrils are prepared by incubating 500 µL of 100 µM α-synuclein monomer in 20 mM sodium phosphate buffer with 150 mM NaCl in a 1.5 mL Eppendorf tube that is incubated at 37 • C for 14 days. The solution is shaken at 1200 rpm with a 1 mm diameter glass bead using a Eppendorf Ther-moMixer. All subsequent fibril solutions are prepared from 5% equivalent monomer mass seeding from the original fibril sample. 500 µL of 100 µM α-synuclein monomer, in a 1.5 mL Eppendorf tube is incubated at 37 • C in 20 mM sodium phosphate buffer with 150 mM NaCl for 3 days. The solution is shaken at 1200 rpm with a 1 mm diameter glass bead. In order to prepare the seeds, α-synuclein fibrils were sonicated for 5 minutes (25 minute cycle, 2s sonication, 8s break) on ice using a MS72 probe sonicator at 10% amplitude. |
6720011f98c8527d9ea8fe4a | 4 | Bulk aggregation kinetics of α-synuclein amyloid fibrils were performed in a FLUOstar Omega fluorescence microplate reader (BMG Labtech, Germany). All aggregation experiments were performed in 20 mM sodium phosphate buffer with 150 mM NaCl at pH 7.4. Amyloid aggregation was monitored under quiescent conditions at 25 • C by exciting the sample at 440 nm and recording the emission at 480 nm every 10 minutes. α-synuclein amyloid fibrils were grown at 50µM monomer concentration with 5% sonicated amyloid seeds. Aggregation was monitored for two of the six wells by addition of 100 µM Thioflavin T (ThT) and fibrils were grown until a plateau was reached. Half the wells were incubated with 200 nM proteinase K (PK) for 30 minutes. 100 µM PMSF is added to all wells as a PK inhibitor and incubated overnight. 50 µM monomer is added to each well and 100 µM ThT to all ThT-free wells and the monitoring of the aggregation reaction is resumed until a plateau in ThT fluorescence is reached. |
6720011f98c8527d9ea8fe4a | 5 | Capillary flocculation assays were performed using square glass capillaries of 0.4 mm inner diameter and walls of 0.2 mm (Vitro-Com, Mountain Lakes, NJ, USA). Homogenized α-synuclein amyloid seeds of 25 µM equivalent monomer concentration in 20 mM sodium phosphate buffer, 150 mM NaCl and 50 mM ThT, were mixed with 20 nM PK, loaded into a capillary and monitored for 18 hrs at 5 min intervals. The capillary was imaged using an inbuilt ThT channel in the Zeiss Axio vert. A1 microscope (Zeiss, Germany) with a 10X objective lens equipped with a CFP filter cube (model no. 424931, ex. 436/20, beam splitter 455, emission 480/40), illuminated using a Visitron Cool LED pE100 (Visitron Systems, Germany) operating at 440 nm. Flocculation is quantified from relative standard deviation of ThT intensity across the area of the inner capillary. |
6720011f98c8527d9ea8fe4a | 6 | α-synuclein fibrils at 100 µM equivalent monomer mass are treated with 2-iminothiolane at 0.15 mg/mL for 5 minutes prior to being incubated with an UV-ozone activated QCM gold-sensor for 1 hr. The sensors are rinsed in buffer and then incubated with 1 vol. % mPEG-thiol in buffer for 30 min. Sensors are rinsed with miliQ water and dried prior to being inserted into the instrument. All treatment of fibrils in the QCM instrument is performed at 20 mM sodium phosphate buffer with 150 mM NaCl. We measure the elongation rate of amyoid fibrils by injecting 3 cell volumes (60 µl) of monomeric protein solutions at 50µM in 20 mM sodium phosphate buffer with 150 mM NaCl and 1 mM PMSF into different sensor chambers and monitor the third overtone frequency. PMSF is added to ensure that residual active Proteinase K is inhibited. PMSF is added to the α-synuclein solutions from a 0.2 M stock dissolved in ethanol, leaving a residual volume fraction of 0.5 % ethanol. Fibril "shaving" is achieved by injecting 3 cell volumes of 20 nM Proteinase K and subsequent incubation. Proteinase is washed out by flushing the cell with at least 20 cell volumes of buffer. To inhibit remaining proteinase, 3 cell volumes of 5 mM PMSF are injected then and incubated for at least 10 minutes. The cell is then flushed with at least 20 cell volumes of buffer. |
6720011f98c8527d9ea8fe4a | 7 | Human serum Protein (HsP) binding experiments are performed by injecting 3 cell volumes of 1 % HsP solution into the sensor chamber and monitoring the third overtone frequency. Complex media growth experiments are measured by injecting 3 cell volumes of 1 % Hsp + 50 µM α-synuclein into the sensor chamber and monitoring the third overtone frequency. |
6720011f98c8527d9ea8fe4a | 8 | No surface binding or secondary nucleation takes place in this model. In addition to elongation, Model B also allows monomers to bind to the surface when no disordered flanks are present. In Model C, surface binding enables secondary nucleation and subsequent growth. After every proteinase K degradation step new nucleation-viable binding sites become available. Model D assumes that protofilaments can be proteolytically fragmented by Proteinase K degradation. Formal descriptions of the models are provided in the SI. |
6720011f98c8527d9ea8fe4a | 9 | In order to investigate the significance of the fuzzy coat (disordered flanks) on the kinetic parameters of α-synuclein amyloid growth, we have performed bulk ThT aggregation experiments. We grew fibrils in quiescent conditions to maximize fibril length, and hence the ratio of surface area to the number of ends, prior to protease treatment. After "shaving" with proteinase K (pK) and subsequent inhibition of pK activity by PMSF we add fresh monomer to the samples that had ThT from the beginning, and ThT and fresh monomers were added to the samples that had been grown in the absence of ThT. Initial growth rates after the addition of fresh monomer are similar for all samples, however the increase in ThT fluorescence intensity of shaven fibrils ceases rather abruptly, and the ThT fluorescence decreases afterwards (Fig. ). In order to investigate whether the observed apparent cessation of fibril growth stems from a genuine decrease of the elongation rate constant, decreased ThT sensitivity or rather from increased higher order assembly of shaven fibrils, we compared the colloidal behavior of unmodified and shaven fibrils in a microscopic capillary assay. The fibrils were subjected to probe sonication right before the measurement, in order to start from a homogeneous fluorescence distribution, whereby no granularity was observed with our microscope. |
6720011f98c8527d9ea8fe4a | 10 | We find that the ThT intensity of the shaven fibrils decreases by over 50 % within the shaving phase (100 min). Additionally we find that the shaven fibrils are more prone to higher order assembly (flocculation) than non-shaven fibrils, inducing larger spatial variability in ThT intensity across the microcapillary, compared to the sample without pK treatment (Fig. ). We quantify the degree of flocculation through the spatial standard deviation of fluorescence intensity of the entire field of view of the capillary. This higher order assembly is likely to affect the accessibility of the fibril ends for monomer. The strong tendency of shaven α-synuclein fibrils to undergo higher order assembly is unsurprising, given the highly negatively charged and disordered C-terminal tail is removed, which otherwise provides electrostatic and steric repulsion of other molecules. It is reasonable to assume that removal of the fuzzy coat not only affects the higher order assembly of fibrils and interactions with small molecules, such as ThT, but surface interactions with the monomeric protein as well, as charges are removed and the hydrophobic core becomes more accessible. These various effects conspire to render a quantitative comparison of the intrinsic elongation rates of intact and shaven fibrils in such bulk solution assays very challenging. In addition, we can-not fully exclude residual activity of the inhibited PK in such a bulk solution assay. |
6720011f98c8527d9ea8fe4a | 11 | Changes in structure and material properties of the surface-bound layer, such as viscoelasticity, can be inferred from changes in energy dissipation to the environment (∆D). To characterize amyloid elongation, we monitored multiple recurring growth phases of α-synuclein fibrils immobilized on a QCM-D sensor (Fig. A & B). Upon injection a rapid change in frequency and dissipation is observed. This effect originates from the viscosity and density change of solution, as the α-synuclein solution contains 0.5 v/v% ethanol (from addition of PMSF, see Methods). Such a water-ethanol mixture is expected to have an approximately 10 % higher viscosity than pure water . The frequency and dissipation perturbation is reversible when returning to pure buffer conditions (Fig. ). In the incubation period the frequency and dissipation change linearly in time, demonstrating that both dissipation and frequency changes can be used as reporters of fibril growth. Amyloid growth rates are expected to be constant throughout each growth phase, as monomer is not significantly depleted in the cell within the time of the measurements at the monomer concentrations employed . However, during the last growth phase the rate of the frequency response decreases throughout the measurement, while the rate of dissipation change remains constant. The behaviour can be explained by fibrils growing beyond their persistence length and increasingly extend away from the sensor surface where the mass sensitivity decreases. The dissipation response appears to be less affected by this phenomenon compared to the frequency response. |
6720011f98c8527d9ea8fe4a | 12 | To investigate the effect of the fuzzy coat, we utilized a fluid scheme of a single standardised growth phase prior to a shaving step followed by two subsequent growth phases. In all growth phases we inject full length WT α-synuclein. During fibril shaving we observe a fast release of mass, which decays as the bulk of PK-degradable material, the fuzzy coat, becomes increasingly unavailable. The frequency response during monomer incubation, post shaving, is approximately 80% accelerated by frequency and 35% by dissipation and demonstrates the same slow decrease in frequency response as the regular growth experiments. This behavior occurs in both post shaving growth phases, however in the last growth phase the dissipation response is accelerated by 80%. It is not immediately obvious whether the accelerated deposition of mass to the sensor surface originates from monomer binding to the exposed fibril core or accelerated fibril elongation. However, the fibril growth, as measured by the dissipation response, is constant throughout the majority of the measurement. This suggest that the signal is dominated by fibril growth rather than monomer binding, which would be expected to saturate as the shaved fibril surface becomes increasingly covered. In addition, the ratio of dissipation to frequency response during the growth phases is greater than what we find during the shaving phase (approximately -0.3 10 ( -6)/Hz and -0.15 10 ( -6)/Hz respectively), indicating that the material properties of the removed mass by Proteinase K treatment are different from the predominant mass contribution during fibril growth post-shaving. It is reasonable to expect that the material properties of bound monomer would be comparable to that of the fuzzy coat rather than the full fibril, suggesting that monomer binding is not contributing significantly to the response post shaving. |
6720011f98c8527d9ea8fe4a | 13 | To further elucidate if binding of protein to the fibril surface can be distinguished from amyloid elongation, we investigated the characteristic QCM-D response from fibril growth and protein. We exposed α-synuclein amyloid fibrils to 1% Human Serum Protein (HsP) that binds to the fibrils causing a decrease in frequency of -12 Hz, while the dissipation increases only 0.3 10 -6 giving a distinctly different response from fibril growth (Fig. A, the HsP binding is not unspecific binding to the sensor surface, see Fig. ). As a result, when fibrils are exposed to a 1% HsP 50 µM α-synuclein solution, fibril growth can be monitored through the dissipation response, which remains constant throughout the measurement. Meanwhile the frequency response is convoluted by protein binding to the fibrils and fibril growth. |
6720011f98c8527d9ea8fe4a | 14 | Plotting ∆D versus ∆F, which is commonly used to obtain structural information of adsorbed molecules , clearly demonstrates that HsP binding is characterized by a single phase, fibril growth features two phases (injection perturbation and fibril growth) and the mixture of HsP and α-synuclein shows three distinct phases (injection perturbation, HsP binding dominated phase and fibril growth dominated-phase). The fibril growthdominated phase is characterized by a linear correlation between ∆D and ∆F. The ∆D-∆F slope in both conditions are similar (-0.20 10 -6 /Hz ± 0.001 for pure α-synuclein and -0.21 10 -6 /Hz ± 0.0003 for α-synuclein + HsP), showing that the measurement is dominated by the same type of material deposition in these phases. As such we can conclude that the dissipation response of the HsP + α-synuclein incubation, which is constant throughout the entire measurement, reports almost exclusively on the addition of fibrillar material. The ∆D/∆F relationship can be plotted in the time-domain from the numerical derivatives, which provides insight into the duration of each of the phases. From this analysis we find that the pure α-synuclein incubation reports truthfully on fibril growth approx. 200 s after the start in the injection. The HsP + α-synuclein converges to pure fibril growth after 5 minutes of incubation. |
6720011f98c8527d9ea8fe4a | 15 | By comparing the different overtones, we find a frequency overtone dependency of the fibril growth response, while no such dependency is present for HsP binding (Fig. ). This indicates that the dominating source of mass for fibril growth originates from liquid contributions, whereas the HsP binding frequency response is dominated by the dry mass , consistent with the ∆D to ∆F ratio of each mechanism. Our observations agree with earlier findings that the QCM frequency response can be attributed in large parts to water trapped in the fibril matrix . |
6720011f98c8527d9ea8fe4a | 16 | To further demonstrate that monomer binding can be observed during fibril growth, we have measured amyloid elongation of α-synuclein at pH 5.5, where the protein is close to its isoelectric point and surface binding of monomer can be expected, as secondary nucleation is strongly enhanced under these conditions . While no clear bi-phasic behavior is evident in the raw data, the DF-plot reveals a low ∆D to ∆F phase in the early time steps of the pH 5.5 growth phase, which becomes exacerbated after Proteinase K treatment (Fig. ). The time-resolved DF-plots show a lower dissipation response during the first 200 seconds of the incubation period before the DF-response converges to the signature of amyloid elongation. Compared to injections at pH 7.4, where the elongation signature is achieved within 1 minute of incubation, it is evident that some non-fibrillar mass deposition can be identified as a perturbation in the DF-plots(Fig. ). |
6720011f98c8527d9ea8fe4a | 17 | By employing the above described approach we can demonstrate that the material properties of the growth signal following Proteinase K incubation is similar to the non-treated fibrils (Fig. ) and show that the primary mass deposited to the surface is fib-rillar in nature in all growth phases. We find no indication that monomer binding accounts for a significant fraction of the deposited material in any of the measurements as the DF response is effectively continuous between each growth phase (ignoring the first minute of solution volume exchange). Furthermore, we find that the frequency dependence on overtones are comparable between the shaven and non-shaven fibrils, further demonstrating that the nature of the deposited material is similar between the samples (Fig. ). |
6720011f98c8527d9ea8fe4a | 18 | To further understand the role of the fibril surface in the observed behavior and the nature of the Proteinase K modification, we studied fibrils grown over extensive incubation periods. We employed a single standardization growth step followed by several shaving and growth steps, incubating the fibrils for 15 hours with monomer per shaving step. Post shaving, the long fibrils demonstrate a significantly enhanced elongation rate sustained increasingly as the fibrils have grown. During the final growth phase the accelerated growth is sustained for over 10 hours. DFanalysis of the growth phases reveals that, while some structural reorganization of the fibril matrix may occur in the early timepoints of the growth phases, the dominant source of added mass originates from amyloid fibril elongation (Fig. ). It is noticeable that the accelerated growth is additive, suggesting that the source of accelerated growth must be the fibril surface, rather than any structural modification of the fibril end. |
6720011f98c8527d9ea8fe4a | 19 | Each shaving step is maintained until a constant low rate of frequency increase is reached, thereby ensuring that the accessible disordered regions are mostly degraded. The remaining release of mass at low constant rate is likely to originate from fibril shortening from the ends, where the accessibility by Proteinase K is likely to be highest. By investigating the DF plots of Proteinase K incubations, we find that the ∆D to ∆F relationship of the shaving phase converges towards a dissipation insensitive reaction, as the fibrils are grown for longer periods of time prior to shaving (Fig. ). This is consistent with the observation that Proteinase K treatment predominantly removes non-core material, but also slowly shortens the fibrils from the end. For longer fibrils, the Proteinase K treatment is increasingly dominated by the removal of non-core material as the surface-to-end ratio is increased. We find a linear relationship between the frequency response of mass added through fibril elongation and the frequency response of the shaving reaction (Fig. ) in agreement with removal of the entire available fuzzy coat and fibril depolymerization not significantly contributing to the degradation response. The shaven material corresponds to 16.2 ± 2.7 % of the frequency response of the grown fibrils. This, however, does not mean that the Proteinase K necessarily degrades 16 % of the fibril, which would only correspond to approximately 30 % of the fuzzy coat. The frequency response of fibril growth has a significant liquid con-tribution, which is demonstrated by the high ∆D to ∆F ratio and overtone dependency. However, this is not the case for the shaven material, which has a low ∆D to ∆F ratio and no significant overtone divergence (Fig. ). If the liquid contribution of the full fibril accounts for at least 50 % of the frequency response, it follows that the Proteinase K treatment removes at least 65 % of the fuzzy coat: |
6720011f98c8527d9ea8fe4a | 20 | Given that we find little evidence of binding of monomeric protein contributing significantly to the mass deposition in any experiment at pH 7.4, an additive growth acceleration with each shaving step, as well as a linear correlation between the shaven mass and accumulated fibril surface, the accelerated fibril growth observations presented here can only be rationalized by the formation of new fibril ends on the sensor surface. |
6720011f98c8527d9ea8fe4a | 21 | To explain the observed changes in frequency in the QCM-D results, we created kinetic models of several different possible scenarios at the molecular level (Fig. ). Each of these models makes different assumptions on how full length monomers can interact with amyloid fibrils (with and without flanks), and how those interactions translate into changes in growth rate observed in the QCM-D experiments. The null model (Model 0), describes elongation of non-modified amyloid fibrils. Model A is the simplest model, which describes amyloid fibril elongation without the repulsive charges of the fuzzy coat. The growth rate very quickly returns to baseline as recruited monomer reintroduces charges to the fibril ends. In addition to elongation, Model B also allows monomers to bind to the surface when no disordered flanks are present, leading to an acute increase in surfaceassociated mass, and an initial increase in fibril elongation rates. In Model C, surface binding enables secondary nucleation and subsequent growth on the shaven fibril surface. After every proteinase K degradation step new nucleation-viable binding sites become available, which leads to a sustained increase in growth rate in subsequent growth phases. Notably, fibril growth accelerates throughout the growth phase, as new fibrils nucleate. Model D assumes that protofilaments can be proteolytically fragmented by Proteinase K degradation leading to an increase in the number of fibril ends during the shaving period as the newly accessible proto-filament ends are capable of recruiting monomers. |
6720011f98c8527d9ea8fe4a | 22 | The null model demonstrates behavior similar to fibril growth without any shaving phases (Fig. A and Fig. ). Both models A and B predict increased rates of mass addition only for very short periods of time, which is not what is observed. Rather the accelerated growth rates are maintained for many hours. Accelerated mass association to the sensor surface for times extending beyond tens of minutes are only expected in kinetic models, where the formation of new fibril ends occurs (Model C and D). Finally, only model D captures the immediate acceleration in fibril growth rates, which is observed in our experiments, suggesting that the additional fibril ends are made available during the Proteinase K degradation phase. |
6720011f98c8527d9ea8fe4a | 23 | In this work, we have started to elucidate the aggregation behavior of proteinase-treated α-synuclein amyloid fibrils. Using bulk solution measurements we demonstrate several challenges with biochemical analysis of proteinase treated fibrils by traditional methods. Shaven fibrils demonstrate low colloidal stability resulting in extensive flocculation. Furthermore, shaving of the fibrils strongly alters their ThT sensitivity. QCM-D has been demonstrated to be a well-adapted label-free methodology for the study of amyloid fibril growth and has been employed for this purpose for over fifteen years . Here we demonstrate that the QCM-D can discriminate between monomer binding to the fibril surface and fibril elongation, enabling not only measurements of fibril elongation rates in complex media, but provides detailed insights into the molecular nature of deposited material. Utilizing this experimental platform we demonstrate that shaven fibrils are fully capable of recruiting intact monomer. |
6720011f98c8527d9ea8fe4a | 24 | It has been speculated that intrinsically disordered sequence regions can facilitate molecular interactions and recognition through dynamically sampling a large spatial region and therefore increasing the probability of encounter, a process dubbed flycasting . In the case of α-synuclein, it has been proposed that monomers are recruited to the fibril end through interactions with the fuzzy-coat . Our finding that removal of the fuzzy coat does not inhibit the elongation rate of fibrils indicates that the fuzzy coat of α-synuclein fibrils does not play a significant role in recruitment of monomers to the fibril end. |
6720011f98c8527d9ea8fe4a | 25 | The highly negatively charged C-terminal and slightly positively charged N-terminal of α-synuclein fibrils are suspected to decorate the fibril surface as a polymer brush . The grafting of charged polymer brushes to colloids increases their colloidal stability against association . It has been shown that the higher order association of α-synuclein amyloid fibrils strongly depends on pH and ionic strength . Furthermore, electrostatic and steric repulsion between the fibril surface and monomer at neutral pH are suspected to largely prevent monomer-dependent secondary nucleation on the fibril surface . Upon shaving of the fibrils, repulsion between fibrils decreases, which results in a decrease in colloidal stability as demonstrated here. Furthermore, the electrostatic and steric repulsion between the fibril surface and monomer should decrease. However, we find that the primary signal related to mass addition at pH 7.4 originates from fibril elongation, independently of the presence of the fuzzy coat. Monomers remain highly charged and binding to the fibril surface may still be limited by electrostatic repulsion between the monomers. In accordance, we find that at pH 5.5, close to the isoelectric point and where secondary nucleation has been demonstrated to play an important role , removal of the fuzzy coat indeed leads to increased monomer binding. We therefore find no clear evidence that removing the fuzzy coat of the fibril at pH 7.4 enables significant monomer binding to the fibril surface. Furthermore, we do not observe continued acceleration of mass addition rates to the sensor surface during long incubation periods with monomer solutions post-shaving. If the increase in the number of growing fibril ends were caused by monomerdependent secondary nucleation, new fibrils ends should constantly be formed during incubation with monomer and therefore an acceleration, rather than a slowing down of mass deposition would be expected. Instead, it appears that new fibril ends are only created during the incubation with pK. Thus, the experimental results combined with the kinetic models suggest that the proteolytic modification of fibrils can act as a source of secondary growth sites. Monomer-dependent secondary nucleation of α-synuclein is strongly enhanced at mildly acidic pH , restraining efficient amplification of fibrils by secondary pathways to environments such as endosomes and lysosomes . However, our findings, that secondary reactions may occur at the surface of proteolytically modified α-synuclein fibrils even at neutral pH, suggests an alternative pathway for amyloid amplification in the cellular environment. Since the QCM-D instrument is a surfacebased biosensor, the detected accelerated rate of mass addition can only originate from material anchored to the surface. Hence, any secondary fibrils must adhere to the surface of the sensor in order to be detected. As the sensor surface itself has been passivated with a self-assembled monolayer of PEG, the newly formed fibril ends are likely to adhere to the surfaces of the initial population of seed fibrils. It follows that secondary fibrils detected here must effectively be anchored to the hydrophobic core of the shaven fibrils by non-covalent forces without detaching. This observation is similar to findings in Aβ 42, where daughter fibrils formed through secondary nucleation have been observed to adhere to the fibril surface for extended periods of time . |
6720011f98c8527d9ea8fe4a | 26 | It is intriguing to speculate that changes in cohesive properties of α-synuclein fibrils upon proteolytic truncation of the fuzzy coat, such as colloidal stability and formation of strongly adhering secondary daughter fibrils, may play an important role in disease and formation of LBs. Flocculation and higher order assembly of fibrils and further growth by elongation and secondary pathways could indeed form structures similar to LBs, with a dense core and radiating filaments . However, our study does not investigate if LB-like structures form from these conditions. It may be the case that the fuzzy-coat generally could inhibit and slow down the aggregation of proteins and formation of toxic inclusion bodies, and hence be an evolved protective feature. QCM-D proves to be a particularly useful platform for studying the properties of modified fibrils as the flow-cell platform allows for easy removal of modifying enzymes post treatment. The instrument does not rely on reporter-molecules such as fluorescent dyes, avoiding complications of alterations in ThT sensitivity upon modification as observed in this study and can provide insights into the nature of the interactions. It is clear that the QCM-D proves an interesting platform for studying relative binding affinities to other species commonly found in LBs, such as lipids and other proteins. Systematic study of PTMs and fibril binding affinities may allow for a bottom-up approach to the study of LB formation, not easily available by other in vitro platforms. |
6720011f98c8527d9ea8fe4a | 27 | In summary, we have illustrated that the surface-based biosensing technique QCM-D can provide insights into amyloid growth mechanisms of modified fibrils which may be unavailable in bulk measurements due to flocculation, changes in dyesensitivity and persistent modifying agents. We demonstrate that dissipation-frequency (DF)-plots allow the discrimination of addition/removal of mass originating from fibril core and non-core regions as well as surface binding of proteins. Combining our QCM-D results with kinetic modeling, we show that proteolytic truncation of the disordered flanking regions of fibrils does not abolish their ability to elongate through the addition of unmodified monomer. Furthermore, such proteolysis leads to the generation of additional growth-competent ends at neutral pH, where monomer-dependent secondary nucleation is very inefficient, and hence suggests the existence of a new secondary pathway with potential relevance for the proliferation of amyloid pathology in vivo. |
668937bdc9c6a5c07a733145 | 0 | Heteroanionic materials, which contain two or more anionic species, offer compositional and structural flexibility not found in otherwise analogous homoanionic materials . As such, controlling the relative stoichiometries and crystallographic arrangements of the anion species in heteroanionic materials allows their properties to be tuned . This compositional and structural versatility means that heteroanionic materials find applications across a range of critical technologies, including thermoelectrics , photocatalysis , and energy storage . |
668937bdc9c6a5c07a733145 | 1 | The properties of heteroanionic materials depend on their chemical composition; specifically the identities and relative stoichiometries of their constituent anions, and on their structure, particularly the arrangement of these anions within their host crystal structure. While some heteroanionic materials are crystallographically ordered, with their constituent anion species arranged in a regular, repeating pattern, others are crystallographically disordered, with their anion species randomly distributed across crystallographically equivalent sites. In these anion-disordered systems, at long range, the site occupations of these anion species are uncorrelated. At short range, however, these different anion species often exhibit short-range ordering, characterised by one or more local configurations of anions appearing more frequently than in a fully uncorrelated (maximum entropy) anion distribution. |
668937bdc9c6a5c07a733145 | 2 | While experimental techniques that probe long-range correlations between atoms, such as X-ray or neutron Bragg scattering, can determine the average crystal structure of anion-disordered materials, these methods cannot resolve any short-range ordering, if present. Instead, these methods yield only an effective unit cell where each anion site is occupied by a statistical average of the constituent anion species. Short-range structural information can be obtained from scattering experiments in the form of pair-distribution function (PDF) data . However, for heteroanionic materials containing anions with similar X-ray or neutron scattering factors, such as oxyfluorides, it is often not possible to assign anion site occupations based solely on PDF data, and alternative methods must be used to resolve the anionic structure of these materials. |
668937bdc9c6a5c07a733145 | 3 | One method that has proven effective for studying the short-range structure of heteroanionic materials is solidstate nuclear magnetic resonance (NMR) spectroscopy, which provides direct information about the local chemical environments of individual chemical species. In the case of oxyfluorides, the use of NMR spectroscopy is facilitated by the high gyromagnetic ratio and broad chemical shift range of the sole natural isotope of fluorine, 19 F, and several previous studies have used 19 F NMR spectroscopy to study oxygen-fluorine ordering in oxyfluorides . However, using 19 F NMR data alone to unambiguously determine O/F ordering in disordered oxyfluorides can be challenging, due to the large number of possible anion permutations that might need to be considered; as a consequence, complementary experimental or computational data are often required to fully solve the anion structure. |
668937bdc9c6a5c07a733145 | 4 | Another approach to probing short-range order in heteroanionic materials is to use computational electronic structure methods, such as Density Functional Theory (DFT) . By calculating the relative energies of structures with varied anion configurations, low-energy anion structures can be identified directly. The high computational cost of electronic structure methods, however, limits their use to relatively small computational cells and to relatively small numbers of possible anion orderings, making it difficult to fully characterise the anion substructure in partially disordered materials. In these cases, it is necessary to use alternative computational methods that accurately describe correlations in anion site occupations at length scales beyond those typical of electronic structure calculations, and ideally allow for rapid evaluation of possible anion arrangements. |
668937bdc9c6a5c07a733145 | 5 | Here, we report an investigation of the anionic structure in the anion-disordered transition-metal oxyfluoride, cubic (ReO 3 -type) TiOF 2 . Cubic TiOF 2 has previously been studied as a lithium-ion electrode material and as a photocatalyst . The average structure of cubic TiOF 2 consists of corner-sharing Ti[O, F] 6 octahedra within a cubic P m 3m spacegroup (Fig. ). A previous X-ray diffraction study of cubic TiOF 2 found no evidence for anion ordering, and, on this basis, it was suggested that oxygen and fluorine are fully disordered (uncorrelated) over the available Wyckoff 3d sites . Short-range ordering of oxygen and fluorine anions, however, is known in other ReO 3 -type transition-metal oxyfluorides, such as NbO 2 F and TaO 2 F , and it is therefore reasonable to ask whether ReO 3 -type TiOF 2 might also exhibit short-range anion ordering. |
668937bdc9c6a5c07a733145 | 6 | Using a combination of DFT calculations and cluster expansion modelling, we predict strong short-range ordering in ReO 3 -type TiOF 2 , characterised by an absence of collinear O-Ti-O units and a preference for polar cis-TiO 2 F 4 titanium coordination. This polar coordination around titanium allows shorter Ti-O bonds and longer Ti-F bonds relative to the conventional P m 3m structure, which gives increased net bonding relative to trans-TiO 2 F 4 titanium coordination. The preferential cis-TiO 2 F 4 coordination also results in correlated anion disorder , which gives uncorrelated anion siteoccupations at longer distances, in agreement with the average P m 3m structure assigned from long-length-scale diffraction data . |
668937bdc9c6a5c07a733145 | 7 | To validate our computationally predicted structural model, we use a genetic algorithm (GA) to generate structures with partial thermal disorder, which we use as structural models for as-synthesised ReO 3 -type TiOF 2 . We then compute simulated PDF and 19 F NMR data for these GA-predicted structures, and compare these to corresponding experimental PDF and 19 F NMR data. To generate our simulated 19 F NMR spectra, we convert from DFT-calculated magnetic shieldings to (calculated) chemical shifts using an empirical transformation function that we derive by fitting calculated magnetic shielding data for TiF 4 to previously reported experimental 19 F NMR data . For both the PDF and 19 F NMR data, we observe good agreement between our simulated data for the GA-predicted structural model and our experimental data, supporting our computationally-predicted structural model. |
668937bdc9c6a5c07a733145 | 8 | We also perform additional DFT calculations to evaluate how the degree of oxygen/fluorine ordering in ReO 3type TiOF 2 affects its lithium-intercalation properties. We find that increasing anion disorder makes lithium intercalation more favourable by, on average, up to 2 eV. This result suggests that the electrochemical properties of ReO 3 -type TiOF 2 , and potentially other heteroanionic intercalation electrode materials, can be controlled through synthesis protocols designed to produce samples with specific degrees of short-range anion order. |
668937bdc9c6a5c07a733145 | 9 | X-ray powder diffraction analysis was performed using a Rigaku Ultima IV X-ray diffractometer equipped with a Cu Kα radiation source (λ = 1.540 59 Å). Xray total scattering data were collected at the 11-ID-B beamline at the Advanced Photon Source, Argonne National Laboratory, using high energy X-rays (λ = 0.2128 Å) up to a high momentum transfer value, Q max = 18 Å-1 . The raw total scattering data were processed using Fit2D . Pair-distribution function (PDF) data, G(r), were derived by Fourier transformation after eliminating Kapton and background contributions using PDFgetX2 . Refinement of the PDF data was performed using PDFgui , setting the Q damp parameter at 0.04. The refined parameters included the lattice parameter, the scale factor, s ratio -the correction for the low-r to high-r PDF peak ratio due to correlated motion of bonded atoms -and isotropic atomic displacement factors. 19 F solid-state magic angle spinning (MAS) NMR experiments were performed on a Bruker Avance III spectrometer operating at 7.0 T ( 19 F Larmor frequency of 282.2 MHz), using a 1.3 mm CP-MAS probe head. Room temperature 19 F MAS spectra were recorded using a Hahn echo sequence with an interpulse delay equal to one rotor period. The 90 • pulse length was set to 1.55 µs and the recycle delay was set to 20 s. 19 F spectra were referenced to CFCl 3 and fitted using the DMFit software . |
668937bdc9c6a5c07a733145 | 10 | To model the relative energies of competing O/F anion configurations within the ReO 3 -type TiOF 2 structure, we fitted a cluster expansion model to DFT-calculated energies of 65 symmetry inequivalent 2 × 2 × 2 supercells. These 65 supercells were sampled from the complete set of 2664 symmetry inequivalent 2 × 2 × 2 supercells of ReO 3 -type TiOF 2 , which we enumerated using bsym . These DFT calculations were performed using the VASP code , with a plane-wave cutoff energy of 700 eV and a 4 × 4 × 4 Monkhorst-Pack k-point grid. The interactions between core and valence electrons were described using the projector augmented wave method , with cores configurations of [Mg] for Ti, [He] for O, and [He] for F; for Li, all electrons were treated as valence. These calculations used the revised Perdew-Burke-Ernzerhof generalized gradient approximation (GGA) functional (PBEsol) , with a Dudarev +U correction applied to the Ti d states (GGA+U ) . A value of U Ti,d = 4.2 eV was used, as for previous calculations on TiO 2 , Liintercalated TiO 2 , and Ti-deficient hydroxyfluorinated anatase TiO 2 . |
668937bdc9c6a5c07a733145 | 11 | For our cluster expansion model training set, we performed full geometry optimisations, allowing changes to the cell shape and volume as well as internal atomic coordinates. Each geometry optimisation was deemed converged when all atomic forces were smaller than 0.01 eV Å-1 . Our cluster expansion model was fitted us-ing the MAPS component of the ATAT code , which produced a model with 10 non-zero ECIs and a cross-validation score of 0.013 eV per structure . Additional information about the cluster expansion fitting and resulting model are provided in the Supporting Information. |
668937bdc9c6a5c07a733145 | 12 | To validate our structural model, we generated larger TiOF 2 structures (4 × 4 × 4 supercells) using a structureprediction genetic algorithm (GA). Our genetic algorithm used a combination of elitist and proportionate selection, with selection probabilities based on a Boltzmann fitness function and energies of competing configurations calculated using our DFT-derived cluster expansion model. Full details of this genetic algorithm are provided in the Supporting Information. |
668937bdc9c6a5c07a733145 | 13 | Using our GA structure-prediction protocol, we generated four TiOF 2 4×4×4 supercells for validation against our experimental PDF and 19 F NMR data. For each supercell, we initially relaxed atomic positions and cell volume (fixed cell shape) in VASP, using the parameters described above and a 2 × 2 × 2 Monkhorst-Pack k-point grid. For input structures for 19 F NMR spectra calculations, we then performed a full optimisation (atomic positions, cell volume, and cell shape) using CP2K , using the PBE GGA exchange-correlation functional and the DFT-D3 dispersion-correction method of Grimme et al. , which corrects for the overestimation of bondlengths and cell volumes found for typical PBE calculations . The CP2K calculations used Goedecker-Teter-Hutter (GTH) pseudopotentials and TZVP Gaussian basis sets (MOLOPT library), with charge density plane-wave expansion energy cutoff of 720 Ry. |
668937bdc9c6a5c07a733145 | 14 | The 19 F magnetic-shielding tensors for these geometryoptimised TiOF 2 4 × 4 × 4 supercells were calculated using the GIPAW approach within VASP , using the PBE GGA exchange-correlation functional with a 550 eV plane-wave cutoff and a 2×2×2 Monkhorst-Pack k-point grid. The simulated 19 F MAS NMR spectra were constructed from the DFT-calculated 19 F magnetic shielding data using the procedure described in Ref. . For each fluorine atom, a MAS NMR spectrum was simulated, given the relevant experimental values of spin rate (64 kHz) and the magnetic field (7 T). The full MAS NMR spectrum for a given structural model was then obtained by summing the spectra for all the constituent fluorine atoms. The simulated 19 F NMR spectra were computed using the fpNMR package . |
668937bdc9c6a5c07a733145 | 15 | Suitable transformation functions were obtained by fitting linear models for σ iso → δ iso and for σ csa → δ csa using linear least-squares regression between calculated magnetic shielding data TiF 4 and corresponding chemical shift data previously reported . A full discussion of the derivation of these transformation functions is given in Section III D 4. |
668937bdc9c6a5c07a733145 | 16 | The calculation of the 19 F magnetic shielding tensors for TiF 4 was performed using the GIPAW approach within VASP . The TiF 4 input structure for these reference magnetic shielding calculations was obtained from a geometry optimisation performed in VASP , where only atomic coordinates were relaxed, keeping the cell parameters fixed to experimental values). This geometry optimisation used Ti [Ne] and F [He] pseudopotentials and the Perdew-Burke-Ernzerhof (PBE) GGA functional, augmented with the DFT-D3 correction of Grimme et al. to account for dispersion interactions between the isolated columns of corner-linked TiF 6 octahedra that comprise the TiF 4 structure . Both the geometry optimisation calculation and the subsequent 19 F NMR calculation used a plane-wave cutoff of 550 eV and a 1 × 6 × 3 Monkhorst-Pack k-point grid. |
668937bdc9c6a5c07a733145 | 17 | Additional results from calculations of the 19 F magnetic shielding tensors for TiF 4 , performed with optimisation of atomic positions but without the DFT-D3 correction, using CASTEP (to replicate the previous calculations of Murakami et al. ) and VASP, are provided in the Supporting Information. The Supporting Information includes additional details on the effect that different optimisation and relaxation protocols have on the resulting TiF 4 structure. |
668937bdc9c6a5c07a733145 | 18 | Lithium intercalation calculations were performed for three exemplar 4 × 4 × 4 TiOF 2 supercells: one geneticalgorithm-predicted structure, a 4×4×4 supercell special quasi-random structure, and a 4 × 4 × 4 expansion of the the DFT-predicted lowest-energy 2 × 2 × 2 structure. For each structure, we considered lithium intercalation at all non-equivalent cubic interstitial sites, and performed geometry relaxations with lattice parameters fixed to those of the corresponding stoichiometric TiOF 2 model. These calculations used a cutoff energy of 500 eV and a 2×2 ×2 Monkhorst-Pack k-point grid. To calculate lithium intercalation energies, elemental (metallic) lithium was modelled using a Li 2 cell, with a cutoff energy of 500 eV and a 16 × 16 × 16 Monkhorst-Pack k-point grid. |
668937bdc9c6a5c07a733145 | 19 | Our X-ray diffraction data index to a P m 3m structure (Fig. ) with a cell parameter of a 0 = (3.8076 ± 0.0001) Å, consistent with the value of a 0 = (3.798 ± 0.005) Å reported by Vorres and Donohue . This result corresponds to a ReO 3 -type structural model, com- |
668937bdc9c6a5c07a733145 | 20 | Given the different formal charges of O 2-and F -, these anions are expected to exhibit differentiated bonding with Ti, resulting in distinct Ti--O and Ti--F bondlengths. In an anion-ordered system, differences in Ti-O and Ti-F bond lengths should theoretically be observable in the long-range diffraction data as a reduction in crystal symmetry from P m 3m. However, the absence of any observable deviation from P m 3m symmetry in our X-ray diffraction data indicates that at long-ranges, the positions of oxygen and fluorine are uncorrelated, giving an average high-symmetry P m 3m structure. This observation aligns with the previous study of Vorres and Donohue , wherein the absence of long-range O/F correlations in ReO 3 -type TiOF 2 was interpreted as evidence that O and F are randomly distributed across the available Wyckoff 3d positions. |
668937bdc9c6a5c07a733145 | 21 | To better understand the anionic substructure of ReO 3 -type TiOF 2 , we consider the pair-distribution function obtained from X-ray total scattering data. For interatomic distances between 8 Å and 40 Å, the PDF data are well described by a cubic P m 3m model (R w = 13.2 % (Fig. ), in agreement with the X-ray diffraction analysis above. The experimental PDF for r < 8 Å, however, gives a poor fit (R w = 31.2 %) when modelled with a cubic ReO 3 -type (P m 3m) structure (Fig. ), indicating deviations from the average ReO 3 -type structure at short range. Notably, we observe apparent splittings in the nearest-neighbour Ti-X peak at ∼ 1.9 Å and in the next-nearest-neighbour peak at ∼ 3.8 Å, suggesting distinct bonding environments. Based on the expectation that Ti-O bonding will, in general, be stronger than Ti-F bonding , we preliminarily assign the peaks at 1.71 Å and 1.94 Å to Ti-O and Ti-F nearest-neighbour pairs, respectively, and the peaks at 19 F MAS (64 kHz) NMR spectrum (Fig. ). Averaging over lines 1 and 2 gives a weighted average for bridging Ti-F-Ti environments ⟨δiso⟩ = 17.6 ppm. |
668937bdc9c6a5c07a733145 | 22 | B. 19 F NMR Fig. shows the 19 F MAS solid-state NMR spectrum for ReO 3 -type TiOF 2 , which provides additional information about the local environments of the F -anions. The spectrum shows a broad, slightly asymmetric main feature. We have reconstructed the experimental spectrum using two resonances (lines 1 and 2), which we assign to bridging Ti-F-Ti fluorine atoms. Additionally, our reconstruction reveals a broader and less intense line (line 3) at δ iso ≈ 170 ppm, which we attribute to Ti-F-□ "non-bridging" fluorine atoms, where one adjacent titanium site is vacant . |
668937bdc9c6a5c07a733145 | 23 | The asymmetry and width of the main peak in the 19 F NMR spectrum suggest that our ReO 3 -type TiOF 2 sample contains multiple distinct fluoride-ion environments. Previous studies have reported that TiOF 2 prepared by aqueous solution synthesis contains hydroxyl defects and metal vacancies . The relative intensities of the fitted 19 F NMR resonances (Table ) indicate that only 2.2 % of F -ions are "non-bridging" in our sample, implying a relatively high stoichiometric purity, with a Ti vacancy concentration of ≲ 0.5 % . This low Ti-vacancy concentration is insufficient to explain the asymmetry and breadth of the main peak in the 19 F NMR data. Instead, we interpret these features as indicative of O/F disorder, which is expected to produce a range of local fluoride-ion environments and a corresponding distribution of Ti-F bond lengths. |
668937bdc9c6a5c07a733145 | 24 | To further explore the nature of O/F disorder in ReO 3type TiOF 2 , we conducted a computational analysis of all possible 2 × 2 × 2 TiOF 2 supercells (Ti 8 O 8 F 16 ), consisting of 2664 distinct symmetry inequivalent O/F configurations. To efficiently compute the relative energies of all 2664 structures, we first performed DFT calculations on a subset of 65 structures and used these results to fit a cluster-expansion model, as described in the Methods section. This cluster-expansion model was then used to calculate the energies of all 2664 2 × 2 × 2 supercells. The resulting configurational density of states for all 2 × 2 × 2 TiOF 2 supercells (Fig. ) reveals an energy difference of 0.94 eV per formula unit between the configurations with the lowest and highest energies. This energy variation between different anion configurations indicates a strong energetic preference for certain short-range anion configurations over others, consistent with short-range ordering. This finding contradicts the previously proposed structural model that O and F positions in ReO 3 -type TiOF 2 are completely uncorrelated , which instead implies equal energies for all 2 × 2 × 2 TiOF 2 cells. |
668937bdc9c6a5c07a733145 | 25 | ReO 3 -type TiOF 2 is chemically and structurally similar to ReO 3 -type NbO 2 F and TaO 2 F. In these materials, it has been proposed that collinear F-M -F units are disfavoured and that oxygen and fluorine anions preferentially adopt short-range orderings that give asymmetric F-M -O units . This coordination asymmetry allows the central cation to shift off-centre to form shorter Ti-O bonds, which has been suggested to increase the overall bonding strength of the M (O/F) 6 unit. By analogy, we might anticipate that in TiOF 2 , collinear O-Ti-O units are disfavoured compared to asymmetric collinear O-Ti-F units. Fig. shows the distribution of energies for all 2 × 2 × 2 supercells, grouped by the number of collinear O-Ti-O units in each structure, out of a maximum of 8 possible for this supercell size. In general, structures with a greater number of collinear O-Ti-O units have higher configurational energies, while the lowest energy structures have no collinear O-Ti-O subunits. This observed correlation supports the hypothesis that in ReO 3type TiOF 2 , oxygen and fluorine preferentially organise to give asymmetric collinear O-Ti-F units. |
668937bdc9c6a5c07a733145 | 26 | Fig. displays the three lowest energy 2 × 2 × 2 TiOF 2 structures, each of which are comprised entirely of cis-Ti-[O 2 F 4 ] sub-units. This local coordination achieves local electroneutrality, in accordance with Pauling's second rule , while also avoiding collinear O-Ti-O subunits. The energy difference between these three low-energy structures is only 12 meV, and all 2×2×2 structures containing only cis-Ti-[O 2 F 4 ] units are within 80 meV of the lowest energy structure. Consequently, our calculations predict that ReO 3 -type TiOF 2 exhibits a preference for polar cis-Ti-[O 2 F 4 ] coordination, but these cis-Ti-[O 2 F 4 ] units expected to adopt a variety of different relative arrangements within the crystal structure, resulting in correlated anion disorder . To further quantify the relationship between the Ti-(O/F) bond lengths and bonding strength, we calculated integrated crystal-orbital Hamilton populations (iCOHPs) for each Ti-(O/F) bond in our DFT dataset, which are plotted against corresponding bond length in Fig. . iCOHP values serve as indicators of bonding strength, with more negative values attributed to stronger and more covalent bonding . Both Ti-O and Ti-F bonds are predicted to become stronger with as the Ti-(O/F) distance decreases. Moreover, both plots of bond strength versus bond length are concave, which indicates that Ti centres with a mix of shorter-than-average and longer-than-average bonds Ti-X bonds have greater net bond strength than Ti centres where all six Ti-X bonds are of equal length. The DFT and cluster-expansion analysis detailed above (Section III C) predicts that ReO 3 -type TiOF 2 exhibits short-range anion order characterised by preferential cis-Ti[O 2 F 4 ] coordination. Our calculations also provide an explanation for this preference: anion configurations that avoid collinear O-Ti-O units allow local distortions from TiX 6 coordination with six equallength Ti-X bonds, resulting in shorter Ti-O (and consequently longer Ti-F) bonds, thereby increasing the net Ti-X bonding. |
668937bdc9c6a5c07a733145 | 27 | The energetic preference for cis-Ti[O 2 F 4 ] short-range ordering implies a ground-state structure with 100 % cis-Ti[O 2 F 4 ] coordination, which is consistent with the lowest energy 2 × 2 × 2 structures shown in Fig. , each of which exhibits 100 % ordered cis-Ti[O 2 F 4 ] units. The calculated configurational density of states (cDOS) (Fig. ), however, does not show a clear energy gap above the lowest energy 2 × 2 × 2 structure, and we instead predict multiple low-energy structures that may be expected to be competitive under synthesis . Consequently, as-synthesised samples of ReO 3 -type TiOF 2 are expected to exhibit some degree of partial disorder, while still demonstrating a general preference for local cis-Ti[O 2 F 4 ] coordination. |
668937bdc9c6a5c07a733145 | 28 | To create structural models that incorporate this partial disorder, we used a genetic algorithm (GA) structureprediction scheme to generate a set of exemplar 4 × 4 × 4 supercells. For each structure-prediction calculation, we initialised the genetic algorithm with a starting population of 40 4×4×4 TiOF 2 supercells, each with random O and F anion configurations. The genetic algorithm used a combination of elitist selection and proportional selection, employing a Boltzmann-weighted fitness function, f i ∝ exp(E i /kT ), to select structures from each generation for seeding the next generation (full details of the GA algorithm are given in the Supporting Information). The energies for each structure considered by the algorithm were calculated using our DFT-derived cluster expansion model. This genetic algorithm structure-prediction scheme is conceptually similar to running multiple concurrent Monte-Carlo-based simulated annealing simulations, where structural motifs associated with low energy configurations at any point can be shared across simulations. The GA-algorithm quickly filters out high-energy, less probable structures to produce a pool of structures with energetically reasonable O/F anion configurations (Fig. ). |
668937bdc9c6a5c07a733145 | 29 | Using this GA procedure, we generated four 4×4×4 supercells, with each selected as the lowest energy structure after 100 GA generations. The resulting 4 × 4 × 4 structures feature no collinear O-Ti-O units and predominantly exhibit cis-TiO 2 F 4 coordination (93.0 %), with some fac-TiO 3 F 3 (3.5 %) and TiOF 5 (3.5 %) coordination (Fig. ). Additional analysis of the TiX 6 coordination geometries in these 4 × 4 × 4 models (see the Supporting Information) shows that the anions show small average deviations from ideal octahedra, while the mean Ti-O and Ti-F distances are significantly different, due to large off-centre displacements of the Ti cations (averaging 0.20 Å). |
668937bdc9c6a5c07a733145 | 30 | Like cubic TiOF 2 , NbO 2 F also adopts an average ReO 3 -type structure with oxygen and fluorine distributed over the Wyckoff 3d positions. NbO 2 F is believed to exhibit short-range ordering somewhat analogous to that predicted here for TiOF 2 , with collinear F-Ti-F units disfavoured . Electron diffraction data for NbO 2 F, however, show hk 1 3 * sheets of diffuse intensity, which has been attributed to anion ordering in one-dimensional strings along each of the three ⟨001⟩ directions . To explain this experimental observation, Brink et al. proposed a structural model for NbO 2 F in which oxygen and fluorine are ordered along ⟨001⟩ strings in repeating [F-O-O-F] sequences, but are uncorrelated between pairs of ⟨001⟩ strings, whether these strings are aligned along the same or different ⟨001⟩ directions . |
668937bdc9c6a5c07a733145 | 31 | Motivated by this evidence for [F-O-O-F] anion ordering along ⟨100⟩ strings in ReO 3 -structured NbO 2 F [97], we next consider whether ReO 3 -type TiOF 2 is predicted to exhibit analogous [O-F-F-O] ordering. To explore this possibility, we performed two sets of calculations. First, we used our GA-structure-prediction scheme with our DFT-derived cluster-expansion model to generate a 6 × 6 × 6 supercell, with this supercell size chosen to accommodate anion orderings with a ×3 unit cell repeat distance. We then analysed the resulting structure to determine the relative prevalence of different ⟨100⟩ orderings. Second, we computed DFT geometry-optimised energies for three sets of TiOF 2 structures with different supercell sizes (2 × 2 × 2, 3 × 3 × 3, and 4 × 4 × 4) and different ⟨100⟩ anion orderings, to determine whether any of these ⟨100⟩ anion orderings is sufficiently energetically favoured to predict general anion ordering. |
668937bdc9c6a5c07a733145 | 32 | The first calculation, using our GA-structureprediction scheme, yielded a 6 × 6 × 6 supercell with the same local coordination preferences as for the GApredicted 4×4×4 supercells. The resulting structure contains no collinear O-Ti-O units, and the TiX 6 coordination octahedra are predominantly cis-TiO 2 F 4 (89.8 %), with small proportions of fac-TiO 3 F 3 (5.1 %) and TiOF 5 (5.1 %). This distribution of TiO x F 6-x coordination octahedra differs significantly from that predicted by mapping the Brink et al. ] anion sequences (23.1 %) is higher than that expected for an equivalent supercell with a fully random arrangement of anions (6.6 %). We attribute this effect to a second-order consequence of the short-range ordering in TiOF 2 , where collinear O-Ti-O units are strongly disfavoured, resulting in (partial) anion correlations at intermediate length scales. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.